Patent Expiration - Tayloe Mixer / Quadrature Sampling Detector

[allenc]

Twenty years ago a patent application for "Product detector and method therefore" was filed by Motorola with Daniel Richard Tayloe (N7VE) as the inventor (1998-10-15).  Two and a half years later the US patent was granted and the number 6230000 assigned.  Given the rules in force, then and now, this patent should now be expired.  (In the US... not sure about the WO patent grant.)

There has been plenty of controversy about the novelty of this invention, some of it here on EEVblog.  Quadrature sampling had previously been used for commutating mixers and filters, SSB detectors and other designs, some apparently going back to the earliest days of audio and radio electronics depending on how far one wishes to stretch it.  In any case, it has expired so perhaps this technique can finally be discussed without devolving into another squabble about patent law and prior art.

The patent was not used to stifle experimenters and many projects inspired by the QSD technique have been pursued.  In 2002 Gerald Youngblood (K5SDR,) founder of Flex-Radio Systems published a four part series in ARRL's QEX; A Software-Defined Radio for the Masses[1].  The series details the theory, practical circuit and signal processing software involved in modulating and demodulating RF using an audio ADC/DAC.  The earliest Flex-Radio models were based on the QSD.  Other widely know products have incorporated variants of the technique, including classics such as the Softrock transceivers, high performance systems such as the Elecraft KX3 and recent designs such as HobbyPCB's RS-HFIQ.

The name of the device is somewhat problematic.  Informally "Tayloe mixer" is used frequently.  In presentations and papers one can find quadrature sampling detector[2], quadrature product detector[3], Tayloe quadrature detector[4] and others.  Forgive my choice if it doesn't match your view.

[1] https://sites.google.com/site/thesdrinstitute/A-Software-Defined-Radio-for-the-Masses
[2] http://www.wb5rvz.com/sdr/ensemble_rx_ii_vhf/06_qsd.htm, https://qrp-labs.com/images/news/dayton2018/seminar.pdf
[3] http://www.norcalqrp.org/files/Tayloe_mixer_x3a.pdf
[4] http://www.norcalqrp.org/files/NC2030/NC2030_v5.pdf

[Howardlong]

Without doubt, the Gerald Youngblood series of articles "A Software-Defined Radio for the Masses" are one of the best introductions to SDR.

I found that by far most of the other attempts are too abstract and academically based, lacking any real spelt-out practical application.

The other essential text is Richard Lyons' "Understanding Digital Signal Processing", and in particular the chapter on quadrature signals.

[coppice]

The name of the device is somewhat problematic.  Informally "Tayloe mixer" is used frequently.  In presentations and papers one can find quadrature sampling detector[2], quadrature product detector[3], Tayloe quadrature detector[4] and others.  Forgive my choice if it doesn't match your view.

Isn't the patent about Tayloe's specific simple way to implement quadrature sampling detection? There is no way someone could have got a generalised patent on quadrature sampled detection 20 years ago.

[coppice]

The other essential text is Richard Lyons' "Understanding Digital Signal Processing", and in particular the chapter on quadrature signals.

That is a very well regarded beginner's book on DSP.

[Yansi]

Ping! This may become an interesting thread.

[RoGeorge]

https://patents.google.com/patent/US6230000

[coppice]

https://patents.google.com/patent/US6230000

The patent is specifically about the use of 4 synchronously excited single pole LPFs, used to create medium term averages at the I and Q points of the carrier. For a system where the bandwidth of the baseband is a small fraction of the carrier frequency this cheap simple scheme works well.

[mark03]

The patent was not used to stifle experimenters and many projects inspired by the QSD technique have been pursued.

Not to nit-pick but AFAIK patents cannot be used to stifle experimenters in any case.  The whole point of a patent is to give the inventor an incentive to fully describe their invention to the public, thereby spurring further innovation.  In exchange they get exclusive rights in the commercial sphere, but they can't stop experimenters from using the patent, discussing it, sharing implementations of it, blogging about it, etc.  Otherwise I'd be in trouble... I'm using a patented idea right now in my field-mill project, and I fully intend to write it up afterwards, with some minor critiques.

[ebastler]

Twenty years ago a patent application for "Product detector and method therefore" was filed by Motorola with Daniel Richard Tayloe (N7VE) as the inventor (1998-10-15).  Two and a half years later the US patent was granted and the number 6230000 assigned.  Given the rules in force, then and now, this patent should now be expired.  (In the US... not sure about the WO patent grant.)

Not to nitpick further, but the maximum term of a patent is 20 years from the date it was granted. Since the maintenance fees have been paid regularly on US 6,230,000, it will be valid until May 2021. As mentioned by Mark03, that can't stop you from using the technology for amateur purposes or research. But don't sell that design or product yet, if you want to stay on the safe side!

Sorry, that was wrong -- see coppice's correction below. For the US 6,230,000 patent, since it was filed after 1995, the term is indeed 20 years from the filing (priority) date.

[coppice]

Not to nitpick further, but the maximum term of a patent is 20 years from the date it was granted. Since the maintenance fees have been paid regularly on US 6,230,000, it will be valid until May 2021.

In the US the term for patents is 20 years from the date of initial application, or 17 years from issue. Many other countries have slightly shorter terms.

[ebastler]

In the US the term for patents is 20 years from the date of initial application, or 17 years from issue. Many other countries have slightly shorter terms.

Yes, you are right of course. Sorry, I got confused with the (still somewhat recent) change from "17 years after grant" to "20 years from filing date". I will correct my post above.

[allenc]

Has the quadrature product/sampling detector appeared in any commercial applications aside from amateur radio products?

I don't believe anyone has integrated the device, meaning integrating the clock, bus switch, integrating capacitors, op amps, etc. into a single device.  I'd be interested in learning otherwise.

[rhb]

I just started a thread on monitoring the amateur HF allocation, displaying it as a waterfall and storing the IQ streams for each band to disk.  Central is an Si5332 8 output clock driving 8 Tayloe mixers which then get low pass filtered and fed to 1.6 MSa/S 16 bit ADCs.  Those digital  IQ streams then get processed as desired.

[in3otd]

Has the quadrature product/sampling detector appeared in any commercial applications aside from amateur radio products?

Not really useful for experimenting but it has been likely in every cellphone made in the past 15+ years, also in much more sophisticated forms. See e.g. this /https://www.researchgate.net/publication/224070273_A_low_noise_quadrature_subsampling_mixer (behind paywall, unfortunately)

[Yansi]

I just started a thread on monitoring the amateur HF allocation, displaying it as a waterfall and storing the IQ streams for each band to disk.  Central is an Si5332 8 output clock driving 8 Tayloe mixers which then get low pass filtered and fed to 1.6 MSa/S 16 bit ADCs.  Those digital  IQ streams then get processed as desired.

I would have thought that direct sampling would yield way better result, than having a quadrature mixer with miserable balance and spur rejection.  You can always get your I and Q signal in post processing, with way more precision that the HW can ever produce.

[rhb]

I just started a thread on monitoring the amateur HF allocation, displaying it as a waterfall and storing the IQ streams for each band to disk.  Central is an Si5332 8 output clock driving 8 Tayloe mixers which then get low pass filtered and fed to 1.6 MSa/S 16 bit ADCs.  Those digital  IQ streams then get processed as desired.

I would have thought that direct sampling would yield way better result, than having a quadrature mixer with miserable balance and spur rejection.  You can always get your I and Q signal in post processing, with way more precision that the HW can ever produce.

Direct sampling in a single passband from 160 - 1.25 m is possible, but it's not cheap.  You also have the issue of the dynamic range loss due to out of ham band signals.  I think miserable balance and spur rejection is more the responsibility of the builder.  There are some very high performance receivers built around a quadrature mixer that are entirely analog.

www.norcalqrp.org/files/NC2030/NC2030_v5.pdf

Any errors in direct sampling will appear in the analytic signal which is what I & Q form.  I can see no fundamental reason that a quadrature mixer has to be better or worse than any other mixer type.

https://essay.utwente.nl/58276/1/scriptie_Soer.pdf

Sampling analog IQ streams is really just a way of interleaving 2 ADCs in order to meet the Shannon-Nyquist sampling requirements.  There are lots of advantages to having the analytic signal, but as you point out, that's easy to obtain once you're in the digital domain.

[Yansi]

Direct sampling in a single passband from 160 - 1.25 m is possible, but it's not cheap.  You also have the issue of the dynamic range loss due to out of ham band signals.  I think miserable balance and spur rejection is more the responsibility of the builder.  There are some very high performance receivers built around a quadrature mixer that are entirely analog.

Wait what? I thought we are in 2018 and that even sampling the whole HF spectrum (0-30MHz) comes quite cheap these days.   

And btw, I did NOT say single passband. Instead of using cruel quadrature mixers with all sorts of errors, just downcovert to whatever suitable common IF frequency (i.e. 450kHz with desirable bandwidth), and sample that.

You still need 8 mixers for 8 bands, but just 8 ADCs instead of 16 of them. And also simple mixers, that will not introduce other problems, than linearity issues for strong signals.

Maybe it comes to how do you define cheap. I doubt that your 8 quadrature mixers plus 16 ADCs and all associated circuitry will come significantly cheaper, than sampling the whole HF band.

[rhb]

The amateur allocation is 4.272 MHz below 30 MHz and another 12 MHz between 50 and 225 MHz.  You'd need 100 MSa/S to get to 30 MHz.  If you just covered a part of 10 m you could do most of HF with just an STM32F4xx.

As for ADCs, the STM32F4 has 7.2 MSa/S at 12 bits across 24 channels and the LPC4370 has 80 MSa/S at 12 bits across 4 channels.  I've not found any ADCs with similar specs at that low a price.  If you know of some, I'd love to hear about them.

An STM32F410 is ~$10 as is an LPC4370.  And dev boards are about $20.  The good BP filters will probably cost as much as the rest of the hardware.  There are 15 allocations below 225 MHz.  Two LF/VLF, 11 HF bands and two VHF bands.  I plan to skip the LF/VLF allocations as I don't have room for antennas for those.

One could as easily use the ADCs with a regular mixer.  That would actually save an LPC4370.  That only has a 4 channel MUX, so one can only sample 2 of the 10 - 1.25 m allocations in IQ form.  Hence the need for two LPC4370s if you sample I & Q.

The FET switches for a Tayloe mixer are fairly cheap.  The big advantage is being able to use a $10 Si5332 to generate the square wave clocks for 11 quadrature mixers.  I still haven't identified a good clock for 2 and 1.25 m which exceeds the maximum output of the Si5332.

You really won't get very far with a 455 KHz IF and a 4 MHz wide band.

One could select a common IF of say 70 MHz and line up all the bands across a 16 MHz wide IF and then sample that.  If you just did 160 - 6 m you could just feed the IF to an SDRplay RSP2.

Previously I was always inclined to a single ADC.

I'm going to get some FET switches and dead bug a mixer as I now have the test gear for that which was not the case this time last year.

I can't find it at present, but there's an example of a modular, very high performance 40 m transceiver that has something like 120-140dB of image rejection using a Tayloe mixer.The builder changes out the modules when he finds a better design for that section.  So it is constantly changing.

[Yansi]

You seem a bit mislead, or not?

There are only 3 ADCs in the STM32F4. To achieve 7.2MSps, you need to use all three in an interleave mode. So only single channel can be sampled. And at 7Msps (2.4M per ADC), don't really expect 12 ENOB.  Especially on older STM32F4 devices (40x, 42x), the ADC is pure CRAP! Not sure how others like (41x, 46x,  etc), but I would not expect any much better.  So the maximum here you get is 1 channel at 7.2Msp, 3 channels of 2.4Msps or say 6 at 1.2Msps. Well, better than nothing, but...

I am not sure how have you thought to utilize these to sample the whole HF. I do not have any clue 

Regarding standalone fast ADCs, I do know a few, for example AD9283. Only 8 bit, up to 100Msps and to be considered cheap. Can be obtained even cheaper if you know where to look. I know, 8 bit is not much to cover HF, but certainly cheap and gets you somewhere.

Then there was a time, where MAX1420 was available, now obsolete and not available anymore.  60MSps, 12bits. Not really for a full HF spectrum at once, but better than nothing, uh? Was not even that expensive back then.  Still having a few pieces laying somewhere.

Then there is an interesting one: AD9609, available at $10, up to 80Msps and 10bits.  That might start to do something. Interesting to note its huge analog input bandwidth.

The real king of ADCs here would be LTC2216, but this one is what I consider a f*king EXPENSIVE one.  This one I think is used in the WebSDR at U Twente in Netherlands.

I do not know from head any others, these were what I know about, tried to use (AD9283) or have available.

I think that even in the $10 category that something interesting would be possible with the AD9609.
Still, you need to add the FPGA cost to it

You really won't get very far with a 455 KHz IF and a 4 MHz wide band.

Which HAM band is 4MHz wide? I think none of the HF ones are. Like a fraction of a MHz, 300k at most I think.  Or if the BW is so much issue, just select higher IF frequency. It could then even be undersampled (if the ADC has enough input analog bandwidth).

[rhb]

10 m is 1.7 MHz and 6 m is 4 MHz.   I lumped 6 m in with HF.   If you define VHF as "above 30 MHz" then that's not correct.  But it doesn't change the overall picture.

6 m is the band that seems would be the most interesting to be able to review a waterfall replayed at 30x real time.  One gets long distance ducting, particularly over water which is unpredictable.  If you saw there was an opening at 4 am across the Atlantic or Pacific last night, it might be worth setting an alarm clock to see if it repeats tonight.

The MCUs have multiplexers built in.  So you can sample up to 24 channels with the 3 ADCs in the STM32F4 which, as I read the reference manual and datasheet, can be used in any combination of 1, 2 or 3.

That gives six 600 KHz sub-bands with the three ADCs.  Or four 600 KHz sub-bands and four 300 KHz sub-bands for 60, 30, 17 and 12 m. So 2 ADCs are sampling 2 channels alternately, and one is sampling 4 channels using every 4th sample for each channel. So that covers 160 to 12 m using just the STM32F4 and  regular mixers. For a quadrature mixer version, one doubles the number of channels and halves the sample rates on each channel.

One attractive aspect of using the MCUs is that you can apply a digital filter and trim the BW back to the allocation and small guard bands before passing it to an FPGA or CPU for further processing.

A 4K monitor implies 4K FFT lengths which should not be difficult to do.  Even shorter FFTs if you transform each band separately.  So a fast ARM board might be able to do all the processing without requiring an FPGA.

Completely off topic, did you ever resolve the problem of getting good forward and reverse port isolation from a DIY directional coupler?  I was surprised and frustrated that I could not find much information about designing directional couplers.

[Yansi]

I do not think that you can use the interleave mode with multiplexing.  Channel would not get switched until the last conversion of the triplet ends. It will result in non-symmetrically triggered samples, lower overall sample rate and other troubles. But that doesn't matter.

I would not consider the 6m band (50MHz) a HF band. It is not much used here in EU as it might in other places and if my memory serves me correctly 6m is only 2MHz wide in EU (ITU region 1).  At least I do not hear about 50MHz band here almost at all. But it very well may depend on those I go out with. But that does not matter either.

One attractive aspect of using the MCUs is that you can apply a digital filter and trim the BW back to the allocation and small guard bands before passing it to an FPGA or CPU for further processing.

Why would one use MCU as the frontend for an FPGA (or any other MCU)? Especially such as STM32F4 with extremely horrible ADCs?  Does not make any sense to me.  Why not getting a proper beefy dedicated ADC with reasonable performance and sticking it right to the FPGA? Does not matter if one 100MSps ADC or a couple of "slower" ones.

I am not sure if a decent SDR could be done with such small ARM MCU. Yes, I have seen a lot of different builds of such, even with STM32 but they topped at like 200ksps (using 192kHz audio ADCs as the frontend). Cheap ARM MCUs are really neat, but ARM is extremely inefficient when handling a lot of data (memory intensive tasks). And it also sucks, that ST can not still figure out how to implement DDR memory controller to their chips. The 100MHz only SDR memory interface is a nasty bottleneck for MANY applications. But that is a completely another matter (and certainly not the only one) to rant about 

Thank you for asking, I left the directional coupler project alone. Both from the reasons  I could not get any interesting performance out of it and that I haven't had any suitable RF generator handy to test these back then. Now owning a crusty (but functional!) Agilent E4432B I could probably return to many projects and have a second go with them.  Having some proper test equipment available eases a lot of frustration!

[rhb]

I am not sure if a decent SDR could be done with such small ARM MCU. Yes, I have seen a lot of different builds of such, even with STM32 but they topped at like 200ksps (using 192kHz audio ADCs as the frontend). Cheap ARM MCUs are really neat, but ARM is extremely inefficient when handling a lot of data (memory intensive tasks). And it also sucks, that ST can not still figure out how to implement DDR memory controller to their chips. The 100MHz only SDR memory interface is a nasty bottleneck for MANY applications. But that is a completely another matter (and certainly not the only one) to rant about 

Thank you for asking, I left the directional coupler project alone. Both from the reasons  I could not get any interesting performance out of it and that I haven't had any suitable RF generator handy to test these back then. Now owning a crusty (but functional!) Agilent E4432B I could probably return to many projects and have a second go with them.  Having some proper test equipment available eases a lot of frustration!

Very important point, this is not intended as an SDR.  The sole function is monitoring all the allocations for band openings on a 24 x 7 basis.  All I expect the MCUs to do is sample, a bit of cleanup and downsampling and then write the datastream via GPIO to whatever does the bulk of the work such as an RPi,  BeagleBone or Zynq.   I only intend to compute the power spectra, stuff it in the video buffer and scribble the datastreams to disk for use with an SDR at a later time.  So no demodulation, etc.  Just a 4-6 TB disk drive to buffer the last 2-3 days of activity.  I got fascinated by watching waterfalls on my SDRplay RSP2.

I finally overcame my habitual frugality last fall and spent the price of a small car on test gear.  At 65 I figured it was time.  If the MCU ADCs aren't adequate I'll get something else.  I already have everything except the BP filters and the Si5332 clock.  So I plan to prototype operation on one or two bands,  evaluate the performance and take it from there.  I now have the luxury of an 8560A w/ TG, 8648C, 33622A, 5386A, 438A w/ 8481D and 8482A sensors, 2 x 34401A, VNWA 3E, Tek 485 and a LeCroy DDA-125/LC684DLX among other things.  All in excellent working condition.  So I hope to get busy once I get everything setup.  The next major hassle is finding space to mount an 8 port KVM switch as I've outgrown my 4 port.  Oh and I got an FT-991A, so after that is designing and  building an antenna and getting relicensed.  And then there's the GPSDOs for the RSP2 and VNWA as well as the HP gear (most of which have the OXO option)

I'm beginning to wonder if the way to address the coupler isn't by simulating designs with openEMS. I came across a coupler design using copper pipe fittings in the 1972 ARRL VHF handbook.  Trying to fit a capable test bench and 6-7 computers into a 2 x 3 meter closet is not easy.  I'm rather proud of the arrangement, but am *really* tired of working on it as I can't do any electronics because of the general mess.

If you find any professional papers on coupler design please let me know.

A close friend likes to say I'm 3 standard deviations out from the norm.  To save money on an antenna I bought $800 of books on antenna design and numerical EMS.

[Yansi]

Okay, so no realtime SDR, just raw data recorder. Got it.

The STM32F4 ADCs, especially F42x are really poor (same as F40x and F20x - those are mostly the same). They are extremely noise susceptible. There is a built-in problem in those chips, that noise gets coupled in to the ADC from the digital circuitry. 

For noise free signal capture, I'd suggest using either different  MCU type (maybe try the newer ones like F446?) where the issue should be solved (should!) or just use a dedicated clean external ADC, which would be I think a correct way to go.

Unfortunately can't afford almost any test equipment, even used. I have to work with what I got, or try building things myself. Good on ya you have money well spent.

[radioactive]

You might find this project an interesting reference.  I ran across it a while back.  Good medium-frequency (MF) SDR performance on an stm32F4xx utilizing the on-board ADC.  The code should be interesting for your project.  It utilizes fast-convolutional filters via FFT, NCO tuning, and other DSP techniques to achieve a very respectable performance with little more than adding a simple external amp / low-pass anti-aliasing filter to the discovery board.

http://www.sdradio.eu/weaksignals/armradio/

[KE5FX]

You might find this project an interesting reference.  I ran across it a while back.  Good medium-frequency (MF) SDR performance on an stm32F4xx utilizing the on-board ADC.  The code should be interesting for your project.  It utilizes fast-convolutional filters via FFT, NCO tuning, and other DSP techniques to achieve a very respectable performance with little more than adding a simple external amp / low-pass anti-aliasing filter to the discovery board.

http://www.sdradio.eu/weaksignals/armradio/

That's really cool.  His code is unusually clean and readable by the usual standards of this sort of project    Good demonstration of some of Rick Lyons' nifty DSP hacks.

[SeoulBigChris]

Not to nit-pick but AFAIK patents cannot be used to stifle experimenters in any case.  The whole point of a patent is to give the inventor an incentive to fully describe their invention to the public, thereby spurring further innovation.  In exchange they get exclusive rights in the commercial sphere, but they can't stop experimenters from using the patent, discussing it, sharing implementations of it, blogging about it, etc.  Otherwise I'd be in trouble... I'm using a patented idea right now in my field-mill project, and I fully intend to write it up afterwards, with some minor critiques.

I personally ran into this years ago, and to my shock, it is indeed true that a patent DOES grant the holder the right to stifle experiments. Technically speaking, a patent holder can block ANY usage of his patent, even in labs, school classrooms, etc.  Unlike copyright, there is no such thing as "fair use".  Practically speaking, many inventors do allow such benign usage. (Disclaimer, this is what my client's lawyers, Sony's lawyers, and a lawyer friend of mine from college told me maybe 15 years ago - they maybe lied to me or things have changed).

What brought this to my attention was a client who had an unusual idea that was going to be very difficult (expensive) to develop and build. Furthermore it seemed quite likely that somebody had already done it before.  I searched around for any existing products, and ran into a patent assigned to Sony that was exactly what we needed. I contacted the inventor and said we were interested in his technology for a client, and asked if there were any products already on the market which we could purchase.  A week later we received a really nasty letter from Sony telling us to quit doing any lab experiments with this technology or face a big lawsuit.

[Yansi]

Interesting. I have seen this SDR radio on youtube, but did not notice it has all resources available in such a nice form.

Somebody really wants me to have a go with SDR, uh?    (if I had the damn free time to spend here, I would...)

What I do not understand, is what is the reason for the FFT? Aren't there a more efficient ways to filter the baseband signal?

[ogden]

What I do not understand, is what is the reason for the FFT? Aren't there a more efficient ways to filter the baseband signal?

It depends. If you want very sharp filter that would require many FIR taps (>= 64), then FFT convolution filter is more efficient (than FIR). Obviously particular SDR is the case.

[Yansi]

Maybe a stupid question I did not found the answer for, but why can't you use a non-linear phase filter for the baseband filtering?  With a couple of biquads (IIR) you could get a very steep rolloff with a little computational complexity.

[coppice]

Maybe a stupid question I did not found the answer for, but why can't you use a non-linear phase filter for the baseband filtering?  With a couple of biquads (IIR) you could get a very steep rolloff with a little computational complexity.

Once you have scrambled the phase of the signal, what will be you next step in the signal chain?

[Yansi]

Then I have to ask obvious: There are many receiver constructions out there utilizing I and Q signal paths, but processing them in analog domain. There is none linear phase filter in the analog world, is it? How possible, that in analog world one can do with non-linear phase and in digital world, this does not work?

[rhb]

Okay, so no realtime SDR, just raw data recorder. Got it.

The STM32F4 ADCs, especially F42x are really poor (same as F40x and F20x - those are mostly the same). They are extremely noise susceptible. There is a built-in problem in those chips, that noise gets coupled in to the ADC from the digital circuitry. 

For noise free signal capture, I'd suggest using either different  MCU type (maybe try the newer ones like F446?) where the issue should be solved (should!) or just use a dedicated clean external ADC, which would be I think a correct way to go.

Unfortunately can't afford almost any test equipment, even used. I have to work with what I got, or try building things myself. Good on ya you have money well spent.

I was in the same boat over the cost of test gear for so long and was so used to it, that it was only watching my brother in-law rapidly deteriorate from Parkinson's and having another friend diagnosed with cancer after a couple of others died from it and two more died of heart attacks out of the blue that loosened my purse strings.  I was almost obsessively frugal during my working career saving for my old age.  I decided that 65 was old enough to start spending some.  Almost all my gear is 20 years old, but service data is available and most parts as well.  I managed after a *lot* of searching to get the 8560A for under $1500 with shipping.  I had bought a Siglent SSA3021X, but returned it because it was so buggy.

 But I did learn quite a lot struggling, though much nicer not to have to struggle. It also left me with a strong obsession with designing and building low cost test gear.

I've been very impressed by your projects.  Very nice work.

I found the ST documents about the ADC noise issues and their suggested mitigation strategies.  Thanks for the heads up on that.  I had no idea it was such a mess.  It makes me wonder if it's actually possible to do what I want.  By the time I stop doing anything to sample at full capacity, I may not be able to handle the processing.

I have the STM32F429 board that Alberto used and that's been my candidate for prototyping.  But for actually building the gadget I was planning to just use bare MCUs for the sampling.  So I'll make a point of looking at all the available chips and the ADC implementations to pick one that is not too bad.

I have found searching for ADCs by specification a huge pain both using Digikey and AD.

[Leo Bodnar]

Google patents https://patents.google.com/patent/US6230000B1/en now list this as properly expired
 

Application US09/173,030 events
1998-10-15  Application filed by Motorola Solutions Inc
1998-10-15  Priority to US09/173,030
2001-05-08  Application granted
2001-05-08  Publication of US6230000B1
2018-10-15  Anticipated expiration
2020-01-01  Application status is Expired - Lifetime

[German_EE]

Until Verizon killed it off Dan Tayloe was a regular contributor to the Experimental Methods of RF Design Yahoo Group. He always seemed to have something useful to say.

[unitedatoms]

Has the quadrature product/sampling detector appeared in any commercial applications aside from amateur radio products?

I don't believe anyone has integrated the device, meaning integrating the clock, bus switch, integrating capacitors, op amps, etc. into a single device.  I'd be interested in learning otherwise.

One can look at service manual of HP8275A4275A circa 1979 (Yokogawa HP) and see commercial application. Look at phase detector schematics. I am not a patent guru, so in my humble opinion, Motorola's patent is bullsh*t. And b.t.w. mechanical synchronous rectifiers existed for century at that point, so HP is not perfectly original either.

Edit: sorry typo in model name

[Leo Bodnar]

I have a Rubidium oscillator that uses a synchronous detector based on CD4016 CMOS switches driven by CD4013 divided LO.  It has been designed, made and sold in 1974.

It is as if someone patented a novel application of four diodes in the form of a bridge rectifier.

Leo

[ogden]

Why you guys are so cryptic? Any pointers to Rubidium Clock and maybe it's schematics/service.manual? Also - when searching for HP8275A (supposedly it is RF instrument?), search results are automotive brake pads for Mazda 

[unitedatoms]

Nothing cryptic, I assumed that the references to published HP schematics can be accepted by name, and not necessarily by URL.

This service manual
https://doc.xdevs.com/doc/HP_Agilent_Keysight/HP%204275A%20Service.pdf

The figure 8-55. On pdf it is approximately page 283. Part of the diagram labeled "Phase Detector".


The reason I am so aware about this particular detector is that I am trying to reproduce something like this (high end 10MHz LCR Meter) using contemporary parts.

[coppice]

I have a Rubidium oscillator that uses a synchronous detector based on CD4016 CMOS switches driven by CD4013 divided LO.  It has been designed, made and sold in 1974.

It is as if someone patented a novel application of four diodes in the form of a bridge rectifier.

Leo

Try looking back earlier in this thread to what the Tayloe patent is actually about.

[ogden]

Nothing cryptic, I assumed that the references to published HP schematics can be accepted by name

When you provide correct name then yes indeed. Seems you noticed typo & fixed 

[ogden]

I have a Rubidium oscillator that uses a synchronous detector based on CD4016 CMOS switches driven by CD4013 divided LO.  It has been designed, made and sold in 1974.

It is as if someone patented a novel application of four diodes in the form of a bridge rectifier.

Leo

Try looking back earlier in this thread to what the Tayloe patent is actually about.

Isn't it about "synchronous detector based on CMOS switches"? Please elaborate

[coppice]

I have a Rubidium oscillator that uses a synchronous detector based on CD4016 CMOS switches driven by CD4013 divided LO.  It has been designed, made and sold in 1974.

It is as if someone patented a novel application of four diodes in the form of a bridge rectifier.

Leo

Try looking back earlier in this thread to what the Tayloe patent is actually about.

Isn't it about "synchronous detector based on CMOS switches"? Please elaborate

To repeat what I said earlier in this thread:

The Tayloe patent is specifically about the use of 4 synchronously excited single pole LPFs, used to create medium term averages at the I and Q points of the carrier. For a system where the bandwidth of the baseband is a small fraction of the carrier frequency this cheap simple scheme works well.

[Leo Bodnar]

http://bama.edebris.com/manuals/efratom/frk

[Leo Bodnar]

We probably have to agree to disagree but this is such an obvious use of components in a way they were intended to be used.
I read the patent several times and still don't understand what makes it patentable.
Peace!
Leo

The Tayloe patent is specifically about the use of 4 synchronously excited single pole LPFs, used to create medium term averages at the I and Q points of the carrier. For a system where the bandwidth of the baseband is a small fraction of the carrier frequency this cheap simple scheme works well.

[chris_leyson]

In the paper "Single sideband modulation using sequence asymmetric polyphase networks"
M.J. Gingell, January 1973. I'm sure there was a diagram of a four way commutator followed by a polyphase filter or it might have been the other way around, polyphase filter followed by commutator. It's not obvious from his US patent US3559042

Of course this could be false memory and there was no commutator in the paper cited above.  Then again it could have been another paper by M.J. Gingell and D.R. Barber published either '73 or '74.
I think STC published the paper in one of their magazines so it wouldn't have been widely circulated.
Gingle and Barber have a patent for "N-Path Frequency Translation System" US3562556.

I might still have a copy of the Gingell, Barber paper somewhere. From a polyphase filter point of view you can trace this back to the early 1970's. My Google-Fu isn't working today so N-Path polyphase or asymetric sequence filters might be a better search term.

[coppice]

In the paper "Single sideband modulation using sequence asymmetric polyphase networks"
M.J. Gingell, January 1973. I'm sure there was a diagram of a four way commutator followed by a polyphase filter or it might have been the other way around, polyphase filter followed by commutator. It's not obvious from his US patent US3559042

That patent was filed in 1969, so it was 4 years before the paper you cited. He might have taken his ideas a lot further in 4 years. However, nothing in US3559042 is like the Tayloe configuration, although its not that far away. It has the 4 single pole filters, but they are sampling in a different way. Looking at those patent diagrams it seems strange that he didn't add one with the very simple and symmetric configuration Tayloe used. but then most things seems obvious once someone has pointed them out.