Replacement for HP8640B signal generator

[Alan_G3XAQ]

I have two HP8640B sig gens. Yes they are wonderful when working. No they don't keep working for long. Old age is getting to them. Allan has published a long history recently of the things that go wrong, and his experience is just a sample.

One of mine has a failed bespoke HP ECL divider chip so it only works above 8MHz. The chips are hard to find but I've managed to source one from France. The other produces RF but has an as-yet undiagnosed failure in the counter display. And no, it's nothing as simple as the EXT pushbutton being pressed or the toggle switch at the back.

Sooo, I'm wondering if there is an affordable sign gen on the second hand market for a (very) few hundred quid with a noise performance anywhere near as good as the 8640B? Or am I in cloud cuckoo land?

73, Alan G3XAQ

[G0HZU_JMR]

Probably the nearest alternative in a similar price bracket would be the old Marconi 2017. However, I wouldn't recommend it unless you can live with the size and weight. I'd expect it to be more reliable than an 8640B but that isn't saying much really.

The Marconi 2018 and 2019 generators have fairly good phase noise across the 2-30MHz HF band. Not as clean as an 8640B but still quite good. The prices tend to be quite low and the 2018s are often below £200. However, these generators aren't very reliable. They should be easier and quicker to fix than an 8640B but you may find yourself inside the generator every year or two to sort intermittent issues.

If you want low phase noise and good reliability then you might have to consider the Marconi 2023. These will probably sell for >£500. The phase noise is a bit better than the 2019 but not quite as good as an 8640b. However, the very close in phase noise/jitter within maybe 100Hz will be much better than an 8640B because it is fully synthesised.

[Radio Wrangler]

Stewart of reading (I know, sounds like a hair salon) keep offering 8662A generators for about £750. I have a few t work that get used where phase noise is more likely to be a problem. Two get used as an intermod setup. They're fairly good on wide phase noise pedestal which is the bane of this test.

They're old, but not as old as the 8640.
They break down, but seem fixable no placcy gears!

OK not quite an 8663 with the discriminator loop option, but do quite well.

David

[Skywave]

Alan - first, please excuse me if this suggestion is unworkable for you, but it is an obvious thought. Can you not make one good 8640B out of the two faulty ones?

Al.

[G0HZU_JMR]

That would certainly buy some time and the best advice I can offer is to be patient and wait for a bargain to pop up as a future replacement. That's the way I've always bought or salvaged test gear and this has worked very well for me for many years.

[Alan_G3XAQ]

Thanks for the suggestions chaps. I'd like to have two sig gens so I can make intermod measurements although in fairness I don't do that very often. As you've said, maybe my best option is to use one 8640B as a donor to keep the other working although which to use as donor and which as recipient is a dice throw. Still, that makes buying one 8662A instead of two a bit more palatable. Yes, I've been happy with Stewart of Reading. One of these sig gens came from them and also my Tek 2445A scope.

Oh, and being too poor (yeah, I know you are all sobbing) to buy a new Rigol or Siglen spectrum analyser I'm also in the market for suggestions for a replacement for my old clunker TF2370. It came from John's Radio in Birstall (remember him?) for £100 but only goes to 110MHz and these days almost everything I want to see is above there. Well, I'd rather not see it but you know what I mean.

73, Alan G3XAQ

[G0HZU_JMR]

A lot depends on what kind of work you want to do with the analyser. At work we had a TF2370 back in the early 1990s. Not many people wanted to use it because it looked so retro and it had that odd milky display and the control was very 'manual'.

However, I recall it was a high performance spectrum analyser. One or two older members of staff were highly protective of it. I didn't use it that often but the two tone IMD dynamic range was quite impressive on it and it had quite a range of RBW filters. Therefore, you may find yourself to be disappointed with a replacement unless you buy something old but high end.

If you have the space and can live with the size, the weight and the fan noise (and the heat in the summer) then I guess the benchmark would be the classic HP8568B. Sadly it will be getting harder and harder to find a reliable and unmolested example and the CRT display will often be quite tired on these analysers. Prices start at about £500 and go up to £1000 but you would have to be very patient and very picky to find something that would be reliable and niggle free. Plus there is the misery of the loud fan noise they produce. The fan noise is fine in a typical and busy lab but it will be an issue if used in a spare bedroom for example.

Anything that starts with HP859x is a low performance analyser especially by today's standards. They are plentiful and quite cheap and they look nice and fresh at first glance but don't be deceived. By all means buy one if the price is right but these analysers are a bit like buying a (reliable) Austin Maestro. They have limited log range and limited RBW range and a poor display. If you want to do critical IMD testing then the limited log range of these analysers might be a problem. Some of them only go down to 1kHz RBW which is another significant limitation.

The next level up is the HP8560/1/2/3. These are nice lab grade analysers and they look quite similar to the HP859x series. The prices for the 2.9GHz HP8560A(E) with a CRT display are tumbling but you are unlikely to get one below about £900. We still have a few of these at work.

The alternative is to spend less and get one of the classic low/mid class machines like the HP 8553, 141 or to buy something from Takeda Riken (Advantest) or Anritsu.

My advice for most people would be to either dig deep and buy the fresh and modern Siglent (with all its modern features and decent performance and the big display) or to be patient and try and grab an HP8560E for a bit less. The A version will be a bit cheaper still but the E version offers digital RBW filtering below 300Hz and this is a very nice feature to have.

If you do check out the Siglent then make sure you can live with the high phase noise on spans of about 1-2MHz. You will be looking at -102dBc/Hz at 100kHz offset. The HP8560E will probably be 20dB better than this with its (fragile?) YIG based LO.

[Radio Wrangler]

The 856x portable models use the same CRT as the 8568/8566. It's completely unobtainable as a spare and there will be people actively looking.

Can be used with an XYZ display... it's vector written, not raster.

There is a rare version of the 856x portables with a colour LCD.

The 859x family started as a low cost range, and the performance was fairly poor. The 8590A was just the works of the 8558 plug in for the 180 scope repackaged with some added DACs and ADCs and a CPU added. They drift like the devil is after them.

The later 859x variants are a lot better. They still only have 80dB display range, but the logger is better than this and, yes you can do reasonable intermod tests with them. The 6x family is still a higher performer, but the 9x has closed the gap a lot. The 9x family got a raster scan display and an analogue TV output for monitor/video recorder. You can get built-in trackers as well on some variants. Narrow RBW is an option board in the 9x, standard fitment in the 6x

Max RBWis 1MHz in the 6x, 3MHz in the 9x but you can press RBW 5 MHz on the 9x and get it wider, though no longer a 5th synch shape. This is actually rather valuable in the day-job, but not the ham shack.

The 9x family got a big boost in functionality and interfacing as the 9xE range. Successful to the point that it had to be introduced to the 6x range.

At home I have a 9x prototype that was never released. Too likely to hit 6x sales. And an 8566B 'Doomsday box' (its original R&D project name). Between them, they do what I need.

David

[G0HZU_JMR]

There is a rare version of the 856x portables with a colour LCD.

We have quite a few of these at work, mainly the 26.5GHz 8563EC version. They are largely unloved because the LCD display isn't nice to view. The trace and the graticule often clash with poor contrast between them. They make the CRT version look very dated but I still prefer the CRT version.

The 856xEC range will generally command high prices so will sadly be way out of budget in this case.

[G0HZU_JMR]

The big old Advantest TR4172 (1800MHz) would be another contender although it is even bigger and heavier than the HP8568B. It does have a tracking generator and can be configured to work as a basic VNA and impedance analyser when used with an external bridge. This requires the impedance measurement option to be fitted but I think many of them have this.

This is a very high performance analyser with low phase noise and really good dynamic range. RBW range is 7Hz to 1MHz. The CRT display is very good with 1000 x 1000 datapoints in the display. Prices will probably start at about £500. However, I can't really recommend this analyser because it is so huge and heavy and there are a few reliability issues and spares will be very difficult to find.

[Alan_G3XAQ]

I confess to being put off a bit, chaps. Something bigger/heavier than my TF2370 and equally hard to fix isn't attractive. All the HP stuff is either big, expensive, or has less IMD DR and noise performance than my old clunker. Or all three.

The same goes for the HP 8662A sig gen: a bit dear, too heavy to go on the shelf, and the description of fault finding here is a bit, erm, daunting.


https://whatever.sdfa3.org/hp8662a-s...or-part-1.html

For my interest in HF and unmodulated CW maybe a homebrew Si570 and switched attenuator might be less adventurous. Or am I being naive as usual?

73, Alan

[G0HZU_JMR]

Yes, some of this old stuff is really big and heavy. The HP 8662A weighs something like 30kg. The HP8642B is slightly heavier at something like 33kg.
The Advantest TR4172 spectrum analyser weighs about 50kg!

I made a vow a few years ago to not let any more large and heavy test gear into the house because it takes up so much space and is hard to move around. At some point I'm going to get rid of a lot of my old test gear. I haven't used some of it for a long time and it just takes up space.

For my interest in HF and unmodulated CW maybe a homebrew Si570 and switched attenuator might be less adventurous. Or am I being naive as usual?

I think it would become a bit tedious to use a setup like that unless you only use it occasionally and you don't do lots of tests that require lots of changes in amplitude.

A working Marconi 2018 would probably be the cheapest lab sig gen that offers fairly good phase noise across the HF bands. Prices are usually in the £100 to £200 range. Do you really need the phase noise performance of the HP 8640B?

If you want to go really cheap (and small) on the spectrum analyser then consider the little RSP1A USB SDR. These are about £100 and there is some spectrum analyser software that can be used with it. However, it's only really useful as a narrowband signal analyser with spans up to a couple of MHz. When zoomed in to narrow spans of <20kHz to look at narrowband signals it can often outperform all of the analysers so far suggested. It won't be that good for IMD testing but probably good enough for typical amateur radio testing.

Where it will be fairly hopeless will be when you want to look at the whole of the HF band or maybe want to sweep 0-500MHz. Then it will display lots of spurious and it will be slow to sweep. It is then more like a toy rather than a serious instrument.

[Alan_G3XAQ]

I like to make and modify HF receivers. An average performing receiver might have 100dB IMD DR in 2.7KHz. If I have my sums right (often not true) the noise from even a "good" sig gen at -150dBc/Hz is -116dBc in 2.7KHz and so is only just good enough not to mask DR measurements with noise. And ideally the noise spec is needed down to 2KHz carrier offset when testing CW receivers. If a receiver has a really quiet LO then finding a sig gen to measure reciprocal mixing is problematic.

I'm not much interested in SSB/FT8. My main focus is CW (I suppose that's a bit like the confessions at AA meetings). This means I don't often look at the output IMD of anything. Harmonics (down at least 60dB) and their filters are of interest as are parasitics. I do sometimes look at the close-in keying spectrum if a transmitter has a dodgy looking time-domain keying envelope.

So the test gear I have mostly satisfies my spec, with the noted exception of specanz bandwidth, but I would dearly love for it to be smaller and less heavy and for it NOT TO KEEP BREAKING DOWN.

Yes, I know: I don't want much for my few hundred quid do I?

73, Alan

[G0HZU_JMR]

Those are fairly serious requirements!

At work we have a lot of experience/history with high performance receiver design and that is partly why we have always had high performance spectrum analysers in the RF labs. Often we connect a spectrum analyser in place of the digital IF when doing development work on a downconverter design.

This requires a really good spectrum analyser. Over the years we first used the TF2370 at the digital IF, then the 400MHz Marconi Mi2382, then we were donated a HP8568B and this was followed by an Advantest TR4172 when we reached the dynamic range limits of the HP 8568B.

There wasn't much between the Mi2382 and the HP8568B but the TR4172 was a useful step up in dynamic range and it was especially good for measuring IMD and harmonic distortion. The first LO runs at about +23dBm and the front end mixer has 8 diodes in it. The typical mixer IP3 was +22dBm across the HF band and into VHF. The phase noise at 100kHz offset was about -128dBc/Hz. The 2HI performance was also a class apart from the other analysers. However, I don't think this analyser would meet your requirements!

[Radio Wrangler]

I never did understand why HP stuck with the antiparallel diode pair mixer even in the low band of their analysers. Significantly better mixers are possible.

Things went badly wrong at HP after the 8568/8566 analysers. They were good, but nothing better was brought out for a period of DECADES. R&D projects were run, one after the other, each ending in a 'business decision' to can it. They did not understand that doing nothing was worse than doing the wrong thing. The 6x portables were cut down derivatives of the 8566 (not the 8568, which is interestingly different)

8566 are OK, main failure is tants in the coil driver boards where they get cooked.

David

[G0HZU_JMR]

In terms of the sig gen phase noise, it's >30 years since I last used an HP8640B but the manual states that the typical phase noise at a 2kHz offset is about -105dBc/Hz up at the main UHF oscillator. This covers 256-512MHz.

So by the time this gets divided by 100 to cover 2.56-5.12MHz on the HF band the noise will be 40dB cleaner. So it may well manage better than -145dBc/Hz at a 2kHz offset down at these frequencies. This assumes that the signal chain doesn't limit the noise floor. A lot of lab sig gens have a noise floor at around -145dBc/Hz. Some are worse than this.

In comparison, the old Marconi 2018/9 used a suite of 4 low noise VCOs to cover over an octave up at UHF. The 2018/9 uses a similar divider + filter system down to about 2MHz so the noise performance won't be that far off the HP8640B. It might be within about 10dB across the HF bands but I can't remember now. The problem with the 2018/19 is that this sig gen produces a lot of low down spurious terms and a bumpy far out noise floor. This issue affects a lot of modern sig gens and it's as if the designers decided that nobody cares about this aspect of the performance any more. This allows the sig gens to be smaller and much lighter.

I can dig out one of my 2019s tomorrow and test it for close in phase noise on my Tek spectrum analyser. The analyser won't be able to test it down to its ultimate noise floor but it should be able to show the close to carrier phase noise for carrier frequencies across the HF band. I think the results will be quite good but probably about 10dB worse than an HP8640B at a guess. At 2.5MHz carrier frequency I think the noise will be about -135dBc/Hz at a 2kHz offset because I remember measuring it a while back for someone else on this forum.

[G0HZU_JMR]

I just dug out the service manual for the Marconi 2019A. The four VCOs cover:
260-309MHz
309-368MHz
368-437MHz
437-520MHz

It might actually be more interesting for me to measure the close to carrier phase noise when the sig gen is set to these frequencies. From memory, the main PLL in the 2019 has a loop BW of only about 150Hz so the close to carrier phase noise is going to be quite high up at UHF. I could probably measure the phase noise fairly well out to offsets of about 5kHz or maybe even 10kHz.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

it's as if the designers decided that nobody cares about this aspect of the performance any more.

Therein lies the crux

Since the rise and rise of the cellphone, wifi, bluetooth etc, the vast bulk of RF hardware is only needed to do trivially short ranges.

Low power, high levels of integration and cheapness are wanted at all costs.

There are still people doing high performance RF engineering for niche applications, but the money they have to spend on test gear is trivial in comparison.

Add in modern business approaches that tell you to design your products to suit say, 90% of uses, and then you totally ignore the rest. You'd make a loss on them if you went for the remainder, and it would take more (and better) engineers from serving that 90%

You need sig gens somewhat better than the things you're working on, but as the things being developed are less stressed in terms of dynamic range, phase noise etc, the top model sig gens didn't get refreshed. Where development did go was into quadrature modulation, arbitrary modulation generators and the ability to create complex test signals.

The sig gens with I/Q modulators tend to run those modulators at low levels which is bad for noise floor, and the modulators are broadband for maximum verastility.... you can see where this is going!

Cheapness says to use a single frac-N loop and this is inevitably a noise shaping design, so you get a significant pedestal of noise extending maybe 100kHz around the carrier. Plus the broadband noise from all the broadband signal processing and amplification. Cheapness says you also don't do much division with its attendant bandpass filtering, you instead use a het band to what we see as fairly high frequencies. So the phase noise improvement of division doesn't go down very far.

There is still a market for sig gens one notch better than the single-loopers, where you get PLL top loop and then a frac-N bottom loop to generate high resolution.

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.

David

[cmjones01]

What about the HP 8656/7 A/B generators? They seem to be reasonably plentiful and the documentation available is good. I got an unloved example of the 8657B for about £130 from a disposal auction a couple of years ago. It's served time at the Nokia plant in Bochum (the presets still have GSM-related frequencies in!). It came with the bottom cover missing, and I think there's something wrong with the frequency doubler so amplitude of signals above 1GHz isn't right, but I find it very useful. I don't do high performance receiver design but the specs look sensible on paper. I leave it to those with greater RF experience to judge whether it's actually any good or not.

Chris

[Alan_G3XAQ]

Originally Posted by Radio Wrangler

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.

That's a pity. the HP8640B is getting harder to fix. There are solutions for the nasty plastic gears but some of the bespoke components like the own-brand ECL is becoming unobtainable and my limited experience suggests those chips are nearing end of life, I suppose from metalisation migration or packaging imperfections.

A friend has offered me his partly-working 8640B so I'm thinking maybe I can make two runners out of the three. I'll need some fancy footwork to get "yet more junk" past the missus though.

73, Alan