UK Vintage Radio Repair and Restoration Powered By Google Custom Search Vintage Radio and TV Service Data

Go Back   UK Vintage Radio Repair and Restoration Discussion Forum > Specific Vintage Equipment > Vintage Test Gear and Workshop Equipment

Notices

Vintage Test Gear and Workshop Equipment For discussions about vintage test gear and workshop equipment such as coil winders.

Closed Thread
 
Thread Tools
Old 18th Sep 2020, 11:23 pm   #161
Skywave
Rest in Peace
 
Join Date: Jun 2006
Location: Chard, South Somerset, UK.
Posts: 7,457
Arrow Re: Replacement for HP8640B signal generator

Alan G3XAQ: you have a P.M.

Al. (Skywave)
Skywave is offline  
Old 25th Sep 2020, 10:24 am   #162
trh01uk
Octode
 
trh01uk's Avatar
 
Join Date: Nov 2010
Location: Ross-on-Wye, Herefordshire, UK.
Posts: 1,648
Default Re: Replacement for HP8640B signal generator

I've just read this entire thread having missing most of it when it was written back in June this year. Fascinating diversions into HP history, spectrum analysers and tracking generators.....but I note that answers on what the alternative is to an 8640B are limited, presuming you really do want the low phase noise and freedom from spurs that it provides.

Perhaps the Marconi 2017 is good enough on phase noise (but less good on other spurs), and the HP8662A is also a good substitute - but I wonder how often working examples turn up, given that Ebay list of sold examples are all "parts only"?

I can see a 2017 for sale at the moment for around £1000 apparently fully working (but not calibrated in years) its still looking like a pricey option. I've got a Marconi 2018, but its showing the unreliability problems (display/controls) that I first observed in the one I bought new at work in the 1980s when it was a mere 5 years old.

I think Radio Wrangler summed up the situation back in post #18, where he said:

"So, the low-noise high RF performance sig gens are with very few exceptions, a thing of the past, and the ones available are getting older. If we want them, then we just have to grit our teeth and keep fixing them."

I guess its going to keep this forum busy for a while yet!

Richard
trh01uk is offline  
Old 6th Jan 2021, 10:26 pm   #163
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Well, my #1 and #2 HP8640Bs seem to be working (this week, touch wood). I've done the Shift-W error correction on my HP98568B spectrum analyser and tried all the performance checks I can do without needing esoteric testgear and all seems near enough for my modest needs, with errors mostly under 1dB. I've repurposed a scrap Boonton 92B millivoltmeter as an AD8307 dBm meter. And I've made an Arduino speak Morse code from a PC using the Winkeyer protocol. So now what, while waiting behind the 13 million ahead of me in the Covid vaccination queue?

I've been reading about the disappointing performance of the AD9910 DDS here

https://martein.home.xs4all.nl/pa3ak...0_amnoise.html

but the author shows a graph for the older AD9951 (thumbnail attached)
that claims with a giant decoupling cap on the DAC voltage reference the noise performance is pretty decent: -147dBm/Hz at 2KHz from a 14.3MHz output. Yes, there are the usual DDS sprogs but you have to accept them with only a 14-bit DAC. But they are discrete frequencies and not too loud so might not be a huge problem in a signal generator? Or even a receiver local oscillator??

Cheapish kits are available, for example here

https://www.minikits.com.au/Basic-AD9951-DDS

So go on, tell me the approach will be a disaster...

Alan
Attached Thumbnails
Click image for larger version

Name:	dds_ad9951_bypass.gif
Views:	74
Size:	26.4 KB
ID:	224024  
Alan_G3XAQ is offline  
Old 7th Jan 2021, 1:37 am   #164
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Replacement for HP8640B signal generator

The AD9910 has a 'clock multiplier'

Think GHz PLL with an on-chip VCO. So a phase noise generator. This noises-up the clock to the whole DDS.

Solves the problem of getting a GHz clock, but wrecks the performance of the whole thing.

What were they thinking?

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 7th Jan 2021, 11:40 am   #165
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Yes, David, but I don't believe you have to use the multiplier. That PA3AKE guy says he gets -154dBc/Hz at 1KHz offset at a 16MHz output frequency (7MHz receiver, 9MHz IF) if he creates the 1GHz reference from a 100MHz low noise oscillator module and multiplies+filters it ten times:-

https://martein.home.xs4all.nl/pa3ak..._refclock.html

For a signal generator the big problem with the 9910 is the AM noise when generating sine waves. For a receiver mixer application the AM noise is scrubbed off in the square wave LO driver.

But yes, you do make a good point about using the older 9951 chip: it still needs a good external reference clock.

Alan
Alan_G3XAQ is offline  
Old 7th Jan 2021, 12:12 pm   #166
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Replacement for HP8640B signal generator

Watch out for some chips with multipliers. SOME of them have a multiplier out function which leaves the PLL in the path, but just set to x1.

They aren't very clear in datasheets. Almost as if they didn't want you to know.

Beware also of crystal oscillators with PLLs in them enhancing their profitability (don't need to make as many different thichnesses of quartz) Noisy little beggers.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 7th Jan 2021, 11:02 pm   #167
G0HZU_JMR
Dekatron
 
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
Default Re: Replacement for HP8640B signal generator

Over the years I've collected up quite a few old DDS eval boards from Analog Devices and in amongst them all I think I still have their official 400MSPS 14bit AD9954 eval board. Looking on google this has two AD9954 devices on it and I think it is very similar to the AD9951. The 9954 can do some extra tricks via onboard RAM etc but otherwise I think the core of the DDS is the same as the AD9951.

https://www.element14.com/community/...al-synthesizer

I haven't used it for years and I think it must be well over 15 years old now. These are old devices but the performance was quite good. The eval board requires an external 400MHz clock and I think I used my Mi2024 sig gen for this the last time I powered it up. This is obviously not an ideal choice for the external clock but it worked OK for my needs at the time. I think I also have the official eval boards for AD9850, AD9851, AD9910 and AD9912 but these will be lost in the loft somewhere. I've also got an early (pre official release) eval board for the AD9914 but this is at work.

I can do a few tests with the AD9954 if you want to see what the spurious look like for certain test frequencies in the HF band if that helps? I will struggle to look at phase noise though. At work I have access to an Agilent E5052A signal source analyser but I'm working from home and probably will be for a couple of months. Otherwise I could quickly measure the phase noise at work for you whilst trying different bypass caps and also with a better 400MHz clock than the Mi2024.

One thing to watch out for is the relatively poor far out phase noise of these devices. A reasonably well designed LC oscillator will often outperform these DDS chips at offsets beyond about 10kHz. By 100kHz offset the LC oscillator should be much cleaner, possibly 15-20dB cleaner.

I think the MiniKits AD9951 board is configured to run with the x4 multiplier but hopefully there should be an easy way to get it to run directly from an external 400MHz clock.
__________________
Regards, Jeremy G0HZU
G0HZU_JMR is offline  
Old 8th Jan 2021, 12:24 am   #168
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Replacement for HP8640B signal generator

Sorry, Jeremy, it's the phase noise that the clock multipliers mutilate.

There was a paper by Jezry-Popeil about high-width DDSs flooring at higher levels of spurs than DACs etc would explain. 20 years ago there was a report in HP on it and suggesting a digital state-variable filter structure (just on the threshold oscillating) as a better alternative to the phase register/lookup approach.

Did anything come of this or was it a red herring?

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 8th Jan 2021, 3:21 pm   #169
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Hello Jeremy. Happy New Year. Yes, the lockdown is a pain, even for us retired old geezers.

If you look at the plot attached my post #163 it seems the AD9951 with 100uF bypass has monotonically decreasing phase noise all the way out to 500KHz. It only goes below the "best possible IMD3 receiver performance line" around 150KHz out so it's never going to win any prizes but on the other hand it looks fairly easy to apply using something like the I0CG 400MHz clock board. For a more "average" good receiver the 9951 performance looks like a pretty good match to the likely IMD dynamic range.

But of course none of this buys me HP8640B noise performance...

Alan

Last edited by Alan_G3XAQ; 8th Jan 2021 at 3:49 pm.
Alan_G3XAQ is offline  
Old 8th Jan 2021, 8:01 pm   #170
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Replacement for HP8640B signal generator

If you only want to test IMD on one band, building a couple of crystal oscillators is a hard act to beat. It's comparable to the old remark "Never overlook the bandwidth of a UPS van full of CD ROMs"

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 8th Jan 2021, 8:21 pm   #171
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Yes, until I start wanting to measure IMD over a range of tone spacings. Then it's back to praying neither of my HP8640Bs are poorly, plus cascode post- generator amplifiers and a drawer full of attenuators.

But your point about crystal oscillators does point the way to simplify receiver phase noise measurements. Just one quiet oscillator is all that is needed. Is there a cookbook recipe for one that delivers no more than -160dBm/Hz noise at 1kHz on, say, 7MHz using an off the shelf AT-cut crystal? There seems to be a lot of black magic surrounding crystal oscillators.
Alan_G3XAQ is offline  
Old 8th Jan 2021, 9:08 pm   #172
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,799
Default Re: Replacement for HP8640B signal generator

I made a little unit with a couple of crystal oscillators in it, battery powered with 1 Watt drivers into isolating attenuators and a combiner. It ran on a PP3. You could tell a lot about a transceiver with a quick listen. I once went round the RSGB convention at the NEC trying anything I could have a go with It was slightly depressing how few importers twigged what I was doing.

You can always change one rystal to shift the spacing.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 9th Jan 2021, 2:05 pm   #173
G0HZU_JMR
Dekatron
 
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
Default Re: Replacement for HP8640B signal generator

Quote:
Sorry, Jeremy, it's the phase noise that the clock multipliers mutilate.
Agreed but with the AD9954 it is possible to bypass the multiplier/PLL with a 400MHz external clock. Sadly I can only use a Mi2024 sig gen to generate this at the moment and this will probably degrade the close in phase noise of the 9954 DDS by a few dB.

I managed to find the old AD9954 eval board and it does have two AD9954 devices on it and they both appear to be working. In theory, it should be possible to arrange them to be in quadrature because the eval software from AD allows 14 bit control of the phase. It is still possible to download the eval software from AD but I already have a copy here and I think it is still the latest version 2.0.0.

This AD9954 eval board is ancient and it uses a parallel/Centronics interface! I've dug out an old laptop with a parallel port and managed to find an old parallel/Centronics cable in the loft. My initial fumblings with the eval board suggest it is all working fine. I should be able to use the two outputs to 'self' measure the phase noise using a diode ring mixer and a Tek (RTSA) spectrum analyser. The schematic for the eval board indicates that there is a 1uF cap on the DACBP pin if that is of any interest?

The RTSA has low phase noise but not low enough to measure the phase noise of the AD9954 at 14MHz. However, I should be able to lash something up with the two DDS outputs in quadrature.
__________________
Regards, Jeremy G0HZU
G0HZU_JMR is offline  
Old 9th Jan 2021, 2:23 pm   #174
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

That's great, Jeremy. The PA3AKE published graphs suggest going form 1uF to 100uF improves phase noise by just a few dBc/Hz, maybe 5dB at most.

Gosh: a Centronix parallel port? I've not owned a PC with one of those since the 1980s or early 90s!

Alan
Alan_G3XAQ is offline  
Old 9th Jan 2021, 2:28 pm   #175
G0HZU_JMR
Dekatron
 
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
Default Re: Replacement for HP8640B signal generator

Quote:
Is there a cookbook recipe for one that delivers no more than -160dBm/Hz noise at 1kHz on, say, 7MHz using an off the shelf AT-cut crystal? There seems to be a lot of black magic surrounding crystal oscillators.
I have fairly limited experience of designing crystal oscillators and I find that it is quite difficult to accurately predict the phase noise performance of a classic/basic Colpitts 'cookbook' circuit. This is because the active device only conducts for a tiny fraction of the cycle in a typical Colpitts circuit and also it isn't easy to predict/prove the power in the resonator.

However, if I abandon the Colpitts circuit and simulate using a classic class A feedback amplifier and I enter a typical loaded Q for an AT cut resonator and I assume -10dBm in the resonator and allow for device NF and losses in the resonator I get the phase noise plot below. This is an incredibly dodgy simulation but I think the phase noise looks reasonable.

To improve on this would require more power through the resonator and there will be a limit to what an AT crystal can tolerate. However, I guess you won't be concerned about ageing effects so maybe you can push the limits a bit? At higher resonator drive levels the simulator predicts that it will be possible to get close to -160dBc/Hz at 1kHz but this will require a fair bit of drive level (abuse?) to the AT crystal. I think SC cut crystals can be driven harder but they might not offer the same loaded Q performance.
Attached Thumbnails
Click image for larger version

Name:	AT_7MHz.jpg
Views:	39
Size:	71.2 KB
ID:	224193  
__________________
Regards, Jeremy G0HZU
G0HZU_JMR is offline  
Old 9th Jan 2021, 2:55 pm   #176
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Thanks, Jeremy. If you could check the PA3AKE phase noise figures without too much effort that would be great. Your close in noise estimate for a crystal oscillator is not hugely better, maybe 4 or 5dB. But I do take your point about possible differences with a Colpitts class C type oscillator. David might have an opinion on this when he's not busy chasing neddies round a field with a foot of snow.

Alan
Alan_G3XAQ is offline  
Old 9th Jan 2021, 3:03 pm   #177
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Some of the very lastest megabux amateur receivers are coming up with PN figures above -150dBc/Hz but more affordable radios are in the low 140s. If the PA3AKE figures are accurate it strikes me that an AD9951 might be just good enough for the basis of a signal generator for measurements on the Elecraft K3 and similar. I'm thinking the switched attenuator from my #3 HP8640B might be a suitable donor and I have a 0-10 1db-step attenuator in a box somewhere in the garage.

Alan
Alan_G3XAQ is offline  
Old 9th Jan 2021, 4:37 pm   #178
G0HZU_JMR
Dekatron
 
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
Default Re: Replacement for HP8640B signal generator

Ideally, the phase noise of the test generator should be 10-14dB better than the device under test so this will be quite challenging to meet for some modern HF receivers.

I've started playing with my 9954 DDS board and I think my overall test system is still a bit compromised. The Mi2024 has fairly good phase noise at 400MHz but since I last played with the DDS board I've added an old and scruffy IFR 3416 vector signal generator to my pile of test gear. This has slightly better phase noise at 400MHz when I measure it on one of my analysers. I get -110dBc/Hz at 2kHz offset. Looking at some library (E5052A) phase noise plots for one of the IFR 3413 generators at work at 432MHz this looks to be correct and it should manage -110dBc/Hz at 2kHz offset, -120dBc/Hz at 10kHz offset and about -142dBc/Hz at 100kHz offset at 432MHz.

If I do my tests at 14.3MHz then 20*log 400/14.3 = 29dB. Therefore, this might limit the DDS performance for a 14.3MHz test signal.

I'm also going to be near to the noise floor limits of the Tek RTSA at audio frequencies. I think the noise floor is typically -155dBm/Hz. I'm going to need to add a preamp ahead of it to squeeze a bit more performance from the analyser. The DDS chucks out -5.5dBm at 14.3MHz. The mixer drops this by about 6dB and I also have to include a 3dB factor for the summing of both DDS noise levels.

I need to optimise the calibration a bit but my initial attempts to measure the phase noise of the AD9954 at 14.3MHz show a response slightly worse than the green trace in post #163. However, I think I'm partly exploring the limits of the current test setup.
__________________
Regards, Jeremy G0HZU

Last edited by G0HZU_JMR; 9th Jan 2021 at 4:43 pm.
G0HZU_JMR is offline  
Old 9th Jan 2021, 5:06 pm   #179
Alan_G3XAQ
Pentode
 
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
Default Re: Replacement for HP8640B signal generator

Blimey, I'm going to have to buy you a beer after all this.

This 9951 family is already looking pretty good to me. Once you are happy with your test setup I wonder if you can dob on a 100uF cap at DACBP and confirm or disprove the claimed improvement?

If I buy the kit from Australia I'm going to puzzle over the Supplier VAT that came into force in the New Year. Maybe sidestep it by trying to persuade them to sell via eBay.

Alan
Alan_G3XAQ is offline  
Old 9th Jan 2021, 6:21 pm   #180
G0HZU_JMR
Dekatron
 
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
Default Re: Replacement for HP8640B signal generator

At the moment I'm seeing the results below at 14.3MHz from the AD9954:

-139dBc/Hz at 2kHz offset
-145dBc/Hz at 10kHz offset

This is 2dB worse than PA3AKE was measuring with a 1uF bypass cap but I think it may be due to the limitations of my test setup. In terms of the 400MHz external clock I do seem to get slightly better results with the Mi2024 sig gen than the IFR3416 which is a bit puzzling. I'm not really sure how the overall noise/jitter from the 400MHz clock affects the phase noise but the IFR3416 is several dB cleaner at a 2kHz offset but it gives inferior results at the DDS output by a couple of dB. I'm not sure why...

I think I need to improve the test setup before trying the 100uF bypass cap. I might also try using a low noise OCXO as the LO for the mixer rather than one of the AD9954 outputs. This would mean testing at 10MHz but it might give better results. I also have a CLC425A amplifier somewhere that would be ideal as a baseband amplifier. This is an ultra low noise opamp that runs from split +/-5V supplies and it has really good performance. This would lift the output well above the noise floor of the Tek RTSA.

Note that the AD9954 datasheet suggests the noise performance at 9.5MHz is as below when using an external 400MHz clock:

-142dBc/Hz at 2kHz offset
-150dBc/Hz at 10kHz offset

This is similar to the results from PA3AKE at 14.3MHz although the datasheet performance is slightly better at the 10KHz offset.
__________________
Regards, Jeremy G0HZU
G0HZU_JMR is offline  
Closed Thread

Thread Tools



All times are GMT +1. The time now is 3:43 pm.


All information and advice on this forum is subject to the WARNING AND DISCLAIMER located at https://www.vintage-radio.net/rules.html.
Failure to heed this warning may result in death or serious injury to yourself and/or others.


Powered by vBulletin®
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright ©2002 - 2023, Paul Stenning.