Pages: 1 [2] 3 4   Go Down

Author Topic: Baseband FFT system - any h/w and s/w experience?  (Read 706 times)

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #15 on: October 04, 2021, 01:46:45 AM »

I knew I had some plots for the 2024 as well.  This is actually a late model Aeroflex 2023, which is the direct equivalent of the 1.2 GHz SML-01.



Swings and roundabouts for a given user, I think.  Given the number of 2024 (2.4 GHz) units around for about the same price as the SML, the extra frequency coverage might trump the slightly better phase noise in some shacks.

73, Peter.

Logged

HAMHOCK75

  • Member
  • Posts: 1297
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #16 on: October 04, 2021, 05:04:40 AM »

The phase noise plot of the 2023 as a function of frequency is typical of most synthesizers and signal generators. The lowest frequency band is usually created by mixing the highest band against a fixed frequency to down convert which creates a low band with the least mixing spurious but the phase noise suffers because of it.

For VK6HP's application, one solution is to simply take the lowest noise band and continue down with frequency dividers rather than use the products down-converter to get lower frequencies. The phase noise will then be limited only by the performance of the dividers as shown in the M&F appendix.

G0HZU's plot reminds me that one way to get a very clean spurious free plot is to not use a signal generator or synthesizer as the source. For instance, I used crystal oscillators sometimes, but I had the advantage that HP made their own precision quartz crystals from raw quartz. I never measured a crystal Q below 1 million but the cost was high too. The quartz oscillator could be run off batteries to free it from line related spurious pickup.

For these purposes and the expected measurement range, I suspect a quartz crystal even with a crystal with poor ESR might work well as the source oscillator or a higher frequency quartz crystal divided down to the final frequency might be a less costly solution if a spurious free measurement is desired.
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #17 on: October 04, 2021, 11:09:08 AM »

I have to be careful when connecting regular clock sources to the Tek analyser in case they have any residual DC offset. The baseband mode works from 40MHz right down to DC so if I don't DC block the input it complains about DC levels.

I dug out a DC block and tried measuring the 10MHz ref out from some other instruments. The best one had slightly higher phase noise at 10Hz offset but the phase noise floor from 100kHz onwards on the Tek display was up to 6dB lower. The 100Hz, 200Hz spurs were gone.

Many years ago I built up a phase noise library using the company E5052A SSA and see below for a Marconi 2024 at 14MHz. The odd little bump at 1.5kHz offset was the reason I measured this particular 2024. Other ones don't have it. The -136dBc blip at 60kHz was something either in the E5052A or something in the building as it was present on every make/model sig gen I tested that day. I think I measured the phase noise at 10, 14, 20, 28, 50, 70, 144, 250, 433, 500, 750, 1000, 1300 and 2400MHz for all sig gens that could cover this range.



Note that to get the best noise floor out of a 2024 that has the 25dBm output power option fitted it pays to exploit the attenuator lock feature. Otherwise the noise floor can be quite high. The plot above is for a standard 2024 with the standard 10MHz oscillator. The phase noise below 10MHz would have been better with the OCXO option fitted.

My 2024 isn't this good because it is the 25dBm version and I have to use high power and an external attenuator and attenuator lock to get a noise floor close to -150dBm. That's why the noise floor isn't quite as low as it could be on the earlier plot I took with the Tek 3408A. Some of the noise is from the Mi 2024 even at >100kHz offsets
« Last Edit: October 04, 2021, 11:12:13 AM by G0HZU »
Logged
Regards
Jeremy

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #18 on: October 04, 2021, 11:27:57 AM »

I had a rummage and managed to find the phase noise plot of the 2024 at 1GHz. It seems to agree very well with the plot you posted up Peter.



This 2024 only has the standard oscillator and not the OCXO option so that probably explains the higher phase noise below about 100Hz. The pesky blip at 1.5kHz is still there. I do have some plots somewhere of a healthy 2024 and the blip is gone.

The R&S  SMKL-01 looks to be a very good signal generator. I saw a phase noise plot in a glossy brochure at about 9MHz at it was impressive.
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #19 on: October 05, 2021, 04:29:03 AM »

Interesting plots, Jeremy, and certainly a reminder that the devil is in the details in this game.  The behaviour of the SML-01 looks a bit like the old resonator generators but there is a plateau further out near 10 kHz. There is a passing reference in the service manual to a ceramic resonator in the UHF range but no further details are provided, although there is a promise of block diagrams and PLL description in some literature I don't yet have. The 2024A is really a textbook phase lock plot and your generator does indeed look very close to the advertised plot, at least beyond a few hundred hertz offset. 

My general thought was to run the SML-01 in divider mode and, for critical measurements below 10 MHz to use an external divider along the lines mentioned by HH75.  It is true that for the MF power amplifier measurements a crystal LO could be another possibility but, on the other hand, these measurements are probably the least demanding, even with a significant boost in performance over my existing raw spectrum analyser characterizations.


73, Peter.
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #20 on: October 07, 2021, 03:36:49 PM »

One other option for the LO is to make a decent free running LC oscillator. This would be at its best foruse below 30MHz and would provide low phase nosie at offsets in the 10kHz to 100kHz range. It should outclass the sig gens in this range and should be better than a typical crystal oscillator at these offsets. It won't be as good at offsets below about 5kHz but it should perform much better than a basic crystal oscillator or a sig gen at a 100kHz offset.

The phase noise simulation below is based on Leeson's equation for a basic LC oscillator with low flicker corner frequency and a loaded Q of 42 and about 5mW resonator power at 10.7MHz. The phase noise beyond 10kHz offset is going to be a lot better than any lab sig gen I've used. The downside will be that it will be prone to warmup drift and also to interference pickup. It would ideally need to be in a screened box.



I'm not sure about using this technique for the LO beyond the HF bands though. It may become a bit too drifty.
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #21 on: October 08, 2021, 03:23:23 AM »

That's an interesting plot, Jeremy, and happens to strike me at a weak moment when I'm considering whether I can find space for an old HP606B signal generator on offer locally.  Your example bears close thought - interesting to see how good such a simple unit might be beyond a few kHz offset.

For my immediate MF non-linear PA application I do need to get at least to 10 Hz offset but not, as I mentioned, down to incredibly low noise floors.  There are various physical mechanisms involved in the PA but, in essence, it comes down to assessing additive noise, where the original driver signal can be used as the reference.  That's good, largely because it removes issues of drift and the need for any phase locking.

More generally, your comments about drift led me to wonder about how essential the phase lock is in modern practice.  I guess if the drift is slow/small enough to allow a few stable spectra to be acquired, and allow good calibration of the levels measured on the baseband analyser, that may be good enough.  The locked system is undoubtedly more desirable but with, for example , a reference synthesizer and a unit under test both locked to (e.g.) a typical GPSDO, it occurs to me that one might get away with no phase locking for at least some measurements.  (The SML-01 has a carrier phase control which might be useful for setting the optimum operating point for the mixer).

73, Peter
« Last Edit: October 08, 2021, 03:37:16 AM by VK6HP »
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #22 on: October 08, 2021, 06:22:18 AM »

As HH75 says the other option is a crystal oscillator. I think some care is needed in terms of selection because a basic crystal oscillator might only run at 50uW crystal power. This is good for stability but not for far out phase noise. I've often seen OCXO phase noise plots where the phase noise floor plateaus at -155dBc/Hz out to 100kHz.

Golledge do have some remarkably good crystal oscillator modules like the GXO-7506 range...

https://www.golledge.com/products/gxo-7506l-ultra-low-jitter-3.3v-oscillator/c-26/p-139

Phase noise (typ @ 24.576MHz)   
-101dBc/Hz @ 10Hz
-132dBc/Hz @ 100Hz
-155dBc/Hz @ 1kHz
-166dBc/Hz @ 10kHz
-174dBc/Hz @ 100kHz
-175dBc/Hz @ 1MHz
-175dBc/Hz @ 5MHz

I'm not sure how they achieve this but it is very impressive and it would make a good benchmark as a  'gold' test standard to have on the bench. If I punch some basic numbers into Leeson's equation for a crystal oscillator it is the crystal resonator power of 50uW to 100uW that limits the far out phase noise regardless of the Q. I have seen some crystal oscillators where the plateau is as high as -145dBc/Hz out to 100kHz. I guess these oscillators run with a lower RF power at the resonator.
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #23 on: October 11, 2021, 12:02:12 AM »

In terms of a credible lab/bench "standard" the Golledge modules look good and I suspect that (for a price) they may even be able to provide a measurement on a given module.  Just a little behind those modules is something like this Crystek version:

https://www.ebay.com/itm/393248253528?hash=item5b8f6c2658:g:meUAAOSwMU1geUQr

which is pretty cheap (USD60) but possibly not quite as robust in specification, at least if it comes from the Chinese eBay supplier.  The plots are not easy to read but here's my addition to Jeremy's table, based on the Crystek 24.576 MHz module (but note the few dB better performance of their 45.1584 MHz module, even before a dvide-by-two):

Golledge (Crystek) Phase noise (typ @ 24.576MHz) 
 
-101  (-98)  dBc/Hz @ 10Hz
-132  (-126) dBc/Hz @ 100Hz
-155  (-148) dBc/Hz @ 1kHz
-166  (-162)  dBc/Hz @ 10kHz
-174  (-170) dBc/Hz @ 100kHz
-175  (-170) dBc/Hz @ 1MHz
-175   (-170?) dBc/Hz @ 5MHz

Out of interest I sent through an enquiry to Golledge, just to see what their prices are like for small quantities.

I was tempted to make a condescending remark about the golden-eared audio crowd until I realised that their DAQ and clock kit could actually end up being quite useful!

73, Peter.


« Last Edit: October 11, 2021, 12:08:34 AM by VK6HP »
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #24 on: October 11, 2021, 09:33:43 AM »

I might also ask for samples from Golledge to see if they have any surplus 7506 oscillators.

I found this paper written by WJ on the web that describes the use of diode mixers as phase detectors. I've only skimmed it so far but it looks to have some useful info in it.

https://www.jlab.org/uspas11/Reading/RF/Mixers%20-%20phase%20detectors.pdf

I haven't looked at this stuff for years but I still have an old Minicircuits SRA-3 mixer in a metal box with BNC ports on it. In its day this was a decent VHF mixer. If it goes well I'll try and knock up a quadrature network at 10.7MHz and see how it performs. I've also got a lot of other Minicircuits mixers in various eval fixtures but I suspect the old SRA-3 will perform the best here.

Logged
Regards
Jeremy

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #25 on: October 11, 2021, 02:00:21 PM »

I tried measuring the phase response of the SRA-3 mixer over >360degrees and got the plot below. This was at 10.7MHz with a 7dBm LO and -6dBm RF drive after a LPF and 10dB attenuator.  i.e. RF sig gen>>+4dBm>> LPF >> 10dB_ATTN >> -6dBm to mixer RF port.

I took a phase detector output measurement for every degree step in phase. All ports were quite well terminated for this test and RF harmonics into the mixer would have been lower than -80dBc.

I did this using a VB program to control a couple of sig gens, the SRA3 mixer and a bench DMM. The sig gens have programmable phase when locked to the same reference so the data should be quite good.

It looks like it outputs up to +/- 95mV across all phase angles

 
« Last Edit: October 11, 2021, 02:09:27 PM by G0HZU »
Logged
Regards
Jeremy

VK6HP

  • Member
  • Posts: 1222
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #26 on: October 12, 2021, 05:57:46 AM »

That's an encouraging plot, Jeremy.  With decent isolation and optimum drive levels you're seeing good performance with just about minimum dc offset.  I hadn't seen the WJ Tech Note but, based on what was available on eBay at the time and some general principles, I chose a WJ M1 DBM for my test system.

Golledge were polite but unable to help and, while it'd be rude to reproduce the correspondence publicly, I can pass on that one would need to be thinking quantities in the low thousands in order to raise any interest.  The Crystek modules (https://www.crystek.com/home/oscillator/ultralowphasenoise.aspx) are about USD38 each at Mouser and I think I'll get one of the  ~50 MHz versions, calibrate it at work, and use it as my home bench DUT "standard" (and more, no doubt).

73, Peter.
« Last Edit: October 12, 2021, 06:13:58 AM by VK6HP »
Logged

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #27 on: October 13, 2021, 09:46:03 AM »

The Crystek module is a high performance oscillator so definitely a contender. Golledge used to be keen to hand out samples but I've seen a lot of change throughout the whole electronics industry. The vendors only want to sell products in huge quantity these days.

I found this IEEE note about an ultra low phase noise 10MHz crystal oscillator.

https://eprints.whiterose.ac.uk/141563/1/08540461.pdf

The performance looks to be amazing at offsets below 1kHz. The differential amp is useful because it provides an easy way to tap out the signal efficiently. This IEEE paper was a serious attempt at achieving low phase noise very close to carrier.

Right down at the other end of the scale I can remember designing basic crystal oscillators in the 1990s at work and this was usually done by consulting a couple of classic books. One was by Matthys, and the other was by Parzen. I still have these old books but there is very little in them about designing for low phase noise.
Logged
Regards
Jeremy

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #28 on: October 13, 2021, 12:08:04 PM »

I dug out an old and very simple crystal oscillator board that was designed using the chart and tables in Matthys. This was a 2.5MHz crystal oscillator and it uses table and chart values for a process 50 JFET. I used a BF256B for this circuit but Matthys used a 2N4416.

I just plugged it into the Tek analyser and the phase noise of this oscillator is probably nearing the limits of the analyser because it shows -115dBc/Hz at 10Hz and close to -130dBc/Hz at 100Hz offset. It is probably better than this. However, I suspect the far out phase noise floor will not be that great as this is a low power circuit.

 
Logged
Regards
Jeremy

G0HZU

  • Member
  • Posts: 345
Re: Baseband FFT system - any h/w and s/w experience?
« Reply #29 on: October 13, 2021, 02:44:49 PM »

For a bit of fun I had another rummage and found a couple of bags of 10 of these cheap Chinese 9.8304MHz crystals from RS Components. RS part 814-9419.
Here in the UK you can get ten crystals for about £2.

I just stuck one on my VNA and took two port s-parameters across 9.780 9.880MHz with 1601 data points.

It looks like the ESR is about 15 ohms the Lm is about 23.841mH and the Cm is 11fF and Co = 3.45pF. However, I haven't spent much time reverse engineering these with any great accuracy.

I've uploaded the s parameter file here:
https://www.qsl.net/g/g0hzu//RF%20Transistor%20data/MISC/9830.S2P

If I put the crystal into an open loop 50R MMIC feedback/oscillator circuit I see a loaded Q of 13,300. If the crystal resonator can cope with -6dBm drive then Leeson's equation predicts the phase noise plot below.


This looks quite good! If the IEEE circuit with the low noise diff amp was used it might all drop a few dB.

I originally bought two bags of 10 of these crystals because they were so cheap and maybe they could be used to make a decent crystal filter. If so, this filter could be used to test the phase noise of the oscillator using a spectrum analyser. This could get interesting...

I also entered the parameters for the IEEE design into my Leeson's spreadsheet and it had the same result as their simulation.
Logged
Regards
Jeremy
Pages: 1 [2] 3 4   Go Up