EMRFD Message Archive 12921

Message Date From Subject
12921 2016-06-06 18:56:45 k1rf_digital_stev... Using 1/2 Tayloe detector with downconversion/image rejection in sof
Hi all. I have been toying with a concept and need some thoughts on whether this would work or I'm all wet. A standard Tayloe detector provides I and Q output which can be processed via a soundcard and software to  remove the image.  The image rejection is not ideal but can be reduced via careful hardware adjustments in the audio circuitry for amplitude balance and phase balance, and perhaps use software to eliminate any remaining errors.  But I've been thinking - why not use only one half of a Tayloe detector, say, the I channel, and do the downconversion in software. Let's say, for example, the bandwidth of the "I" cnannel-only Tayloe detector is set for a bandwidth of 50 KHz. You can follow it with analog filters to keep frequencies, say, in the 25 to 50K region. Now feed this into one channel of a sound card with a 192 KHz sampling rate.  Develop I and Q by generating synchronized software sin and cosine waves at 50 KHz and mixing these two in software with two software mixers (multipliers) with the in-phase input signal from the 1/2 Tayloe detector to baseband.  You now have I and Q at baseband with very accurate amplitude and phase of synthesized I and Q signals. 

Wouldn't this technique essentially eliminate the ampliude and phase errors between I and Q for all practical purposes since the digital processing is very accurate?  This IQ can be further processed for any type of filtering or demodulation desired. I'm not looking for wideband signals for a spectrum display.  I'm just looking for bandwidths adequate for a single SSB or CW signal. This technique could take advantage of the Tayloe Detector's excellent dynamic range and inherent filtering but provide better IQ accuracy.  Thoughts?

"Digital Steve", K1RF

Regards,
"Digital Steve", K1RF


12922 2016-06-06 19:17:13 Bill Carver Re: Using 1/2 Tayloe detector with downconversion/image rejection in
If you downconvert from HF to a 25-50 KHz wide band using an oscillator frequency "f", the output will be F+25 to f+50, at the same time overlaid with f-25 to f-50.     IE, you have no image rejection.
W7AAZ


12923 2016-06-06 21:38:58 bob_ledoux Re: Using 1/2 Tayloe detector with downconversion/image rejection in
The Tayloe detector produces four quadrants of the incoming sine wave signal.   These are provided from a VFO using either a 90 degree phase shifter or VFO signal at four times the receive frequency.  The VFO signal is turned into a two bit counter used to advance the detector chip switches.  The detector chip outputs four 90 degree components of the incoming signal.

If you keep the four times frequency, but remove one pair of outputs you are only getting two instead of four of the 90 degree quadrants so you realize about 3db of loss.

It might be possible to reduce the VFO to twice the receive frequency and use it to drive a pair, instead of four switches.  Each output would be 180 degrees of the received sine wave, which, when moved into the two inputs of an op amp would provide a detected signal with much less loss, perhaps down to less than 1db.

The high detector Q results from a low pass filter consisting of the series resistance of detector components and input impedance combined with the shunting integrating capacitor.  With the VFO running at 2 times instead of 4 times the receive frequency the capacitor has twice as much time to accumulate its charge, so its size is adjusted.

As a side note, come confuse the Tayloe detector with other commutating detectors because they fail to realize the impact of the integrating capacitor, its ability to accumulate charge over time, and its role in generating the high Q low pass filter.

I've considered the possibility of using the Tayloe detector as a detector for a superhet but it scrambles my brain.  In other words, its output would feed a crystal filter at an IF frequency.  it would be nice to maintain the high Q and minimal signal loss.  But it is a low pass filter, which makes its appear more feasible at lower (audio) frequencies.  I may be wrong, but I think the Q would drop significantly as the output frequency is increased.  I'm also having difficulty understanding the detector function when output is in multiple megahertz.

bob-N7SUR
12924 2016-06-06 21:56:15 Bill Carver Re: Using 1/2 Tayloe detector with downconversion/image rejection in
The Tayloe makes a great product detector: great dynamic range, low noise, and 3dB-important: RF-IF noise from the opposite sideband is rejected by 40 dB or so. Conversion to a "zero IF" is its strength.

But it would have to be configured as an image-reject mixer in the proposed RF-to-low IF application. To get image rejeciton you'd need all the phases and all the switches and you'd have to combine them to reject the RF image. At that point (if you're successful) it's really not a "Tayloe" mixer any more, it's just an image-reject RF-to-IF mixer.

W7AAZ


12935 2016-06-08 13:42:53 k1rf_digital_stev... Re: Using 1/2 Tayloe detector with downconversion/image rejection in
Thanks all for your comments on this.

Here’s my full idea. I’m developing a very high performance QRP CW transceiver. Low rig input power is not an issue for me. It will operate only on 40 meters and 20 meters since these are my workhorse QRP CW bands.
 
1. Front end filters: 80M and 40M bandpass filters developed by YU1LM. See http://yu1lm.qrpradio.com/HF-6m%20BP%20YU1LM%20ver1.pdf
These are not as sharp as some others but have good return loss, good attenuation at 1/2 and twice the center frequency, and use standard capacitor values.  They also knock down broadcast band strong signals for the 40 and 20 meter bands of operation.
 
2. First mixer – a KISS mixer as described by Chris Trask here: http://www.mikrocontroller.net/attachment/146369/Mixer_Musings.pdf which includes a NC7WZ04 based squarer.
 
3. First Local oscillator – an SI570 for relatively good phase noise for the price  driving a divide by 2 flipflop for good output symmetry + 3dB phase noise reduction
 
4. Roofing filter: A 12 MHz twin T diplexer driving a quadrature coupled dual 3-crystal 12 MHz 1 KHz wide filter.  This is relatively broad for CW but provides a good wideband impedance match for the mixer as well as knocking down a lot of undesired stuff for the next stages.  I simulated this in LT Spice with the crystal parameters of my 12 MHz crystals and it looks great. Similar to concept in EMRFD figure 6-93 except I’m using L-networks for end matching.  The exact bandwidth and crystals I have on hand come very close to 50 ohms anyway with no matching.  The actual CW filtering will be done further down the chain well within the roofing filter bandwidth.
 
5. Post-mixer/roofing filter amplifier: W7IUV preamp based on the 2N5109.  See http://www.mtmscientific.com/preamp.pdf. This preamp is relatively straightforward to build, has very good input return loss, better than 6dB noise figure, and OIP3 of about +41dBm and about 20dB of gain. Follow this by a TBD resistive attrenuator.
 
OK. So at this point, I have a receiver front end that is sensitive enough for the bands in question, high dynamic range with a 12MHz, fairly narrow IF out.  The question now is, what to do to allow as much processing as possible in DSP and eliminate extra mixers at relatively low cost. I want to try to stick to 1 mixer to eliminate as much spuri as possible.  Sure, I can add a second conventional mixer with an audio diplexer and process at baseband with a sound card.  Or perhaps a conventional Tayloe detector? Or perhaps I could use under-sampling with a low cost DSP chip like the STM32L4, which has 12 bit, 5 magasample/sec ADCs and has a development board available for $14.00. I want to make the best of what a combination of what analog and DSP have to offer.   High performance at low cost.  I'll describe my transmitter concept separately in a separate post.

Appreciate your thoughts.
"Digital Steve", K1RF
12937 2016-06-08 18:56:03 Gary Tennyson Re: Using 1/2 Tayloe detector with downconversion/image rejection in

Steve,

The part about doing the audio (or low tens of kHz) processing in DSP is a goal that I currently have.  Writing the FIR code to implement a Hilbert transform is -- in theory -- doable.  I'm having trouble getting the basic I/O.  I started with the STM32F4, but the setup was daunting.  I'm currently trying to use the Teensy 3.2 paired with the same vendor's 16-bit stereo codec board.  While I can write code that runs directly on the Teensy well enough, I haven't gotten to the point that I can communicate with the codec. 

If you get to the point that you can identify a low-cost, easily-implemented DSP development platform, please let me know.

Thanks

Gary

KO4CY


12938 2016-06-08 19:17:33 Dana Myers Re: Using 1/2 Tayloe detector with downconversion/image rejection in
12939 2016-06-08 21:25:50 AD7ZU Re: Using 1/2 Tayloe detector with downconversion/image rejection in
Gary,

you might have a look at an architecture that uses the quickfilter technology parts .. in particular the QF1d512 which can interface directly with a codec.  its a 512 tap or 256 tap filter (32 bit coefficients 24 bit data) depending upon whether the coefficients are symmetric.  A 32bit fixed point fir kernel has significantly more precision than single precision floating point.
There is a tool suite provided by quickfilter that will calculate the filter coefficients or import the coefficients from an external source.  this way there is no dsp sw to write .. only load the coefficients via a SPI interface.

both a 90 degree phase shift and bandpass filtering can be done in a single fir kernel.  you need 2 fir filters, one for each lane with identical bandpass characteristics but a 90 degree phase shift between the lanes.   email me off line if you would like the C source to compute the 32 bit filter kernels and i can send you enough of a code snip to calculate the filter kernel coefficients and the windowing function. .. its very straight forward... i was going to do something with an in GUI with it java or python but havent yet.

the quickfilter chips are available or they sell a 14pin dip board dipster.. otherwise the qf1d512 comes in a 3mm qfn16 which is difficult to proto ..least for me it was.  the quickfilter chips will sample up to 500Khz and consume under 10ma at 3v so battery operation is not a problem.  i used a teensy board to load the filter coefficients.  am getting another board soon with more memory and more horsepower (m4) to calculate new filter kernels in situ.

just a couple of thoughts

Randy
AD7ZU

From: "Gary Tennys
12940 2016-06-09 03:07:28 Alberto I2PHD Re: Using 1/2 Tayloe detector with downconversion/image rejection in

The part about doing the audio (or low tens of kHz) processing in DSP is a goal that I currently have.  Writing the FIR code to implement a Hilbert transform is -- in theory -- doable.  I'm having trouble getting the basic I/O.  I started with the STM32F4, but the setup was daunting.  I'm currently trying to use the Teensy 3.2 paired with the same vendor's 16-bit stereo codec board.  While I can write code that runs directly on the Teensy well enough, I haven't gotten to the point that I can communicate with the codec. 

If you get to the point that you can identify a low-cost, easily-implemented DSP development platform, please let me know.


I recently published on Internet a project of mine based on the Discovery board from STM with the ARM STM32M4F chip,
and distributed the source code as open source. You can download it and use it for some examples on how to do filtering
and demodulation using the features of the ARM chip, together with the CMSIS library.

http://armradio.weaksignals.com

--
73 Alberto I2PHD
Credo Ut Intelligam


12941 2016-06-09 05:11:57 k1rf_digital_stev... Re: Using 1/2 Tayloe detector with downconversion/image rejection in
Hi Gary. I've been  putting it off but yes that's my next big step in actually implementing the DSP code.  A fundamental hurdle is to select a DSP platform with a low cost (read that free) development environment and then try to build on existing open source code that others developed.  I'm not a software expert but have written simple code for arduino processors.   I have looked at some of the code out there including:
SDR-to-go from the AQRP group, the follow-on STM32 SDR, Alberto I2PHD's Arm radio, and a few others.  I would like to do it all in software and avoid FPGAs and FPGA coding which, by comparison, is generally more difficult to develop and maintain, with more costly hardware.  I may use a fixed digital downconversion approach using what is known as the "Fs/4" technique which puts less burden on a software-based DSP.  A simplified explanation of this can be found here: http://dspguru.com/dsp/tricks/complex-downconverters-at-fs-over-4

Regards,
"Digital Steve", K1RF

 

12943 2016-06-09 10:09:44 Gary & Janice Ten... Re: Using 1/2 Tayloe detector with downconversion/image rejection in

Dana,

I'm aware of the ARM CMSIS-DSP library.  I have a user ID on the ARM site to download it, but it wasn't clear to me how to use it with the Teensy.   (The Tennsy 3.2 has an ARM Cortex M4 core with fixed-point DSP instructions, no FPU.)

If I revert back to the STM32F4, I'll get into that eventually.

Gary


12944 2016-06-09 10:37:28 Dana Myers Re: Using 1/2 Tayloe detector with downconversion/image rejection in
12945 2016-06-09 12:32:43 Gary and Janice T... Re: Using 1/2 Tayloe detector with downconversion/image rejection in
Thanks for the info. I didn't know about the PSoC 5LP. Looks good.

Gary
KO4CY

12946 2016-06-09 12:38:45 Dana Myers Re: Using 1/2 Tayloe detector with downconversion/image rejection in