Tektronix SG503 can't reach the low frequency end of some ranges

[Bill Woodbridge]

I’ve been fault-finding / tinkering with some TM500 calibration plug-ins for many months now prior to using them to calibrate my 2465B, prior to using that for some real work …

One problem that’s continuing to elude me is an SG503 unit that works beautifully apart from not quite being able to deliver the low end of the 5-10 and 10-25 MHz ranges when the variable frequency knob is turned towards its low limit.  Those ranges ‘cut off’ at 5.05 and 11.3 MHz respectively, at which points the oscillator output collapses to nothing, and of course the display starts blinking to indicate unlevelled.  All the other ranges, including the highest 100-250 MHz served by the Q140 oscillator, can go both higher and lower than their nominal ranges as suggested by Table 3-3 in the service manual.  It’s not a big problem of course as the ‘missing’ frequencies can almost all be generated instead by using the high end of the next range down, but it’s annoying and it’s out of spec.

So far I’ve tried:

1.   Swapping Q130 and Q140 (they’re both the same Motorola part in this unit) which gave a very slight improvement to the cutoff points.  I put this down to slight variation in components, or maybe a slight variation in their position with respect to the PCB when I relocated them.  The VHF (Q140) circuit continued to work perfectly with both transistors.

2.   Cleaning the variable capacitor, in fact a complete strip down, ultrasonic clean and rebuild.  No appreciable difference in performance.

3.   Cleaning all the switch contacts and buzzing out each individual oscillator configuration appropriate to each range.  No problems found, and no appreciable difference in performance.

4.   Tweaking the coil slugs for the affected ranges (I know I’ve got to revisit that in order to do the harmonic suppression routine eventually).  I saw some weird distortions, but it didn’t help the cutoff problems.

5.   Removing the levelling feedback transistor Q300 and temporarily replacing with a pot, as the service manual says.  All I saw was the oscillator output collapsing slightly more gradually as I lowered the frequency, since the levelling wasn’t there to try to rescue it.  The levelling itself seems to be working fine, but of course it can’t ‘level’ a completely dead oscillator circuit.

What’s really defeating me is that I can find no components common to the two affected ranges but not used in the other working ranges, apart from the fact that the affected ranges are the highest two before the rear (largest) section of the variable capacitor is switched out altogether.  But the rear section is working fine across its whole travel for all the other ranges.  Also I could understand there might be some weird HF gremlin coming to light as the rear VC is pushed towards the highest frequencies it must handle in each different oscillator configuration, but the problem is at the low end of the ranges!  It feels almost as if there is ‘suddenly’ too much capacitance being dialled-in by the VC non-linearly in this region which the lower range oscillator configurations can cope with, but the affected ranges can’t?  But how?

I thought I saw a reference in one of the forums (but can’t find it again now) to the fact that the ranges weren’t guaranteed to level outside their nominal values, and that the variable frequency knob’s travel might be capable of going past the point of unlevelling.  Which mine does, but it also fails to reach to the nominal lower values too, which is more of a problem.

Any hints or help gratefully received, as always.

Thanks, Bill

[CaptDon]

My SG503 was always flakey!!! The biggest difference for me to keep it oscillating was adjusting the symmetry pot if I remember correctly the name. It was a trade off between oscillation collapsing and poor waveshape. These units are really poor in their design. The fan in my RTM506 actually FM modulates the SG503 from vibration. The unit is so microphonic that you can get audio acoustic feedback when using it as a signal source for VHF commercial walkie-talkies when you turn up the receiver volume. They are as unstable as a cork in the ocean and vary like crazy when you change the output 'Amplitude Multiplier' or the Vernier. Sadly the SG503 is pretty much useless for a bench signal generator. I hope others chime in with a cure because I have been very disappointed with mine since the day I bought it. The FG504 is a beautiful feature packed unit, but also very unstable and best used below maybe 10mhz. I have considered selling my RTM506 and upgrading to a 5006 so I could also upgrade to some of the PLL synthesized generators. Best wishes, Cheers mate!!

[alm]

My SG503 was always flakey!!! The biggest difference for me to keep it oscillating was adjusting the symmetry pot if I remember correctly the name. It was a trade off between oscillation collapsing and poor waveshape. These units are really poor in their design. The fan in my RTM506 actually FM modulates the SG503 from vibration. The unit is so microphonic that you can get audio acoustic feedback when using it as a signal source for VHF commercial walkie-talkies when you turn up the receiver volume. They are as unstable as a cork in the ocean and vary like crazy when you change the output 'Amplitude Multiplier' or the Vernier.

What stability are you talking about? Frequency? Amplitude? Distortion? Mine has been stable in amplitude across amplitude multiplier settings, and apart from some dirty contacts in the range switch that require exercising once in a while, is as stable I expected from an oscilloscope vertical amplitude calibration generator.

Sadly the SG503 is pretty much useless for a bench signal generator. I hope others chime in with a cure because I have been very disappointed with mine since the day I bought it.

What were your expectations? It's not designed as an RF signal generator with good frequency stability and low harmonics like say an HP 8640A. It was designed to deliver a sineish signal of roughly the set frequency (note that the frequency indicator is 3 digits max) with a very stable amplitude in a good 50 Ohm load. Mine has done that and it seemed to match its amplitude specifications as far as I could tell. I don't think frequency stability matters much for its intended usage: the frequency roll-off of a scope or probe isn't going to be much difference between 99 MHz and 101 Mhz.

The FG504 is a beautiful feature packed unit, but also very unstable and best used below maybe 10mhz. I have considered selling my RTM506 and upgrading to a 5006 so I could also upgrade to some of the PLL synthesized generators. Best wishes, Cheers mate!!

Do you mean 10 mHz or 10 MHz? There's 9 orders of magnitude difference . Assuming 10 MHz, well, I don't expect a 40 MHz function generator to perform great at 40 MHz, especially for none-sinusoidal signals. What was the problem? Not meeting its rise time spec for square waves? Excessive harmonics (> -20 dB) in sine waves? Or are you expecting synthesizer-like frequency stability from an RC-oscillator based function generator?

[Jeff eelcr]

Bill I did a quick check on mine which is an early seriel number.
At freq set   Bottom     Top
   .25           .221         .544
   .5             .432         1.08
    1             .943         2.69
    2.5          2.14         5.38
    5             4.56         11.5
    10           8.79         26.2
    25          23.3          51.8
    50          40.0          105
  100          94.8          270
All settings are stable and values shown are close to actual freq.
Measured with Fluke 1911A and Tektronix 2815.
Jeff

[David Hess]

The SG503 is optimized for amplitude stability and nothing else.  Frequency stability is not required for oscilloscope calibration.

My guess is that at the low end of the 10 and 25 MHz ranges, the resonate Q drops too low for the amplifier to maintain oscillation.  Besides checking the passives around Q130, check its decoupling and look for cracked solder joints.  I would try outright replacing Q130 with a higher performance part.

Maybe Q300 is not able to supply enough current?  I am not real impressed with that part of the design.  What limits the current from Q300?

Check that the regulated voltage from Q160 is not sagging.

[Bill Woodbridge]

Thanks very much to everyone.  I do entirely take the point that the SG503 is really all about the levelling, and frequency precision / stability / low distortion isn’t the design aim.  Having said that, I do find my unit ‘stable’ in the sense that amplitude, frequency and distortion stay exactly where they’re put once they’re dialled-in, and the levelling is always rock-solid (apart from my fault).  And also that having a few black spots of missing frequencies isn’t a showstopper for the intended application.  It’s just … annoying.

Jeff – thanks for taking the trouble to compile that table.  It’s similar to my own measurements, except of course that you get good tail-ends at the bottom of 5-10 and 10-25 MHz as well as all the others.

David – I will get cracking on those suggestions, many thanks for being much more specific than the usual SG503 generic online chatter which revolves around Deoxit and recapping.  Q160 voltage looks strong and stable, so I’m now looking at your Q300 suggestion: from the circuit, it looks like R140 (75 ohms) is supposed to limit the DC current supplied by Q300, assuming its base is being driven negative enough by the feedback loop.  I’m very hazy about understanding and faultfinding at the transistor level, but I had a look and the maximum voltage across R140 (at the point of frequency cutoff, presumably when the levelling feedback is trying to sink as much current as possible from Q130) was around 1.1V, ie around 15mA current assuming R140 really is 75 ohms.  But rather promisingly, I also noticed that just the act of prodding the R140 pads with the voltmeter probes lowered the cutoff noticeably on both bad ranges.  This seems very suggestive of bad joints as you said, perhaps limiting the current sinking capability of Q300, so off comes the attenuator daughter board yet again for some proper inspection and rework around the whole Q300 / R140 current source, which I haven’t looked at in any previous investigations.  If I can’t find anything obvious, I might shunt R140 a little to try to get more current into the oscillator to see if I’m on the right lines.  Fingers crossed and I’ll report back.

Re-reading the service manual, I notice there are select on test resistors (R118 and R116) associated with the 1-2.5 and 0.5-10 Mhz ranges’ coils which must be selected for ‘oscillator stability’, but not for my two bad ranges (the next two up).  It seems that the designers were perhaps a bit close to the edge with all those many variations on the circuit topology which get switched in and out across the ranges!

Thanks again,
Bill

[CaptDon]

Granted Amplitude stability is superior and I guess that was the goal of the design. Yes, Mhz and not mHz obvious by the FG-504 being a 40Mhz unit. At the price point these things were sold for I expected much greater frequency stability. O.K. so they are a drift-o-matic, that is one part of the equation, but the shift with changing the Amplitude Multiplier and Vernier were not expected as one would assume several buffer stages between the oscillator and output circuit and that the leveling would be done beyond at least the first buffer so as not to upset the oscillator. I love the features of the FG-504, very handy for audio testing and even filter alignment of 262Khz, 455Khz and even 10.7Mhz I.F. strips. Sadly it isn't really suited for aligning the front ends of short wave radios since the drift of both the SG-503 and FG-504 is often as wide as the -6Db points of the I.F. strip. It is interesting when using the SG-503 with VHF radios I can simply 'talk' to the SG-503 from about 3 inches away and the voice comes clearly through on narrowband fm radios expecting 5Khz deviation as 100% audio output. "Can you hear me?" "Yes Mr. Edison, Loud and clear!" "I think you have invented frequency modulation".

[David Hess]

so I’m now looking at your Q300 suggestion: from the circuit, it looks like R140 (75 ohms) is supposed to limit the DC current supplied by Q300, assuming its base is being driven negative enough by the feedback loop.  I’m very hazy about understanding and faultfinding at the transistor level, but I had a look and the maximum voltage across R140 (at the point of frequency cutoff, presumably when the levelling feedback is trying to sink as much current as possible from Q130) was around 1.1V, ie around 15mA current assuming R140 really is 75 ohms.

My next suggestion was going to be to check if Q300 is saturating, but with only 1.1 volts across R140 that is not happening.

The voltage at the base of Q130 is suppose to be -19.4 volts.  The voltage at the emitter of Q300 is -22 volts.  So after subtracting the roughly 0.7 volt Vbe of Q300, that leaves 1.9 volts total between R140 and the collector voltage of Q300.  (1) You measured 1.1 volts across R140 so that leaves 0.8 volts at the collector of Q300.  That is sufficient for a transistor with good saturation characteristics but Q300 might have aged to the point where that is not enough. (2)

Now we know how Tektronix limited the current from Q300; its compliance voltage is just barely adequate.

Q300 is a Tektronix 151-0302-00 which is a 2N2222A in a real TO-18 metal can and everything like mother used to make.  It is specified to have a 100 millivolt saturation at 50 milliamps so a 2N3904 would not be a good replacement for it, but a 2N4401 would be.

(1) Good bipolar transistors can support a *lower* collector voltage than their base voltage.  I usually assume a minimum collector voltage of half of the minimum base voltage, or 0.3 volts, but in some circuits it can be lower, or even zero volts in translinear circuits.

(2) I would put Q300 on a curve tracer to see if it has become "weak"; I have found quite a few old transistors where they still worked, but had degraded saturation which shows up as a curved instead of straight line at low Vce on a curve tracer.  The difference can be 0.3 volts on a good part and 3 volts on a "weak" one which is a huge difference, but many circuits would still work.

But rather promisingly, I also noticed that just the act of prodding the R140 pads with the voltmeter probes lowered the cutoff noticeably on both bad ranges.  This seems very suggestive of bad joints as you said, perhaps limiting the current sinking capability of Q300, so off comes the attenuator daughter board yet again for some proper inspection and rework around the whole Q300 / R140 current source, which I haven’t looked at in any previous investigations.  If I can’t find anything obvious, I might shunt R140 a little to try to get more current into the oscillator to see if I’m on the right lines.  Fingers crossed and I’ll report back.

I think shunting R140 is a good idea.  I would not go for half measures; drop it to 66% of its design value, which would match R150 on Q140, if not 50%.

Re-reading the service manual, I notice there are select on test resistors (R118 and R116) associated with the 1-2.5 and 0.5-10 Mhz ranges’ coils which must be selected for ‘oscillator stability’, but not for my two bad ranges (the next two up).  It seems that the designers were perhaps a bit close to the edge with all those many variations on the circuit topology which get switched in and out across the ranges!

I think those lower the Q of those coils to suppress spurious modes.  Wide range LC oscillators are tricky.

[David Hess]

Granted Amplitude stability is superior and I guess that was the goal of the design. Yes, Mhz and not mHz obvious by the FG-504 being a 40Mhz unit. At the price point these things were sold for I expected much greater frequency stability. O.K. so they are a drift-o-matic, that is one part of the equation, but the shift with changing the Amplitude Multiplier and Vernier were not expected as one would assume several buffer stages between the oscillator and output circuit and that the leveling would be done beyond at least the first buffer so as not to upset the oscillator.

I have used my SG503 as an RF source and after warming up, drift was in the 100Hz range, but adjustment is tricky and changing the output amplitude pulled the frequency.

I love the features of the FG-504, very handy for audio testing and even filter alignment of 262Khz, 455Khz and even 10.7Mhz I.F. strips. Sadly it isn't really suited for aligning the front ends of short wave radios since the drift of both the SG-503 and FG-504 is often as wide as the -6Db points of the I.F. strip.

I think Tektronix published examples of frequency locking the FG-504 and similar function generators with a VCO input to make a synthesized source.  HP made synthesized function generators which did exactly this internally.

[TERRA Operative]

I'm currently replacing an intermittent crystal with a Vectron OCXO in my unit (just for the fun and better accuracy), so once I have it done, I can provide some measurements for reference if needed.

[jonpaul]

Clean and exercise all range sw contacts

Beware method/type cleaner  in range sw


Check serv manual on BAMA or TEK WiKi

Follow adjust/CAL proc

[Bill Woodbridge]

David - thanks again, that all makes sense (although did you mean ’the roughly 0.7 volt Vbe of Q130’ rather than Q300?).  I haven’t got a curve tracer, but the first thing I did was swap out Q300 with one from another SG503 awaiting investigation (which has a problem both with power supplies and another somewhere inside that TTL bird’s nest of counter logic - ugh).  The change of transistor made no difference at all to the cutoff frequency.

I remade the joints around R140, both of which looked poor, but again sadly no difference to cutoff frequency.  And everything tested OK with the correct DC resistances when I probed through.

Then I tried shunting R140 with a parallel 112 ohm resistor, thus bringing the total down to 45 ohms.  Very slight difference to cutoff frequency.  Here are my measured DC voltages in millivolts across R140 with and without the shunt (I’m purely concentrating on the low end of the 10 – 25 MHz range now).  The voltages don’t rise any more beyond these maximum values as the frequency control is turned down further.

Frequency    V (R140, 75 ohm)      V (R140, shunted to 45 ohm)

12.0               232               377
11.5               286               448
11.0               410               640
10.9               429               682
10.8               469               779
10.7               529               943
10.6               660             1201
10.5               775             1289 (blinking, unlevelled)
10.4                  1248 (blinking, unlevelled)

With that sudden ‘rush’ to the maximum voltage as the cutoff frequency is reached, I am coming back around to thinking that Q300 is performing correctly and doing its best, but it can’t do any more to rescue the collapsing oscillator circuit once it gets to around 1.9-1.248 = 0.65V Vce.  In other words, Q300 bottoming-out is a symptom of the oscillator collapse, not the cause.

TERRA Operative - many thanks for the offer, although at the moment I can't think of what else to measure in a good reference unit that would give me an insight into my faulty unit.

jonpaul - thanks, I've cleaned the range wafer switches and the big S100 contacts many times, and I'm pretty convinced they are both OK - everything is completely consistent and stable as I twirl the range knob around.  I've got the service manual.  But your comment on the adjust / CAL procedure led me to fiddle again with the coil slug for this range (L112), partly in desperation.  I had done this before, several months ago, and seen such crazy waveforms at the oscillator output that I got frightened and thought it best to leave well alone.  But now I'm looking at the final output into 50 ohms with the famous 'precision' cable, (which I wasn't doing before),  I can see a respectable looking sine wave even after I managed to move the slug to keep the oscillator alive at 9.90 MHz!  I should have realised that all that heavy post-oscillator filtering is there for a reason.

But now of course I need to break out the spectrum analyser (the best I've got is FFT on a 200MHz Keysight DSOX, which I hope will be enough for this frequency range at least) and calibrate properly for harmonic distortion.  I just hope it's not going to be too much of a trade-off between oscillator bandwidth and harmonic distortion, as per a previous comment.

[David Hess]

David - thanks again, that all makes sense (although did you mean ’the roughly 0.7 volt Vbe of Q130’ rather than Q300?).

Yes, sorry, that should be Q130.

I haven’t got a curve tracer, but the first thing I did was swap out Q300 with one from another SG503 awaiting investigation (which has a problem both with power supplies and another somewhere inside that TTL bird’s nest of counter logic - ugh).  The change of transistor made no difference at all to the cutoff frequency.

It would not surprise me of all similar transistors of that age have degraded.  When I rebuilt my DC505, half of the many 2N3565 transistors were weak displaying high saturation voltages.

Then I tried shunting R140 with a parallel 112 ohm resistor, thus bringing the total down to 45 ohms.  Very slight difference to cutoff frequency.  Here are my measured DC voltages in millivolts across R140 with and without the shunt (I’m purely concentrating on the low end of the 10 – 25 MHz range now).  The voltages don’t rise any more beyond these maximum values as the frequency control is turned down further.

Frequency    V (R140, 75 ohm)      V (R140, shunted to 45 ohm)

12.0               232               377
11.5               286               448
11.0               410               640
10.9               429               682
10.8               469               779
10.7               529               943
10.6               660             1201
10.5               775             1289 (blinking, unlevelled)
10.4                  1248 (blinking, unlevelled)

So lowering the value of R140 increased the voltage across it?

With that sudden ‘rush’ to the maximum voltage as the cutoff frequency is reached, I am coming back around to thinking that Q300 is performing correctly and doing its best, but it can’t do any more to rescue the collapsing oscillator circuit once it gets to around 1.9-1.248 = 0.65V Vce.  In other words, Q300 bottoming-out is a symptom of the oscillator collapse, not the cause.

Q300 should always have been bottoming out when the RF output cannot be maintained.  The question was whether Q300 had good enough saturation to actually bottom out at a reasonable value.  Apparently that is the case and Q300 is not the problem.

[Bill Woodbridge]

Yes, shunting/lowering R140 resulted in larger voltage drops across it for the same oscillator frequency.  This is where my lifelong poor grasp of transistor circuit design confuses me yet again – assuming Q130 always requires the same emitter current to deliver the required levelled amplitude at any particular frequency, how can that same current through a smaller resistance drop more voltage rather than less?

And on your last point – apologies, point taken.  I temporarily lost the plot that establishing this was the whole purpose of the investigation.  Thanks again.

[Bill Woodbridge]

Just to report that I think the repair journey is now successfully complete, and to thank everyone again for their advice.

I think in the end the breakthrough was a combination of swapping both Q130 and Q300 to slightly better-performing examples, but mostly no longer being afraid to tweak the relevant coil slugs.  By doing this, I was finally able to get both the 5-10 and 10-25 MHz ranges to level beyond their nominal end-points.  The 10-25 in particular didn’t have much headroom in what it would achieve at either end (9.82 to 25.1 MHz) but as long as it can span the nominal values it’s within spec, I guess.  And, to my surprise, the unit easily achieved the harmonic suppression specification right from 50kHz to 250MHz (I enjoyed exploring my DSOX3014’s FFT capabilities, which worked very well) and also met the 1% / 3% levelling specification.

The only casualty was one of the plastic side cover clips, which eventually shattered after 50 years of use.  They seem to be unobtanium, so the plan is now Araldite and a made-up metal pin to hold the remains together.

Onto the next SG503, with the counter problem … thanks again.

[alm]

Congrats on the repair! Problems like this, where nothing is really broken, but just not performing quite as well as they should, or needing some tuning, can be quite tough.

[Jeff eelcr]

Bill If those are the plastic locking tab in the rear center of the side case?
I did have 2 of those that broke the only unit I have that uses them so how
to fix them?
I used a small drill bit and put a center hole thru both halves then taped a
2-56 thread. Then screwed them together. The fine thread allows some type
of taughtness adjustment. A nylon screw can be used.
Jeff 

[Bill Woodbridge]

Yes, that's the one.  Thanks for the tip - I'd thought of a screw, but not a nylon one which sounds promising.  Just need to find a miniature size.

[frantisek]

I'd return the Q130 and 300! They certainly didn't cause this problem!! MM I have SG 503 two and if there was ever a problem with not covering the bottom of the range, always! a fine tuning of the appropriate coil solved it. Tried and tested. I had a worse problem in the 0.5 to 1 MHz range, when the positive and negative peaks of the sine wave were sort of wobbly, or AM modulated. But even that could be solved elegantly.
By replacing the Q130 and Q 300, all ranges have been tuned! The transistors are common for all ranges. If their replacement brought an improvement, it was purely accidental... There is only one way to fix it, and that is to tune the respective coil.