Tektronix 7A26 problems

[ColinTheAmpMan1]

Hi,

I have posted this on the Tekscopes forum with unfortunately insufficient help and advice to enable a fix. This surprises me somewhat, as the 7A26 is believed to have been the most successful dual trace amplifier that Tektronix produced for the 7000-series (120,00 units, according to Tekwiki) and so I would have expected this problem to have cropped up before now.

The gist of it is that I have one of my two 7A26s that has gain problems on both channels. CH2 can just about be adjusted to give calibrated gain, but with only a little headroom. On the other hand, CH1 can only be adjusted to give a little less than the calibrated gain. The Tek custom ICs (155-0078-10) all seem to be OK; swapping the first ICs between the two channels shows no difference and putting the IC from the poor CH1 into the "good" 7A26 doesn't mess anything up. So, I'm fairly sure that the problem is not with either of the first ICs, but with the circuitry.

I have checked the voltages on the ICs in the "good" 7A26 and the faulty one and there are significant differences. I am not at all familiar with these differential amplifier circuits and am thus rather baffled . Has anyone experience of such a problem, or can anyone offer some advice to help me fix this?

It bothers me that the gain of both channels is low, but not to the same extent, or is this a red herring? Clutching at straws, should I replace the tantalum caps that I can see in the faulty 7A26, but not in the good one?

TIA, Colin.

[MrBungle]

Interesting. I'd start shove 20mV p-p in with all attenuators switched out, then start signal tracing and comparing to the good plugin. If it's misbehaving on both channels it may be part of or in the channel switching side of things. There's steps for that in the service manual for the plugin with expected waveforms I think.

That is incredibly difficult if you don't have two extenders though (one reason I don't have a 7000 any more!)

The actual ASICs they use for the diff amps are pretty simple. Unfortunately the application in the faster plugins is quite complicated. Both the current source and the load resistors have complex frequency and thermal compensation. If that's gone wonky I imagine it would be hell to fix. I never tried. Last thing I fixed was a 7A18 and that was considerably easier and was a cracked resistor!

[Refugee]

I had two of those with bandwidth limiter slide switches that were dirty and intermittent.
You should see all the tool marks on those diff amp chips where the twiddler had been there before me.
All the chips survived despite the scars.
Quite a bit of work can be done on them with an external dual rail 15 volt power supply.

[dave cox]

I had one example where the 'bias' resistors had gone high, IIRC I had 4 identical ones in one unit effected but my other 7A26's were fine with a different brand.

dc

EDIT
I should say it was the same symptom, not enough gain.
I can take a quick look and see which ones I replaced when I get home.

[ColinTheAmpMan1]

Originally Posted by MrBungle

That is incredibly difficult if you don't have two extenders though (one reason I don't have a 7000 any more!)

I have a total of four extenders - two Tek flexible, one Tek rigid and a John Griessen flexible kit waiting to be completed.

Colin.

[MrBungle]

Perfect!

[dave cox]

These resistors, you might need to remove the IC to check them in place ...

dc

[Craig Sawyers]

I've had problems with poor contact of the dual FET associated with the input circuit in its socket. That gives strange gain anomalies.

Craig

[ColinTheAmpMan1]

The biasing resistors check out to be within spec and I have tried wiggling the dual fets - all to no avail.

However, I have checked the voltages on the ICs and there's certainly something amiss. I did this with no input signal and the 7A26 settings as in the Service Manual. While the first IC (the 2X gain amplifier U1350) in the faulty CH1 seems to be multiplying by two, a number of the DC voltages are out of spec., which I think may be putting subsequent stages out of kilter. As an instance, the voltage on pins 4 & 10 (substrate) should be about -7.5 VDC, they are about -4.7 VDC. At the outputs, pins 5 & 9 should be +3.2 VDC and are +0.9 VDC; pins 6 & 8 should be +4.4 VDC and are +2.07 VDC. I can see that the voltages on the next stage's input pins are consequently wrong, too. Perhaps I should next lift ends of R1327, R1329 and C1327, which are on the substrate and check if they are within spec. I have checked those resistors in circuit without any sensible results. I should probably also do what Mr Bungle suggested and do a bit of signal tracing, but I would like to know why there is this discrepancy with the substrate voltage on U1350.

I will be a happy bunny when (if?) I can get this plug-in working correctly and calibrated.

Colin.

[RogerEvans]

One of the difficulies with DC coupled amplifier chains is finding out whether any discrepancies are due to a preceding or following stage. Did you check voltages on pins 2 and 3 of U1550 on the input side? On the output side did you swap U1450 or U1550 with their partners in channel 2? Since everyone is supicious of resin dipped tantalums you could check the voltages around Q150A&B where there are C140 and C158 decoupling the +/- 15V rails. On my 7A26 Q150 appears to be soldered but as Craig says, if yours is socketed then check that poor contacts are not upsetting the DC operating point or gain.

You said at one point that the bandwidth switch moved the trace around, that is a little surprisng since they are after Ch1 and Ch2 are combined so a fault there should affect both channels. Q820 840, 860, 880 are usually socketed so easy to check.

Hope your persistence is rewarded,

Roger

[dave cox]

My analysis is the OC outputs (9,6,8,5) are either :-
(1) delivering too much current, pulling the voltage low. Controlled by R1320 & R1323 (681), R1339 & R1338 (36)
(2) their load is too high a resistance, R1347 (280R), R1343 & R1341 (56), R1403 & R1401 (60R4)

Gilbert cell 155-0078-00

dc

[MrBungle]

I may have a spare one of those ICs floating around if it comes to that... let me know and I’ll get digging.

[Craig Sawyers]

For what it's worth Barrie Gilbert is still alive and kicking. He was also responsible for the readout system on 7000 series scopes.

Another anecdote. He set up a bet with Widlar in the late 60's to design a minimum count current mirror. The best that Gilbert did was a three transistor mirror, and Widlar came up with the (now standard) two transistor mirror and won the bet.

Craig

[Argus25]

Originally Posted by dave cox

My analysis is the OC outputs (9,6,8,5) are either :-
(1) delivering too much current, pulling the voltage low. Controlled by R1320 & R1323 (681), R1339 & R1338 (36)
(2) their load is too high a resistance, R1347 (280R), R1343 & R1341 (56), R1403 & R1401 (60R4)

Gilbert cell 155-0078-00

dc

It should really be named a Jones Cell, Jones invented it in the early 1960's and did not patent his work. The Nagra tape recorder company also used the 4 quadrant multiplier as a phase detector in their machines to crystal lock the tape speed, prior to the Gilbert patent too. It makes for a great scope input amplifier as a monolithic IC (temp drifts cancel) and it is extremely easy to invert a channel by altering a reference voltage on one of its ports.

[Craig Sawyers]

This is a lecture by Gilbert in 1991 describing the translinear principle that is the core of the 7A26 (and 7A24) amplifier chips.

https://www.youtube.com/watch?v=LQNJVtcFrCc

Craig

[Argus25]

Here are some useful remarks on the 4 quadrant multiplier explaining how it works and correctly gives the credit for its invention to Jones:

https://ipfs.io/ipfs/QmXoypizjW3WknF...bert_cell.html

[dave cox]

You'll struggle to google it using 'Jones Cell" !

But yes, its wrongly attributed to the patent holder.
Does 'prior art' not invalidate the patent ?

dc

[Craig Sawyers]

All you have to do is identify an "inventive step". The core Gilbert patent was originally filed in 1968, and was multiply abandoned and material brought into later filings.

The final patent was filed in 1976 and granted in 1978. None of the cited references include work by Jones.

But there is a whole lot of circuit topologies included in US patent 4,075,574. And there are multiple correction sheets at the end - so even that very clever Gilbert had a few issued getting his head around his own circuit!

Craig

[ColinTheAmpMan1]

To respond to some of these observations:

Roger - by U1550 do you mean U1350? If so the DC voltages on pins 2 & 3 are both -0.8, which is correct, not the +0.8 that is shown on the CH1 amplifier circuit. I swapped U1350 in CH1 with U2350 in CH2 with no difference. I also tried the U1350 in another 7A26's channel 1 with no problem moving with that IC. I haven't tried swapping U1450 and/or U1550 with their counterparts in CH2, but I will do so. Perhaps I ought to check the voltages around Q150A/B too and compare them with the voltages on my good 7A26. Originally, I though that the problem might lie in the Switching and Output section and I re-seated Q820, Q840, Q860 and Q880 with no change. I did note that the DC voltages at the output terminals were +0.04 (R890/R846 junction) and +0.02 (R892/R846) rather than 0V. I lifted one end of some of the resistors to check and found that R890, R892 (each 1k0) and R846 (120R) were within spec. Resoldering R890 and R892 at their -15V end improved the voltage balance somewhat.

Dave - Regarding the fact that pins 2 & 3 on U1350 are correct, I think that R1320, R1323, R1339 and R1338 are probably OK, but I haven't checked. Likewise, I haven't checked R1347, R1343, R1341, R1403 and R1401. Perhaps I should. However, the suspect DC voltages on the second IC in CH1 (U1450) are pin 1 0.9V (properly 3.2V), pins 2&3 both 0.18V, pins 5,6,8 & 9 7.7V (6.9V), pins 13,14 & 16 0.9V. Maybe this throws some light on the circuitry between U1350 and U1450.

MrBungle - Thanks for the offer, but I purchased four (tested) ICs from a member of this forum who also trades on eBay.

I may go a little quiet next week (5th - 10th August) as I am volunteering at the Great British Beer Festival in Olympia. Why not come along?

Thanks to all, Colin.

[dave cox]

Errrr, no!

The voltage at the emitters (pin 2,3) is determined by the voltage at 13/16 for pin 2 and 14/1 for pin 3. So the emitter current is then defined by the tail resistors. This current then ends up passing through the 4 paths 56R (*2)+ 280R and 60R4 (*2) + 280R. The dc voltages on pin 11 & 12 determine how the split occurs (with the input signal, if it is present).

For the 1st stage (X2) pins 11/12 switch the gain between X1 and X2 with some trimming to get exactly double the gain in the X2 setting. For the 2nd stage pins 11/12 do the gain trim and the VAR gain control. For the 3rd stage its just used for ON/OFF. The other 3rd stage drives the timebase trigger, again with ON/OFF switching. There is probably some voltage gain in there as well

Your 2nd stage is PROBABLY getting starved of emitter current since the 1st outputs are too low voltage (as you measured). Now, no matter how hard it tries it can't steer enough current into the output to generate enough (current) gain.

Even with a signal present there is not much to see in the voltage domain, that's what makes this amplifier so fast since it is not trying to charge any capasitors up!

dc