Solartron OS101 audio signal generator.

[Jez1234]

This was acquired along with an Airmec 701 RF generator (still in the to do list)
from member "wheelyjon" who was acting on behalf of the widow of an electronics enthusiast to help clear her garage of boatanchors. I strongly suspect there were rather more "desirable" (lighter!) items in the silent key sell off and that the two items I got were the ones which non but the bravest of enthusiasts would try to get past the distaff side The price was attractive, ie FOC, and with he help of "Ictutor" and "Radio Wrangler" the FCS did it's job.

Now the first and most overwhelmingly salient point about the Solartron OS101 is it is huge... HUGE indeed and no opportunity to add mass or girth to this unit was spared by the manufacturer. Everything you may reasonably expect to be aluminium is steel. The outer casework is way bigger then it needs to be and yes is entirely steel. Just the outer casework must weigh around 15lbs! The whole thing is half a hundred weight.... It's also 22" wide, 15" deep including the handles and 11" high...did I mention that it's just an audio sine wave generator? It is reassuringly solid...

First I did the obvious safety checks on mains cable, verifying earth continuity etc. I then wound it up on the variac over a few hours and was pleasantly surprised that it worked! Large "body tip and spot" anode load resistors were getting rather warm (3 x 2" long or so 6k8's in parallel for the output buffer) and wax getting molten and as completely expected "that capacitor" x 2 were to blame, one in the buffer and one in the Wien bridge oscillator. The two usual suspect paper caps were replaced by modern polypropylene caps which were very easy to hide as the originals were fitted underneath the turret boards, which are on tall stand-offs. Replacing the awful paper caps raised the HT by 50V! The output waveform now looked great.

Outer casework was thoroughly washed and scrubbed in the bath as were all removable screening compartments and lids etc. The innards were cleaned of coal soot and dust and the front panel given a good clean, all knobs removed and cleaned etc.

As it got some use in testing functionality, calibration etc the originally smooth Muirhead slow motion drive began making awful graunching sounds and so it was removed, cleaned out and everything lubricated. This seemed to do the trick.

Switches, valve bases etc were treated to some contact cleaner "just in case", they seemed OK in fact. A few hours after doing this I switched it back on only to be greeted by hardly any output and what was present badly distorted. I then noticed the purple glow from the Z77 (GEC version of EF91). For the second time in a few months a valve which was working perfectly had gone soft a few hours after being removed from the socket, having its pins cleaned, and the socket treated to a squirt of contact cleaner.... a pattern seems to be emerging here... or more likely a complete coincidence! (no I hadn't "bent the pins straight" or anything like that).
Luckily I had a few NOS Mullard EF91's and normal service was resumed.

I left it on for few hours after this to see if anything else was going to fail but all was well.

A few resistors were replaced which had gone high but not so much as to cause any problem and some of the precision 1% resistors in the frequency determining networks which had gone high to the degree of +2 - 2.5% were replaced by selected for <0.1% modern metal film resistors. Most of the originals were still within +/- 0.5% which is pretty impressive considering the date on the main smoothing caps is Feb 1952! These were left in place. All electrolytics tested fine other than one had been replaced with 1974 dated Plessey green one's, 3 x 10uF in parallel, to replace a 32uF original. I replaced these with a modern one. So it's been repaired before, after 1974 at the earliest....

Calibration was the next step and this is where things got interesting... I've had to accept for now that it's "close enough for jazz" but could foreseeably be better. I have a circuit diagram and parts list but no service manual with calibration procedure you see...

The user manul states that the max output is "100mW into 600R" which is of course 7.75V RMS but the actual output with the front panel level control set to max was around 11.7V RMS. More confusingly its own output meter was showing the correct 7.75V RMS!

The meter is a Sangamo Weston AC reading instrument, 2&1/4" and shown on the schematic as "0 - 10V AC meter. Resistor provided". This refers to the ballast resistor which calibrates it and the original had 19.8K written on it. I checked this, 20.1K so drifted a bit not THAT much! I presumed dodgy internal diodes so opened it up to find a bridge made up of GEC GEX34 germanium diodes. These tested between an expected 0.37V and a more silicon like 0.57 V so I replaced them all with small HP schottky diodes and reassembled the meter. Upon refitting it I was greeted by the same reading of 7.75V! I have to assume that the meter has lost sensitivity over the years, maybe the magnet has lost some of its original strength!?? Using a 100K trimmer across the original resistor I came up with a suitable value for a fixed resistor which would give good accuracy and duly made one up from three resistors to give the precise required value. I now had the front panel meter agreeing exactly with 3 true RMS DMM's and the scope as to the true output level.

As a relevant aside the unit is made up of a two valve Wien bridge oscillator using a Z77 (GEC EF91) and an N78 (a high slope "pocket rocket" of a B7G output valve. Think 6BW6 with 10mA/V Gm) which is followed by the same line up again as a unity gain buffer with an output Z of only 10R and very low distortion. The rectifier is a 5Z4G. Now, the meter reads the output from this 10R output impedance buffer directly, so it seems there can be no ambiguity over what is supposed be being read here. My conclusion being that yes the meter had gone out of tolerance by something like 35%!!

So why was it reading correctly as the unit was first recommissioned? ie the manual says max output 7.75V and this is what the meter read? Well as I said above the actual output was about 11.7V and I reckon the last owner had simply adjusted the preset fine level control in the oscillator to show the "correct" figure on the internal meter which was reading around 35% low! Oh did all this have me tearing my hair out! In fact things were further complicated by the fact that the oscillator cannot be set for as low an output as 7.75V! Around 8.8V is the lowest it will give and the range is about 8.8V - 13.4V. All the components have been tested and anything out of tolerance replaced and the valves have been proved OK by substitution (unless both N78's are exactly equally knackered which I rather doubt) so I'm left to conclude that the thermistor in the Wien bridge has rather drifted in value over the years. It's set between paxolin sheets sandwiched together and mounted in, not behind, the front panel behind the main dial to try and keep it away from internal heat sources BTW. Yep it's pretty much no expense spared for a UK made 1952 AF sine generator and very well made.

As it won't go as low as specified I took the decision that decades are good and set the max level to 10V RMS.

You know I'm still having moments of wondering if I missed a trick here and everything was as intended, which is why it indicated 7.75V as the manual says.... and there is some good reason it was exactly this high in order to compensate for something... Like I said I don't have the factory calibration procedure available. All common sense says though that yes it was out by a mile in terms of both meter accuracy and output level. The manual does state that "the meter reads Volts RMS directly at the output of the low impedance buffer" and going by this then yes I'm as sure as I can be in the absence of a service manual that it is now as correct as it can be. If anyone knows differently then please speak up

Next was frequency calibration... I've had to make some "best guesses" here without the service manual and have ended up with it within spec (+/- 1%) for around 70% of the frequency range and within 2.5% at the very worst spot on the dial on the highest frequency range. Not bad. But I'm sure it can be got considerably better and the spec of +/- 1% was probably actually within more like 0.3% in practice. Laboratory quality instruments are normally very conservatively rated.

The fact that the best calibration I've been able to achieve is with all the trimmer capacitors fully unmeshed hints that there is "a rabbit away" somewhere here...

Now the lowest frequency range of 25Hz - 250Hz has no trimming available for the low frequency end so I checked this first. Sure enough with the main tuning cap fully meshed and the dial set at its end stop, when 25Hz was set on the dial it was near as damn it correct. About 24.3Hz in fact. By experimenting I soon found that with the high end trimmer cap for this range almost completely unmeshed the inaccuracy at the high end of this scale was removed and the low end read precisely 25Hz (the adjustment for the HF end of a range does have a slight effect at the low end). So this was the lowest 25Hz to 250Hz scale that has no preset adjustment for the lowest frequency now reading correct to within about 0.2Hz across the range. Pretty damn good I reckon and giving me no reason to doubt that things are correct and it should be possible now to calibrate the higher frequency ranges accurately. After all, the lowest range can't be adjusted to get 25Hz correct... so as it was reading correctly I assumed this was proof that the basic set up, correlation of dial markings and tuning capacitor etc, must be correct.

After quite some time trying to get the higher ranges calibrated I've got it so it's within the specified 1% on the 250Hz - 2500Hz range, within 1% from 2500Hz to 15KHz on the 2500Hz to 25KHz range and worst case about 1.8% on this range, and within 1% between 25KHz and 100KHz on the highest range but worst points about 2.5% at the high end of this range.

As i hinted above the thing that's got me scratching my head is that this optimum calibration has been achieved with all but the lowest frequency ranges' HF end trimmer capacitor, which is maybe 15% meshed, completely unmeshed... Note that non of these is in parallel with the main tuning capacitor. They are switched in by a wafer on the range set switch and a separate 50pF Jackson Brothers variable is selected for each range and is in series with the bottom half of the actual Wien bridge to the input of the bridge. On all but the lowest frequency range there is also a Colvern pot in series with range set precision resistor at the other end of the Wien bridge which gives a few % adjustment of predominantly the low frequency end of that range. Somewhat counter intuitively, increasing the capacitance of any of these trimmers increases the output frequency... it's too high unless they are completely unmeshed.

So I'm wondering if maybe light alloy furring of the vanes of the main tuning cap has put it's value out a bit? There is a further pair of Jackson variables as trimmers of the overall value of the two sections of this main tuning cap but I haven't touched these... Or is it meant to be calibrated with the dial offset a few degrees from everything against the end stop?? There is a letter C with a single dial marking which doesn't seem to have any relevance to any other scale markings etc. It's off the scale in fact. Does the C mean Calibrate here?? Ah the luxury of having a service manual

So... it's very very close to correct calibration on frequency and at a known and predictable output level, a little higher than the manual specifies but in many ways a better figure. Personally in this day and age I'll go for 10, 1, 0.1 over 7.75, 0.775, 0.0775V output all day long and I can set the level control to show the 7.75V calibrate mark on the level meter dead easy if I want to
It does show a slight but noticeable difference in level from one end of the dial to the other which I spent yet more time looking into before I decided to actually measure the error on a meter accurately calibrated in dB and found it was well within the specified +/- 1dB amplitude accuracy. About +/- 0.3dB in fact.

The only real "design error" (more a difficult to avoid problem at the time I think) I have found pertains to mains hum. Now it is in fact low by the standards of the day but it rears it's ugly head by the fact that if the main hum component is bucked out using the 50R wirewound variable across the heater supply (wiper to ground) it leaves a hum component which beats with the oscillators output when the unit is actually set to around the 50Hz region, hence there is a small but noticeable amplitude modulation going on between the two which gets to zero beat as the unit is set to precisely 50Hz. It can be "tuned out" by use of the hum bucking pot but the best setting for this is at the other end of the pots range from the one that gives minimum hum everywhere else! It is definitely coming from the AC heater supply on the oscillator section itself. I've checked every other possibility. There are very high impedances in play in the oscillator and in fact with the screening "can" removed it's basically unusable on the 25Hz to 250Hz range!

I'm pondering a smoothed DC heater supply to just the oscillator to cure this but in two minds due to originality considerations. I do like to actually use my vintage test gear in my work but as it's only really a problem between 40 and 60Hz....

Overall then it's a very well made piece of kit from 1952 that is completely free of the perennial issue of moving around the bench as you press buttons (), has very good frequency stability and adequate (<1dB) amplitude stability across the frequency range. It also has an attenuator which gives 0 - 60dB in 1dB steps with excellent accuracy and with the built in (and now accurate) level meter you can set 0dB anywhere between 1 and 10V RMS.
THD is a very good for 1952 0.2% (measured) and with the attenuator switched out you have a 1 - 10V RMS output with a source impedance of only 10R. Nice, if hernia inducing!

[trobbins]

It may be possible to link in a battery supply to the actual valve heater terminals, but leave the heater wiring in place and energised, to confirm that it is the heater wiring doing the coupling and not the valves per se.

If it is the valves doing the coupling then swapping valves may show some change, or perhaps easy to try an elevated tuned humdinger to see if the leakage resistance of the heater-cathode can be raised.

[Jez1234]

The optimum point of the hum bucker did change when I tried swapping valves around so indeed I had assumed at least some of it is coming from AC within the valves but I see no way of killing the hum beast completely dead other than giving the oscillator itself a smoothed DC supply. I've already simmed it and Schottky rectifiers have a low enough forward drop to get about 8V DC on a first smoothing cap of 10,000uF 16V and a further two of these capacitors with 0.68R between sections gets around 6.1V DC with only 5mV or so ripple.

Whilst I haven't done precise measurements on the hum level I have little reason to doubt that it's within original spec of <0.1% of max output but this is still 10mV of hum on the 10V RMS output, so 60dB down. Not awful but still somewhat offends my modern sensibilities on such things.

I find myself very torn between originality and improving performance when I'm working on vintage test equipment! If some important aspect of performance can be improved by a factor of 10 or even 100 by simply adding something like a 3 pin voltage regulator or by replacing an original 500uF 12V capacitor with a modern 10,000uF 16V that easily fits then do you or don't you? That's the $64,000 question!

My own rules on this are that anything done should be reversible and not involve any butchery such as drilling holes through a chassis etc.

I generally collect test gear from between around 1962-3 to 1980 as any earlier and the performance can be too poor for it to work for a living and even by 1980 the scourge of microprocessors, software and programability was setting in along with unobtainium ASIC's and EPROM's etc which obviously had to be programmed by the manufacturer. In the case of equipment built to full on laboratory grade with little expense spared, such as this Solartron and also the Airmec 701 I got with it, then I'm happy to go a little earlier

[trobbins]

DC elevation of the heater may be enough to noticeably increase the level of Rhk, and hence lower hum leakage into the cathodes. Elevation means at least 20Vdc difference from the cathode DC levels.

https://dalmura.com.au/static/Hum article.pdf

[Jez1234]

Thanks but there are two mechanisms causing hum and the fact that the input impedance of the oscillator section is several megohms doesn't help. Hence if I do anything about it then it will be the smoothed DC supply which will hopefully be a total cure.