Hi Jan, this is precisely the instrument that is lacking at a hobbyist price. It would be excellent to have the possibility of measuring phase noise. Can you anticipate the features of the Sampling DMTD? Can it be used with Timelab? We are waiting for your new ones. Luciano Da "time-nuts" time-nuts-bounces@lists.febo.com A "Discussion of precise time and frequency measurement" time-nuts@lists.febo.com Cc Data Sat, 14 Sep 2019 14:25:48 +0200 Oggetto Re: [time-nuts] A simple sampling DMTD Update: I have finished routing the board (placement diagram at http://www.lartmaker.nl/time-nuts/DMTD%20rev1.00%20assembly.pdf ) and ordered a few prototype PCBs. After the earlier discussions on the list I've grown sufficiently concerned about the impact of 1/f converter noise that I have added headers to the board to allow me to replace the D-flipflop sampler with an FPGA-based I/Q downconverter. While the main PCBs are in production I'll draw a simple daughterboard with dual ice40 UltraPlus FPGAs, If the FPGA solution turns out to be necessary (or a noticeable improvement), I'll redraw the main PCB. To be continued, JDB. On Sun, Sep 1, 2019 at 2:09 AM Jan-Derk Bakker <jdbakker@gmail.com> wrote: > Dear all, > > I've been working on a design for a (relatively) simple, standalone > sampling DMTD. Very rough preliminary schematics can be found at > http://www.lartmaker.nl/time-nuts/DMTD_rev0.99.pdf . > > Design goals are: > - ps-level accuracy > - comparison of frequencies between at least 10 and 50MHz, preferably > between 1 and 100MHz > - comparison of (selected) different frequencies (in my case: 10MHz vs > 50MHz) > - standalone operation, field-portable > - option for raw data sampling / (post)processing on a PC > - option for generating a tuning voltage to lock the measured oscillator > to the reference, so the DMTD can act as a PLL in phase noise test setups > > Context: you may remember that a year or two ago I posted to time-nuts > about a GPSDO-design geared for mobile applications, which I was working on > for an SDR-platform my students are working with. This SDR-platform has now > grown to include a 100-channel phased array receiver. To validate the > reference clock distribution in this array (amongst other things) I would > like to have a DMTD. As the commercial offerings are outside the budget of > our lab, I was planning to roll my own. > > The core of the system is a transformer-coupled LTC2140-14 dual 14-bit > ADC, sampling at an offset frequency of nominally 10MHz+10Hz generated by a > VCTCXO (with an option for an OCXO). The ADC was chosen for its large input > bandwidth and small aperture jitter. Simulations of a simple software ZCD > consisting of a digital filter and least-squares fitting showed that > 100ksps would be more than enough to get the desired accuracy. As the ADC > design is unable to achieve sample rates lower than 1MSPS, D-flipflops are > used to decimate the samples. These DFFs are also used to multiplex the > 2x14-bit samples to an 8-bit data bus going into one of the GPIO-ports of > an XMega. The XMega runs the ZCD, and generates a tuning voltage for the > offset oscillator. Communication to a logging PC is done with a > galvanically isolated FT2232H, which has both an ASCII COM-port for the ZCD > data and a control interface and an asynchronous FIFO to transfer raw > samples. System power comes from the isolated USB bus or a barrel jack; BOM > cost in qty10+ is around 100US$. > > (The DMTD has a few more power rails than I would have liked. Originally I > had planned to use the LTC2295 and have a 3v3-only system, but after > re-reading the NIST paper on SDR-as-a-DMTD I concluded that the single > clocking path of the 2140 would likely have better aperture jitter > correlation between the channels. As a 1.8V/10MHz XMega would only be > borderline able to handle the computations, I ended up with this design. > LVC logic is used to go from 3.3V->1.8V, LV1T translators for the opposite > direction.) > > Design decisions and/or non-goals: > - I considered putting a small FPGA (specifically a Lattice ice40 > UltraPlus) between the ADC and the processor. This was rejected because the > performance of the decimator appeared to be sufficient, and I wasn't > certain that I could get DDR mode + a CORDIC working in this FPGA. > - Especially when I found the necessity to move part of the system to 1.8V > I considered moving to an ARM. I stuck with the XMega as performance was > sufficient, and I am very familiar with both the CPU and the peripherals > (particularly time-stamping counters and the Event system) that would ease > the ZCD implementation and issues like synchronization between processor > and sampling system. > - I looked into integrating a phase noise measurement, but could find no > easy way that wouldn't degrade DMTD operation in the process. The tuning > voltage output is an inexpensive compromise (as I still had a DAC and > enough cycles to spare) > - The main thing I'm unsure about is the effect of the balun on phase > performance wrt temperature and termination matching. I've kept to the > baluns as they add less noise than a fully differential amplifier would. > > While I've made this design for my own purposes, I would be more than > happy to put it under an Open Hardware-license and/or work with TAPR or > other parties to get it distributed, should there be interest. > > Thoughts? > > with kind regards, > > Jan-Derk Bakker > [planning to start board layout tomorrow; looks like this should > definitely fit on a 100x160mm Eurocard inside a Hammond 1455-series box] > _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.

Hi You are not going to get useful phase noise data out of a normal DMTD …. The DMTD is targeted at ADEV and similar long term / high accuracy measurements. If a DMTD “makes it” to a 10 Hz beat note that is about as far as most people take it. Bob > On Sep 16, 2019, at 10:41 AM, timeok@timeok.it wrote: > > > Hi Jan, > > this is precisely the instrument that is lacking at a hobbyist price. > > It would be excellent to have the possibility of measuring phase noise. > > Can you anticipate the features of the Sampling DMTD? > > Can it be used with Timelab? > > We are waiting for your new ones. > > Luciano > > > Da "time-nuts" time-nuts-bounces@lists.febo.com > A "Discussion of precise time and frequency measurement" time-nuts@lists.febo.com > Cc > Data Sat, 14 Sep 2019 14:25:48 +0200 > Oggetto Re: [time-nuts] A simple sampling DMTD > Update: I have finished routing the board (placement diagram at > http://www.lartmaker.nl/time-nuts/DMTD%20rev1.00%20assembly.pdf ) and > ordered a few prototype PCBs. > > After the earlier discussions on the list I've grown sufficiently concerned > about the impact of 1/f converter noise that I have added headers to the > board to allow me to replace the D-flipflop sampler with an FPGA-based I/Q > downconverter. While the main PCBs are in production I'll draw a simple > daughterboard with dual ice40 UltraPlus FPGAs, If the FPGA solution turns > out to be necessary (or a noticeable improvement), I'll redraw the main PCB. > > To be continued, > > JDB. > > On Sun, Sep 1, 2019 at 2:09 AM Jan-Derk Bakker <jdbakker@gmail.com> wrote: > >> Dear all, >> >> I've been working on a design for a (relatively) simple, standalone >> sampling DMTD. Very rough preliminary schematics can be found at >> http://www.lartmaker.nl/time-nuts/DMTD_rev0.99.pdf . >> >> Design goals are: >> - ps-level accuracy >> - comparison of frequencies between at least 10 and 50MHz, preferably >> between 1 and 100MHz >> - comparison of (selected) different frequencies (in my case: 10MHz vs >> 50MHz) >> - standalone operation, field-portable >> - option for raw data sampling / (post)processing on a PC >> - option for generating a tuning voltage to lock the measured oscillator >> to the reference, so the DMTD can act as a PLL in phase noise test setups >> >> Context: you may remember that a year or two ago I posted to time-nuts >> about a GPSDO-design geared for mobile applications, which I was working on >> for an SDR-platform my students are working with. This SDR-platform has now >> grown to include a 100-channel phased array receiver. To validate the >> reference clock distribution in this array (amongst other things) I would >> like to have a DMTD. As the commercial offerings are outside the budget of >> our lab, I was planning to roll my own. >> >> The core of the system is a transformer-coupled LTC2140-14 dual 14-bit >> ADC, sampling at an offset frequency of nominally 10MHz+10Hz generated by a >> VCTCXO (with an option for an OCXO). The ADC was chosen for its large input >> bandwidth and small aperture jitter. Simulations of a simple software ZCD >> consisting of a digital filter and least-squares fitting showed that >> 100ksps would be more than enough to get the desired accuracy. As the ADC >> design is unable to achieve sample rates lower than 1MSPS, D-flipflops are >> used to decimate the samples. These DFFs are also used to multiplex the >> 2x14-bit samples to an 8-bit data bus going into one of the GPIO-ports of >> an XMega. The XMega runs the ZCD, and generates a tuning voltage for the >> offset oscillator. Communication to a logging PC is done with a >> galvanically isolated FT2232H, which has both an ASCII COM-port for the ZCD >> data and a control interface and an asynchronous FIFO to transfer raw >> samples. System power comes from the isolated USB bus or a barrel jack; BOM >> cost in qty10+ is around 100US$. >> >> (The DMTD has a few more power rails than I would have liked. Originally I >> had planned to use the LTC2295 and have a 3v3-only system, but after >> re-reading the NIST paper on SDR-as-a-DMTD I concluded that the single >> clocking path of the 2140 would likely have better aperture jitter >> correlation between the channels. As a 1.8V/10MHz XMega would only be >> borderline able to handle the computations, I ended up with this design. >> LVC logic is used to go from 3.3V->1.8V, LV1T translators for the opposite >> direction.) >> >> Design decisions and/or non-goals: >> - I considered putting a small FPGA (specifically a Lattice ice40 >> UltraPlus) between the ADC and the processor. This was rejected because the >> performance of the decimator appeared to be sufficient, and I wasn't >> certain that I could get DDR mode + a CORDIC working in this FPGA. >> - Especially when I found the necessity to move part of the system to 1.8V >> I considered moving to an ARM. I stuck with the XMega as performance was >> sufficient, and I am very familiar with both the CPU and the peripherals >> (particularly time-stamping counters and the Event system) that would ease >> the ZCD implementation and issues like synchronization between processor >> and sampling system. >> - I looked into integrating a phase noise measurement, but could find no >> easy way that wouldn't degrade DMTD operation in the process. The tuning >> voltage output is an inexpensive compromise (as I still had a DAC and >> enough cycles to spare) >> - The main thing I'm unsure about is the effect of the balun on phase >> performance wrt temperature and termination matching. I've kept to the >> baluns as they add less noise than a fully differential amplifier would. >> >> While I've made this design for my own purposes, I would be more than >> happy to put it under an Open Hardware-license and/or work with TAPR or >> other parties to get it distributed, should there be interest. >> >> Thoughts? >> >> with kind regards, >> >> Jan-Derk Bakker >> [planning to start board layout tomorrow; looks like this should >> definitely fit on a 100x160mm Eurocard inside a Hammond 1455-series box] >> > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there.

I wonder if anyone can shed any light on this question, since this forum is loaded with those who REALLY understand crystals. I am modeling crystal filters (VHF)  in SPICE. There are some specific acoustic mode models for SPICE in some Post Doctorial papers, very interesting, they would be the best but rather painful to use. However I using simplified Rm, Lm, Cm, Cs, Cp, Ccase etc My question is, how does Rm vary with overtone number ? My assumptions are Lm stays the same, Cm reduces proportionally to the square of the overtone number.  Those assumptions are close enough and canon. I of course need the Rm number to acurately model loss. 73 glen english VK1XX

Hi Bottom line - if you are designing a filter, you need the real values for the Cm, Lm and C0. Guessing at them is likely to lead to trouble if it is a reasonably complex filter. Rm generally goes as the overtone. It can deviate quite a bit from that (as can the other parameters) depending on how the blank is shaped and plated. ( If I want a good 3rd overtone, it will be designed to work well there. It may be pretty bad on the fundamental ….). Bob > On Sep 18, 2019, at 2:35 PM, Glen English VK1XX <glenlist@pacificmedia.com.au> wrote: > > I wonder if anyone can shed any light on this question, since this forum is loaded with those who REALLY understand crystals. > > I am modeling crystal filters (VHF) in SPICE. There are some specific acoustic mode models for SPICE in some Post Doctorial papers, very interesting, they would be the best but rather painful to use. > > However I using simplified Rm, Lm, Cm, Cs, Cp, Ccase etc > > My question is, how does Rm vary with overtone number ? > > My assumptions are Lm stays the same, Cm reduces proportionally to the square of the overtone number. Those assumptions are close enough and canon. > > I of course need the Rm number to acurately model loss. > > 73 > > glen english > > VK1XX > > > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there.

Hi Bob thanks for the insight. OK so the Rm for an overtone crystal , measured at the fundamental might be a bad indicator of the overtone Rm.  I have found empirically, a loose relationship of Rm proportional to overtone number   from the fundamental. But loose- I mean +/- 50% which, as you point out, may be optimized for the overtone, which is why my numbers are so far out cheers. On 19/09/2019 10:48 AM, Bob kb8tq wrote: > Hi > > Bottom line - if you are designing a filter, you need the real values for > the Cm, Lm and C0. Guessing at them is likely to lead to trouble if it is > a reasonably complex filter. > > Rm generally goes as the overtone. It can deviate quite a bit from that > (as can the other parameters) depending on how the blank is shaped and > plated. ( If I want a good 3rd overtone, it will be designed to work well there. > It may be pretty bad on the fundamental ….). > > Bob > >> On Sep 18, 2019, at 2:35 PM, Glen English VK1XX <glenlist@pacificmedia.com.au> wrote: >> >> I wonder if anyone can shed any light on this question, since this forum is loaded with those who REALLY understand crystals. >> >> I am modeling crystal filters (VHF) in SPICE. There are some specific acoustic mode models for SPICE in some Post Doctorial papers, very interesting, they would be the best but rather painful to use. >> >> However I using simplified Rm, Lm, Cm, Cs, Cp, Ccase etc >> >> My question is, how does Rm vary with overtone number ? >> >> My assumptions are Lm stays the same, Cm reduces proportionally to the square of the overtone number. Those assumptions are close enough and canon. >> >> I of course need the Rm number to acurately model loss. >> >> 73 >> >> glen english >> >> VK1XX >> >> >> >>

Hi, There is nothing like a fixed ratio between R1 at 3rd or 5th overtone the R1 at fundamental mode. The best approach through C1 and Q. C1 reduces with the square of overtone N (for an infinite crystal plate). In reality C1(3rds about 85% of C1(fund/N^2. For the 5th and higher OT it is about 75 to 70% of C1(fund)/N^2.. Now Q comes into the game: The Q of a crystal designed for 3rd overtone is approximately such that Q*f = 2E12 (f in Hz), for 5th OT is may be 3 to 5*E12 . Fundamental mode crystals have lower Q*f , around 1E12 This all is for a plano-plano AT-cut crystal plates and is only a rule of thumb. It finally depends on the crystal size and shape and some design details. Small crystals with convex shape will have better Q*f at fundamental, mode, crystals with well polished and plaon-parallel surfaces may be better at 3rd OT and have lower Q*f at fundamental mode. C1 at fundamental mode is given by the equation C1(fF) = 0.15*Ael^2*f, with Ael = electrode are in mm^2 and frequency f in MHz With that and with the above estimated values for Q, the resistance can be calculated fundamentally from R1 =1/(2*PI*freq*C1*Q) - here is freq in Hz and C1 in F Have fun Best regards Bernd __________________ I wonder if anyone can shed any light on this question, since this forum is loaded with those who REALLY understand crystals. I am modeling crystal filters (VHF) in SPICE. There are some specific acoustic mode models for SPICE in some Post Doctorial papers, very interesting, they would be the best but rather painful to use. However I using simplified Rm, Lm, Cm, Cs, Cp, Ccase etc My question is, how does Rm vary with overtone number ? My assumptions are Lm stays the same, Cm reduces proportionally to the square of the overtone number. Those assumptions are close enough and canon. I of course need the Rm number to acurately model loss. 73 glen english VK1XX _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.

Sorry, I forgot the static capacitance C0, which is (without holder and strays) C0 (pF) = 0.024*Ael^2*f/N with electrode area Ael in mm^2 and f in MHz Best regards Bernd -----Ursprüngliche Nachricht----- Von: Bernd Neubig [mailto:BNeubig@t-online.de] Gesendet: Samstag, 21. September 2019 08:42 An: 'Discussion of precise time and frequency measurement' <time-nuts@lists.febo.com> Betreff: AW: [time-nuts] overtone crystal question Hi, There is nothing like a fixed ratio between R1 at 3rd or 5th overtone the R1 at fundamental mode. The best approach through C1 and Q. C1 reduces with the square of overtone N (for an infinite crystal plate). In reality C1(3rds about 85% of C1(fund/N^2. For the 5th and higher OT it is about 75 to 70% of C1(fund)/N^2.. Now Q comes into the game: The Q of a crystal designed for 3rd overtone is approximately such that Q*f = 2E12 (f in Hz), for 5th OT is may be 3 to 5*E12 . Fundamental mode crystals have lower Q*f , around 1E12 This all is for a plano-plano AT-cut crystal plates and is only a rule of thumb. It finally depends on the crystal size and shape and some design details. Small crystals with convex shape will have better Q*f at fundamental, mode, crystals with well polished and plaon-parallel surfaces may be better at 3rd OT and have lower Q*f at fundamental mode. C1 at fundamental mode is given by the equation C1(fF) = 0.15*Ael^2*f, with Ael = electrode are in mm^2 and frequency f in MHz With that and with the above estimated values for Q, the resistance can be calculated fundamentally from R1 =1/(2*PI*freq*C1*Q) - here is freq in Hz and C1 in F Have fun Best regards Bernd __________________ I wonder if anyone can shed any light on this question, since this forum is loaded with those who REALLY understand crystals. I am modeling crystal filters (VHF) in SPICE. There are some specific acoustic mode models for SPICE in some Post Doctorial papers, very interesting, they would be the best but rather painful to use. However I using simplified Rm, Lm, Cm, Cs, Cp, Ccase etc My question is, how does Rm vary with overtone number ? My assumptions are Lm stays the same, Cm reduces proportionally to the square of the overtone number. Those assumptions are close enough and canon. I of course need the Rm number to acurately model loss. 73 glen english VK1XX _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.

Bernd, thanks for the excellent treatment on the subject glen On 21/09/2019 4:42 PM, Bernd Neubig wrote: > Hi, > > There is nothing like a fixed ratio between R1 at 3rd or 5th overtone the R1 at fundamental mode. The best approach through C1 and Q. > C1 reduces with the square of overtone N (for an infinite crystal plate). In reality C1(3rds about 85% of C1(fund/N^2. For the 5th and higher OT it is about 75 to 70% of C1(fund)/N^2.. > Now Q comes into the game: > The Q of a crystal designed for 3rd overtone is approximately s