Message	Date	From	Subject
3640	2009-10-23 21:58:24	Gary	How Receivers Sound
I'm interested in subjective impressions about how various receivers sound. I have limited experience, having only owned a few rigs. My R2Pro is simply superb. I can hear the smallest nuance on a keyed CW signal, clearly identifying small amounts of hum and even a slight click on the rise or fall -- and am able to tell which edge is which. Distortion, equalization, and various kinds of speech processing are plainly audible on SSB. On very strong signals, I can hear background noise, microphone proximity effects, and other acoustic artifacts. Noise is also handled gracefully, never being intermodulated with actual signals, thus making it fully incoherent and easier for the ear to reject. This is truly high fidelity. My other long-term experience was a Drake TR-4, many years ago, with only a 2.8 kHz filter. I recall that having good fidelity as well. I currently have in my hands a borrowed Elecraft K3 (thanks to Ron, W6WO). This is my first experience with a modern, DSP rig. It's technical capabilities are remarkable, especially the infinitely adjustable filtering, noise reduction, and so forth. I can extract single signals in horrible QRM conditions. But sonically? I have mostly listened to CW signals so far, and it sounds truly bizarre. Even with a wide filter, the character of the CW note is inscrutable. As the filter is sharpened, noise takes on an almost cyclic, whirling, and distracting character. I personally find it quite fatiguing to listen for an extended period (something that cannot be quantified). "Why do you care what the CW note sounds like anyway? It's just on or off." Well, no, it's not, it does have bandwidth, and after years
3642	2009-10-24 04:55:53	Tim	Re: How Receivers Sound
I believe that a huge factor that favors simple receivers sounding good is that the phase delay across the passband is smooth (if not linear). Like you, I'm a CW guy, and my opinions and beliefs are strongly biased by my experience. Many many years ago I picked up the Op-Amp cookbooks and decided to build an insanely narrow audio filter for CW. The result sounded like crap for a number of reasons, but the biggest was that the thing rung horribly in response to band noise. It was only after I built it, that I realized that dialing in an insanely high Q was not always a virtue. The R2PRO filters do not have especially steep skirts. I don't believe this is a compromise but a well made choice. For good phasing you don't want a lot of radical phase shifts anywhere up to and including the skirts of your filter, and because the skirts are relatively gentle (contrasted with an umpteen pole SSB crystal filter or DSP implementation of the same) in the R2PRO that means the phase delay is nice and smooth for a very broad bandwidth. For the excellent phasing in the R2PRO the phase shifts must be incredibly well controlled over the whole passband, and the easiest way to make them well controlled is to have them vary smoothly rather than radically. DSP is not necessarily bad, but most who choose filter configurations seem to emphasize sharp skirts - and insanely sharp skirts are just a matter of tweaking numbers in the DSP which gives the poor sound. I suspect that there's some tweak to the K3 to give it filter characteristics that result in linear phase delays; the skirts won't be so sharp but the sound will be a lot better. I don't know the K3 but I do know the guys who built the K1 and K2 would not design a rig to be purposefully bad on CW. All that said, I don't think it'll ever sound as good as our R2PRO's because the R2PRO doesn't carry around the baggage of a double conversion superhet. I asked Wes a while back as to why GPLA and the other EMRFD tools don't make it easy to plot phase delay. It doesn't seem to be a big priority and I do observe that much of the ham community and press doesn't care about anything except sharp skirts. Right now to see phase delay, I make phase shift plots for filter designs with LtSpice or whatever other Spice variant is handy, and then take the derivative to get to phase delay.
3645	2009-10-24 06:42:35	Russell Shaw	Re: How Receivers Sound
Tim wrote: > I believe that a huge factor that favors simple receivers sounding good is that the phase delay across the passband is smooth (if not linear). > > Like you, I'm a CW guy, and my opinions and beliefs are strongly biased by my experience. > > Many many years ago I picked up the Op-Amp cookbooks and decided to build an insanely narrow audio filter for CW. The result sounded like crap for a number of reasons, but the biggest was that the thing rung horribly in response to band noise. It was only after I built it, that I realized that dialing in an insanely high Q was not always a virtue. > > The R2PRO filters do not have especially steep skirts. I don't believe this is a compromise but a well made choice. For good phasing you don't want a lot of radical phase shifts anywhere up to and including the skirts of your filter, and because the skirts are relatively gentle (contrasted with an umpteen pole SSB crystal filter or DSP implementation of the same) in the R2PRO that means the phase delay is nice and smooth for a very broad bandwidth. For the excellent phasing in the R2PRO the phase shifts must be incredibly well controlled over the whole passband, and the easiest way to make them well controlled is to have them vary smoothly rather than radically. > > DSP is not necessarily bad, but most who choose filter configurations seem to emphasize sharp skirts - and insanely sharp skirts are just a matter of tweaking numbers in the DSP which gives the poor sound. > > I suspect that there's some tweak to the K3 to give it filter characteristics that result in linear phase delays; the skirts won't be so sharp but the sound will be a lot better. I don't know the K3 but I do know the guys who built the K1 and K2 would not design a rig to be purposefully bad on CW. All that said, I don't think it'll ever sound as good as our R2PRO's because the R2PRO doesn't carry around the baggage of a double conversion superhet. > > I asked Wes a while back as to why GPLA and the other EMRFD tools don't make it easy to plot phase delay. It doesn't seem to be a big priority and I do observe that much of the ham community and press doesn't care about anything except sharp skirts. Right now to see phase delay, I make phase shift plots for filter designs with LtSpice or whatever other Spice variant is handy, and then take the derivative to get to phase delay. > >
3647	2009-10-24 06:57:45	Chris Trask	Re: How Receivers Sound
> > I believe that a huge factor that favors simple receivers > sounding good is that the phase delay across the passband > is smooth (if not linear). > Actually, the quilty party is group delay (degrees/Hz), and it can make a real mess out of audio quality, same as it does for bit error rate in digital communication systems. Group delay equalization is a science unto itself, and you could easily make a career doing nothing more than that. About 40 years ago, a fellow working for COMSAT devised a narrow bandpass cavity filter that had a single transmission zero to either side of the passband while having perfectly flat group delay. We called these things pseudo-elliptical filters, and they could be realized in microstrip and stripline. The trick was that there were two sections that were cross-coupled. You can easily make first- and second-order all-pass sections in active filters. I once made a very narrow-band baseband filter with a quad switched capacitor filter IC using three sections for the filter and the fourth for a group delay equalizer. The engineering manager, an obnoxious kid from Canada, was furious as he had not had the opportunity to take the idea to the company owners and claim it as being his own. Such was life when working for a living. They fired him six months after I left and the comapny failed a year later. Chris Regards, Chris
3656	2009-10-26 02:44:21	Tim	Re: How Receivers Sound

3657	2009-10-26 03:21:06	Alberto I2PHD	Re: How Receivers Sound
Russell Shaw wrote: > Top posting is bad. > Also bottom posting without trimming away all the unneeded stuff is quite bad... :-) 73 Alberto I2PHD
3658	2009-10-26 06:23:27	Chris Trask	Re: How Receivers Sound
> > > , same as it does for bit error rate in digital > > communication systems. Group delay equalization is a science > > unto itself, and you could easily make a career doing nothing > > more than that. > > I don't think you have to make it sound too complicated. To a > CW guy like me and many others here, the real thing we are trying > to avoid is RINGING in the filter. > Which is a direct result of the group delay. > > I think that paying > extreme attention to a few specs at the lossage of the others > is the gotcha in CW receiver or filter design. > You mean such as ignoring group delay entirely and focusing on skirt selectivity. That's what got you into the ringing problem in the first place. > > I don't think you have to make it sound too complicated. > If it wasn't complicated, people would readily include it in their designs. Instead, it's routinely ignored because people just don't want to have to deal with the heavy-duty math that is required to include it in the design. Chris ,----------------------. High Performance Mixers and / What's all this \ Amplifiers for RF Communications / extinct stuff, anyhow? / \ _______,--------------' Chris Trask / N7ZWY _
3659	2009-10-26 08:03:03	Russell Shaw	Re: How Receivers Sound
Alberto I2PHD wrote: > Russell Shaw wrote: >> Top posting is bad. >> > Also bottom posting without trimming away all the unneeded stuff is > quite bad... :-) I got tired of doing that, so i posted the full thing to demonstrate the point.
3660	2009-10-26 11:44:42	Tim	Re: How Receivers Sound
> If it wasn't complicated, people would readily include it in their > designs. Instead, it's routinely ignored because people just don't want to > have to deal with the heavy-duty math that is required to include it in the > design. It's not heavy-duty math. It's trivially easy to plot group delay with Spice etc. And it's trivially easy to hear how bad it sounds if you have group delay varying in the passband. I only academically understood the relations until I saw the group delay for a Chesbyshev filter plotted in EMRFD. Then I knew exactly how to plot what I had been hearing since I was a kid.
3661	2009-10-26 12:00:38	Niels A. Moseley	Re: How Receivers Sound
Tim wrote: >> If it wasn't complicated, people would readily include it in their >> designs. Instead, it's routinely ignored because people just don't want to >> have to deal with the heavy-duty math that is required to include it in the >> design. > > It's not heavy-duty math. It's trivially easy to plot group delay with Spice etc. And it's trivially easy to hear how bad it sounds if you have group delay varying in the passband. > > I only academically understood the relations until I saw the group delay for a Chesbyshev filter plotted in EMRFD. Then I knew exactly how to plot what I had been hearing since I was a kid. Plotting the group delay is the easy part. Designing a receiver with a "good sounding" group delay requires the heavy-duty math, especially if you want to do group delay shaping/equalization. 73, Niels PA1DSP.
3662	2009-10-26 12:38:32	Tim	Re: How Receivers Sound

3663	2009-10-26 12:50:19	w4zcb	Re: How Receivers Sound
The world is not doomed by fear of math to using bad sounding filters. Tim N3QE Good on you Tim, I've made crystal filters in the same manner, have an 18 pole 5.2 MHz one that doesn't ring. And in the DSP world, FIR filters don't ring either. Just stay away from IIR filters. Never try to make things more complicated than they are. W4ZCB
3664	2009-10-26 13:04:10	Alberto I2PHD	Re: How Receivers Sound
Niels A. Moseley wrote: > Plotting the group delay is the easy part. Designing a receiver with a > "good sounding" group delay requires the heavy-duty math, especially if > you want to do group delay shaping/equalization. > Just do it in software, using a FIR filter with symmetric coefficients. It will have a linear phase response, i.e. a constant group delay. That's why an SDR sounds always so much better than a radio with xtal filters.... unless the SDR designer did choose to use IIR filters, but in that case 50 years of jail would not be enough.... :-) 73 Alberto I2PHD [Non-text portions of this message have been removed]
3665	2009-10-26 13:12:20	Niels A. Moseley	Re: How Receivers Sound
Alberto I2PHD wrote: > Niels A. Moseley wrote: >> Plotting the group delay is the easy part. Designing a receiver with a >> "good sounding" group delay requires the heavy-duty math, especially if >> you want to do group delay shaping/equalization. >> > Just do it in software, using a FIR filter with symmetric coefficients. > It will have a linear > phase response, i.e. a constant group delay. > > That's why an SDR sounds always so much better than a radio with xtal > filters.... > unless the SDR designer did choose to use IIR filters, but in that case > 50 years > of jail would not be enough.... :-) There are ways to use IIR filters and still get a perfectly linear phase response. Also, as long as the group delay isn't varying wildly, non-linear phase IIR filters can still sound pretty good. Besides.. a speaker/headphones will impose an IIR response to your audio signal. 73, Niels PA1DSP.
3666	2009-10-26 13:21:07	Chris Trask	Re: How Receivers Sound
> > > If it wasn't complicated, people would readily include it in > > their designs. Instead, it's routinely ignored because people > > just don't want to have to deal with the heavy-duty math that is > > required to include it in the design. > > It's not heavy-duty math. It's trivially easy to plot group delay > with Spice etc. And it's trivially easy to hear how bad it sounds > if you have group delay varying in the passband. > That part is easy. The difficult part is designing the group delay equalizer itself. Many years ago, I resorted to writing a curve-fitting routine that generated a rational polynomial to fit the group delay curve for telecommunications low-pass filters. Then that was inverted and a subsequent routine fitted a first-order equalizer and then as many second-order equalizers as were needed to achieve the required group delay flatness. It wasn't sophisticated enough to include predistortion, so that part had to be thumbnailed. Chris
3667	2009-10-26 13:43:39	Chris Trask	Re: How Receivers Sound
One of the things that I have found lacking in radio design has to do with phase-generated (aka Hartley) SSB modulation and demodulation. Typically, the phase distortion of the SSB generator and demodulator is ignored, and after passing through the transmitter and then the receiver, the voice at the microphone end is unrecognizable at the speaker end due to the phase distortion. I hadn't considered this until designing a phasing type adjacent chennel interference canceller and realized that the music from SWBC stations would suffer. So, I simply added an additional phase shifter after the canceller that had the poles and zeroes reversed. Chris
3669	2009-10-26 17:36:21	Russell Shaw	Re: How Receivers Sound
w4zcb wrote: > The world is not doomed by fear of math to using bad sounding filters. > > Tim N3QE > > Good on you Tim, I've made crystal filters in the same manner, have an > 18 pole 5.2 MHz one that doesn't ring. And in the DSP world, FIR > filters don't ring either. Just stay away from IIR filters. That's news to me;) I'd like to see a "square" FIR LPF that doesn't have a ringing step response. For certain filter specs, IIR filters with little/no ringing can be made with less resources than an equivalent FIR filter. > Never try to make things more complicated than they are.
3671	2009-10-27 06:13:27	k5nwa	Re: How Receivers Sound
At 04:44 AM 10/26/2009, Tim wrote: > > >
3672	2009-10-27 10:42:42	Tim	Re: How Receivers Sound

3676	2009-10-28 10:53:47	boblarkin02	Re: How Receivers Sound
Hi All - Just looking at the filtering issue, there are two topics, the group delay distortion, and the filter amplitude response. I realize these have been tossed around quite a bit on this thread, but some specifics might be helpful. During the writing of EMRFD, I looked at the transient response of filters. Here are my notes from that Matlab simulation: "The first look was the impulse response of two similar LP filters, one minimum-phase (LC Chebychev) and the other a linear-phase 155 Tap FIR (10 kHz sample rate). They both had 3 dB points at 175 Hz, -40 at around 325 Hz and 0.01 to 0.02 dB ripple. Not exactly the same filters, as they never can be, but close enough. The LC had a peak unit-impulse undershoot of about 0.29 and the FIR was about 0.1. The delay for the LC was 4 msec and 7.8 msec for the FIR. It takes about 10 msec for the LC ringing to be down to 0.1, the FIR value. The FIR ringing stops completely at 7.8 msec past the peak. "Conclusion on this one is that the linear-phase response has a substantial improvement in its ringing characteristics." This was with a LP filter, but if you work it through with a BP filter the conclusion would be the same. Also, the term, "linear phase" is synonymous with "constant group delay." The term "minimum phase" comes from the analytical side of electronics, but in hardware terms, essentially means that there is a single path from the input of the filter to the output, like many crystal or LC filters. It also means that if the filter is minimum phase, the phase (or group delay) response is completely determined by the amplitude response. So to improve the transient response, you either compromise the amplitude response, or must leave the world of minimum phase, two examples of which are delay equalizers and FIR filters. And, of course, all of this can be implemented in either analog or DSP circuitry (p3-29 and p10-18). But, as has been pointed out, most FIR filters with flat group delay still ring. They just ring less than some others. An aside is that part of what we call ringing is the real thing on signals. The other part is the sound of noise when the spectral components are (almost) all of a single frequency. The latter can be wearing, but is not really ringing. So, you want zero signal ringing on a narrow filter. Look at the foot note on page 10-19 relative to W8MQW's filter. The coefficients of this FIR filter with a "sin(x)/x" frequency response are a segment of a sine wave, say 10 cycles. They stop abruptly, and so any signal going through the filter also stops abruptly, i.e., no ringing. In keeping with "free lunches," the frequency response has side lobes that are only 13 dB down, it is easy to mis-tune to a side lobe that works poorly (you really need a tuning aid of some sort), and the out of band rejection is poor. That said, if you are working on a frequency that is relatively free of QRM, and your main object is to improve S/N, this filter is the only one that works for my ears and mind. It is close to amazing! I have tried this with other people, and they agree, the darned thing works. And it does reject QRM, just not a brick wall. By the way, it is normally designed to be a matched filter for a dot, and the dashes do OK. This means the length of the sine wave in the FIR coefficients is that of a dot at the intended WPM. Each code speed sort of needs a different filter. There is a related topic, of digital distortion, but the filtering part is tough enough to wrap one's mind around. 73, Bob W7PUA
3677	2009-10-28 13:59:43	Roelof Bakker	Re: How Receivers Sound
Hello all, > So, you want zero signal ringing on a narrow filter. Look at the foot > note on page 10-19 relative to W8MQW's filter. The foot note is in my book and mentiones an article on the CD by the book, which happens not to be on my copy. Can anyone help me out? As I am active at LF with aural copy of weak signals, which has much in common with EME, any help will be much appreciated. One last question: what is the bandwidth of this filter? 73, Roelof, pa0rdt
3678	2009-10-28 21:28:38	boblarkin02	Re: How Receivers Sound
Roelof, the LF application that you are working on should be a fine place to experiment with this filter idea. I have experience with its use at VHF through microwaves, only. I put a plot of the MacCluer non-ringing filter frequency response at http://www.proaxis.com/~boblark/MatchDSP10.gif This is the filter implemented in the DSP-10 and except for sample rates, the one Chuck MacCluer implemented, as well. The 6 dB bandwidth for this one is 47 Hz. I incorrectly stated the design description. After getting myself back up to speed: This is a "matched filter" for a very high speed CW dot. His intention was to give up some S/N for slower CW, in order to prevent the sound from being "mushy." This apparently helps the human mind to separate out dots from dashes from key up. He suggests that one could increase the number of cycles of sine wave from the 8 used here to as many as 20. I have not experimented with this, but somebody should. That would narrow the filter inverse proportionally. A matched filter is defined to be that response which maximizes the peak S/N at the filter output. This is not quite what this is, and so I added "pseudo" to the plot. I note the copyright for the original article is held by the ARRL, so maybe we can make this available. The material that was indicated to be on the CD was the DSP-10 transceiver program. Let us know if this fits into your experiments> 73, Bob W7PUA
3679	2009-10-29 04:31:08	Roelof Bakker	Re: How Receivers Sound
Bob, thank you for explanation. This is an interesting thread. The last 20 years, I have been playing with direct conversion receivers and audio filters and have always been pleasantly surprised by the clear audio. The filters used were passive high pass and low pass designs to synthesize a band pass filter and were derived from a Ham Radio article by Stefan Newiadomsky. Ironically the filters I use for weak signal work at LF are 12 Hz and 6 Hz wide. The 12 Hz filter is still wide enough to pass 12 WPM morse code, but don't sound anywhere near good, as all signals sound the same. The main reason I use them is to dig very weak signals out of the noise. These filters are simple four section analogue designs. The bandwidth of each section is around 5 Hz and I use "stagger tuning" to align them for a wider bandwidth. This is a simple way to tweak the bandwidth at the desired value. These filters do ring of course, but fed with a low level signal (read low noise), this is not really a problem. Keeping the overall audio level low helps a great deal. The only drawback is that one needs a quiet room. It will certainly be possible to implement such a filter in the digital domain. However, I have not seen any readily available yet. My present receiver (PERSEUS SDR) comes close with a minimum bandwidth of 22 Hz, but there is still room for improvement. 73, Roelof, pa0rdt
3680	2009-10-29 07:10:24	Chris Trask	Re: How Receivers Sound
> > Bob, thank you for explanation. This is an interesting thread. > The last 20 years, I have been playing with direct conversion > receivers and audio filters and have always been pleasantly > surprised by the clear audio. The filters used were passive > high pass and low pass designs to synthesize a band pass > filter and were derived from a Ham Radio article by Stefan > Newiadomsky. > A direct conversion (DC) receiver presents an interesting opportunity for addressing the ringing problem being discussed here. In a conversion receiver, the multiple bandpass sections in the IF coupled with any lowpass filtering in the baseband makes the problem of delay equalization a bit tricky, as well as expensive. With the filtering confined to the baseband, one could easily make a multi-section bandpass filter such as you have described using monolithic switched capacitor filters. I had earlier mentioned my experience in making such a filter that had a 2-pole delay equalizer, which had a centre frequency of 9.6kHz. Using one of the National Semiconductor devices, the filter had a 2-pole lowpass, 2-pole highpass, and 2-pole bandpass section. The fourth function in the IC was used to make a 2-pole allpass section for the delay equalization. What's nice about those switched capacitor filter ICs is that you can change the cutoff (or centre) frequencies by adjusting the clock generator. The filter that I made was designed to use a single four-section IC, but it I were to make a more serious one I would probably partition the high-order filter(s) into 2-pole sections, using one section of the IC for the filter section and another for equalizing that section. This would be much simpler than trying to equalize it all at once. Question now would be if such a thing could be implemented in DSP. Chris ,----------------------. High Performance Mixers and / What's all this \ Amplifiers for RF Communications / extinct stuff, anyhow? / \ _______,--------------' Chris Trask / N7ZWY _
3681	2009-10-29 08:01:05	Tom H	Re: How Receivers Sound (FT-817 experiences)
All of the discussion, so far, is centered on non-linear phase-frequency response. I'll add my 2 cents to these much higher-value contributions but on other factors affecting the sound of receivers that I ran into with my FT-817ND. Immediately on using it, I was bothered by crackling distortion on CW and peak distortion on voice in SSB, AM and WBFM modes. CW was the most obvious and that led to the discovery of a manufacturing error affecting many production lots where a resistor value affecting the gain of the 1st audio amp for CW/SSB was 1/3 of the design value. Ultimately, I increased the values of all resistors for all modes to add about 6dB of headroom over the design values because there was still residual distortion that bothered me. That didn't clear it all. I found a couple more causes that I was able to ameliorate through LTspice simulation of the original circuit and modified ones. A weakness in the FT-817 design is that the AGC detector, AM detector and product detector are always connected in parallel to the last IF stage, regardless of the mode in use. And the IF amp is not a low-Z source. The AGC detector diode is an ever-present non-linear load, especially around its conduction threshold, i.e., the AGC threshold. Its input resistor was only 1K, about the same as the IF amp source resistance, so it was causing distortion of the IF signal feeding the modulation detectors. This was reduced with changes in the AGC circuit that allowed raising the input resistor by a factor of 10 and changes in the IF amplifier to lower its source resistance by a factor of 2-3. I couldn't do anything practical with the ever-present AM detector possibly affecting CW/SSB distortion. Ideally, only the detectors in use should be connected to the IF. A subtle cause of distortion was found with the product detector - the beat frequency modulated the amplitude of the IF input - another non-linear load. Reducing the resistor at its input along with the reduced IF amp resistance lowered the overall IF source resistance and the depth of this modulation along with the consequent distortion. AGC distortion of speech persisted and is still not as good as I would like as I may be close to having modified that area of the pcb to within a hair of its life and am wary of going any further. The 817 has two AGC speeds (plus Off which merely disconnects the output of the AGC). AGC control that tracks modulation distorts the modulation so the Slow speed is the one to use with speech. Others have found, as I have, the 817's Slow to be not slow enough and have strapped in a large capacitor in parallel with less-than-expected benefit. The problem is that the Fast time constant is always in circuit with Slow being switched in parallel. The 100-200 ohm resistance of the CMOS switch causes the AGC release to initially follow the fast time constant so you cannot get below a certain level of modulation tracking, no matter how large the Slow cap is. I've written and documented more on the foregoing in the FT817 Yahoo Group for anyone interested. I have only the stock 2.7kHz filter and long to add either the 500Hz or 300 Hz options. I read contradictory opinions about which is better. It sure would be a help to have comparative tests including impulse response and group delay measurement. I daresay there are other subjective variables that differ among operators, e.g., how much of the world do you want to be aware of, how coloured do you mind the noise to be? And then there are lots of other factors, maybe not affecting sound as much but limiting digital mode performance such as phase jitter. My Radio Shack DX-394A receiver sucked as a tuner for an audio DSP demodulator for DRM and synchronous AM until I inserted a buffer amp between the 2nd LO and the 2nd Mixer, as in the B revision. The gain of the mixer when varied, under either manual or AGC, changed the load on the oscillator and pulled its frequency. Hope these comments are of interest to some, 73, Tom VE3MEO
3682	2009-10-29 08:32:10	Sam Morgan	Re: How Receivers Sound (FT-817 experiences)
Tom H wrote: > I'll add my 2 cents to these much higher-value contributions but on > other factors affecting the sound of receivers that I ran into with my > FT-817ND. > snip many mods > Hope these comments are of interest to some, > well I sure appreciate them Tom, they will keep me from ever wasting my funds on a FT-817. If the user has to make that many modes, just to have a useful rig, I'd say it was a piece of junk to begin with. -- GB & 73 K5OAI Sam Morgan
3683	2009-10-29 09:24:52	Tom H	Re: How Receivers Sound (FT-817 experiences)

3684	2009-10-29 09:30:46	Leon	Re: How Receivers Sound (FT-817 experiences)

3685	2009-10-29 10:12:52	Alberto I2PHD	Re: How Receivers Sound
> > The last 20 years, I have been playing with direct conversion > receivers and audio filters and have always been pleasantly > surprised by the clear audio. My suggestion is to find a nearby OM with a Softrock board driving a sound card on a PC with one of the many SDR programs running. You will be amazed by the quality of the audio, despite the steepness of the passband filters. And you can buy the Softrock kit for a few tens of dollars. The various software programs are all free. 73 Alberto I2PHD
3686	2009-10-29 11:16:39	Glen Leinweber	Re: How Receivers Sound
The W8MQW filter that Bob mentions is an interesting one. I believe this is a bandpass FIR filter. Its lowpass counterpart has great interest too... all of its coefficients are "1.0" and it also has ends that sharply cut off (to "0.0"). It has a sin(x)/x response, centered at 0 Hz. It is known as the moving-average filter. Its amplitude settling response has no equal - on a step function it rises to maximum in the least possible time (in exactly the length of the filter) and has no overshoot whatsoever - no ringing. Given a filter length, no other type is as effective at reducing noise amplitude. On the downside, stopband response is poor - many other filter types reject signals much better. However, because of the sin(x)/x response, there ARE many null points in the stopband frequency range where interfering carriers are fully rejected. The first one has a period equal to the filter length It is remarkable that a digital filter so easily coded is so pleasing to the ear. I have coded such filters into simple PIC processors, both lowpass and bandpass: <http://epic.mcmaster.ca/~glen/4a4/filt_det.pdf> This was for detecting signals, not for listening. I wouldn't suggest that the 10-bit A-to-D in a PIC is appropriate for serious audio use.
3687	2009-10-29 12:50:05	Roelof Bakker	Re: How Receivers Sound
Hello Alberto, Thank you for suggesting a Softrock kit. I must confess that one has been lying around here for a year. So now it must be build! 73, Roelof, pa0rdt
3688	2009-10-29 17:30:32	Russell Shaw	Re: How Receivers Sound
Glen Leinweber wrote: > The W8MQW filter that Bob mentions is an interesting > one. I believe this is a bandpass FIR filter. Its lowpass > counterpart has great interest too... all of its coefficients > are "1.0" and it also has ends that sharply cut off > (to "0.0"). It has a sin(x)/x response, centered at 0 Hz. > It is known as the moving-average filter. > Its amplitude settling response has no equal - on a > step function it rises to maximum in the least possible > time (in exactly the length of the filter) The length of time that the step response output is changing is the same for any FIR filter of the same length as long as the taps are non-zero. A 10 tap FIR filter with the first nine taps zero has a step response rise time of one sample, unlike the moving average filter which has a rise time of 10 samples. > and has no > overshoot whatsoever - no ringing. Neither does many other FIR filters such as the latter one. > Given a filter length, no other type is as effective at > reducing noise amplitude. The ratio of secondary lobes to the main f=0 lobe is better for a gaussion FIR filter than the moving average one, therefore having better SNR enhancement. > On the downside, stopband response is poor - many > other filter types reject signals much better. However, > because of the sin(x)/x response, there ARE many > null points in the stopband frequency range where > interfering carriers are fully rejected. The first one has > a period equal to the filter length All FIR filters of the same length have the same nulls at the same frequencies. An advantage of the moving-average filter is that the calculations are simple, but at the cost of less than optimal performance in many cases except where it happens to be the matched filter of choice for the right length pulse waveforms. > It is remarkable that a digital filter so easily coded is so > pleasing to the ear. I have coded such filters into simple > PIC processors, both lowpass and bandpass: > <http://epic.mcmaster.ca/~glen/4a4/filt_det.pdf> > This was for detecting signals, not for listening. I > wouldn't suggest that the 10-bit A-to-D in a PIC > is appropriate for serious audio use. There are many woeful DSP filters out there. They may be implemented well, but the maths is not well understood by the implementors, and so are not well suited for the application. A lot are just copy-and-paste code from the web. It just shows how little lengths one needs to go to get above the average.
3689	2009-10-29 17:57:54	Russell Shaw	Re: How Receivers Sound
Russell Shaw wrote: > Glen Leinweber wrote: ... > All FIR filters of the same length have the same nulls at the > same frequencies. Correction, not true. If eg the last few taps of the moving average filter were zero, then the nulls would be further apart.
3690	2009-10-29 18:15:52	Russell Shaw	Re: How Receivers Sound
Russell Shaw wrote: > Glen Leinweber wrote: ... >> Given a filter length, no other type is as effective at >> reducing noise amplitude. > > The ratio of secondary lobes to the main f=0 lobe is better > for a gaussion FIR filter than the moving average one, therefore > having better SNR enhancement. I should qualify this. For voice with most of the energy increasing towards f=0, lower stop-band lobes would be better. For CW, a moving average filter has higher stop-band lobes, but these capture the energy from the pulse-like CW waveform. This matched-filter operation is more optimal. One could have a "CW" control on their receiver that adjusts the FIR length for "fast" or "slow" morse. >> On the downside, stopband response is poor - many >> other filter types reject signals much better. However, >> because of the sin(x)/x response, there ARE many >> null points in the stopband frequency range where >> interfering carriers are fully rejected. The first one has >> a period equal to the filter length
3691	2009-10-30 08:58:31	c6alk	Re: How Receivers Sound (FT-817 experiences)
Hi Tom and Group: I wonder if these mods were documented in any way. I would be very interested in at least checking if my 817 has any of the described issues, and if so, then fixing them as Tom describes. If there any drawings etc., even
3692	2009-10-30 09:09:54	Roelof Bakker	Re: How Receivers Sound
Hello all, Bob has been so kind to provide MacCluer's article which makes an interesting read, especially the part on aural copy techniques used by experienced EME oparators. It is stated that "the higher tones mask the lower ones and not vica versa". This confirms my findings that everything being equal a filter with a centre frequency of 500 Hz seems to ring less than a filter centered around 600 Hz or higher. Some operators prefer a wide bandwidth (2000 Hz), whilst others prefer a filter bandwidth of 250 Hz. My finding is that on a quiet band a wide bandwidth seems to work best. The human brain can simulate a 50 Hz wide filter and this seems to work best in a wide band of equally distributed noise. The use of very narrow audio filters makes only sense when a bandwidth smaller than 50 Hz is called for. The reason can be twofold: digging weaker signals out of the noise or getting rid of adjacent signals. The first reason will be on hand by weak signal aural EME-work; the second won't happen much in normal amateur CW operation. I have analysed the band occupancy during the CW CQWW contest and it is amazing to find how much frequency space is still unoccupied when looking from the perspective of the use of narrow filters. Under these conditions a wider filter will do and is preferred by many as a lot of stations don't answer your call on exact the same frequency. How your receiver sound will be important to prevent listening fatique working a major contest. I believe that on a good sounding receiver a pile up should have musical qualities and it should be easy to pick out the individual station. At present my main amateur radio occupation is listening to non directional beacons. These operate from 190 kHz to 1740 kHz, but the majority can be found below 450 kHz. The station's identifier is given in slow morse (5 to 10 w.p.m.) and repeated every 5 to 30 seconds. Most NDB's use AM modulation with a keyed tone of 400 or 1020 Hz. Canadian NDB's transmit only the carrier and the upper sideband 400 Hz and French NDB's use a keyed carrier. NDB's are allocated on a 1 kHz grid (In Europe also .5 kHz) and it is easy to imagine that the carrier of one station is only 20 Hz away of a station on the adjacent channel which uses 1020 Hz modulation. NDB's are not high tech devices and often the modulation is derived from a simple not all too stable RC oscillator. As a result carrier and modulation can be only a few Hz apart and that is where the need for a very narrow filter arises. The circuit diagram and a few filter plots can be found here: http://www.ndb.demon.nl/pa0rdt/ The plots have been made by feeding noise from a noise generator into a receiver and then into the filter. The noise generator is from SSD, page 168, fig. 62. Getting enough noise through the filter to create a nice plot takes a little time though: some 20 minutes. To set a bandwidth of 10 Hz, stage 1 and 3 are peaked at 495 Hz and stage 2 and 4 at 505 Hz. This is a bit tricky, but I use a selective level meter and tracking generator, which makes it easy to do. For signals 10 dB above the noise, only the top of the filter comes into play. However the ultimate attenuation is about 60 dB. How does this filter sound? When I used it for the first time, I thought it to be completely useless. However setting the audio level low, gave some results and after a few hours my brain ear system had learned to cope with it. There is one weird phenomenon I won't withheld: when there is no signal in the passband the noise is clearly audible. As soon as the weakest audible signal appears, there is silence as the noise seems to drop away. Only the signal is standing out. And no, I don't use AGC. I hope some of you might find this of interest. 73, Roelof, pa0rdt
3695	2009-10-30 12:01:33	Tom H	Re: How Receivers Sound (FT-817 experiences)
The FT817 distortion, analysis and mods are extensively documented here 0Problems%20-%20Receiver/> . If that link does not come through, then try: http://groups.yahoo.com/group/FT817/files/Audio%20Distortion%2C%20AGC%20\ Problems%20-%20Receiver/ 0Problems%20-%20Receiver/> , else go to: http://groups.yahoo.com/group/FT817/ then > Files , then > Audio Distortion, AGC Problems - Receiver r/> . The contents are reasonably up-to-date but I got hung up for a time with a suspected fault due the accidental use of acid flux that turned out to be something else entirely, moving house, etc. I know the page _refs_re_ft817_rx_distortion.htm 3LwGgMnMtIkrJqg9f349Ozo0peXVNa4N51Yo1E97ApuYDsT5SEs90/Audio%20Distortion\ %2C%20AGC%20Problems%20-%20Receiver/_refs_re_ft817_rx_distortion.htm> is not complete. There are A-B audio samples that attempt to illustrate the distortion. If you can hear the difference, then you should be able to tell whether your rig behaves similarly. Good luck! Tom VE3MEO
3697	2009-10-30 12:11:58	Tom H	Re: How Receivers Sound (FT-817 experiences)
Ouch! Yahoo's Rich Text Format message editor is messing up. Here it is again in plain text: The FT817 distortion, analysis and mods are extensively documented here: http://groups.yahoo.com/group/FT817/files/Audio%20Distortion%2C%20AGC%20Problems%20-%20Receiver/ If that link does not come through, then go to: http://groups.yahoo.com/group/FT817/ then Files then folder Audio Distortion, AGC Problems - Receiver. The contents are reasonably up-to-date but I got hung up for a time with a suspected fault due the accidental use of acid flux that turned out to be something else entirely, moving house, etc. I know the page _refs_re_ft817_rx_distortion.htm is not complete. There are A-B audio samples that attempt to illustrate the distortion. If you can hear the difference, then you should be able to tell whether> your rig behaves similarly. Good luck! Tom VE3MEO >
3710	2009-10-31 10:10:10	c6alk	Re: How Receivers Sound (FT-817 experiences)
Tom: Thanks very much for the info, got it all now. I appreciate you sharing it. Brian K7RE

Message

Date

From

Subject

3640

2009-10-23 21:58:24

Gary

How Receivers Sound

I'm interested in subjective impressions about how various receivers sound. I have limited experience, having only owned a few rigs. My R2Pro is simply superb. I can hear the smallest nuance on a keyed CW signal, clearly identifying small amounts of hum and even a slight click on the rise or fall -- and am able to tell which edge is which. Distortion, equalization, and various kinds of speech processing are plainly audible on SSB. On very strong signals, I can hear background noise, microphone proximity effects, and other acoustic artifacts. Noise is also handled gracefully, never being intermodulated with actual signals, thus making it fully incoherent and easier for the ear to reject. This is truly high fidelity. My other long-term experience was a Drake TR-4, many years ago, with only a 2.8 kHz filter. I recall that having good fidelity as well.

I currently have in my hands a borrowed Elecraft K3 (thanks to Ron, W6WO). This is my first experience with a modern, DSP rig. It's technical capabilities are remarkable, especially the infinitely adjustable filtering, noise reduction, and so forth. I can extract single signals in horrible QRM conditions. But sonically? I have mostly listened to CW signals so far, and it sounds truly bizarre. Even with a wide filter, the character of the CW note is inscrutable. As the filter is sharpened, noise takes on an almost cyclic, whirling, and distracting character. I personally find it quite fatiguing to listen for an extended period (something that cannot be quantified). "Why do you care what the CW note sounds like anyway? It's just on or off." Well, no, it's not, it does have bandwidth, and after years

3642

2009-10-24 04:55:53

Tim

Re: How Receivers Sound

I believe that a huge factor that favors simple receivers sounding good is that the phase delay across the passband is smooth (if not linear).

Like you, I'm a CW guy, and my opinions and beliefs are strongly biased by my experience.

Many many years ago I picked up the Op-Amp cookbooks and decided to build an insanely narrow audio filter for CW. The result sounded like crap for a number of reasons, but the biggest was that the thing rung horribly in response to band noise. It was only after I built it, that I realized that dialing in an insanely high Q was not always a virtue.

The R2PRO filters do not have especially steep skirts. I don't believe this is a compromise but a well made choice. For good phasing you don't want a lot of radical phase shifts anywhere up to and including the skirts of your filter, and because the skirts are relatively gentle (contrasted with an umpteen pole SSB crystal filter or DSP implementation of the same) in the R2PRO that means the phase delay is nice and smooth for a very broad bandwidth. For the excellent phasing in the R2PRO the phase shifts must be incredibly well controlled over the whole passband, and the easiest way to make them well controlled is to have them vary smoothly rather than radically.

DSP is not necessarily bad, but most who choose filter configurations seem to emphasize sharp skirts - and insanely sharp skirts are just a matter of tweaking numbers in the DSP which gives the poor sound.

I suspect that there's some tweak to the K3 to give it filter characteristics that result in linear phase delays; the skirts won't be so sharp but the sound will be a lot better. I don't know the K3 but I do know the guys who built the K1 and K2 would not design a rig to be purposefully bad on CW. All that said, I don't think it'll ever sound as good as our R2PRO's because the R2PRO doesn't carry around the baggage of a double conversion superhet.

I asked Wes a while back as to why GPLA and the other EMRFD tools don't make it easy to plot phase delay. It doesn't seem to be a big priority and I do observe that much of the ham community and press doesn't care about anything except sharp skirts. Right now to see phase delay, I make phase shift plots for filter designs with LtSpice or whatever other Spice variant is handy, and then take the derivative to get to phase delay.

3645

2009-10-24 06:42:35

Russell Shaw

Re: How Receivers Sound

Tim wrote:

> I believe that a huge factor that favors simple receivers sounding good is that the phase delay across the passband is smooth (if not linear).
>
> Like you, I'm a CW guy, and my opinions and beliefs are strongly biased by my experience.
>
> Many many years ago I picked up the Op-Amp cookbooks and decided to build an insanely narrow audio filter for CW. The result sounded like crap for a number of reasons, but the biggest was that the thing rung horribly in response to band noise. It was only after I built it, that I realized that dialing in an insanely high Q was not always a virtue.
>
> The R2PRO filters do not have especially steep skirts. I don't believe this is a compromise but a well made choice. For good phasing you don't want a lot of radical phase shifts anywhere up to and including the skirts of your filter, and because the skirts are relatively gentle (contrasted with an umpteen pole SSB crystal filter or DSP implementation of the same) in the R2PRO that means the phase delay is nice and smooth for a very broad bandwidth. For the excellent phasing in the R2PRO the phase shifts must be incredibly well controlled over the whole passband, and the easiest way to make them well controlled is to have them vary smoothly rather than radically.
>
> DSP is not necessarily bad, but most who choose filter configurations seem to emphasize sharp skirts - and insanely sharp skirts are just a matter of tweaking numbers in the DSP which gives the poor sound.
>
> I suspect that there's some tweak to the K3 to give it filter characteristics that result in linear phase delays; the skirts won't be so sharp but the sound will be a lot better. I don't know the K3 but I do know the guys who built the K1 and K2 would not design a rig to be purposefully bad on CW. All that said, I don't think it'll ever sound as good as our R2PRO's because the R2PRO doesn't carry around the baggage of a double conversion superhet.
>
> I asked Wes a while back as to why GPLA and the other EMRFD tools don't make it easy to plot phase delay. It doesn't seem to be a big priority and I do observe that much of the ham community and press doesn't care about anything except sharp skirts. Right now to see phase delay, I make phase shift plots for filter designs with LtSpice or whatever other Spice variant is handy, and then take the derivative to get to phase delay.
>
>

3647

2009-10-24 06:57:45

Chris Trask

Re: How Receivers Sound

>
> I believe that a huge factor that favors simple receivers
> sounding good is that the phase delay across the passband
> is smooth (if not linear).
>

Actually, the quilty party is group delay (degrees/Hz), and it can make a real mess out of audio quality, same as it does for bit error rate in digital communication systems. Group delay equalization is a science unto itself, and you could easily make a career doing nothing more than that.

About 40 years ago, a fellow working for COMSAT devised a narrow bandpass cavity filter that had a single transmission zero to either side of the passband while having perfectly flat group delay. We called these things pseudo-elliptical filters, and they could be realized in microstrip and stripline. The trick was that there were two sections that were cross-coupled.

You can easily make first- and second-order all-pass sections in active filters. I once made a very narrow-band baseband filter with a quad switched capacitor filter IC using three sections for the filter and the fourth for a group delay equalizer. The engineering manager, an obnoxious kid from Canada, was furious as he had not had the opportunity to take the idea to the company owners and claim it as being his own. Such was life when working for a living. They fired him six months after I left and the comapny failed a year later.

Chris

Regards,
Chris

3656

2009-10-26 02:44:21

Tim

Re: How Receivers Sound

3657

2009-10-26 03:21:06

Alberto I2PHD

Re: How Receivers Sound

Russell Shaw wrote:

> Top posting is bad.
>

Also bottom posting without trimming away all the unneeded stuff is
quite bad... :-)

73 Alberto I2PHD

3658

2009-10-26 06:23:27

Chris Trask

Re: How Receivers Sound

>
> > , same as it does for bit error rate in digital
> > communication systems. Group delay equalization is a science
> > unto itself, and you could easily make a career doing nothing
> > more than that.
>
> I don't think you have to make it sound too complicated. To a
> CW guy like me and many others here, the real thing we are trying
> to avoid is RINGING in the filter.
>

Which is a direct result of the group delay.

>
> I think that paying
> extreme attention to a few specs at the lossage of the others
> is the gotcha in CW receiver or filter design.
>

You mean such as ignoring group delay entirely and focusing on skirt
selectivity. That's what got you into the ringing problem in the first
place.

>
> I don't think you have to make it sound too complicated.
>

If it wasn't complicated, people would readily include it in their
designs. Instead, it's routinely ignored because people just don't want to
have to deal with the heavy-duty math that is required to include it in the
design.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_

3659

2009-10-26 08:03:03

Russell Shaw

Re: How Receivers Sound

Alberto I2PHD wrote:

> Russell Shaw wrote:
>> Top posting is bad.
>>
> Also bottom posting without trimming away all the unneeded stuff is
> quite bad... :-)

I got tired of doing that, so i posted the full thing to
demonstrate the point.

3660

2009-10-26 11:44:42

Tim

Re: How Receivers Sound

> If it wasn't complicated, people would readily include it in their
> designs. Instead, it's routinely ignored because people just don't want to
> have to deal with the heavy-duty math that is required to include it in the
> design.

It's not heavy-duty math. It's trivially easy to plot group delay with Spice etc. And it's trivially easy to hear how bad it sounds if you have group delay varying in the passband.

I only academically understood the relations until I saw the group delay for a Chesbyshev filter plotted in EMRFD. Then I knew exactly how to plot what I had been hearing since I was a kid.

3661

2009-10-26 12:00:38

Niels A. Moseley

Re: How Receivers Sound

Tim wrote:

>> If it wasn't complicated, people would readily include it in their
>> designs. Instead, it's routinely ignored because people just don't want to
>> have to deal with the heavy-duty math that is required to include it in the
>> design.
>
> It's not heavy-duty math. It's trivially easy to plot group delay with Spice etc. And it's trivially easy to hear how bad it sounds if you have group delay varying in the passband.
>
> I only academically understood the relations until I saw the group delay for a Chesbyshev filter plotted in EMRFD. Then I knew exactly how to plot what I had been hearing since I was a kid.

Plotting the group delay is the easy part. Designing a receiver with a
"good sounding" group delay requires the heavy-duty math, especially if
you want to do group delay shaping/equalization.

73,
Niels PA1DSP.

3662

2009-10-26 12:38:32

Tim

Re: How Receivers Sound

3663

2009-10-26 12:50:19

w4zcb

Re: How Receivers Sound

The world is not doomed by fear of math to using bad sounding filters.

Tim N3QE

Good on you Tim, I've made crystal filters in the same manner, have an
18 pole 5.2 MHz one that doesn't ring. And in the DSP world, FIR
filters don't ring either. Just stay away from IIR filters.

Never try to make things more complicated than they are.

W4ZCB

3664

2009-10-26 13:04:10

Alberto I2PHD

Re: How Receivers Sound

Niels A. Moseley wrote:

> Plotting the group delay is the easy part. Designing a receiver with a
> "good sounding" group delay requires the heavy-duty math, especially if
> you want to do group delay shaping/equalization.
>

Just do it in software, using a FIR filter with symmetric coefficients.
It will have a linear
phase response, i.e. a constant group delay.

That's why an SDR sounds always so much better than a radio with xtal
filters....
unless the SDR designer did choose to use IIR filters, but in that case
50 years
of jail would not be enough.... :-)

73 Alberto I2PHD

[Non-text portions of this message have been removed]

3665

2009-10-26 13:12:20

Niels A. Moseley

Re: How Receivers Sound

Alberto I2PHD wrote:

> Niels A. Moseley wrote:
>> Plotting the group delay is the easy part. Designing a receiver with a
>> "good sounding" group delay requires the heavy-duty math, especially if
>> you want to do group delay shaping/equalization.
>>
> Just do it in software, using a FIR filter with symmetric coefficients.
> It will have a linear
> phase response, i.e. a constant group delay.
>
> That's why an SDR sounds always so much better than a radio with xtal
> filters....
> unless the SDR designer did choose to use IIR filters, but in that case
> 50 years
> of jail would not be enough.... :-)

There are ways to use IIR filters and still get a perfectly linear phase
response. Also, as long as the group delay isn't varying wildly,
non-linear phase IIR filters can still sound pretty good.

Besides.. a speaker/headphones will impose an IIR response to your audio
signal.

73,
Niels PA1DSP.

3666

2009-10-26 13:21:07

Chris Trask

Re: How Receivers Sound

>
> > If it wasn't complicated, people would readily include it in
> > their designs. Instead, it's routinely ignored because people
> > just don't want to have to deal with the heavy-duty math that is
> > required to include it in the design.
>
> It's not heavy-duty math. It's trivially easy to plot group delay
> with Spice etc. And it's trivially easy to hear how bad it sounds
> if you have group delay varying in the passband.
>

That part is easy. The difficult part is designing the group delay equalizer itself. Many years ago, I resorted to writing a curve-fitting routine that generated a rational polynomial to fit the group delay curve for telecommunications low-pass filters. Then that was inverted and a subsequent routine fitted a first-order equalizer and then as many second-order equalizers as were needed to achieve the required group delay flatness. It wasn't sophisticated enough to include predistortion, so that part had to be thumbnailed.

Chris

3667

2009-10-26 13:43:39

Chris Trask

Re: How Receivers Sound

One of the things that I have found lacking in radio design has to do with phase-generated (aka Hartley) SSB modulation and demodulation. Typically, the phase distortion of the SSB generator and demodulator is ignored, and after passing through the transmitter and then the receiver, the voice at the microphone end is unrecognizable at the speaker end due to the phase distortion.

I hadn't considered this until designing a phasing type adjacent chennel interference canceller and realized that the music from SWBC stations would suffer. So, I simply added an additional phase shifter after the canceller that had the poles and zeroes reversed.

Chris

3669

2009-10-26 17:36:21

Russell Shaw

Re: How Receivers Sound

w4zcb wrote:

> The world is not doomed by fear of math to using bad sounding filters.
>
> Tim N3QE
>
> Good on you Tim, I've made crystal filters in the same manner, have an
> 18 pole 5.2 MHz one that doesn't ring. And in the DSP world, FIR
> filters don't ring either. Just stay away from IIR filters.

That's news to me;)

I'd like to see a "square" FIR LPF that doesn't have a ringing
step response.

For certain filter specs, IIR filters with little/no ringing
can be made with less resources than an equivalent FIR filter.

> Never try to make things more complicated than they are.

3671

2009-10-27 06:13:27

k5nwa

Re: How Receivers Sound

At 04:44 AM 10/26/2009, Tim wrote:

>
>
>

3672

2009-10-27 10:42:42

Tim

Re: How Receivers Sound

3676

2009-10-28 10:53:47

boblarkin02

Re: How Receivers Sound

Hi All - Just looking at the filtering issue, there are two topics, the group delay distortion, and the filter amplitude response. I realize these have been tossed around quite a bit on this thread, but some specifics might be helpful.

During the writing of EMRFD, I looked at the transient response of filters. Here are my notes from that Matlab simulation:

"The first look was the impulse response of two similar LP filters, one minimum-phase (LC Chebychev) and the other a linear-phase 155 Tap FIR (10 kHz sample rate). They both had 3 dB points at 175 Hz, -40 at around 325 Hz and 0.01 to 0.02 dB ripple. Not exactly the same filters, as they never can be, but close enough. The LC had a peak unit-impulse undershoot of about 0.29 and the FIR was about 0.1. The delay for the LC was 4 msec and 7.8 msec for the FIR. It takes about 10 msec for the LC ringing to be down to 0.1, the FIR value. The FIR ringing stops completely at 7.8 msec past the peak.

"Conclusion on this one is that the linear-phase response has a substantial improvement in its ringing characteristics."

This was with a LP filter, but if you work it through with a BP filter the conclusion would be the same. Also, the term, "linear phase" is synonymous with "constant group delay."

The term "minimum phase" comes from the analytical side of electronics, but in hardware terms, essentially means that there is a single path from the input of the filter to the output, like many crystal or LC filters. It also means that if the filter is minimum phase, the phase (or group delay) response is completely determined by the amplitude response.

So to improve the transient response, you either compromise the amplitude response, or must leave the world of minimum phase, two examples of which are delay equalizers and FIR filters. And, of course, all of this can be implemented in either analog or DSP circuitry (p3-29 and p10-18).

But, as has been pointed out, most FIR filters with flat group delay still ring. They just ring less than some others. An aside is that part of what we call ringing is the real thing on signals. The other part is the sound of noise when the spectral components are (almost) all of a single frequency. The latter can be wearing, but is not really ringing.

So, you want zero signal ringing on a narrow filter. Look at the foot note on page 10-19 relative to W8MQW's filter. The coefficients of this FIR filter with a "sin(x)/x" frequency response are a segment of a sine wave, say 10 cycles. They stop abruptly, and so any signal going through the filter also stops abruptly, i.e., no ringing. In keeping with "free lunches," the frequency response has side lobes that are only 13 dB down, it is easy to mis-tune to a side lobe that works poorly (you really need a tuning aid of some sort), and the out of band rejection is poor. That said, if you are working on a frequency that is relatively free of QRM, and your main object is to improve S/N, this filter is the only one that works for my ears and mind. It is close to amazing! I have tried this with other people, and they agree, the darned thing works. And it does reject QRM, just not a brick wall.

By the way, it is normally designed to be a matched filter for a dot, and the dashes do OK. This means the length of the sine wave in the FIR coefficients is that of a dot at the intended WPM. Each code speed sort of needs a different filter.

There is a related topic, of digital distortion, but the filtering part is tough enough to wrap one's mind around.

73, Bob W7PUA

3677

2009-10-28 13:59:43

Roelof Bakker

Re: How Receivers Sound

Hello all,

> So, you want zero signal ringing on a narrow filter. Look at the foot
> note on page 10-19 relative to W8MQW's filter.

The foot note is in my book and mentiones an article on the CD by the
book, which happens not to be on my copy.
Can anyone help me out?

As I am active at LF with aural copy of weak signals, which has much in
common with EME, any help will be much appreciated.

One last question: what is the bandwidth of this filter?

73,
Roelof, pa0rdt

3678

2009-10-28 21:28:38

boblarkin02

Re: How Receivers Sound

Roelof, the LF application that you are working on should be a fine place to experiment with this filter idea. I have experience with its use at VHF through microwaves, only.

I put a plot of the MacCluer non-ringing filter frequency response at http://www.proaxis.com/~boblark/MatchDSP10.gif
This is the filter implemented in the DSP-10 and except for sample rates, the one Chuck MacCluer implemented, as well. The 6 dB bandwidth for this one is 47 Hz.

I incorrectly stated the design description. After getting myself back up to speed: This is a "matched filter" for a very high speed CW dot. His intention was to give up some S/N for slower CW, in order to prevent the sound from being "mushy." This apparently helps the human mind to separate out dots from dashes from key up. He suggests that one could increase the number of cycles of sine wave from the 8 used here to as many as 20. I have not experimented with this, but somebody should. That would narrow the filter inverse proportionally.

A matched filter is defined to be that response which maximizes the peak S/N at the filter output. This is not quite what this is, and so I added "pseudo" to the plot.

I note the copyright for the original article is held by the ARRL, so maybe we can make this available. The material that was indicated to be on the CD was the DSP-10 transceiver program.

Let us know if this fits into your experiments>

73, Bob W7PUA

3679

2009-10-29 04:31:08

Roelof Bakker

Re: How Receivers Sound

Bob, thank you for explanation. This is an interesting thread.
The last 20 years, I have been playing with direct conversion receivers
and audio filters and have always been pleasantly surprised by the clear
audio. The filters used were passive high pass and low pass designs to
synthesize a band pass filter and were derived from a Ham Radio article by
Stefan Newiadomsky.

Ironically the filters I use for weak signal work at LF are 12 Hz and 6 Hz
wide. The 12 Hz filter is still wide enough to pass 12 WPM morse code, but
don't sound anywhere near good, as all signals sound the same. The main
reason I use them is to dig very weak signals out of the noise. These
filters are simple four section analogue designs. The bandwidth of each
section is around 5 Hz and I use "stagger tuning" to align them for a
wider bandwidth. This is a simple way to tweak the bandwidth at the
desired value.

These filters do ring of course, but fed with a low level signal (read low
noise), this is not really a problem.
Keeping the overall audio level low helps a great deal. The only drawback
is that one needs a quiet room.

It will certainly be possible to implement such a filter in the digital
domain. However, I have not seen any readily available yet. My present
receiver (PERSEUS SDR) comes close with a minimum bandwidth of 22 Hz, but
there is still room for improvement.

73,
Roelof, pa0rdt

3680

2009-10-29 07:10:24

Chris Trask

Re: How Receivers Sound

>
> Bob, thank you for explanation. This is an interesting thread.
> The last 20 years, I have been playing with direct conversion
> receivers and audio filters and have always been pleasantly
> surprised by the clear audio. The filters used were passive
> high pass and low pass designs to synthesize a band pass
> filter and were derived from a Ham Radio article by Stefan
> Newiadomsky.
>

A direct conversion (DC) receiver presents an interesting opportunity for addressing the ringing problem being discussed here. In a conversion receiver, the multiple bandpass sections in the IF coupled with any lowpass filtering in the baseband makes the problem of delay equalization a bit tricky, as well as expensive.

With the filtering confined to the baseband, one could easily make a multi-section bandpass filter such as you have described using monolithic switched capacitor filters. I had earlier mentioned my experience in making such a filter that had a 2-pole delay equalizer, which had a centre frequency of 9.6kHz. Using one of the National Semiconductor devices, the filter had a 2-pole lowpass, 2-pole highpass, and 2-pole bandpass section. The fourth function in the IC was used to make a 2-pole allpass section for the delay equalization.

What's nice about those switched capacitor filter ICs is that you can change the cutoff (or centre) frequencies by adjusting the clock generator. The filter that I made was designed to use a single four-section IC, but it I were to make a more serious one I would probably partition the high-order filter(s) into 2-pole sections, using one section of the IC for the filter section and another for equalizing that section. This would be much simpler than trying to equalize it all at once.

Question now would be if such a thing could be implemented in DSP.

Chris

,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_

3681

2009-10-29 08:01:05

Tom H

Re: How Receivers Sound (FT-817 experiences)

All of the discussion, so far, is centered on non-linear phase-frequency response. I'll add my 2 cents to these much higher-value contributions but on other factors affecting the sound of receivers that I ran into with my FT-817ND.

Immediately on using it, I was bothered by crackling distortion on CW and peak distortion on voice in SSB, AM and WBFM modes. CW was the most obvious and that led to the discovery of a manufacturing error affecting many production lots where a resistor value affecting the gain of the 1st audio amp for CW/SSB was 1/3 of the design value. Ultimately, I increased the values of all resistors for all modes to add about 6dB of headroom over the design values because there was still residual distortion that bothered me.

That didn't clear it all. I found a couple more causes that I was able to ameliorate through LTspice simulation of the original circuit and modified ones. A weakness in the FT-817 design is that the AGC detector, AM detector and product detector are always connected in parallel to the last IF stage, regardless of the mode in use. And the IF amp is not a low-Z source.

The AGC detector diode is an ever-present non-linear load, especially around its conduction threshold, i.e., the AGC threshold. Its input resistor was only 1K, about the same as the IF amp source resistance, so it was causing distortion of the IF signal feeding the modulation detectors. This was reduced with changes in the AGC circuit that allowed raising the input resistor by a factor of 10 and changes in the IF amplifier to lower its source resistance by a factor of 2-3.

I couldn't do anything practical with the ever-present AM detector possibly affecting CW/SSB distortion. Ideally, only the detectors in use should be connected to the IF.

A subtle cause of distortion was found with the product detector - the beat frequency modulated the amplitude of the IF input - another non-linear load. Reducing the resistor at its input along with the reduced IF amp resistance lowered the overall IF source resistance and the depth of this modulation along with the consequent distortion.

AGC distortion of speech persisted and is still not as good as I would like as I may be close to having modified that area of the pcb to within a hair of its life and am wary of going any further. The 817 has two AGC speeds (plus Off which merely disconnects the output of the AGC). AGC control that tracks modulation distorts the modulation so the Slow speed is the one to use with speech. Others have found, as I have, the 817's Slow to be not slow enough and have strapped in a large capacitor in parallel with less-than-expected benefit. The problem is that the Fast time constant is always in circuit with Slow being switched in parallel. The 100-200 ohm resistance of the CMOS switch causes the AGC release to initially follow the fast time constant so you cannot get below a certain level of modulation tracking, no matter how large the Slow cap is.

I've written and documented more on the foregoing in the FT817 Yahoo Group for anyone interested.

I have only the stock 2.7kHz filter and long to add either the 500Hz or 300 Hz options. I read contradictory opinions about which is better. It sure would be a help to have comparative tests including impulse response and group delay measurement. I daresay there are other subjective variables that differ among operators, e.g., how much of the world do you want to be aware of, how coloured do you mind the noise to be?

And then there are lots of other factors, maybe not affecting sound as much but limiting digital mode performance such as phase jitter. My Radio Shack DX-394A receiver sucked as a tuner for an audio DSP demodulator for DRM and synchronous AM until I inserted a buffer amp between the 2nd LO and the 2nd Mixer, as in the B revision. The gain of the mixer when varied, under either manual or AGC, changed the load on the oscillator and pulled its frequency.

Hope these comments are of interest to some,

73, Tom VE3MEO

3682

2009-10-29 08:32:10

Sam Morgan

Re: How Receivers Sound (FT-817 experiences)

Tom H wrote:

> I'll add my 2 cents to these much higher-value contributions but on
> other factors affecting the sound of receivers that I ran into with my
> FT-817ND.
>

snip many mods

> Hope these comments are of interest to some,
>

well I sure appreciate them Tom, they will keep me from ever wasting my funds on
a FT-817. If the user has to make that many modes, just to have a useful rig,
I'd say it was a piece of junk to begin with.

--
GB & 73
K5OAI
Sam Morgan

3683

2009-10-29 09:24:52

Tom H

Re: How Receivers Sound (FT-817 experiences)

3684

2009-10-29 09:30:46

Leon

Re: How Receivers Sound (FT-817 experiences)

3685

2009-10-29 10:12:52

Alberto I2PHD

Re: How Receivers Sound

>
> The last 20 years, I have been playing with direct conversion
> receivers and audio filters and have always been pleasantly
> surprised by the clear audio.

My suggestion is to find a nearby OM with a Softrock board
driving a sound card on a PC with one of the many SDR programs
running. You will be amazed by the quality of the audio, despite
the steepness of the passband filters. And you can buy the
Softrock kit for a few tens of dollars. The various software
programs are all free.

73 Alberto I2PHD

3686

2009-10-29 11:16:39

Glen Leinweber

Re: How Receivers Sound

The W8MQW filter that Bob mentions is an interesting
one. I believe this is a bandpass FIR filter. Its lowpass
counterpart has great interest too... all of its coefficients
are "1.0" and it also has ends that sharply cut off
(to "0.0"). It has a sin(x)/x response, centered at 0 Hz.
It is known as the moving-average filter.
Its amplitude settling response has no equal - on a
step function it rises to maximum in the least possible
time (in exactly the length of the filter) and has no
overshoot whatsoever - no ringing.
Given a filter length, no other type is as effective at
reducing noise amplitude.
On the downside, stopband response is poor - many
other filter types reject signals much better. However,
because of the sin(x)/x response, there ARE many
null points in the stopband frequency range where
interfering carriers are fully rejected. The first one has
a period equal to the filter length

It is remarkable that a digital filter so easily coded is so
pleasing to the ear. I have coded such filters into simple
PIC processors, both lowpass and bandpass:
<http://epic.mcmaster.ca/~glen/4a4/filt_det.pdf>
This was for detecting signals, not for listening. I
wouldn't suggest that the 10-bit A-to-D in a PIC
is appropriate for serious audio use.

3687

2009-10-29 12:50:05

Roelof Bakker

Re: How Receivers Sound

Hello Alberto,

Thank you for suggesting a Softrock kit.
I must confess that one has been lying around here for a year.
So now it must be build!

73,
Roelof, pa0rdt

3688

2009-10-29 17:30:32

Russell Shaw

Re: How Receivers Sound

Glen Leinweber wrote:

> The W8MQW filter that Bob mentions is an interesting
> one. I believe this is a bandpass FIR filter. Its lowpass
> counterpart has great interest too... all of its coefficients
> are "1.0" and it also has ends that sharply cut off
> (to "0.0"). It has a sin(x)/x response, centered at 0 Hz.
> It is known as the moving-average filter.
> Its amplitude settling response has no equal - on a
> step function it rises to maximum in the least possible
> time (in exactly the length of the filter)

The length of time that the step response output is changing
is the same for *any* FIR filter of the same length as long
as the taps are non-zero.

A 10 tap FIR filter with the first nine taps zero has
a step response rise time of *one* sample, unlike the
moving average filter which has a rise time of 10 samples.

> and has no
> overshoot whatsoever - no ringing.

Neither does many other FIR filters such as the latter one.

> Given a filter length, no other type is as effective at
> reducing noise amplitude.

The ratio of secondary lobes to the main f=0 lobe is better
for a gaussion FIR filter than the moving average one, therefore
having better SNR enhancement.

> On the downside, stopband response is poor - many
> other filter types reject signals much better. However,
> because of the sin(x)/x response, there ARE many
> null points in the stopband frequency range where
> interfering carriers are fully rejected. The first one has
> a period equal to the filter length

All FIR filters of the same length have the same nulls at the
same frequencies.

An advantage of the moving-average filter is that the calculations
are simple, but at the cost of less than optimal performance in
many cases except where it happens to be the matched filter of
choice for the right length pulse waveforms.

> It is remarkable that a digital filter so easily coded is so
> pleasing to the ear. I have coded such filters into simple
> PIC processors, both lowpass and bandpass:
> <http://epic.mcmaster.ca/~glen/4a4/filt_det.pdf>
> This was for detecting signals, not for listening. I
> wouldn't suggest that the 10-bit A-to-D in a PIC
> is appropriate for serious audio use.

There are many woeful DSP filters out there. They may be implemented
well, but the maths is not well understood by the implementors, and so
are not well suited for the application. A lot are just copy-and-paste
code from the web. It just shows how little lengths one needs to go to
get above the average.

3689

2009-10-29 17:57:54

Russell Shaw

Re: How Receivers Sound

Russell Shaw wrote:

> Glen Leinweber wrote:

...

> All FIR filters of the same length have the same nulls at the
> same frequencies.

Correction, not true. If eg the last few taps of the moving average
filter were zero, then the nulls would be further apart.

3690

2009-10-29 18:15:52

Russell Shaw

Re: How Receivers Sound

Russell Shaw wrote:

> Glen Leinweber wrote:

...

>> Given a filter length, no other type is as effective at
>> reducing noise amplitude.
>
> The ratio of secondary lobes to the main f=0 lobe is better
> for a gaussion FIR filter than the moving average one, therefore
> having better SNR enhancement.

I should qualify this. For voice with most of the energy increasing
towards f=0, lower stop-band lobes would be better.

For CW, a moving average filter has higher stop-band lobes,
but these capture the energy from the pulse-like CW waveform.
This matched-filter operation is more optimal.

One could have a "CW" control on their receiver that
adjusts the FIR length for "fast" or "slow" morse.

>> On the downside, stopband response is poor - many
>> other filter types reject signals much better. However,
>> because of the sin(x)/x response, there ARE many
>> null points in the stopband frequency range where
>> interfering carriers are fully rejected. The first one has
>> a period equal to the filter length

3691

2009-10-30 08:58:31

c6alk

Re: How Receivers Sound (FT-817 experiences)

Hi Tom and Group:

I wonder if these mods were documented in any way. I would be very interested in at least checking if my 817 has any of the described issues, and if so, then fixing them as Tom describes. If there any drawings etc., even

3692

2009-10-30 09:09:54

Roelof Bakker

Re: How Receivers Sound

Hello all,

Bob has been so kind to provide MacCluer's article which makes an
interesting read, especially the part on aural copy techniques used by
experienced EME oparators. It is stated that "the higher tones mask the
lower ones and not vica versa". This confirms my findings that everything
being equal a filter with a centre frequency of 500 Hz seems to ring less
than a filter centered around 600 Hz or higher.

Some operators prefer a wide bandwidth (2000 Hz), whilst others prefer a
filter bandwidth of 250 Hz.
My finding is that on a quiet band a wide bandwidth seems to work best.
The human brain can simulate a 50 Hz wide filter and this seems to work
best in a wide band of equally distributed noise.

The use of very narrow audio filters makes only sense when a bandwidth
smaller than 50 Hz is called for.
The reason can be twofold: digging weaker signals out of the noise or
getting rid of adjacent signals.
The first reason will be on hand by weak signal aural EME-work; the second
won't happen much in normal amateur CW operation. I have analysed the band
occupancy during the CW CQWW contest and it is amazing to find how much
frequency space is still unoccupied when looking from the perspective of
the use of narrow filters. Under these conditions a wider filter will do
and is preferred by many as a lot of stations don't answer your call on
exact the same frequency. How your receiver sound will be important to
prevent listening fatique working a major contest. I believe that on a
good sounding receiver a pile up should have musical qualities and it
should be easy to pick out the individual station.

At present my main amateur radio occupation is listening to non
directional beacons. These operate from 190 kHz to 1740 kHz, but the
majority can be found below 450 kHz. The station's identifier is given in
slow morse (5 to 10 w.p.m.) and repeated every 5 to 30 seconds. Most NDB's
use AM modulation with a keyed tone of 400 or 1020 Hz.
Canadian NDB's transmit only the carrier and the upper sideband 400 Hz and
French NDB's use a keyed carrier.
NDB's are allocated on a 1 kHz grid (In Europe also .5 kHz) and it is easy
to imagine that the carrier of one station is only 20 Hz away of a station
on the adjacent channel which uses 1020 Hz modulation.
NDB's are not high tech devices and often the modulation is derived from a
simple not all too stable RC oscillator. As a result carrier and
modulation can be only a few Hz apart and that is where the need for a
very narrow filter arises.

The circuit diagram and a few filter plots can be found here:

http://www.ndb.demon.nl/pa0rdt/

The plots have been made by feeding noise from a noise generator into a
receiver and then into the filter.
The noise generator is from SSD, page 168, fig. 62. Getting enough noise
through the filter to create a nice plot takes a little time though: some
20 minutes.
To set a bandwidth of 10 Hz, stage 1 and 3 are peaked at 495 Hz and stage
2 and 4 at 505 Hz. This is a bit tricky, but I use a selective level meter
and tracking generator, which makes it easy to do.

For signals 10 dB above the noise, only the top of the filter comes into
play. However the ultimate attenuation is about 60 dB.

How does this filter sound? When I used it for the first time, I thought
it to be completely useless. However setting the audio level low, gave
some results and after a few hours my brain ear system had learned to cope
with it. There is one weird phenomenon I won't withheld: when there is no
signal in the passband the noise is clearly audible. As soon as the
weakest audible signal appears, there is silence as the noise seems to
drop away. Only the signal is standing out. And no, I don't use AGC.

I hope some of you might find this of interest.

73,
Roelof, pa0rdt

3695

2009-10-30 12:01:33

Tom H

Re: How Receivers Sound (FT-817 experiences)

The FT817 distortion, analysis and mods are extensively documented here
0Problems%20-%20Receiver/> . If that link does not come through, then
try:

http://groups.yahoo.com/group/FT817/files/Audio%20Distortion%2C%20AGC%20\
Problems%20-%20Receiver/
0Problems%20-%20Receiver/> , else go to:

http://groups.yahoo.com/group/FT817/
then > Files
, then > Audio Distortion, AGC Problems - Receiver
r/> .

The contents are reasonably up-to-date but I got hung up for a time with
a suspected fault due the accidental use of acid flux that turned out to
be something else entirely, moving house, etc. I know the page
_refs_re_ft817_rx_distortion.htm
3LwGgMnMtIkrJqg9f349Ozo0peXVNa4N51Yo1E97ApuYDsT5SEs90/Audio%20Distortion\
%2C%20AGC%20Problems%20-%20Receiver/_refs_re_ft817_rx_distortion.htm>
is not complete.

There are A-B audio samples that attempt to illustrate the distortion.
If you can hear the difference, then you should be able to tell whether
your rig behaves similarly.

Good luck!

Tom VE3MEO

3697

2009-10-30 12:11:58

Tom H

Re: How Receivers Sound (FT-817 experiences)

Ouch! Yahoo's Rich Text Format message editor is messing up. Here it is again in plain text:

The FT817 distortion, analysis and mods are extensively documented here:
http://groups.yahoo.com/group/FT817/files/Audio%20Distortion%2C%20AGC%20Problems%20-%20Receiver/

If that link does not come through, then go to:

http://groups.yahoo.com/group/FT817/
then Files
then folder Audio Distortion, AGC Problems - Receiver.

The contents are reasonably up-to-date but I got hung up for a time with a suspected fault due the accidental use of acid flux that turned out to be something else entirely, moving house, etc. I know the page _refs_re_ft817_rx_distortion.htm is not complete.

There are A-B audio samples that attempt to illustrate the distortion. If you can hear the difference, then you should be able to tell whether> your rig behaves similarly.

Good luck!

Tom VE3MEO

>

3710

2009-10-31 10:10:10

c6alk

Re: How Receivers Sound (FT-817 experiences)

Tom:
Thanks very much for the info, got it all now. I appreciate you sharing it.

Brian K7RE