HP8568B calibrator noise / spurious signals

[G0HZU_JMR]

It sounds like it's worth replacing R39.

I'm trying to make sense of how all of these locations have this spiky issue. Can you check the various reference voltages? eg at TP13. If R39 is jittery then R17 could be too. R17 is near TP13. This reference feeds the DAC for one thing.

Also, what is the reference voltage at TP15 like?

[Chris Wilson]

I just replaced R39, and the issue continues. Seems *something* in that region is unstable, can diodes or Zener diodes go funny? I also changed Q5 and Q6 as a matter of course. On board A22 I know that C10 and C24 show highish ESR, but I only have a 1uF bead tant for the 1.5uF C24, and a correct capacity bead tant for C10. Can I use bead tants?

TP15 is stable, no spikes. TP13 also seems stable. I am sure it's something on this board though. I believe it's something physically near R39, light tapping sends it wild. I don't have a riser card, but could possibly make one. Cheers Jeremy, got to go out soon, but really appreciate the help.


TP15 is ultra clean compared to TP13 though, see screen shots. Is that to be expected? Same scope and probe settings for both.

[G0HZU_JMR]

It does seem to be a strange one...

One classic trick is to give a suspect area of A22 a few seconds (no more) with a hairdrier on warm blow setting. See if the issue gets much worse. As the PCB is very long you might be able to narrow it down to maybe an area covering a third of the PCB using the hairdrier.

The modern tool for this is a SMD hot air reflow pencil with the temperature turned right down but a hairdrier could still be fine although it can quickly cook and potentially damage things.

Freezer spray can be useful too as long as it is used very lightly.

I'm still concerned about the DAC area after you mentioned it went a bit mad after you soldered a wire there. Was the board still warm from the soldering?

I think you could try those caps but looking at the shape and narrow width of the spikes it's hard to see how a dodgy cap could cause or allow such a sharp spike (and the circuit quickly recharge the voltage) unless it had lost nearly all it's capacitance and/or developed extremely high ESR.

Worth a try though...

[G0HZU_JMR]

I took the cover off mine again with the analyser on its side and tried tapping the A22 card all along its edge after removing the perspex cover. I do see some microphony on a 1.1MHz span on 20MHz CF but only enough to make the filter shape wobble slightly. It's a bit like a muted version of what my earlier video showed. But no spikes were visible.

Looking at the circuit near U18 the -15V rail feeding Q17 and U18 has been given some hefty isolation from the rest of the -15V rail using a 10R resistor and a 100uF cap C17. This must be for a reason so it might be worth checking C17 for ESR if you haven't already. It's hard to keep up with what has and hasn't already been tested.

The ribbon cable to connection point J1 near R39 would be a potential weak spot for an intermittent connection. You could try lightly touching this ribbon cable to see if it has developed a poor connection? Obviously, don't overdo this or you could make it unreliable even if it was previously OK. Normally you can diagnose these by lightly wobbling the ribbon up and down a few mm near the connector. Don't twist the ribbon though.

Also, it could be worth inspecting the PCB soldering in this area to look for poor joints as there are several devices on heatsinks. Maybe 25 years of hot/cold cycling has opened up a poor solder joint somewhere in this part of the board.

can diodes or Zener diodes go funny?

Yes, and so can transistors. My old HP8405A VVM had several BJTs that went intermittent with temperature. The behaviour became more and more erratic the warmer they got.

The intermittent faults are always the worst!

[Chris Wilson]

I am beginning to think I have 2 issues. The side to side shimmying, only seen in over 1MHz spans, and the noise. When the shimmying is bad R39, both the original and the new, is hyper sensitive, as if some other component value change is making it almost microphnic in its sensitivity. Right now there's no jitter and I can tap R39 and gently press down on the adjuster screw and nothing happens. When the jitter starts and becomes bad, tapping or just the lightest press on the R39 adjuster sends the trace sideways many divisions.

I am now wondering if the noise is a separate issue, I occasionally see jitter with no noise, and vice versa. Usually I get both though.

How is R39 actually set, and what should be seen at TP1? I see a rising a falling volatage there, as R39 is adjusted. On a scope there;'s no apparent noise on it.

C17 measures fine Jeremy, thanks for the idea though. I have changed any tants on A22 that measured suspect.

I re-soldered any joints to devices that run hot, or devices nearby. No joy.


I heat things up with my de-soldering iron set to blow rather than suck and set to a low temp. It''s such a fine jet of hot air I can warm an individual resistor, no problem. I need some more freeze spray though. I had to do an emergency plumbing repair for a friend and used it to freeze a water pipe as his stop cock was seized up

I now have some of the tiny co-ax connector adaptors to BNC for the link leads, so if anyone thinks of a test involving them I can now connect to male or female.

[G0HZU_JMR]

I'd have to look in the service manual for any R39 adjustment info.
However, I do have a suggestion. Can you generate a reasonable comb of signals across the 1.01MHz span?

I can do this easily here with a vector sig gen (see below) but for your purposes you don't need every tone to be the same amplitude.

Could you generate a similar comb using AM (or FM) from your sig gen?

The reason I ask is because with a single test signal you are only looking at one tiny segment of the YTO tuning ramp at any one time. If you have lots of test tones you can maybe see a pattern in how this thing is behaving because you can monitor numerous parts of the YTO ramp in (almost) real time. The spike may be hitting the ramp/tones in a pattern. So you may see some tones behaving in one part of the sweep as others get the wiggles.

[Chris Wilson]

What about using a noise source? I have access to one of these:

http://www.g8fek.com/BBGen.htm

Any use? I am not sure how I would get a comb from a sinewave output from my sig gen. There are commercial comb generators available on Ebay, but they are in the USA or Israel.

I see how to be sure R39 is set correctly now, I need to measure the frequency from the rear socket as it's adjusted. I want to check that as I don't want to be unsure that fiddling with it has caused otherunknowns, cheers Jeremy.

[Radio Wrangler]

No, not a noise source. They just give a floor of featureless noise. The point in using a comb is to have markers you can check are evenly spaced. An old type amateur radio 'crystal calibrator' would be good.

David

[G0HZU_JMR]

Can you set your sig gen to FM external modulation and feed in a square wave from a function generator at 100kHz at 1Vpkpk (approx) to the ext mod input of the sig gen?

Then set the sig gen to external FM and set the deviation to a high level on the sig gen (eg 100kHz) deviation.

This should give quite a spread of tones across a 1MHz span on the analyser. The tones won't all be at the same level but you should get 10 or more tones across the 1.01MHz span.

It's not worth buying a comb generator just for this. It might not prove to be a useful test but I'm expecting to see differences across the span in terms of which tones wiggle and when they wiggle.

[Chris Wilson]

Hmmm, I may be able to link a function generator to my sig gen.... I'll have a play. The other possibility is using the xtal calibrator in my FT-902DM, not sure how easy it is to bring the calibrator signal out from it, and I don't want to ****** up anything else


I'll try setting R39 back to the correct figures, then have a look at the function gen / sig gen set up. Thanks David and Jeremy.

[G0HZU_JMR]

I had another look through all your scope plots and although I can't yet explain why the spikes are there I think you may have an issue at TP17.

I think TP17 should provide a linear tuning ramp and you don't seem to have this at TP17. Your ramp waveform is curved and kinked.

It looks to me like the circuit uses Q11 as a constant current generator. The current from Q11 will follow the classic Shockley equation in that the current will increase exponentially with increasing Vbe at Q11. So Vbe at Q11 needs to be controlled very carefully (especially with temperature changes) to deliver the target constant current for a given sweep time.

So the circuit driving Q11 emitter will be arranged that Q11 delivers a specific constant current into C16 for a given sweep time. This will obviously charge C16 such that the voltage across C16 rises in a linear ramp at a rate that corresponds with the desired sweep time. If your TP17 waveform is as per your previous plots then I think this needs to be fixed because it should be a linear ramp.

So either the current delivered into C16 isn't constant across the sweep or something is loading this circuit.

Q17 JFET and Q19 BJT look to be a high impedance unity gain buffer after C16so the voltage at the gate of Q17 should be the same as the voltage at its source if the buffer is working OK (source of Q17=TP17)

Can you look at TP17 for a very wide span (eg default 0-1500MHz) and then alter the sweep time and look at TP17. The ramp should be different for different sweep times and for very slow sweep times it will get very long.

Does it behave itself on some of the sweep time settings or is the ramp always curvy and kinked as per your earlier plots?

[G0HZU_JMR]

Q17 JFET and Q19 BJT look to be a high impedance unity gain buffer after C16so the voltage at the gate of Q17 should be the same as the voltage at its source if the buffer is working OK (source of Q17=TP17)

Note that I mean the amplitude of the ramp waveform will be about the same at the gate and the source but obviously there will be a DC offset of a volt or two between source and gate.

I tried to think of something that could cause the sharp/huge positive spike in your TP17 scope plot in my previous post.

Because the ramp voltage recovers to the approx the same place after the spike event, I think that the spike can't be happening at C16 because C16 resumes charging from the same place after the spike. So one possible cause of this spike at TP17 is that something is intermittent in or around Q17 or Q19. It only takes a brief and tiny spurt of extra/unwanted base current to cause Q19 to briefly spike up the voltage at TP17.

So it might be worth checking the soldering around Q17, Q19 and the various resistors in this area. It is possible that either Q17 and Q19 are playing up but it's worth checking the soldering around here.

Maybe there is a common issue here that contributes to spikes and also the curvy ramp response at TP17.

[Radio Wrangler]

The spectrum analyser needs a wide range of sweep speeds. They don't need to be especially accurate and only a 1-2-5-10 sequence of steps is really needed, so to get the range without switching the timing capacitor, someone has had the idea of using a logarithmic converter, so that a linear series of increments produces a logarithmic series of sweep rates. The logarithmic converter is current in/current out and is formed by U16B and Q11B.

U16B runs with its inputs sitting several volts positive of ground. Circuit block B makes +10v (with a deliberate tempco) to act as a general reference for the whole ramp-current source.

Q11A is used to control an offset to the reference voltage in order to temperature compensate the log converter transistor Q11B which is the other half of a dual part. this offset voltage affects the speeds of all sweeps, so it has the trimmer for adjusting the slowest speed (blurred R66?)

Schematic block 'C' is a bank of switched resistors to ground, modified by a temperature dependent voltage set by the Vbe of Q12, which is used as just a diode.

The resistance to ground of block 'C' modifies the gain of U16B and hence the output voltage of U16B. Note that this is all with respect to the reference voltage. Note also that the 'gain' of U16B is rather low if you think of the grounded ends of block 'C' resistors as being an input to U16B.

The difference between the output of U16B and the reference voltage sets the base-emitter voltage of Q11B. The result is that the collector current of U11B is the anti-log of the current through the switched resistors in 'C' and so it covers an expanded range.

Why on earth did they use such a complicated arrangement, bring ing in all sorts of touchy temperature compensation etc? OK, it saves switching some capacitors but it doesn't use fewer components. There is only one benefit from using a single unstitched capacitor, and that is that if you push a button to change sweep speed part way through a sweep, the speed changes and continues from the place you pushed the button. With switched capacitors for different speeds, the sweep would jump to whatever charge there was left on the new capacitor every time there was a change.

So this explains what all the stuff before C16 is doing, but C16 should have a nice linear ramp on it. No curviness, no spikes.

There are switches to allow the sweep to be halted to make a pause for the frequency controller to operate, of for the signal frequency counter to run. This will put a flat spot on the ramp - just a bit of horizontal time, not a spike.

There is a switch to reset the ramp to the start. This uses U1B as a feedback system to pull C16 voltage down until the output of the buffer amp (Q17, Q18) is held at a voltage set by R109, R110. This seems to be working.

TP17 looks at the ramp after the buffer so your scope doesn't load the ramp capacitor C16 directly. On TP17 you see a curved ramp and spikes. The spikes are fairly fast and would take a lot of current to make C16 move that fast. It's unlikely that this is happening. Besides, as Jeremy says, C16 would show a lasting effect from such a jump.

What about something downstream putting spikes onto everything? Q17 is a JFET and its gate should be a reverse biased diode. If something big enough happened downstream the FET junction could become forward biased, but that doesn't explain the curvature.

I think Q17 isn't happy and it's taking gate current which increases with input voltage, thus slowing the ramp on C16, and curving it. Maybe its glamorous assistant, Q18, isn't good and Q17 is trying to drive the output alone?

This points to replacing Q17 and Q18.

But first, I'd check the voltage on C17 for spikes, and the ESR of C17. I'd check that U18 wasn't trying to do anything during the ramps by probing R110.

HP's usual resistors are long-lived. Their capacitors are showing age, and there are some transistors which definitely have shown long-term issues. Q18 is probably an 1853-0020 (good old 2N3906) and marked 3-020 usually. I'm not going to guess Q17


David

[Chris Wilson]

Wow, great info and analysis, thanks you two! I did manage an hour and a half on it early yesterday, and followed the set up routine for A22 in the manual. Nothing was a mile out save R39. That was out and is also very touchy to get the frequency right from the rear BNC. Not touchy as in noisy or scratchy, touchy as in a tiny amount of turn gives a big shift in frequency. More than I would have said was good design practice. Again this comes back to something in that circuit perhaps having changed? A quick squirt of freeze spray seems to send things a bit daft, too in that area, but hard to say which particular component without a riser board.

After setting up as per the manual it did seem better behaved, but the jitter still returned, and just twice, some spikes.

David: C17 measured fine, out of circuit. 0 Ohms ESR

I also checked all the matt black 0.01% resistors, in circuit, and they showed no obvious value deviation, hot or cold. I am glad you mentioned part numbers, the HP parts list has big gaps where only a HP number is given for some transistors. Is there much chance some are unobtainable?

My gut feeling is there is something "downstream" putting spikes on things, it looks to me, (but what the hell do I know, it's just a *feeling*) too many test points have these seemingly random spikes. They seem so fast as well, just like the spikes on the display. Whilst the jitter seems fairly slow. One tiny clue is if I let it do a self calibration it's common for me to see the spikes straight after....




Jeremy: Should I be probing with the probe at X1 or X10? Nearly all shots are at X10. I don't want to be skewing things with bad measurement practices.



I left it running last night, and will have a look at it in a bit.

I will re-do the scope patterns and post a screen shot. I'll also try with my USB scope, which I now have back after lending it someone for a day, two months ago!

I see the traces in the schematic have a slightly flat top to them, is that significant? Mine re pointed...

Thanks for your patience and fabulous support, this is by far the most complex thing I have ever had to fiddle with.

[Radio Wrangler]

Something downstream could be doing spikes, but the curved ramp points very strongly at those two transistors AND they could also make the spikes themselves.

David

[Chris Wilson]

I'm on it in a few minutes! I'll post screen shots, then `er indoors is putting a wall paper scraper in one of my hands and a list of jobs to do in the other

[Radio Wrangler]

I've got part way through re-felting a wooden stable roof that got shredded by the wind last week. I'm going to be too tarry to go near the computer for a while.

David

Oh, there's a list of HP part numbers versus manufacturer's numbers kicking about the inter web.

[Chris Wilson]

http://www.gatesgarth.com/scope-shots2.zip has today's screen shots.

I took a photos of where I put a probe wire off C17 as I couldn't believe the spikes I was seeing there. X1 probe on the negative side of C17 where it joins R106. That can't be right, surely? Very bad jitter when doing these tests, you can see the different time durations on the TP17 shots. So what's causing the big regular spikes at C17 then??


There are also two shots of probing R110 to R109 junction.

Thanks!

[Chris Wilson]

< 1MHz span, C17, two different V / Div settings, with noise spikes also present on SA screen.

[G0HZU_JMR]

I think we need to fix the TP17 ramp first. It needs to be linear.

This ramp is also used as the reference for the DAC that provides the tuning ramp for the YTO.

From what I can see, for any given frequency span, the DAC is programmed with a static value on its LSB-MSB bus and the reason you get a ramp from the DAC output is because it is referenced to the same ramp generated at TP17.

So the DAC really just acts as a programmable attenuator of the linear ramp signal at TP17. The attenuation is set by the static value on the LSB-MSB data bus and obviously, this value will change with changes to the span selected by the user because you need different degrees of 'attenuation' of the TP17 ramp signal to linearly steer the YTO for given span size.

So if you get spikes at TP17 you will get spikes from the DAC and because the DAC steers the YTO this will cause the YTO coil to 'try' and track these spikes. This will mean big and sharp gulps of current from the -15V supply rail whenever the DAC spits out a spike.

So I think the curvy/kinked ramp at TP17 needs to be fixed as the first priority.

Maybe the spikes will be fixed at the same time? The unity gain buffer comprising JFET Q17+ PNP Q19 is a possible suspect.

The regular switching transients seen at C17 could just be from the ramp resetting