21st Jun 2020, 10:30 am | #21 | |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
Quote:
The A versions were quite poor and I believe it is the 8656A that is a bit of a lemon with poor phase noise and lots of close to carrier spurious terms. This generator probably best described as 'dirty'. The 8657B is much better and it looks to be a good general purpose sig gen with phase noise easily good enough for most applications. It doesn't quite qualify as an 'ultra' low noise (and low spurious) generator for the HF bands though.
__________________
Regards, Jeremy G0HZU |
|
21st Jun 2020, 10:42 am | #22 | |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
Quote:
__________________
Regards, Jeremy G0HZU |
|
21st Jun 2020, 11:02 am | #23 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
See below for a phase noise plot of the main 256-512MHz oscillator used in the HP8640B. I vaguely recall reading that the later versions of the 8640 had slightly improved phase noise. I'm not sure if the plot below is the old plot or the newer version.
Edit: I've managed to find new vs old so both plots are shown below. I've had a quick look at both of my 2019 sig gens and (unsurprisingly) they can't match this despite using 4 separate VCOs to try and compete with the single cavity oscillator in the 8640. The results are still very good though. I do remember at work that the company designed a VCO that was heavily based on the VCO design used in the 2019. Our version looked very similar with a BFR91 and a printed resonator and we covered it with a milled metal screen. However it was much too large (and heavy!) to be used in any of our high performance receiver designs. I think we used it as part of the UHF LO in a satellite downconverter ahead of a modem.
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 21st Jun 2020 at 11:08 am. |
21st Jun 2020, 11:46 am | #24 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
Here's a phase noise plot of one of my 2019s at 262MHz. This can be compared to the 256MHz plot for the HP8640B.
At a 2kHz offset there isn't much in it. The plot shows -104dBc/Hz at a 2kHz offset for the 2019. The later/improved HP8640B is maybe 4dB better here. Above about 5kHz I think the Mi2019 phase noise plot below will be compromised by the analyser phase noise. So the noise at 10kHz offset will be the phase noise of the analyser at -117dBc/Hz. To try and predict what the phase noise of a typical high performance printed VCO might be at a 10kHz offset I did a quick spreadsheet design and the excel spreadsheet shows the predicted phase noise of a reasonably high power VCO. This is what I would expect to achieve if I had to design an old school VCO covering 260-309MHz. I used to be allowed to do design work like this in the 1990s. However the NRE costs of designing a VCO put a stop to that a long time ago. All our VCOs are manufactured for us by the usual big names these days. I suspect the phase noise response in the excel plot below is fairly close to the performance of the 2019 VCO and this is at least 10dB worse than the big cavity oscillator in the HP8640B when measured at a 10kHz offset. I've included the phase noise plot of the later HP8640B below for comparison.
__________________
Regards, Jeremy G0HZU |
21st Jun 2020, 12:02 pm | #25 | |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,894
|
Re: Replacement for HP8640B signal generator
Quote:
I suppose it fitted into the range replacing the old 8660A which was a thermal life-test machine if ever there was one... later 8660 got a HUGE fan bolted on the back. 8657 and B models came along with some useful improvement, and the addition of a doubler. These are reasonable sig gens if you want AM/FM and aren't trying to measure adjacent channel rejection of good radios. Below them in status came the 8648 some years later which was an ultra cheap job aimed at oriental firms making radios and FM tuners as a quick source for a production line worker. Its schtick was a selection of 1-button presets. Single loop, poor phase noise, poor spurs. THe intended replacement for the 8640 was the 8642. Fully synthesised, designed for low phase noise, designed to use about the same amount of rack space. So 8640 users could 'upgrade' and get synth accuracy and programmability. It's quite a good sig gen, but hit the time that everyone put their cellphone goggles on, so didn't sell in high quantities. It is UNBELIEVABLY heavy. How they got the weight without lead or depleted uranium beats me. It's a little heavier than the 8662 but in a smaller case, so the apparent weight is rather noticeable. it seems to have more and sharper corners when carrying one. Ours at work has a 'two person lift' sticker. Anyway, the 8642 parentage is the comms signal generator line. The old 8660 was the synthesiser line primarily intended to make frequencies and be programmable as a lego brick in an ATE rack. The 8662 was a replacement for the 8660 and got built-in AM/FM without needing a plug-in. Along with it came a big step in reliability a big reduction in noise, the 8660 was really early generation os synthesisers went. The 8662 begat the 8663 and the 8663 grew an extra-low phase noise option. This is what went into the phase noise measuring systems. Adjacent channel and intermod tests were considered the hardest tests, but there was another less well known bugbear. If a broadband mixer gets a very broadband signal like say an FDM multiplex, and the LO has broadband noise from amplifiers etc, the resulting noise floor in even a narrowband IF can be most surprising. You begin to think the universe has it in for you! You're into the noise*noise mixing problem. When I had to get through this the 8640 broadband noise from its amplifiers and jitter from those dividers was too much. Guy and Hugh found that only the old generation valve sig gens like the 606/608 were low enough noise. Horses for courses again..... (and I'm riding again, after 12 weeks out of the saddle. Quite entertaining because the neddies are full of sass) David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
|
21st Jun 2020, 12:38 pm | #26 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
At work we have several generators from the flagship PSG (Performance Signal Generator) range from Agilent/Keysight.
These cover up to 20GHz and I think they were incredibly expensive when new. However, I don't think the close in phase noise on the HF bands is that special because these sig gens use a downconverter for output frequencies at VHF or below. https://www.keysight.com/en/pcx-x205...s?cc=US&lc=eng I think I've still got one of our old boards in the garage with a VCO based on the 2019. Maybe this is going way off topic but I'm tempted to try and find it in the huge boneyard box in the garage. From memory, whoever designed (copied it?) it made a few goofs with the layout and choice of components so it probably won't be as good as the original. I also tried measuring the phase noise of the 2019 when set to 4.09MHz. This is 262MHz divided by 64 so there should be a 36dB improvement in phase noise. I got very close to 36dB improvement out to 1kHz but by 2kHz the analyser phase noise was limiting the result. I had to use a different analyser for this test. It showed -136dBc/Hz at 2kHz offset and in theory it should be -140dBc/Hz. However, the internal phase noise of the analyser isn't much better than this at 2kHz offset. So I think it's reasonable to expect that a healthy 2019 can manage a phase noise of about -140dBc/Hz at 2kHz offset for frequencies in the range of the 80m amateur band. The HP8640B might be 4 dB better than this at this offset.
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 21st Jun 2020 at 12:46 pm. |
21st Jun 2020, 1:15 pm | #27 |
Pentode
Join Date: Jun 2017
Location: Basildon, Essex, UK.
Posts: 188
|
Re: Replacement for HP8640B signal generator
There is always the TK2373 frequency extender with would take the 2370 up to the GHz. I know Stewart’s still stock them.....
|
21st Jun 2020, 1:20 pm | #28 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
The other option (for reciprocal mixing testing of a receiver) is to simply make a basic LC test oscillator for the 80m band. It should be stable enough and theory states it should be possible to achieve close to -160dBc/Hz at 10kHz offset. At a 2kHz offset the performance will be limited by the flicker noise of the transistor but if this is chosen carefully it should be possible to achieve about -140dBc/Hz here.
This will cost less than a fiver, will fit in a fag packet and will weigh next to nothing. It would mean making a dedicated oscillator for every amateur band though. At offsets of 100kHz I'd expect the phase noise to be low enough such that it would beat the HP8640B by at least 15dB at 100kHz offset.
__________________
Regards, Jeremy G0HZU |
21st Jun 2020, 1:41 pm | #29 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,894
|
Re: Replacement for HP8640B signal generator
Years ago the 'ZNX dirty tricks department made a little group of three very small diecast boxed screwed together. Two contained crystal oscillators 20kHz apart in the CW end of the 40m band (crystals in that vicinity were readily available for QRP stuff) along with a drive amplifier. The third box was a hybrid combiner with fixed attenuators. The levels were chosen to give an HF rig a good run for its money. Battery powered it went in my pocket, and it was good fun at a rally where one of the main importers had brought along their latest wonder. Connect its short cable to the inpout of a transceiver and tune across the appropriate frequency range and see how audible the 3rd and 5th order intermods were.
Not exactly numerical (think of the laws of S-meters!) but it certainly sorted the men from the boys. Left behind somewhere in Dayton, not sure whether I mis-placed it or it got lifted. If someone did nick it they would have to have known what it was to have thought it worthwhile unless it was just grab it and find out what it was afterwards. Needless to say I wasn't advertising what it was and what it was for. Only tw of the stands had folk who sussed out what I was up to. 'Test signal generator for sensitivity' was mostly truthful. There was no rumour going around the place of a brit with a pocket 2-tone generator. Remember, general purpose sig gens take on a lot of compromise because of that one word, 'general' If you have one specific task, you can outdo them for pocket money. Oscillators based on ceramic resonators can be quite decent VCOs for synths. Low phase noise crystal oscillators are a field in themselves. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
21st Jun 2020, 3:46 pm | #30 |
Pentode
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
|
Re: Replacement for HP8640B signal generator
Yes, Wes Hayward and is acolytes published details of such test oscillators in 1977 in a well-known ARRL book. Gosh, was it really that long ago?
It is sometimes handy to be able to measure IMD versus tone spacing but maybe one working HP8640B plus a crystal oscillator would be a sensible way forward. I'll go collect the donor next week. You might think a batch of three ought to yield one that works. I'm now being distracted by tales of the TK2373 range extender for my TF2370 spectrum analyser for not much money. If a 3-box set weighing 60Kg will cover up to GHz then I'll go with that and forget about esoteric HP replacements. or expensive Rigol/Siglent newcomers. 73, Alan |
21st Jun 2020, 5:02 pm | #31 |
Pentode
Join Date: Sep 2014
Location: Bedford, Bedfordshire, UK.
Posts: 174
|
Re: Replacement for HP8640B signal generator
Alan,
There is always the British designed and built Racal 9087, brief details attached. You don't see many of them about. Racal know a bit about phase locked loops and I guess this was brought out as a rival to HP's 8662 for low phase noise. It's a solidly built beast as they have to be with excellent shielding of course. I have one which doesn't get used enough sadly. I also have another British sig gen, a Farnell PSG1000. I tend to use it more. It doesn't have the performance but is rather more portable! Ian |
21st Jun 2020, 5:16 pm | #32 |
Pentode
Join Date: Apr 2018
Location: Canterbury, Kent, UK.
Posts: 189
|
Re: Replacement for HP8640B signal generator
Hello Ian. OK, so something else to look out for, I don't understand the spec sheet. It says -150dBc/Hz but the graph doesn't go below about -142. What am I missing?
Alan |
21st Jun 2020, 6:09 pm | #33 |
Pentode
Join Date: Sep 2014
Location: Bedford, Bedfordshire, UK.
Posts: 174
|
Re: Replacement for HP8640B signal generator
That graph is for relatively close in to the carrier. The attached clip from the spec sheet shows -150dBc further out.
|
21st Jun 2020, 6:15 pm | #34 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
There will be a degree of specmanship (AKA weasel wording) here. The broadband spec of -150dBc/Hz sounds nice but it may refer to frequency offsets that are much further out than the graph range shows.
All (modernish) lab sig gens that I have used have to be used carefully in order to extract the best/lowest broadband noise floor. To get the best from the sig gen the operator usually has to set the generator to a very high output level and sometimes this also requires some sneaky use of the attenuator lock feature if the sig gen has this feature. This allows the user to explore the upper range of the ALC levelling range and it usually results in a lower noise floor at offsets across maybe 100kHz to several MHz. Even when a basic sig gen is used on the fundamental oscillation range, there will be a lot of circuitry between the oscillator and the final output amplifier and this will usually include an amplitude modulator and levelling circuitry. When the sig gen level is backed off away from the maximum then the sig gen can compromise its own signal to noise ratio. Some 'economy' HP/Agilent sig gens are notoriously bad at this. The HP864x series is one example. At work we bought several of these many years ago. The specs for phase noise looked OK for general use but then we tried to use one to drive a broadband transmit strip up at VHF. This required the sig gen level to be reduced down to -10dBm. The spectrum analyser showed a huge increase in the noise floor. By swapping across to a Marconi 2024 at the same drive level this problem went away. When tested on a phase noise analyser at an output level of -10dBm the noise floor of the HP8648 sig gen was about -125dBc/Hz out to several MHz. These noisy generators caused so much confusion they were banned from the RF labs.
__________________
Regards, Jeremy G0HZU |
21st Jun 2020, 6:33 pm | #35 | |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,894
|
Re: Replacement for HP8640B signal generator
Quote:
R&S seemed to take over the high performance RF instrument mantle for a while, but seemed to suffer badly with operating system woes. HP had an edict from on high that new products must not use homegrown firmware, but must be written on top of Windows. Yup. This made them vulnerable to attack if attached to internet connections. So instruments needed to have malware protection installed and updated regularly. I referred to this as 'The Windows Tax' David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
|
21st Jun 2020, 6:35 pm | #36 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
One way to get experience of the above is to try and do a receiver blocking/mixing test with a modern lab sig gen. As the sig gen level is increased in 1dB steps there will come a point when the relays in the generator's output step attenuator will reduce the attenuation by (say) 10dB and this means that the ALC has to back off the RF level at the output amplifier by 9dB to maintain a +1dB change. I hope I've got that the right way round...
This can cause the receiver under test to suddenly appear to fail the blocking/mixing test because it just got really noisy on that most recent 1dB increase in sig gen drive. But in reality, the sig gen just raised its own noise floor and jammed the receiver with noise. This issue can be offset to a degree by using the attenuator lock feature and this can delay the onset of the step attenuator by up to about 10dB for some generators. However, once this effect is significant it is usually time to turn off the generator and try and find something cleaner. A discretely designed LC oscillator can be hard to beat here as it can easily achieve better than -170dBc/Hz at a 100kHz offset. This will be 20dB better than most lab sig gens at this offset.
__________________
Regards, Jeremy G0HZU Last edited by G0HZU_JMR; 21st Jun 2020 at 6:48 pm. |
21st Jun 2020, 6:43 pm | #37 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
I just dug out the offending phase noise plot from an old archive. It looks like it managed -127dBc/Hz at 1MHz. I think there were earlier plots that went further out and the noise floor was poor across the whole VHF band.
__________________
Regards, Jeremy G0HZU |
21st Jun 2020, 7:57 pm | #38 |
Octode
Join Date: May 2010
Location: Grantham, Lincolnshire, UK.
Posts: 1,176
|
Re: Replacement for HP8640B signal generator
Glad I saw this post, chance to reduce the pile of test equipment I have.
My Marconi 2019A, had it for probably 15 years, scarcely use it now, always works when I turn it on and comes with the original manual. A HP 8568A (so with the 9845 CPU, not the 68000) spectrum analyser with an original set of four HP manuals for the 8566A and the link cables. A GW Instek GSP827 spectrum analyser. This baffles me, the display is about 50dbm too low, but altering the signal by 5 or 10 dBm and the trace alters to match. I thought it was finger trouble, but none of my fingers have owned up. Two Systron Donner frequency counters, one does 18GHz, the other 24GHz, both worked many years ago but the gate time switch had problems on one of them. Enough to make a trip worthwhile, the whole lot for £400. |
21st Jun 2020, 9:03 pm | #39 | ||
Nonode
Join Date: Oct 2008
Location: Warsaw, Poland and Cambridge, UK
Posts: 2,678
|
Re: Replacement for HP8640B signal generator
Quote:
Quote:
Chris
__________________
What's going on in the workshop? http://martin-jones.com/ |
||
21st Jun 2020, 9:11 pm | #40 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Replacement for HP8640B signal generator
Wow that sounds like a really good deal from Woodchips. In case Alan is tempted, I should point out that the 2018/19 generators only do the division trick down to about 2MHz. So for top band (1.9MHz) and across the LW and MW bands the close to carrier phase noise and the far out phase noise will degrade a lot compared to the performance on the 80m band at 3.7MHz. I'm not sure if that rules out the 2019 for Alan.
For the generation of large signals it is easy to overcome this issue for the 2019 by making an external low noise differential amp/squarer (a couple of PNP BJTs) and divide by 10 (74HC390) . This will cost peanuts and the idea would be to run the sig gen at 19MHz and the external circuit divides by 10 to give 1.9MHz. This will restore the noise floor and the close to carrier phase noise. It should then deliver about -140dBc/Hz at 2kHz offset and a decent wideband noise floor. Without this, the spectral purity of the 2019 will not be clean enough on top band. The HP8640 does the division trick right down to 500kHz and I'm not aware of any other make of sig gen that does the same.
__________________
Regards, Jeremy G0HZU |