Pages: [1] 2 3 4   Go Down

Author Topic: Baseband FFT system - any h/w and s/w experience?  (Read 705 times)

VK6HP

  • Member
  • Posts: 1222
Baseband FFT system - any h/w and s/w experience?
« on: September 28, 2021, 01:53:44 AM »

I'm looking to put together a baseband FFT analyser for general use, including the analysis of composite and phase noise from non-linear RF power amplifiers via a low-noise heterodyne mix-down arrangement.  I'm looking for something a bit more integrated than a custom MATLAB or similar environment - I guess the modern equivalent of the venerable HP3561 dynamic signal analyser is something like what I have in mind. Users of those instruments will be aware of the great care that went into the front-end design, especially in terms of low-frequency noise.  Having looked through various forums and interest groups I see my aim is a pretty common one but little consensus on good matches between hardware and software emerges.

I'd be very interest to know if any eHam users are using baseband or "super-audio" systems, with a minimum measurement bandwidth of 100 kHz and preferably 200 kHz.  To get sufficient advantage over rough and ready DSO-based acquisition systems I'd envisage 16-bit quantization, probably with at least 2 channels.  For the type of work I'm doing something like the SIGVIEW package (with good spectral zoom and easy spectral density analysis) looks reasonable but the selection of reliable, easily interfaced, low noise ADCs remains.  SIGVIEW (https://www.sigview.com/?gclid=EAIaIQobChMIjKaWxaKh8wIVyp1LBR2NxAIvEAMYASAAEgJRI_D_BwE) list a few possibilities for real-time data acquisition, including:

NIDAQ and NIDAQmx compatible DAQ devices from National Instruments;
All Measurement Computing® DAQ devices (12/16/24-bit voltage input);
LabJack® U3/U6 and T4/T7 USB/Network DAQ devices;
All MS Windows compatible sound cards (DirectSound and ASIO driver models are supported).

While many devices could no doubt be made to work, I'd be interested to hear of similar systems in actual use, perhaps with an RF interface application.  I expect to do an amount of customization but I'd be happy to get as close to a useful measurement capability with as little fiddling as possible. (Better design and development things to do!).

73, Peter.
« Last Edit: September 28, 2021, 01:56:28 AM by VK6HP »
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #1 on: September 28, 2021, 08:05:06 PM »

VK6HP,

While I am familiar with the measurement, just wondering why you need it automated. At HP several automated systems where produced for production testing but for lab testing, I used the HP3561A which was a big step up from the HP3580A. The HP3580A required much patience.

https://www.hpmemoryproject.org/technics/bench/3048/config_hard_01.htm
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #2 on: September 29, 2021, 01:54:09 AM »

HH75,

Thanks for the reply.  I don't need it automated in any production/test line sense but, for the phase noise/jitter measurements, I would like to see the spectrum as a whole for a few reasons ranging from laziness (!) to performance trade-offs occurring in different carrier offset regions. I'd also like the convenience of doing the measurements at home, rather than imposing on the R&D lab which I built in a former life and which has the Keysight contemporary equivalents to the classic system you referenced.

For HF phase noise measurements the carrier crystal notch filter approach works well enough in conjunction with a good spectrum analyser.  However, some of my Class D and E PAs are at 2200 and 630 m where  crystals have a high ESR, making common deep notch implementations tricky, even if the crystals are available.

 Of course the heterodyne approach is more versatile in any case, and I was lucky enough to recently pick up a mint condition Rohde and Schwarz SML-01 synthesizer for a good price; it will make an excellent LO source. generator.  It's not that there aren't ways of examining the baseband spectrum apart from an FFT system but I figure that I might as well go hard :)  And, in truth, some of the noise products in the non-linear amplifiers are quite close to the carrier making, for example, good resolution and zoom capability worthwhile.

73, Peter.

Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #3 on: September 29, 2021, 12:49:38 PM »

VK6HP,

2200 m and 630 m are about 136 kHz and 476 kHz, respectively. I can't really say what kind of phase noise you might expect from amplifiers in this range, but I can say that when I tried to do amplifier phase noise measurements at HP with the best equipment available in the 1973-1980 time frame, I was not able to measure amplifier phase noise without significant improvement to the measurement system itself. Even the HP3561A was not quiet enough to do the measurement.

I think you mentioned that you have a copy of Motchenbacher & Fitchen's "Low Noise Electronic Design". HP released for public consumption the preamp design used to improve the HP3561A noise figure. It can be found in the appendix of that book. That edition also has data on the phase noise of frequency dividers measured using the heterodyne method.

The heterodyne method is extremely sensitive but I am not sure even it can do the measurement you need without some work.
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #4 on: September 29, 2021, 10:53:06 PM »

HH75,

Yes, sensitive phase or composite noise measurements are definitely a challenge, particularly on a hobby budget.  Modern systems such as the latest generation high-end transceivers are so good that one doesn't even start the race without a measurement system with an ultimate noise floor in the -140 dBc/Hz range, and desirably 10-20 dB better.  You're right in saying that measurements of simple linear components like good quality amplifiers is especially difficult. Fortunately (I suppose) the imperfections in the switching power amplifiers I'm measuring are easier to measure, with quantitative differences showing up at perhaps -100 dBc/Hz at low to medium (kHz) offsets.  These imperfections can have many causes but two of the most common are combined effects in the conversion of the analog drive signal to digital form, and plain old AM caused by ripple on the power supply.  The digitization artifacts are commonly the result of jitter in sub-optimum devices (e.g. CMOS Schmitt triggers), aided and abetted by PM introduced from low-level mains and other interference.

To give you an idea of what can be seen with "normal" equipment, I can reference a common MF push-pull Class D topology, typically driven by a doubler followed by a CMOS divide-by-2 flip-flop (say, an HEF4013 with a Schmitt trigger clock input), ensuring "perfect" complementary drive signals at 475 kHz.  An alternative circuit of my own design requires no doubler, being based on a very fast LT1016 no-hysteresis comparator, quite clearly shows much reduced close-in jitter/phase noise when viewed on a Rigol RSA3015E-TG spectrum analyser operating in real-time FFT mode or, indeed, by the band scope in my TS-890S.  However, as designs get better and trade-offs begin to appear, a more sensitive and convenient analysis system would be useful.

By the way, the example is not a criticism of previous designs, which are adequate for the task in hand.  The comparison is just a reflection of my own interest.

Thanks for reminding me of the M&F appendix, which I'll look up.  Some of the R&S reference material for their signal analysers also makes interesting reading, as do the proceedings of the HP phase noise seminar in 1985 and lots of their application notes.  This UK NPL write-up from 2004 is also very readable and to the point:

https://www.npl.co.uk/special-pages/guides/gpg68_noise.aspx?ext=.

73, Peter.
« Last Edit: September 29, 2021, 11:05:40 PM by VK6HP »
Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #5 on: September 30, 2021, 03:05:15 PM »

VK6HP,

Phase noise in the -100 dBc range is well within the range of the heterodyne method just using off the shelf components and instruments.

Have you considered the HP3561A as the analyzer? It used to cost about $10k but these days they are on eBay for around $600. It looks like a conventional spectrum analyzer but is really a fast Fourier transform analyzer. It has an HPIB ( IEEE488 ) interface for computer control.
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #6 on: September 30, 2021, 07:37:10 PM »

HH75,

Yes, I have thought about looking for a serviceable 3561A or similar.  It's clear from various forums that many engineers, across a number of fields, really miss the old HP stand-alone analysers.  Of course these instruments are rather complex and are getting old, so discussions about things such as failing EPROMs and the like are common. The other problem is that they are large instruments and, with freight as it is nowadays, it costs at least USD300 to ship a used instrument from the US to Australia.  But it is a possibility and a classic HP (or maybe even Anritsu) dynamic signal analyser would not look out of place in any aficionado's workshop.

The other route I'm drawn a little more to is to purchase a copy of (e.g.) SIGVIEW and one of the Measurement Computing DAQ boxes, perhaps one of the 1608 series: https://www.mccdaq.com/data-acquisition/high-speed-daq  In the worst case a low noise (incl. flicker noise) FET pre-amp might be needed, but the devices may be good enough as they come. 

73, Peter.
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #7 on: October 01, 2021, 04:23:08 PM »

Interesting thread, thanks!

I'd like to have a decent phase noise measurement system and keep meaning to make something better than what I currently have.  At the moment I have a couple of microwave band spectrum analysers that have a phase noise measurement (SSA) option fitted but the performance is not quite as good as I'd like. One is generally good enough up at V/UHF and the other (Tektronix 3408A) is often good enough down on the HF bands. I'd expect/hope the Tek is at least as good as the classic HP3561 although I've not tried to compare the two side by side. I've also got some Analog Devices DDS Eval boards here that have programmable phase so they can be used to measure phase noise with the quadrature method but I've only tinkered with them briefly. This could probably beat the Tek analyser for close in phase noise but it is a very fiddly and limited method.

The Measurement Computing products look very good value for money. I'd be interested to see what performance their 16 bit units can offer. The prices look very tempting!
Logged
Regards
Jeremy

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #8 on: October 01, 2021, 04:40:31 PM »

I don't know how available or affordable the HP11848A phase noise interfaces are but this can be used with the HP3561A to achieve good results using the quadrature method.

http://www.hparchive.com/Manuals/HP-3048A_cal_03048_90015.pdf

I've tried doing this with a phase detector and the Tek analyser and the DDS board when locked to the same reference but I recall I had problems with interference pickup.
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #9 on: October 02, 2021, 02:19:23 AM »

Jeremy (G0HZU)

Thanks for your comments. I've seen that you're interested in the topic via discussions in several other forums and I've found your contributions to be useful and interesting.  You mentioned the HP11848A and I've been in the process of comparing it with an HP11729B, of which several examples have recently appeared on eBay.  To cut a long story short the 11848A is the preferred model performance-wise, although both have a lower out-of-the-box signal input limit of 5 MHz, with some level of complication when testing below 20 MHz.  I suspect that one could optimize both for HF use and, with a bit of re-working, I may even be able to coax some MF operation from the equipment.  However, looking at the price of the used HP11729B and the likely effort involved, the resources may well be better directed to a new system.

I suspect the Measurement Computing DAQ boxes would be a viable path.  Looking a bit more closely at the MC site, I see that they have Data Translation DAQ systems which are actually called "dynamic signal analysers" and offer some complementary (but not complimentary) proprietary Fourier analysis software. One 2 or 3-channel system is described at https://www.mccdaq.com/Products/Sound-Vibration-DAQ/DT9847.  It's not a bargain basement price by ham standards (USD 4500 for hardware + advanced Fourier analysis + basic general purpose analysis) but, compared with the original price of the HP3561A mentioned by HH75, it'd certainly be viable for the serious user.  I mention this largely because it's about the closest contemporary substitute I've found for the HP series instruments.  With a maximum baseband analysis bandwidth of 100 kHz the combination of the baseband system and an RF spectrum analyser would complete the measurement suite, as in the HP system.

Exactly how the MC lower cost options would compare withe the Dynamic Signal Analyser hardware is an open question and perhaps some advice from one of their application engineers would be worthwhile.  I note that if one adds up the desirable features in the lower cost DAQs, the difference between those units and the Data Translation DSA is not as great as it first appears.  Desirable features include true simultaneous sampling of channels and a couple of DACs, just to make the box more useful in general DSP applications.

Another open question is whether SIGVIEW works with the Data Translation DSA hardware (as opposed to the MC DAQs).  If so, I suspect that's a lower cost software environment in which to establish oneself.  Still, the software will make or break the integrated system so some careful testing would be in order.  SIGVIEW (at least) offers a trial version, which would certainly be handy.

73, Peter.

Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #10 on: October 02, 2021, 11:54:07 AM »

Thanks Peter. In my case I'd want to use fairly modern or new hardware and hopefully something that doesn't take up much space. At work there are still a couple of Agilent E5052A SSAs that I have access to but I know one of them is prone to failing. Each year it would struggle to get through a full calibration and sometimes it would fail to function at all. I think it used to get sent to Germany when it needed a repair and would sometimes spend months there. I've always hoped one of them would get declared BER so I could buy it cheaply but the huge size and weight and the poor reliability kind of puts me off now.

I know KE5FX and SM5BSZ are heavily into this stuff and SM5BSZ is really pushing for really low phase noise measurements like -180dBc/Hz. I'd just like something that can measure to -150dBc/Hz at a few kHz offset that is easy to set up and will deliver reliable results. -170dBc/Hz would be nice at 100kHz offset but I'm less concerned about achieving that.

It would definitely be worth asking an application engineer at MC what can be achieved. I do think someone will set up a decent and affordable phase noise measuring system soon. That's why I'm prepared to be a bit patient to wait for a modern solution.

At work I'm currently evaluating a receiver design that uses a very fast ADC from AD and I'm setting up methods to see how accurately it can measure noise. Sadly, this won't be measuring phase noise at very low levels. It's a microwave band receiver so the phase noise is likely to be little better than -90dBc/Hz at a few kHz offset.
Regards
Jeremy
Logged
Regards
Jeremy

HAMHOCK75

  • Member
  • Posts: 1297
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #11 on: October 03, 2021, 05:19:25 AM »

Quote from: VK6HP
although both have a lower out-of-the-box signal input limit of 5 MHz,

The only component that I can think of that creates a lower frequency limit is the mixer that converts to baseband. The mixer RF, LO inputs have a lower frequency limit.
Logged

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #12 on: October 03, 2021, 07:00:40 AM »

Thank you both for the informative comments.  With a bit more digging I can see that Sigview comes close to supporting the DT9847, in that the lower speed DAQ in the same family is definitely compatible.  I think my next move is to send off a few queries to Sigview and MC, questions to the latter also referencing comparative noise spectra of the various modules.

Quote from: VK6HP
although both have a lower out-of-the-box signal input limit of 5 MHz,

The only component that I can think of that creates a lower frequency limit is the mixer that converts to baseband. The mixer RF, LO inputs have a lower frequency limit.

Yes, the mixer is undoubtedly the place to start.  I also recall reading some warnings about the need for extra post-mixer filtering with low IFs (below 20 MHz) in order to avoid overloading some internal amplifiers. Incidentally my experiments have used Watkins Johnson M1 (0.2-500 MHz) double balanced mixers with a dc coupled IF port.

At work I'm currently evaluating a receiver design that uses a very fast ADC from AD and I'm setting up methods to see how accurately it can measure noise. Sadly, this won't be measuring phase noise at very low levels. It's a microwave band receiver so the phase noise is likely to be little better than -90dBc/Hz at a few kHz offset.

Jeremy, your example underlines the need for thought as to whether pushing an "audio" DAQ system upwards, or an IF/RF digitizing system downwards, makes more sense.   In some respects dynamic signal analyser modules such as the Data Translation packages are overkill in terms of e.g. resolution in inherently noisy RF applications with post-processing averaging, and their ADC technology might not, in any case, be my first choice in an RF system.  However, as we all already have good RF instruments for offset measurements above 100 kHz, and bearing in mind the general usefulness of a good baseband/audio FFT system, I see profit in my own case in pursuing that line. But we've become so used to radio DSP in various forms it pays to look at what technology (and bottom line performance) is available from those systems, too. 

73, Peter.
« Last Edit: October 03, 2021, 07:03:37 AM by VK6HP »
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #13 on: October 03, 2021, 08:45:55 AM »

Ideally I'd want to be able to measure at offsets up to 1MHz away but 100kHz is probably enough.

The image below shows the limits of the Tek analyser when making a baseband measurement at 10MHz. I've fed it with a Marconi 2024 sig gen set to 10MHz output. The -100dBc spurs are in the sig gen. It needs a service as I think it needs some tired electrolytic caps replacing. It used to be spur free in plots like this.




The Tek analyser seems to have a noise floor limit at around -137dBc/Hz. I'm not sure if the sig gen is limiting the performance at 10Hz offset. However, it manages almost -120dBc/Hz here.

Ideally, I'd want to see another 15-20dB beyond what the Tek analyser can manage across 1kHz to 100kHz offsets.
« Last Edit: October 03, 2021, 08:48:06 AM by G0HZU »
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #14 on: October 04, 2021, 12:47:38 AM »

Very interesting Jeremy.  The RS 3408A looks like a nice instrument and when I do the mental arithemetic from the specification sheet phase noise at 1 GHz I'd expect about -135 dBc/Hz for 10 MHz at 1 kHz offset so it looks like you're doing a little better, give or take the spurs. I couldn't find a spec for less than 1 kHz offset.

It is interesting to think about where the Tek 14-bit analyser's baseband noise limitation comes from. I'm not familiar with the instrument's design but no doubt conversion clock jitter is one noise component. (And that gives pause for thought, too, in the context of alternative hardware for low-frequency FFTs).

I like the Marconi 2024 signal generator, especially since the various user groups have good troubleshooting information available, including full component-level schematics.  However, in the end I went for a used R&S SML-01 because it was available at a good price, had slightly better phase noise in some regions of interest than the Marconi, and an electronic attenuator.  Mine came via a DoD surplus channel and has a few options, including the pulse modulator.  Some manufacturer's phase noise plots below.





The second image is not particularly clear but the log frequency offset axis starts at 1 Hz and the power density axis is 0 to -170 dBc/Hz.  It's not a stellar performance in absolute terms but not bad for the money (about USD 500 for a very well kept unit).

I believe the Marconi also has an "extended divider" mode, on which I'm pinning some of my hopes for HF measurements and a noise floor in the -150 dBc/Hz region for the "fully tunable" HF test set.

73, Peter.


Logged
Pages: [1] 2 3 4   Go Up