EMRFD Message Archive 15460

Message Date From Subject
15460 2019-09-12 16:43:54 jgnoci@gmail.com Just to introduce Myself and some of the radios I built.

Hello to all. I am Joe, call is V51JN, on the west coast of Namibia.

I have built quite a few Direct Conversion receivers, the most ambitious being a 200watt PEP all HF band transceiver built mainly from ideas, design concepts and math lifted from EMRFD.

The transceiver can be seen in my album - Joe's radios - I built all the modules as separate sections , testing each section and connecting them all up on a big prototype 'breadboard', before making final PCB's and fitting all into a nice custom case. A microprocessor controls it all, with a digital display for frequency, RX attenuation setting, ALC, RF Output, SWR, etc. An AD8954 DDS provides the LO signal. Audio filters are switchable LC bandpass or OP-Bandpass, with a selectable final SCAF section. Works nicely and is a joy to use.


I am also involved in Wildlife conservation, specifically in collaring Lions with VHF/GPS/SAT tracking collars as an aid in mitigating growing human/wildlife conflict in the rural regions here. To that end I have built a VHF Superhet CW receiver that works well - an MDS of around -141dBM, which gives me a detection range of about 15km LOS with a 2 element Yagi - the collars emit 10mw into a very poor antenna, embedded in the collar strapping.


I am attempting to built a DC CW receiver to replace the superhet - in the hope of reducing size, power consumption, etc, but my prototype sensitivity eludes me..I have so far been unable to better and MDS of around -129dBM.. The front end preamp/mixer is the same as the superhet. The audio sections are typically EMRFD, and I have really worked hard at a low noise circuit implementation, but I think I am missing some fundamentals... I hope to ask some questions on this forum in the coming days re this subject.

The performance requirements for this receiver are very different from the usual HAM receiver. I have no need of high performance in terms of overload, IMD figure, etc. I need the best MDS and lowest noise possible! These signals are listened for in very rural environements - there are no signal sources around for 10s to hundreds of km, not even cell phone towers! And the collars are very low power. The receiver signal bandwidth needs to be around 2 to 3 Khz - more an SSB bandwidth than CW, because some collars have very 'cheap' transmitters - a crystal osc timed with a low power multivib circuit - a 20ms pulse every 2 seconds. And the frequency output ( from 140 to 150MHz band) can easily drift 1Khz between day and night temperatures. So we need to 'hear' that there is a signal in the vicinity of the selected frequency, and a CW filter might make that difficult.


Thanks - and forgive the long post please!

Joe

V51JN

Namibia...

15461 2019-09-12 22:46:00 jgnoci@gmail.com Re: Just to introduce Myself and some of the radios I built.
Sorry, Not sure if one can edit ones post to correct typo's - if so  I did not find a way yet - 
The frequency band of the collars is 148MHz to 150MHz, not 140MHz as I first typed..
Joe
15462 2019-09-13 00:41:59 Lasse Moell Re: Just to introduce Myself and some of the radios I built.
Joe,
you mention that the super-het receiver, using a CW BW, has -141 dBm MDS and a homdyne receiver with SSB BW only gives -129. Well for a starter, if your BW is 3 kHz rather than 300 Hz will give 10 dB more noise. And do not forget that the DC-receiver has two sidebands, doubling the noise. So a simple calculation tells me that you may end up with 13 dB less sensistivity. -141 - 13= -128 dBm.

/Lasse SM5GLC

12 september 2019 22:40:06 +02:00, skrev jgnoci@gmail.com [emrfd] :
 


Hello to all. I am Joe, call is V51JN, on the west coast of Namibia.

I have built quite a few Direct Conversion receivers, the most ambitious being a 200watt PEP all HF band transceiver built mainly from ideas, design concepts and math lifted from EMRFD.

The transceiver can be seen in my album - Joe's radios - I built all the modules as separate sections , testing each section and connecting them all up on a big prototype 'breadboard', before making final PCB's and fitting all into a nice custom case. A microprocessor controls it all, with a digital display for frequency, RX attenuation setting, ALC, RF Output, SWR, etc. An AD8954 DDS provides the LO signal. Audio filters are switchable LC bandpass or OP-Bandpass, with a selectable final SCAF section. Works nicely and is a joy to use.


I am also involved in Wildlife conservation, specifically in collaring Lions with VHF/GPS/SAT tracking collars as an aid in mitigating growing human/wildlife conflict in the rural regions here. To that end I have built a VHF Superhet CW receiver that works well - an MDS of around -141dBM, which gives me a detection range of about 15km LOS with a 2 element Yagi - the collars emit 10mw into a very poor antenna, embedded in the collar strapping.


I am attempting to built a DC CW receiver to replace the superhet - in the hope of reducing size, power consumption, etc, but my prototype sensitivity eludes me..I have so far been unable to better and MDS of around -129dBM.. The front end preamp/mixer is the same as the superhet. The audio sections are typically EMRFD, and I have really worked hard at a low noise circuit implementation, but I think I am missing some fundamentals... I hope to ask some questions on this forum in the coming days re this subject.

The performance requirements for this receiver are very different from the usual HAM receiver. I have no need of high performance in terms of overload, IMD figure, etc. I need the best MDS and lowest noise possible! These signals are listened for in very rural environements - there are no signal sources around for 10s to hundreds of km, not even cell phone towers! And the collars are very low power. The receiver signal bandwidth needs to be around 2 to 3 Khz - more an SSB bandwidth than CW, because some collars have very 'cheap' transmitters - a crystal osc timed with a low power multivib circuit - a 20ms pulse every 2 seconds. And the frequency output ( from 140 to 150MHz band) can easily drift 1Khz between day and night temperatures. So we need to 'hear' that there is a signal in the vicinity of the selected frequency, and a CW filter might make that difficult.


Thanks - and forgive the long post please!

Joe

V51JN

Namibia...




15463 2019-09-13 08:55:50 jgnoci@gmail.com Re: Just to introduce Myself and some of the radios I built.
Hi there Lasse, 
Thanks for your reply - I fear I may have been lax in my descriptions - I did not want to mix to much deep detail with what I suppose was to be an introduction post. I apologize for that.

What I did not manage to say effectively was that I use it to detect what amounts to a CW signal from the collar. What I failed to give was better info on the receiver..
The receiver has an MRF966 dual gate fet as preamp ( measures around 18dB gain), into an AD831 balanced active mixer. The mixer has a output amplifier with resistor set gains - I have it set to +6dB gain, with a -6dB resistive pad after, feeding a dual gate fet amplifier, more a buffer and impedance converter to a 10.7MHz , 2.4KHz bandwidth, crystal filter ( I have a handful of these removed from surplus equipment). Following is another dual gate fet IF amplifier, into an MC1496 product detector, and then into an audio pre-amp, a switchable bandwidth audio filter ( 3KHz, 2KHz, 1KHz and 500Hz) and then the speaker amplifier. 
The approx -140dB MDS is with the 2KHz AF filter in place. The receiver LO is a PLL, using a Motorola PLL chip. The mixer specs indicate around a 10dB noise figure for the mixer...

I use an HP8591EM Spectrum analyser to measure. The sig-gen is set to -40dBM, and confirmed on the spec analyser. I then use attenuators in line and checked that the analyser shows the drop in level to be as should be per attenuator. I left the sig-gen at the -40dB setting and then keep adding attenuators till I got to the -140dBM level, with the signal just discernible. I did this as I did not really trust trying to read a -140dBM signal on the spec-analyser..

The attempt at a DC version receiver used the same front end and mixer and gave the result indicated. I have just now tested an changed version, with very similar ( poor) results - this version does seem to be quieter, but so is the signal ( lack of audio gain?) - I do not have a sound means of evaluating noise figure and am not too sure how to, even after reading up a lot about it! 
Anyway, the changed version now has the same RF front end, with a TUF-3 mixer, followed by almost straight EMRFD Diplexor and the common base amp, followed by a common emitter amp ( total around 55dB gain, simulated on LTSpice - and verified by means of oscilloscope), then a 2khz bandwidth filter - a peaked LO and HI-pass filter using an LM6622 ( also tried a few other op-amps to see if lower noise would improve things) and then an LM386 speaker amp. The filter has approx 10dB gain in the passband, the audio speaker amp has a 40dB gain capability.

And so I am still at a loss - perhaps this architecture is simply not suited for the application? It's fun though..

Joe
15464 2019-09-13 10:03:49 Ashhar Farhan Re: Just to introduce Myself and some of the radios I built.
Joe,
Perhaps a different approach to your direct conversion will help. I suspect that the high noise figure is either because the filters are not optimally tuned or that the demodulator is lossy.
It might be an idea to try another mosfet as a detector, a la w1fb pattern. You dont have any am or fm broadcast to worry about. I would carefully measure the loss in the rf amplifier itself.  Does the RF amp show up a pip on the spectrum analyzer when fed with a signal below -130 dbm? 
Your work is truely inspiring.
73, f

15465 2019-09-13 10:20:29 John Levreault Re: Just to introduce Myself and some of the radios I built.
+1

JL nb1i

15466 2019-09-13 11:08:56 Bill Carver Re: Just to introduce Myself and some of the radios I built.
My thinking:

From a theoretical viewpoint, if you had a perfect 0 dB noise figure receiver, with a 2.4 Khz bandwidth, your minimum detectable signal (signal + noise is 3 dB higher than noise alone) is -140.2 dBm. A real receiver will never be quite as low as that. Perhaps you have a narrower tone filter following the product detector to meet your -141 dBm number. Or perhaps there is some "leakage" around your external 100 dB attenuator string and the signal is really a little bigger than -141 dBm.

The AD831 mixer can be OK, but at that frequency its noise figure is high. And it can be tricky especially since you have configured it with feedback resistors. I like your diode mixer better: it's passive, not as much to go wrong. Terminating the mixer's output diplexer with a J310 with gate-source transformer feedback can be a perfect 50 ohm load with 12 dB of gain compensating for mixer loss, and 1 dB noise figure all at the same time.

As was pointed out, opposite sideband noise can degrade the minimum detectable signal, and noise figure, by 3 dB. An SDR surrenders 3 dB unless there is both a lot of gain ahead of it and a phasing network to reject the opposite sideband noise. You must have both gain ahead, and phasing cancellation: for example, two of your 18 dB RF amplifiers into a pair of diode mixers with quadrature LOs and a quadrature phasing scheme to combine the two outputs. A bit tricky at VHF.

Personally, I'd stick with the xtal filter since you already have it. I would make best use of it by having a  low noise stage behind it, a dual gate MOSFET for example.

I make these comments only after modelling things on paper. Even though on paper one RF amp into diode mixer _JFET should have a low noise figure, putting more gain the front end can only help. Normally you would not want to add too much RF gain, but without large out-of-passband signals to worry about there is no downside to adding gain. I'd add another 18 dB gain stage ahead of whatever you finally use.

Bill W7AAZ

15467 2019-09-13 13:44:47 qvutik6r37a6umpnd... Re: Just to introduce Myself and some of the radios I built.
Thank you to all replying! And thank you Bill for taking detail time to explain things to me. I followed much of the discussion in EMRFD on the sort of crossover point of complexity versus performance between DC and Superhet receivers, and I suppose I thought I would discover the Holy Grail all by myself...adding all the extras you mention, ie, two of everything, a quadrature LO, etc, makes the receiver far more complex than a well performing simple superhet. The purpose of building this form of receiver is not because I require it - it just became a challenge in the attempt to reduce the size and power consumption by reducing component count - something you have helped me understand will not be!
I am going to add another RF pre-amp at the front end, just to see what happens, but as you say, there are no signals out there to bother me, not even IN passband...

One last thing Bill, could you point me in the direction of the Math and perhaps some explanation guidelines so I may understand the theory behind your first paragraph? Especially how to calculate the theoretical effect of receive channel bandwidth versus MDS, etc.  I also wish to try measure more accurately the signal levels I am stating as the MDS level - My attenuators are SUHNER N-Type, of good quality, so I would like to understand this better.

Thank You.
regardsJoe
15468 2019-09-13 14:13:23 Bill Carver Re: Just to introduce Myself and some of the radios I built.
If you can run a DOS program on your computer I can send you a program I wrote (RX5.EXE) that lets you enter each stage in a receiver and see the results on a stage-by-stage basis (while it's useful, it does not take bandwidth of individual stages, just the effective noise bandwidth of everything).

Bill


15469 2019-09-13 14:22:45 Tayloe, Dan (Noki... Re: Just to introduce Myself and some of the radios I built.

If you are trying to check out sensitivity, you need more audio gain.  Headphone level for good earbuds is about 10 mV.  If you are trying to start with 0.3 uV weak signals (for example) that is a gain of 20*log10(10*10^-3/0.3*10^-6) or about 90 dB of gain.  Your simple DC receiver may be working fine but It will be hard to hear these weak signals with only 55 dB of overall receiver gain.  Speaker level needs roughly another 20 dB of gain.  On my more sensitive DC phasing receivers, I run at about 90 dB of end-to-end gain and I tend to use more sensitive ear bud type headphones.

 

For receiver sensitivity measurement I have pick up used HP AC voltmeters from hamfests.  Who want an AC voltmeter anyway?  I often see them for $10.  You have to have enough audio output so that the “hiss” level can be measured by the voltmeter.  Add enough signal level to get a 3 dB rise in the measured AC voltage.

 

I have seen programs that can make this measurement using the audio input of a computer, but I have not used these much.

 

  • Dan, N7VE

 

15470 2019-09-13 16:14:25 Bill Carver Re: Just to introduce Myself and some of the radios I built.
For correct measurement of SNR you need to measure noise power. HP3400 and HP3403 are true RMS meters, as is the Fluke Fluke 187. Computer-computed numbers can be true RMS if the programmer did an RMS calculating from the samples. But that doesn't happen unless the programmer does it.

Other AC voltmeters actually measure average voltage of one polarity of the input, then tell you what the RMS would be if it is a pure sinusoid. On some HP voltmeters, such as the HP403B, printed on the meter face is "R.M.S. volts", and just below that "average responding".  When used in identical noise bandwidths they're OK to indicate that you've improved your sensitivity. But they do not measure what the actual S+N power is.

W7AAZ


15471 2019-09-13 17:15:23 kerrypwr Re: Just to introduce Myself and some of the radios I built.
Sabin's "Measuring SSB/CW Receiver Sensitivity", which is on the EMRFD CD, is a very good article on the basics of noise & sensitivity; HP/Agilent app note AN57 and its descendants AN57-1, AN57-2 etc are also excellent.

Kerry VK2TIL.
15472 2019-09-14 09:04:15 qvutik6r37a6umpnd... Re: Just to introduce Myself and some of the radios I built.
Thanks Bill.
I have an oldish Fluke 87, which says 'True RMS Voltmeter' on the front face, and a new KeySight U1281A, which also indicates the same on the front face.

I need to read up a lot more and understand more about noise and gain...Not being familiar with the subject, nor schooled in these subjects, I find it difficult to grasp some of the concepts and terms and many texts are aimed at scholars in the field...I will keep at it! I am a digital guy, having designed and manufactured flight control, autopilot and navigation computer and graphics systems for fast jets and slightly slower 'copters...so lack in the Radio field a 'little'.  

Regards
Joe

15473 2019-09-14 09:05:40 qvutik6r37a6umpnd... Re: Just to introduce Myself and some of the radios I built.
Thanks You Dan, 
I fear my use of language may have buried what I was trying to say - again...

Regarding the gain of the various stages - they are , as best I can ascertain, 

Fronet end RF pre-amp - 18dB, Mixer +6db followed by a -6dB pad, so 0dB, then EMRFD common base/comm
15474 2019-09-14 09:29:40 Bill Carver Re: Just to introduce Myself and some of the radios I built.
We're on opposite sides of the street: I'm an analog guy, DEC had just introduced the PDP-8. The definition of MDS is pretty simple, when signal plus noise power is 3 dB higher than noise power alone, then the signal must be MDS. The trick is measuring noise power: that requires a RMS measurement......the very basic definition of the RMS process is it computes a DC voltage that would produce the same heating of the load.

I'm not familiar with the Fluke 87, have a "187" which I know IS true RMS. I think since it says true RMS, you can trust your Fluke for that 3 dB rise accurately indicating you're injecting MDS into the receiver.

If you're able to make a low noise RF amplifier, I'd make one in its own little box and keep it around as a piece of "test equipment".  As a sanity check. Any time you think you've arrived at a scheme that's bear the best sensitivity being able to connect that amplifier and see no improvement is a confirmation you're right. If, on the other hand, connecting the amp produces a significant improvement in sensitivity then you have to scratch your head and figure out why the RX should be quieter but isn't.

You would chuckle at what's bugging me: SOFTWARE, hi. I've been working on an SDR controller using at Atmel AVR, generating "CAT commands" to control the HPSDR program in a PC. I.E. sending "ZZTU1" causes it to transmit, "ZZTU0" causes it to stop transmitting. I have a little box with eight knobs and 24 pushbuttons, and 70 pages of C code sprinkled with bugs!

I'll send a copy of RX5.EXE for your amusement, maybe even the RX5.PAS Turbo Pascal source, but I have to get a floppy drive reconnected to the old DOS machine.

73 - Bill - W7AAZ

15475 2019-09-14 11:37:45 Gary Re: Just to introduce Myself and some of the radios I built.
Bill,
Wow, Turbo Pascal!  My favorite language
and programming environment.
But, you say you are using C too?
Being mostly a digital guy, if you want to
shoot me the code too, I would take a look
at your problems.
73
Gary
WB6OGD 


Sent from my iPhone

15476 2019-09-14 11:50:51 Dana Myers Re: Just to introduce Myself and some of the radios I built.
15477 2019-09-14 11:53:29 Thomas S. Knutsen Re: Just to introduce Myself and some of the radios I built.
The important part when measuring noise is the Crest factor of the meter. All of the "true RMS" reading meters will measure wrong with true white noise input. A basic introduction is found in this article:  https://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1964-01.pdf
In basic, Crest factor is the difference between the peak amplitude and RMS value of a given signal. For a sine, this is the well known sqrt(2), or 1.414, while for a square wave with 50% dutty cycle it is 2. For true gaussian noise it is infinite. As such, the value you measure with a common DMM with Crest factor of 2 has no real value. 

Fluke gives the 87 series crest factor as 3 at full scale, 6 at middle scale. These are quite typical parameters for common high quality meters. Linear Technologies app note 61 by Jim Williams have a quite interesting modern approach to measuring gaussian noise.  
There is a benefit in a receiver, the output is bandwidth limited. As such, if you do the math, you will find that the crest factor reduces, and it will become a real number. Assuming a Crest factor of 10 for most SSB/CW/FM bandwiths should be quite close to work. 

Crest factor is also important when amplifying noise. Lets say a amplifier has 20dBm 1dB compression point, and the noise has a crest factor of 10 (20dB). In this case, the max RMS amplitude the amplifier can have will be 0dBm, since the peaks will be 20dB higher. 

A bandwidth limited software approach doing a fourer transform, and calculation of the RMS from that should be quite good if done correctly, when measuring noise, and signal + noise. 

As you can see, measuring noise accurately isn't easy. In additi
15478 2019-09-14 13:57:37 Vern VanZ Re: Just to introduce Myself and some of the radios I built.
I use the USB floppy drives on both my Windblows 10 and Linux (Debian) boxes here at home, and so far, they've worked flawlessly (less than $10 too!).  Worked great for recovering some of the old C source code, and spreadsheets I built decades ago...

73, Vern N7GTB

15479 2019-09-14 16:54:50 jgnoci@gmail.com Re: Just to introduce Myself and some of the radios I built.
Hi Bill.

I have been a ham for 48 years but have not dug deep into the stuff that really should be backed up by some fundamental education in the field - My first deeply technical build was the HF transceiver in my photo's - that took 2 years to build, with many prototypes along the way. I did learn a heck of a lot during that , but discover that my education is only beginning..!

My schooling is in the digital world, but my Masters is in Aeronautics, with specifics in very fast unmanned aircraft..
Having left that world behind, I am delving into Radio again and enjoying it a lot.

My good wife had her career in the same aeronautics environment as me, and was VP of a large software division, accredited by British Aerospace to do software on their Jets.
So, if you would like some help on software, I have a very good contact....She would be happy to help!

Feels humbling to have retired and realize one has to start learning all over again!
I will persevere with the little VHF-DC receiver, learn some more, and no doubt ask you all a few, or many, more questions!
Thanks to all 

Joe

PS - I have hunted hi and lo, all over the internet, yahoo help forums, etc...Have not found how I can have my posts headed with my name, instead of some odd Yahoo alias, or my email address - you guys all seem to have got it right - How ????

15480 2019-09-14 20:00:56 Bill Carver Re: Just to introduce Myself and some of the radios I built.
I have to install a floppy drive on the DOS machine, it's missing. Once I can make a floppy then the USB floppy drive on this machine will let me read the files into a current machine.
W7AAZ


15481 2019-09-14 20:22:36 Bill Carver Re: Just to introduce Myself and some of the radios I built.
thanks Dana.....as I said in another reply, there is NO I/O except IEE4888 in this DOS PC. I need to install a floppy drive to get files out of it. A buddy (KI7CC) has a USB floppy to read it into one of the Windows machines.
Bill

15482 2019-09-14 20:24:37 Ashhar Farhan Re: Just to introduce Myself and some of the radios I built.
The cascade.exe in the older EMRFD CDs does this work of calculating overall noise figure of cascaded stages.

I would also mention that Rick, KK7B, built a vhf instrumentation receiver for this kind of beacon hunting. It is briefly described in a pdf that is publicly available.

It is not entirely correct that a simple superhet will outstrip a single signal direct conversion with the same level of complexity. A phasing rx needs minimal amount of RF circuitry. The remaining signal processing is all at audio regardless of the rx frequency. The audio complexity is of three op amps and two BC147s (leas noisier than the 2N3904s). Sure, there are a few more resistors than usual, but that is all there is to it.

A Bill suggests, an RF pre amp can do wonders. Mine is on the rooftop to sniff out cubesats that skim the horizon. It is a BF998 job. Check it out in the EMRFD archives. The original was descibed by Wes on www.w7zoi.net for 50 MHz.

73, f



15483 2019-09-14 20:28:25 Bill Carver Re: Just to introduce Myself and some of the radios I built.
Yea Turbo Pascal, or Pascal in general, is a nicer language for me.....it packs a lot of run time checking into the finished code so (for example) when an array pointer exceeds the size of the array it tells you. I've learned C, and grit my teeth at some aspects of it: it's a much earlier design than Pascal, with extremely terse syntax and no run time checking, but I can live with it.

My SDR controller is over 70 pages of code.....nothing anyone should have to deal with. It's running now, found the ATMEL Studio 7 did not default to burn the last compile of my code......it was burning an earlier version. Once I realized that things settled down pretty well.

Just frustrating is all.

Bill W7AAZ


15484 2019-09-15 08:20:54 swift_glen Re: Just to introduce Myself and some of the radios I built.
For doing a MDS measurement, a true-RMS reading is certainly required, since you're comparing a
sinusoidal signal with noise. Fluke 87 passes muster in this regard.

The 3dB difference between pure noise, and signal+noise isn't much...it is a difficult measurement to make. Measuring noise is itself noisy....that Fluke display might display about three readings-per-second, and each display differs from the last. You end up taking a mental average reading. when measuring noise.

Complicating this averaging-of-readings is the signal bandwidth, defined by an upper frequency limit and a lower frequency limit. The lower-frequency limit affects the variance of the Fluke's RMS display. Let's say the lower-frequency limit of the measurement bandwidth is 20 Hz. This is not far above the Fluke's display at 3Hz. The variance from reading-to-reading will be large, and more difficult to "average-out".
15485 2019-09-20 00:09:00 jgnoci@gmail.com Re: Just to introduce Myself and some of the radios I built.
Thanks Ashhar.
I finally got to play with Cascade.exe and it was an eye opener for me, not being that conversant in these subjects.. Very interesting that they key lies in the very first stage of RF amplification. Gain and noise figure play here play the greatest role in MDS and total system noise figure. The deleta in MDS for a front end noise figure of 0.5dB versus 2dB is massive!

If a second stage of RF gain is added, both gain and noise figure can stand much greater margins.

What is very interesting is that for a DC RX lineup, the post mixer AF preamp ( the 2N3904 circuit) noise figure seems , within reason, relatively unimportant - I tried with noise figures up to 6dB, and it made a very small difference to the MDS... If I did that correctly, why the big drive to get that stage as low noise as possible?

I did a basic line up as follows:

Stage                   Gain   Noise figure
1st Rf preamp       10      0.5
Helical Filter           -3      2
2nd Rf preamp      10      4
Mixer (tuff-3)          -7      8
Diplexor                 -3      3
AF preamp           40      8   (!)
AF Filter ( 1KHz)  10      8
Speaker Amp       40      8

MDS = -140dBM
Total Noise = 4.4dB
Totatl Gain = 97dB

Lowering the noise figure of any stage from the 2nd RF preamp on had very little negative effect on MDS or noise figure - Reducing the NF of the AF preamp ( post mixer/diplexor) to 1dB had less than 0.5 dB effect on MDS and total noise figure. I have not been frugal in choice of noise figure for the stages following the Helical filter...Most of those numbers are probably easily improved upon with simple homebrew efforts.

Am I missing something? Is the improvement in NF/MDS measurable in real terms, ie, not just 0.5dB, for what seems to be a lot of effort to reduce noise figures of any stages following, say, the mixer?

It seems to me that 99% of the effort should go into the first RF preamp stage - increase that stage gain from 10dB to 20dB and the total NF drops from 4.4dB to 1dB !

Regards
Joe


15486 2019-09-20 03:16:27 Alberto I2PHD Re: Just to introduce Myself and some of the radios I built.
On 2019-09-20 8:48, jgnoci@gmail.com [emrfd] wrote:

What is very interesting is that for a DC RX lineup, the post mixer AF preamp ( the 2N3904 circuit) noise figure seems , within reason, relatively unimportant - I tried with noise figures up to 6dB, and it made a very small difference to the MDS... If I did that correctly, why the big drive to get that stage as low noise as possible?

As far as I know, what is important for a post-mixer stage is not the noise figure, but rather its linearity, intermodulation-wise...

--
73 Alberto I2PHD
<<< http://www.weaksignals.com >>>


15487 2019-09-20 03:17:21 Thomas S. Knutsen Re: Just to introduce Myself and some of the radios I built.
This analysis, while mostly correct does not take into account what frequency you are operating at or the IMD of the amplifiers. 
Unless you are doing EME on VHF/UHF/SHF, then its worthwhile to trade a couple dB worse NF in better IMD performance.  This is due to the ground temperature (ktb) of the surrounding environment on VHF/UHF, and band noise on HF.

Remember that the while the noise figure of the stages behind the first couple does not matter for the total sensitivity of the receiver, the noise produced by them is still at the output. My ears find it much more relaxing to listen to a receiver that has had all the stages optimized for low NF than just slapping a good pre-amp on a mediocre radio.  A

There is a error in your analysis, but it does not alter the end result much. There is no way a passive network (in this case helical filter) can have less noise figure than its loss. Having such would either require it to have gain, or violate the law of conservation of energy. a more reasonable NF is 3dB for a filter with 3dB loss, that is assuming that it does not introduce any other noise than its own loss.  

73 de Thomas LA3PNA. 

15488 2019-09-20 08:54:10 jgnoci@gmail.com Re: Just to introduce Myself and some of the radios I built.
Thank you for your reply Thomas.
My error in filter noise figure versus insertion loss just show my lack of knowledge on this subject..Thanks for the correction.

Your statement - 
Remember that the while the noise figure of the stages behind the first couple does not matter for the total sensitivity of the receiver, the noise produced by them is still at the output.

Makse intuitive sense, but I fail to grasp what it actually means - I understand that the noise of a stage is amplified by the next stage, etc, but at what point does all that noise make my desired signal undetectable?
From Cascades theoretical output data, I can increase the last 2 or 3 stages noise a lot with a very small ( 1db or so) decrease in MDS. 
Forgive a maybe stupid question, but could you explain what difference do I actually 'hear' in the speaker for the following two scenario's: Note - all previous stage gains and NF remain the same for the 2 cases - signal BW=1KHz

1)
Post mixer AF amp NF=2
AF Filter NF=2
Speaker amp NF=4
Gives system NF=2dB, MDS=-142dBM

2)
Post mixer AF amp NF=8
AF Filter NF=8
Speaker amp NF=12
Gives system NF=3.7dB, MDS=-140dBM

The noise in those last three stages, in 2), is hugely increased - but I presume I can still hear my signal now at -140dBM ( MDS is 2 dB less, yes, but audible at that level and not drowned in all the new noise..).
So what do I actually hear? If the noise ( Hiss in the speaker??) is now vastly increased compared to the signal and the S/N ratio therefore negatively impacted, I should not here my signal anymore, and the MDS is incorrect?

Have not grasped it yet..

As a side, the IMD is deliberately ignored since, as I stated in my original post, the environment where this 'receiver' would be used is electrically dead quiet..I can comfortably  afford maybe 40 to 50dB gain, split between two RF pre-amps up front, with the first having the best NF I could achieve.

Regards
Joe
V51JN / ZS6JGN



15489 2019-09-21 16:28:20 lawrence_joy Re: Just to introduce Myself and some of the radios I built.
Hello,
Whenever I have given a talk about receiver architecture/cascade calculations or asked a trivia question "Where in a receiving system is the best signal to noise ratio (SNR or S/N) found?--A receiving system includes the antenna through to the output device.

In EMRFD Clause 6.1 "RECEIVER FUNDAMENTALS" starting on page 6.9, runs through the basic cascade noise factor/figure calculation. The equivalent noise bandwidth (B) used in Eq 6.9, Pn = k T B, is set by what the demodulator/detector sees and is a noise density, that is it is on a per hertz basis. The noise floor at the standard temperature of 290 K is -174 dBm/Hz. To determine the noise floor of your receiver you use the formula of 10*Log B (the units are dB*Hz) and add this to -174 dBm/Hz (notice that the Hz units cancel and you are adding dB to dBm and thus the answer is a level in dBm). This bandwidth, and thus noise floor, is brought forward through the front end of the receiver and out to the antenna terminals. The answer to the question is "at the antenna terminals". Any stage you add in the cascade deteriorates the SNR.

The equivalent noise bandwidth of a filter is the bandwidth of an ideal (brick wall) filter. For first order approximation you can use the -60 dB bandwidth of the filter and remember that as the number of poles in the filter increases, the steeper the skirts become, and you can use the signal (-3 dB) bandwidth. A cascade calculation includes much more of course, 3rd order intercepts, 2nd order intercepts, 1 dB compression points, birdie frequency calculations, and image noise reduction at the input of a mixer. The cascade calculation program should calculate the SNR at each stage so you can see trends and where improvements could be made.
--73 de WN8P, Larry