Signal Generator Output Stage ... Oscillates, poor high Freq

[KD4CCF]

I have a sweep-audio signal generator from 1973.

It looks like the output state is oscillating, and bandwidth goes south about 6kHz, in that the wave form that should be a sine wave becomes a funky ramp.

Changed all the electrolytics, the 8-pin socket that was suspect and the LM301 op amp. Checked the other components as best I could in-circuit. Even changed the final transistors.

The thing is 46+/- years old so shotgunning it is risky to the circuit traces that want to lift just looking at them.

I don't know if there's a design flaw? Really, the thing shouldn't oscillate, max gain should be about 3 with values shown.

The out stage is page 5 of the article.

[KD4CCF]

PDF didn't attach... here it is

[Chris56000]

Hi!

First of all, switch S5 to the "Sinewave" position and verify you've got a clean sinusoidal waveform coming out of IC7 pin 2.

If there's any nasties or distortion on pin 2 of IC7 ('8038), try another device (there seems to be plenty of 'em in China!), them carefully re–check the "Sine Wave Purity" (the "Sym" controls R38 and R39 in Fig. 5), and the settings of the Duty Cycle preset R41 – the adjustment of R41 is obvious, but adjusting R38 and R39 needs a T.H.D. meter or spectrum analyser for the most accurate results unfortunately, however the original adjustment instructions provided should be OK at a pinch.

There's a slight possibility Q5 (MPF111) and the circuitry further back could be causing problems with Pin 10 on IC4, but lifting the drain lead connected to Pin 10 should help prove this!

If the waveform at IC7 pin 2 is satisfactory, check around the various small compensation and shaping capacitors associated with IC9, these are C21, C22 and C25. If you can find no fault with these components, make a precautionary check to ensure you've not got any revolting sawtooth spurii in the supply–rails due to yakky rectifier diodes or smoothing capacitors in the power supply – if you made the generator on the pcb supplied from the kit mentioned in the article, the psu rectifiers were only half–wave!

A defect in IC4 (741) or the preceding control/sweep circuitry could possibly be causing this by resetting IC7 at random intervals, the easiest way to prove this is to lift pin 6 (O/P) of IC4  if you can (did the kit use chip–holders?) – this now allows the '8038 to deliver a constant frequency sine–wave you can use to more easily check the output stage components in Fig. 6 again!

If you're certain the fault is definitely associated with the output stage, Fig. 6, it's best go go thro' the various transistors and diodes, paying particular attention to the series biassing pair D11 and D12.

The transistors specified aren't typical "run–of–the–mill" types usually assosciated with hobbyist–type American circuits, but if you need to replace the output pair, a BC639 for Q11 and a BC640 for Q12 should perform excellently!

The BC639/BC640 transistors have the layout b, c, e, with the leads pointing towards you, base is on the left, collector in the middle and emitter on the right – this lead layout may suit your PCB better than the e, b, c layout used by the BC327/BC337 type of small power transistor.

(I never got the original component layout & pcb artwork for this, did you get a copy from the original publishers? I wouldn't mind a copy if you can scan it!)

Keep us posted, if you're still stuck I'll help you go thro' the sweep control stages!

Chris Williams

[David Hess]

One of the output transistors could have failed with age.  I would replace them with a 2N4401/2N4403 or BC327/BC337 pair but the BC639/BC640 pair Chris56000 suggests is also good.

Take a close look at D9 and D10 or just remove them for testing.  I am not sure why they were even included unless the design was originally intended for the LM709 operational amplifier which cannot tolerate a large differential input voltage without bad things happening.

[KD4CCF]

The top (clean) trace is the wiper of R60 page 5. Clean up to 100kHz at the wiper.

I removed D9 and D10 and makes no difference . Not sure why they're there since IC9 pin 2 is pretty much a virtual ground. I replaced Q11 and Q12 with 2SD438 and 2SB560 complimentary pair.

I didn't get it way back in 1973. $150 was beyond my budget then. It's a new addition to my workbench. So no, I never got the pcb or component layout. In fact this may have been a factory built unit from MITS since the case is different and has their name on it. There was one that sold that had the manual, but I missed getting that one.

Nice thing is it has both sweep-ramp and (blanking) gate which is missing from most the newer equipment.

Once in a while they show up on ebay (where I got mine) MITS SG-1900. I did see a Popular Electronics one that I had missed.

They go for $100+/- shipped.

I guess I'll have to start changing out passives and the bias diodes. Something is funky. or maybe I'll breadboard a duplicate and see what that does before risking the gossamer pcb-board foil.

[magic]

That thing isn't winning THD contests

More on topic, at 100kHz the oscillation is rather slow in the grand scheme of things, so it seems like a fault of that old opamp. Probe simultaneously the junctions of D11/D12 and R67/R68, I bet you will see that the output stage follows drive signal perfectly and it's the chip that is feeding it garbage.

Check if R63 isn't disconnected.

The opamp looks horribly slew limited. If replacement didn't help, maybe something is not right with the compensation components. I have no idea what LM301 is and how to compensate it, if it's socketed I would throw it to trash and try TL071 instead, with pin1 bent out of the socket.

Try 100nF supply rail caps

[Chris56000]

Hi!

It was an constructional article in another mag I saw that recommended a T.H.D. analyzer for adjusting '8038 Symmetry Presets!

I'll have a go at building this ouput–amplifier on vero, etc., and test it myself – I'm definitely curious to know if MITS made an immortal howler in designing it around the use of a '301!

I wouldn't have thought that the pcb tracking of the original design (which I doubt is available now!) would have been so bad as to cause such appallingly bad slew–rate limiting if this particular circuit–design and pcb went into factory production, but a suspect, defective (or even fake!) LM301 is a distinct possibility!

Chris Williams

[David Hess]

The opamp looks horribly slew limited. If replacement didn't help, maybe something is not right with the compensation components. I have no idea what LM301 is and how to compensate it, if it's socketed I would throw it to trash and try TL071 instead, with pin1 bent out of the socket.

Capacitor C21 implements feedforward compensation on the LM301 so a TL071 is actually slower.  Slew rates are comparable but the LM301 has at least 3 times the gain-bandwidth product when configured this way.

C22 and C25 make up for phase lag in the output transistors so a TL071 might require external compensation between its output and inverting input further slowing it down.

I would have capacitively bypassed R64, D11, and D12 but maybe this would have caused stability issues.

[Chris56000]

Hi!

Looking at the LM301 data-sheet & Fig. 6, the voltage-gain of the output amplifier is set by R63/R61, which is 3.9 times or 11dB approx, (the o/p buffer Q11/Q12 should be transparent to the signal as regards feedback!) so interpolating from Fig. 21 on the datasheet, this amplifier, which is "feed-forward compensated" as Member David Hess suggests, should have a satisfactory response to at least 300-350kHz, more than adequate for the maximum frequency design specs given, and according to the "Project Evaluation" box on page 5 of the article, a unit built as specified appears to be capable of an O/P flat to within 0.25dB op to the 100kHz design. max in sweep mode, so there's obviously a component fault somewhere!

Chris Williams

[duak]

I'd bet that one of the passives, traces or solder joints has gone bad,  Probably more likely on the pull up side.

Looking at the parts list I see that the small value caps are disk ceramics.  If they were monolithic ceramic caps, like the little blue or light brown Kemets, I'd immediately suspect one of them to have a broken connection inside the conformal coating, especially if it were installed too close to the PCB without strain relief.

The LM318 opamp has the feedforward network built in.  It is compensated for unity gain so I'd suspect the LM301 might be a bit faster if it's compensated for a gain of -3.9X.  If you want this to be period correct and if it has better performance, it might be an alternative.  Of course, there are a number of other choices these days.

I remember this article.  Back then, I thought it was an enormously complicated design but I quickly understood how it worked.  After attending college and getting more experience I saw it wasn't at all complex.  While not necessarily a marginal design, it was minimalist and parts efficient.  I've repaired a few Tek, Wavetek and Philips function generators that used far more components.  While they were faster, they also had circuitry to better protect the driver devices.  Even so, the instruments I repaired all had bad output transistors and/or resistors - most likely from having their outputs connected to a power supply by mistake.

[magic]

Now that I looked at it again, slew rate is a pathetic 0.04V/µs upwards, but actually somewhat reasonable downwards. So maybe it is the output stage after all, and the upper half of it in particular. The opamp tries to slew "up", saturates, the output stage slowly catches up, the opamp suddenly recovers and overshoots downwards, repeat.

Since you already replaced Q11, measure R65 and R67 at least in circuit, they certainly mustn't be higher than nominal. Power it up and check if Q11 base tracks the output signal as it should.

Make sure the sawtooth doesn't come from 15V rails or all else is moot

[KD4CCF]

I pulled all the related caps and resistors after this photo. All within tolerances.  Continuity checked socket to related component pads.

The 5pf (C25) is too low for my tester, so I'm going to assume it's ok, but just to be safe, I'm replacing it.

The "mini sawtooth" riding on the big sawtooth is about 50kHz and appears to be an oscillation.

+ and - rails have a few mV of 6oHz ripple as read at the IC socket, but clean otherwise.

All ICs are/were original. Shown is board with new output transistors, new socket w/o LM301 removed . Diodes D9,D10 I clipped to no avail.

Rhetorical question... why is it when you want axial caps you always have radials, and vice-versa?

[magic]

Maybe the output is overloaded by some screwed up resistor somewhere. Either way, I would just probe around the output stage to see what's going on.

Stupid question: did this unit ever work? Because C22 is even smaller than C25 and should be out of your meter's range too, maybe somebody put 3.3nF instead of 3.3pF? I mean, I'm just looking at the schematic of 301 and trying to figure out how the hell could that thing behave the way it does. Also, make sure no caps are shorted or leaky.