Farnell TM39 Microvoltmeter.

[WaveyDipole]

Split from this thread:-

https://www.vintage-radio.net/forum/...d.php?t=157578

I have just recently acquired a Farnell TM39 Microvoltmeter. It seems to a quite well made if rather large instrument and seems to be in working order. Interestingly is has provision for two internal PP9 batteries and has a logging feature, although I'm not sure what this connects to? On the rear there are two sockets for a 'pen recorder'? The movement seems operate freely and does centre correctly when no power is applied. A quick test on the 10v range shows that the deflection is about correct in both directions, but I can't get it to zero when power is applied. The indicator is stable and there is a certain adjustment range of around one major scale division available, but it always stays to the left of the centre position between the 4 and 2 mark.

The fault is probably nothing major.

[Craig Sawyers]

If it is anything like the Fluke 845A Microvoltmeter, it will be based on synchronous detection. In the Fluke there are two neons driven by an 84Hz oscillator. They are coupled to two photocells, and produce a modulated ac signal which is demodulated after amplification.

The Fluke gives DC 1uV to 1000V full scale in a 1, 3 , 10 sequence.

On the Fluke the

Craig

[Radio Wrangler]

It could have quite high input resistance, so for zeroing you'll need to short the inputs. Also there may be appreciable capacitance across the input in order to provide a stable voltage for the chopper amplifier. This means it will hold charge if left not connected.

David

[WaveyDipole]

Originally Posted by Craig Sawyers

If it is anything like the Fluke 845A Microvoltmeter, it will be based on synchronous detection.

Craig, thanks for your info. Your last sentence seems to have got truncated?

While it my have similar features to the Fluke 845 including the 1, 3, 5 sequence, the range is from 100μV to 30v so nowhere near as wide. I could not see any neons, photocells or any form of optoc-coupler devices on the board. It is built using discrete components and there are no ICs. One feature in common with the TM2 is the pair of FETs so I am hoping that the TM2 circuit diagram may provide some guidance. I will investigate further in due course.

Originally Posted by Radio Wrangler

It could have quite high input resistance, so for zeroing you'll need to short the inputs.

Good suggestion, but unfortunately I did try with inputs both open and shorted with same result. I think part of the problem may be related to the switches as the pointer shifts by quite a bit when the meter is switched from the Zero to Read. I will do some further testing to determine whether it still does that when the input is shorted. There are also two zeroing controls - a standard rotary knob and a recessed pot with a slotted shaft. The latter does not seem to do anything at all, not even on the μV range.

[Craig Sawyers]

The Fluke has a front panel 90 degree rotation switch that has "zero" in one position and "opr" (for operate) in the other. The schematic shows that the switch shorts the input in the "zero" position.

Craig

[Craig Sawyers]

I now have absolutely no idea what the truncated sentence was intended to say!

Clearly a senior moment...

Craig

[WaveyDipole]

Ah, I expect that the TM39 does the same thing when switched between Zero to Read mode except that is has push switches instead. That would make sense and should be easy enough to verify.

No worries about the truncated sentence. I am grateful for the information you have provided.

[Radio Wrangler]

There will be a need to correct for charge injection if FET switches are used as the choppers because switching freq leakage int the amp creates an effect indistinguishable ftom a real DC input.

The effect varies on different ranges due to the z presented by the input attenuator.

Some HP sensitive voltmeters used photoresistors with flashing lights as the choppers to eliminate charge injection, but their speed is limited.

Others use FET switches and balancing capacitors.

David

[WaveyDipole]

Originally Posted by Radio Wrangler

There will be a need to correct for charge injection if FET switches are used as the choppers because switching freq leakage int the amp creates an effect indistinguishable from a real DC input.

Having probed about a bit today, it does look like the pair of FETs is being used in a chopper circuit. I have confirmed a square wave of approximately 1khz being fed to their gates from the multi-vibrator circuit.

With the instrument turned on for quite a while, I also noticed, contrary to my previous statement which was based on a very quick power-up test, that the pointer does actually jump around quite a bit and is not as stable as I had initially thought. This does not seem to respond to any particular prodding of tapping of switches, zeroing pots or PCB. The meter also 'froze' a couple of times - by which I mean that the pointer returned to dead centre (as when it is powered off) and wouldn't respond to zeroing or a voltage at the input. Powering it off and back on again seemed to get it going again.

Both of the zeroing pots are in parrallel across +5.10vdc and -5.17vdc rails which appear to be regulated using zener diodes and the wiper voltage changes through the range as they are rotated. However, the meter does not respond at all to rotating what I believe to be the S.C. zero (recessed with slotted shaft) pot. That being the case, I am now beginning to suspect that perhaps the FET associated with that half of the chopper may be faulty. There might also be a faulty resistor or switch contact, but I ran out of time to trace this through properly today. The part number of both FETs is 2N5654. I found them at Cricklewood Electronics but they were £3.50 each. Does anyone know of any equivalents? Also, do they have to be a tightly matched pair?

One other thing that I observed was that the meter pointer kicks hard just about all the way over to the left as it is powered on and slowly comes back to the centre, sometimes overshooting it and then returning slowly back to settle in the middle. Sometimes it undulates around the centre point a handful of times before settling. Is this normal behaviour?

I did squirt some Servisol into the switches, but this seems to have made little difference. The TM2 schematic has been somewhat helpful in that it reveals the design principles employed by the manufacturer and the multivibrator and chopper circuits do look very similar, but there are differences and naturally the part numbers are different, so it is a bit of a challenge to follow. I will probably have another go at it tomorrow or over the weekend.

[Craig Sawyers]

1/f noise causing the erratic motion? Maybe the input was overloaded at some point by a DPO and that partially fried a device....

Craig

[WaveyDipole]

That seems like a possibility at least. I made some progress checking everything in the chopper circuit and all seems fine, so again this seems to point to one of the FETs.

I also discovered that the 'Read' switch has no wiring to it at all! It simply serves as a means of releasing the Zero button! The Zero button simply shorts the source of one of the FETs to ground. I have yet to figure out what the 'Log' button does, but I found that with this depressed, the measurement went off the scale! Maybe it needs the 'Pen Logger' connected (whatever one of those is), of there is a fault with it. Without further information I am working a bit in the dark. Anyone come across a Pen Logger before or know how one operates? A Google search drew a blank.

I would like to ask advice about some of the mains wiring within this unit.

The mains flex is perfectly fine, there is a fuse on the Live side and the case is properly earthed. The mains flex runs into a connector and the problem is that all wiring from that point is done with ordinary hookup wire. This runs within the same loom as the low voltage stuff up to the power switch, neon light and back to the transformer. I am rather surprised that ordinary thin hookup wire has been used.

The switch being used for power on/off does not look like it should be used for mains voltage, but since the transformer is very small, the unit must be drawing a very small current, so perhaps this was deemed adequate. I tried powering the unit from an external PSU via the battery contacts and it drew only 14mA. The mains current must be just a little over 1mA. It would be difficult to connect thicker, mains rated wires to the pins on this switch.

The question is whether I should re-wire the mains AC wiring and route it separately from the low voltage stuff, or leave well alone?

I decided that the best way to get a better understanding of the circuit is to reverse engineer it and draw it out. Having spent a few hours on it today scribbling on paper and then drawing it in Kicad, I have made considerable progress, but there is still much to do!

[TonyDuell]

The 'pen logger' (or 'pen recorder' or 'chart recorder') would consist of a pen moving across a paper chart. The pen is moved by a servomechanism, basically comparing the (analogue voltage) input with the voltage from a slidewire mechanically connected to the pen. The chart is moved at constant speed (originally by a mains-powered synchronous motor, or even a clockwork mechanism, later by a stepper motor electronically controlled). So you end up with a graph of voltage against time.

It appears that your microvoltmeter could be used as a preamplifier for such a unit, allowing graphs of very small voltages to be produced.

I doubt the 'log' button has anything to do with that. A chart recorder has a fairly high input impedance, I would assume the output on this instrument has a much lower source impedance so there would be no need to make any circuit changes when the recorder was connected. My first guess as to the 'log' button's function is that it gives a LOGarithmic response, the meter reading is proportional to the lpgarithm of the input voltage.

[Radio Wrangler]

Now, if that log button is a selector for a logarithmic law. The log of zero volts should be minus infinity on the scale and the pointer should plant itself far off the left end, on its stop.

You might have nothing wrong other than a choppe transistor (or more likely the amplifier input transistor) gone noisy.

David

[Craig Sawyers]

The sort-of-equivalent Fluke 845A does not have a log law switch - it is linear only.

I have two. And one pegged left. It took a little while of detailed diagnostics to discover that a DPO had taken a transistor out of its little socket (presumably for testing) and plugged it back in wrong. Putting it in the right way round fixed the unit.

Peak to peak noise spec is 1uV - 0.2uV, 3uV - 0.25uV, 10uV to 1000V - 0.3uV all with input shorted. The units I have are better than those figures.

Craig

[Radio Wrangler]

DPO?

David

[WaveyDipole]

Originally Posted by TonyDuell

The 'pen logger' (or 'pen recorder' or 'chart recorder') would consist of a pen moving across a paper chart. The pen is moved by a servomechanism, basically comparing the (analogue voltage) input with the voltage from a slidewire mechanically connected to the pen. The chart is moved at constant speed (originally by a mains-powered synchronous motor, or even a clockwork mechanism, later by a stepper motor electronically controlled). So you end up with a graph of voltage against time.

It appears that your microvoltmeter could be used as a preamplifier for such a unit, allowing graphs of very small voltages to be produced.

I wondered whether it was something like that. I'm not sure how old this unit is. It has a calibration sticker from 2016 suggesting that it had been in use until then, but the components look to be typically 70s/80s when such paper logging devices might still have been in use. Serial number is 280, so perhaps there was not a great many of them made?

Originally Posted by TonyDuell

My first guess as to the 'log' button's function is that it gives a LOGarithmic response, the meter reading is proportional to the logarithm of the input voltage.

I hadn't considered that possibility and just assumed that this was related to the Pen Logger. However I had pondered over the need for a separate button to engage the logger. I would be interested to know more about what such a logarithmic function might be used for? I did some Googling (intend to do more later) and came up decibel measurements in this context, but there is no additional logarythmic or decibel scale marked on the dial. I did come across a descrption of some logarythmic amplifier ICs, but this is made completely from discrete components. I have yet to explore and diagram the amplifier section but hopefully will at some point be able to determine where the wires on that button link to.

Originally Posted by Radio Wrangler

Now, if that log button is a selector for a logarithmic law. The log of zero volts should be minus infinity on the scale and the pointer should plant itself far off the left end, on its stop.

It doesn't move to the left at all. It stays pretty much central, but does shift a little (two or three minor marks) when compared to the Read mode. An old 1.5v cell measures off the scale on the 30V range.

Originally Posted by Radio Wrangler

You might have nothing wrong other than a choppe transistor (or more likely the amplifier input transistor) gone noisy.

David

I was hoping that it was only down to the FETs, however, I found some 2N3819's (I was sure that I had purchased some from a member a little while ago and it turned out that I had mis-filed them) and replaced both. The meter still behaves exactly the same, so the search for the culprit continues.

I attach a section that I have traced out already. I'm not 100% sure that the mutivibrator section is without errors, but I am confident about the chopper section. Am I right in thinking that when the ZERO switch is closed, then the left hand (ZERO1) control should have no effect leaving the right hand control (ZERO2) to SC zero the meter? I am puzzled because with the ZERO switch closed, the ZERO1 control continues to work and I have confirmed continuity across the switch when it is closed. As I mentioned, the right hand control (ZERO2) does not seem to work at all even on the most sensitive ranges. This continues to be the case even with the FETs replaced.

A further investigation and drawing out of the amplifier section is next on the list. As you say, a noisy amplifier transistor might explain the sporadic jitter.

[Radio Wrangler]

Yes, popcorny noise from maybe a transistor, maybe a resistor.

David

[WaveyDipole]

During my investigations I discovered that the meter movement is directly connected to the battery 'TEST' switch on the rear. I hadn't cleaned the rear switches and both of them felt a bit rough. A quick squirt of Servisol had them operating smoothly again. For good measure, I also replaced the smoothing caps across the +9V/-9V supply rails. This seems to have made a difference.

The meter still undulates and jitters a bit on warm up, but now seems to settle down after a couple of minutes. Now that the pointer is steady once warmed up, I did some tests on the lower ranges using a handful of resistors (10k + 1k, 100k + 1k, 10M + 1k, 10M + 100Ω) as a voltage divider across a 1.5V cell to check various ranges. The meter pointer did not bounce around like it had before, but I did find that on the μV ranges I got only negative readings despite the probes being connected in the forward direction. Reversing the probes didn't change anything. At 100μV I initially got a reading in the right ball park, which then became erratic, but on the 30μV range the reading was well off and on the μ10V range the meter stayed at zero. Maybe insufficient current by that point?

Of course, the meter was very sensitive to the probe leads being touched while o/c on the lowest two ranges but I did wonder whether the current across the divider might have been too small by that point to register a reading properly. The probes I used were not the best. I had already given the rotary selector the Servisol treatment, but I might try giving it another treatment to be sure. With the SC zero control not operating and the OC zero control still having an effect when the input is shorted, something still doesn't seem quite right with the chopper adjustment.

[WaveyDipole]

I was diverted to another project for a while so had a break from this. Just got back to it today.

It seems that that I had somehow managed to switch the labels of the two zeroing controls on my reverse-engineered diagram. This led me to get stuck on the idea that the slotted control must be the S/C zero. It would now appear that this is, in fact, the O/C zero and the rotary control the S/C zero. Now it starts to makes sense. The slotted control does have an effect, but only when the Zero button is OFF. It would appear that the way to set this up is to first press the Zero button in and zero the meter using the rotary control, then press the Read button in (releasing Zero) and zero any remaining offset using the slotted control. Once done, it doesn't have to be adjusted much again. The meter now seems to stay zeroed regardless of whether it is switched to Zero, Read or Log.

Now having done that, I can now get sensible readings on the μV ranges. Using a divider made up of 10MΩ and 10Ω across my 1.5V cell, I got a reading of something over 10μV on the meter. Switching the lower resistor to 100Ω gave me a reading of around 120μV on the There was quite a bit of jitter on the μV ranges and also on the lowest mV ranges making an accurate reading difficult.

Unfortunately after some time the meter got very erratic again, with the pointer jittering quite a bit even on the higher voltage ranges and with being unable to be zeroed again. I hate intermittent faults!

I would like advice about the leads to use? There are actually three sockets, red [+], black [-] and green [GND (chassis)]. There is a capacitor between negative and ground. Should the lead be screened, perhaps with the screen to chassis? Or should the screen connect to the black?

I have reverse-engineered most of the main board. I have attached two of the drawings for reference for anyone who might need them in the future. I also have a drawing of the amplifier section but still need to fill in some component values. I will post that when it is completed.

In the meantime its back to square one....

[WaveyDipole]

I think the hot weather today did me a favour! As it got hotter, the instrument began playing up continuously which allowed me to track the noise to its source. It turned out to be a small electrolytic cap. Positive bias to the transistor base was supplied via a resistor divider from the main positive supply rail. The cap sat between this lower voltage supply derived from the divider and ground, adjacent to the positive bias resistor. I was able to follow the circuit using the diagram I had reverse engineered and using an oscilloscope probe, from the output right back to this first stage transistor. Since there were two of this type of 20μF caps in the early stages of the amp, I replaced them both.

The meter pointer is now rock steady and can be zeroed quite easily. The meter itself seems pretty accurate being less than one minor mark out on all ranges. My test using a 10MΩ/10Ω divider across the old 1.5V cell (cell actually measures 1.41V on the DMM and about the same on the TM39) now produces a steady reading of about 14μV. The photo shows the meter on its lowest 100μV range.

The DMM had trouble keeping a steady display and fluctuated from around .0009mV to as much as 0.0018mV and sometimes dropped below .0008mV. Isn't that 0.9μV to 1.8μV though? The DMM leads were attached to components leads with rather flimsy clips so perhaps the contact was a little iffy. The measurements were done with one meter attached at a time, not both simultaneously. It seems that I evidently attached the DMM leads the wrong way around as well! In any case this was just a quick test done under shed conditions to check that the meter is in the right ball park.

There is still a part of the meter to diagram and the values of resistors in the amplifier circuit need to be recorded on the diagram, so still a bit of work to do, but some progress made today.