Racal-Dana 9008 modulation meter issue

[Chris Wilson]

I picked up a cheap supposedly good Racal-Dana model 9008 modulation meter at a rally today. It kind of works, but has a slight issue. Power off the analogue meter sits happily at zero. On power up, the meter rises a bit, on the 0 to 10 scale it sits at about 0.3 to 0.4. I have a decent Marconi 2019A sig gen, and inputting 20% AM modulation shows readings that are high by about the same as the meter offset. It works though, with the lock being lost only a few mV over the spec at all frequencies, or at least up to 1024 MHz, which is far as my sig gen goes. I have the service manual for the 9008 off Bama, but it's not immediately apparent how to zero and adjust level readings. Before I get carried away and tear it apart, can anyone offer any newbie tips as to what to look for, apart from checking the power supply levels and ripple. I have a decent `scope and some other basic gear. Thanks.

I have attached, I hope, two photos of the thing in action, with a cheap multi meter voltage reading from the DC (external meter) out, rear socket, showing the levels there are not zero either. Thanks.

[Radio Wrangler]

This sort of mod meter down-converts the signal to a lowish, fixed IF using an oscillator controlled by the frequency in the IF (sometimes a phase-lock, more often a frequency locked loop).
There is a precision detector to convert AM at IF to baseband audio and an AC voltmeter calibrated in % modulation, similarly there is an FM discriminator (often part of the frequency locked loop arrangement as well) which feeds audio to the same AC voltmeter.

The first suspect must be a DC offset in the AC voltmeter section.

What does it read on other %AM scales? and on FM deviation scales?

David

[PETERg0rsq]

First test I would make would be to see if the error is affected by selecting a different range.

If there is no difference the problem is being generated purely within the meter driving circuit, and not the measuring circuits.

The meter driving circuit could then be further analysed to try and see where this error is being introduced.

Disconnecting the output from the attenuators, and grounding at a convenient point may also help, but care would be needed!

My gut feeling for this error would be a leaking FET, possibly Q31 or Q32. (does the error vary when selecting Peak, mean or through modes?)

Again careful grounding of the signal path around this area may help isolate the problem.

The other most likely cause is again leaking FETs in the attenuator (Q39, Q41 and Q43) and I note there is a link before the attenuators for isolating them from the rest of the measuring circuit for testing!

I found a printed copy of this part of the circuit in my notes, so that means i have had problems in this part of the circuit on one of my 9008's sometime in the past!!!

Dont forget the Racal Dana Yahoo user group!

Some of these meters also have an outboard AF filter assembly that is not documented.

I have made a basic drawing of this if you feel that might help at some stage.

Peter

[Chris Wilson]

Wow, two replies already, many thanks!

OK, let's go through the three function positions, with nothing connected to the meter, it might be better if I quote the 0 to 3 scale readout :

AM: 0.07 ish

Level: 0.12 ish

FM: 0.1

All positions show a VERY slight waver in the needle.

The meter range position switch does not change these readings once it re-settles.


The reading DOES change through the 3 positions "Peak, Mean and Trough". It rises from the first to the last by about 0.1 overall.



I checked the power supply volts and they looked fine.

After the PSU checks the service manual gives some "Lock PCB" checks. I checked the voltage as instructed on the lock PCB pins. With respect to pin 18 I saw the correct voltages on pins 17 and 10. It then says to adjust R25 on the Sampler Assembly to give 4.00 V on the junction of R11 and R12. Now I was seeing 2.28 volts. The multi turn pot R25 allowed me to up it to 4.00 V but it took several turns. I am not sure if this test point SHOULD be showing 2.28 V, and needs raising for the subsequent tests, or if it should ALWAYS be 4 V. The adjustment to give 4 V has seemed to drop the idling reading of the meter. I suspect that if the test point at the R11 / R12 joint should have been 4 V then something is amiss. It wasn't just a tad out.


The meter has a front panel AF filter switch, it attenuates the correct frequencies, seemingly the right amount, when I alter them from the sig gen, with the AF filter set to "in".

Thanks for the ideas, it appears this meter is familiar to some of you gentlemen.


I assume most semiconductor components are available, or suitable substitutes for them?


EDIT: The PCB Measuring board voltage tests at the DC out rear BNC are wrong, all figures are a bit high, both AM modulation and following their FM test with 50kHz deviation. Not sure whether to try adjusting as they specify or if it may be an issue as you describe above. Will hopefully await for instruction

[John_BS]

I don't know anything about this meter, but looking at the schematic (if it's not a leaking FET or something earlier), try replacing the four 47uF capacitors in the meter drive area with tants (C35, 36, 38 & 41).

John

[Chris Wilson]

Thanks John, I disconnected the attenuator isolating link and the meter still reads above zero, exactly as it did with the link connected. I am trying to work out if I can further isolate the problem, as I do not have any new tantalums, only ordinary electrolytics. Cheers.

[Alistair D]

I would check the supply rails for ripple.

I have one of these meters so can make measurements on it if it will help you. PM me the link to the schematic so I know what I'm doing.

Al

[Chris Wilson]

Thanks Alistair. PM sent. I did a quick freeze spray poke about, cooling Q49 immediately gives a negative voltage on the DC out BNC and the panel meter drops below zero on all meter ranges. Could you measure the voltage on the DC out BNC on the back panel with meter set to AM, range toggle switch set to 1.5 - 50 and meter range slide switch set to 100 / 50 please? Do you see 0.0 V and does the panel meter sit at zero? I see about 20 mV on the DC output socket.


I have done a full measurement PCB alignment and it's quite a lot better, but not quite spot on.

Thanks!

[Alistair D]

Hi Chris, there is 18.5mV on the DC socket and the meter is sitting one small division above zero.

Al

[Chris Wilson]

Hmm, perhaps this is OK now I did the measurement board cal then..... Am I getting paranoid in my old age? This sits 1 division above zero now, too.

Thanks Alistair. I am not happy about Q49 though, it's VERY temp sensitive, I may just change it. Thanks for taking the trouble to check this Alistair.

[Alistair D]

Just given Q49 a quick puff of freezer. The meter needle didn't move.

Al

[Chris Wilson]

OK, that's really good of you to pull the cover, I have ordered some BC109's and will swap the two suspect ones (Q49 and Q42. Thanks, will report back when they arrive. I might have had something that would do, ut for the price have ordered the real McCoy. Thanks again. John_BS, the ripple was fine, but thanks for the idea!

[G0HZU_JMR]

Hi Chris

From my interpretation of the schematic (and I have no experience of servicing this meter) the detectors Q49 and Q50 should be biased so they have close to zero volts DC at their emitter legs with no test signal because any DC offset here at the emitter will find its way to the meter via the buffer op amp IC1 and the meter will rise above its zero point. So Q49 and Q50 need about 0.7V at their base to get 0.00V at the emitter.

Q49 gets its wanted 0.7V bias at its base via the two 'diodes' Q47 and Q48. They form a fairly stiff voltage reference at around 0.7V.

Q50 gets its 0.7V bias at its base via the setting of the 10K trimpot R120 that straddles these 'diodes'. So presumably the correct setting for this 10K trimpot will be somewhere near its mid point.

Having looked at the circuit I think one potential weakness is C35. This must have a very low leakage spec because otherwise it can override the bias reference fed to Q49 from Q47 and Q48 and upset Q49 operation on positive peak detection.

Can you measure the current direction in resistor R116? Ideally this should be feeding in a tiny bias current to Q49. Possibly less than 0.5uA so there will be just a few mV drop in R116 (22K) at the base of Q49 on a healthy unit.

However, if C35 is leaking then it could be the case that R116 will have a tiny current flowing the other way and the detector Q49 will behave erratically. T

My interpretation of the circuit is that the zero position of the meter is set by R171 for the positive peak detector and the zero point is for the trough detector is set by R120. However, the alignment procedure suggests otherwise and this is a bit confusing.

To get zero volts out of the opamp IC1 there needs to be zero volts at it input pin 3 and therefore there has to be zero volts at the emitter of Q49 (and Q50).

On first glance I assume the meter scaling is a straight line function like Y = MX+C where (for positive peak mode) C is the DC offset set by R171.
C is the DC offset set by R120 (for negaive peaks) ?
M is the setting of R134 and sets the 'gain' of the meter.

But the alignment procedure seems to suggest you adjust R171 etc a different way. This is a bit confusing but I only skimmed through the manual so may be missing something.

Regards
Jeremy

[G0HZU_JMR]

Originally Posted by Chris Wilson

Hmm, perhaps this is OK now I did the measurement board cal then..... Am I getting paranoid in my old age? This sits 1 division above zero now, too.

Thanks Alistair. I am not happy about Q49 though, it's VERY temp sensitive, I may just change it. Thanks for taking the trouble to check this Alistair.

On a healthy unit the temperature stability of Q49 should be controlled by the tracking reference 'diode' Q48 which is just another BC109.

It's therefore slightly unfair on the circuit design to just freeze Q49 and not Q48 because in the way the circuit is designed, each B-E junction should track the same in terms of mV per degree C change and the temperature drift should cancel.

So, in theory at least, I'd expect to see a tiny, tiny bit of meter movement with freezer spray on Q49 (but with Q48 still at room temp) provided the meter is set to positive peak mode or mean mode.

However, if it's moving a LOT with the freezer then maybe Q49 is faulty or maybe C35 is leaky. BC109's are cheap and plentiful but you may find that these parts were hand matched by Racal.

eg Q47,48, 49 and 50 may have been selected for similar Vbe behaviour over temperature.

[Chris Wilson]

I have some 47uF tantalums coming, so will change this first, and see what happens. I am out tonight, but will see if I can measure the current direction in R116 tomorrow.
Certainly the measurement board alignment has improved things and it's basically a functional meter, but now I have started I'd like to see if it can be improved. So before changing the BC109's I'll do more checks. Thanks Jeremy! Is there an off the board test for leaky tants?

[G0HZU_JMR]

Yes, you can check the caps for leakage externally and this would be a good thing to try.

I would probably just do it with a DVM on the resistance range with the correct voltage polarity from the meter but that is a bit crude. I'd imagine there are plenty of people on here who do these tests a better way.


Also, you may already know this but:
When you come to calibrate it properly you can use your HP3325A function generator + Mi 2019 and the HP8568B to get a really accurate setting of the 'test' deviation.

i.e. set up the above to get a Bessel null on the analyser. You need to inject an accurate modulation frequency (from the HP function gen) to the ext mod on the Mi 2019 and then fine trim the sig gen deviation for a known Bessel null on the analyser. You then have a very accurate 'test' deviation.

The AF/MOD generator inside the Mi2019 won't be accurate enough in frequency because it isn't synthesised from the internal 10MHz reference. It's just a basic audio oscillator at approx the right frequency so it would cause an error in the Bessel null. So you need to use that big and accurate HP3325A generator at 1Vpkpk fed to the ext mod connector on the 2019 to get the exact audio frequency

[Chris Wilson]

Steady on Jeremy, my head's hurting Using any of the above test gear independently is not too bad, combining all three.... Well, I am a bit confused, need to read up on Bessell nulls first, sorry!

I have grabbed a few minutes now I am back and this is what a cheap multimeter "sees" across R116. The readings were very low and with the function switch set to "level" I saw a reasonable reading both ways.

The voltage with the positive meter probe on the C35 side of r116 shows -0.004 volts

The voltage with the negative meter probe on the C35 side of R116 shows a stable 0.003 volts.

Using the AM positions of the "function" switch the reading is NOT stable and moves about from roughly -0.007 to -0.002 with the positive meter probe on the C35 side of R116 quite rapidly. Does this help? Thanks.

[G0HZU_JMR]

Well here's what I'm expecting to see at Q49 assuming the -10V rail is reasonably close to -10V.

We know there is a 220k resistor from -10V to the emitter of Q49 and the emitter is at about 0V because the meter hardly moves.
So the Q49 collector current must be about 10/220k = 45uA.
If we assume the hfe of the BC109 is about 100 (could be a lot higher) then the current through R116 should be about (45e-6)/100 = approx 0.5uA.
So there should be about (0.5e-6)* 22k = about 10mV drop across R116


You measured about 4mV which is kind of in the right ballpark. But on reflection maybe the best way to test for leakage of C35 is to simply lift one leg of it and see if it affects the meter zero significantly.

You can see that any leakage current from C35 must be very low because the normal base current for Q49 is less than 1uA (assuming I've got my sums correct LOL)

Also, I wouldn't be too worried if the meter doesn't sit exactly on zero. I'd be more worried about fitting 'new' parts to it that aren't as well suited as the originals. It could be that your meter just needs a good alignment although the temp sensitivity of Q49 is a bit of a worry.

In peak mode I'd expect the emitter voltage at Q49 to vary by about 2mV per degree C change if only Q49 was frozen (and Q48 was left at ambient)
So if you froze it by 20degrees then maybe the detector voltage would go down by 40mV and the meter would fall a bit?

The other thing to bear in mind is that in order to calibrate this meter you really do have to make sure your signal generator has very low modulation distortion and very low spurious on the modulation. I think this is because the 9008 uses a peak detector in the final voltmeter section. Also it must keep any RF spurious well suppressed and ideally this should be -50dBc or better.
The main issue is that the relative phase of any harmonic distortion on your modulation source will cause a measurement uncertainty in the peak detector and this will be affected by the phase of the modulation harmonics relative to the fundamental modulation tone.

Therefore, I'm not sure a Mi2019 is the ideal tool here but it's probably close enough. But unless you use a lab standard signal source then the measurement uncertainties from your test gear will creep up. However, it depends if you just want to use it as a general purpose 'indicator' or as a semi serious bit of kit.

I've got a similar meter here (slightly different Racal model) and I dug it out and mine doesn't quite sit on zero volts but it agrees extremely well with a
Marconi 2024 signal generator and the Mi2024 does have very clean and accurate modulation. Maybe not lab standard but probably good enough to
calibrate the meter. However, I've never ever tried to go inside and calibrate it because it already agrees so well with the Mi2024 on both AM and FM.

Note that the peak detector in the 9008 probably won't be very linear with a very tiny AC signal fed to it so anything you do read down near zero meter deflection is suspect anyway. So it might be normal to have it riding just above zero. Tweaking R171 and R120 for a precise zero here might actually degrade the accuracy across other parts of the meter dial. But that really is just a guess

[Chris Wilson]

The user manual does say to try and operate in the top third of the meter scale, so you are quite correct about the lower end. I lifted the leg of C35 and it made no difference whatsoever. I am beginning to think I am worrying needlessly. I have already ordered some BC109's so will pull the suspect one (carefully, I have a great Pace de-soldering tool that works superbly), and put it on my Peak transistor tester and compare it with a new one. I'll only change it if it is obviously very different.

You have a good memory recalling I have an HP3325 (it's a B suffix model, buy hey, I'm impressed!). I will try and make time to check out your earlier post re Bessel Nulls and all that. Sounds daunting to someone mathematically challenged though

I'll see what Q49 looks like against a new one and decide whether to leave well alone now, or change it. Thanks Jeremy and all who have weighed in with great advice.

[Herald1360]

You may need maths to understand the FM deviation spectrum, but not to use Jeremy's suggestion. Well, no more than a bit of arithmetic anyway.

There's another description of the method here:

http://www.fmsystems-inc.com/manuals/BESSELart.pdf

which doesn't seem too daunting and adds a bit of explanation.