Agilent 33120A negative half of output signal clips and looks unhealthy

[FuzzyOtter]

Hi all,

I am working on this 33120A and wanted to get some thoughts on what might be wrong here. Here is a video of the initial clipping behavior:



I opened the device up and had a brief look around. The 33120A output amplifier has a maximum full scale output of +/-10v. When I initially received the device, I put my scope on the amplifier output (between output transistors Q713 and Q714, before the attenuator) and walked the amplitude up, I saw the signal begin clipping at -4.5v. The top half of the waveform looked fine. It does not matter which waveform I select (sine/square/ramp/arb), the bottom part of the signal would always clip at -4.5v. The attenuator works fine and I can cruise through the attenuator steps without any problems.

I did an initial visual inspection and found a burned resistor, R771. This resistor is specified as 26.1 ohms, but I measured the charred resistor as 172 ohms. I measured all of the transistors in the amplifier circuit with my meter in diode mode. They all tested OK (except for an oddity with Q718 which I will get to later) and not seeing an immediate reason for the resistor to burn, I replaced it.

Upon replacement, the clipping still occurs, but it has shifted from occurring at -4.5v to -7.0v. Now that more of the signal is back, it is also more obvious that the entire bottom half of the waveform is distorted and generally looks sloppy, with some minor glitchiness around the zero crossing. The distortion becomes worse as the frequency is increased.

I continued poking around and took note of the following:
-The -18v linear regulator U1001 gets very hot. I cannot keep my finger on it's heatsink. Measured at 65C with my thermocouple.
-The three resistors R701, R702, and R703, all measure way below their specified values. They do not look burnt or damaged in any way.
-I measured across every other resistor in the amplifier circuit and they all measured exactly to spec.
-Measuring across the inputs to U702, I see a 22k resistance. I'm not sure if this is expected.
-I can't measure a voltage drop between the emitter-base or collector-base junctions on PNP output transistor Q718. My meter will only do 1v in diode mode, and all other transistors in the circuit had measurable 0.6v to 0.7v drops. Obviously it's doing *something* because the negative portion of the signal is still "there"... does this line up with a possible failure mode for a transistor like this?

Any advice here? Is this just an immediate "replace Q718" problem? (which seems like sort of a pain as the 2N5583 doesn't look easy to get). It also leaves me unable to explain why resistor R771 was burnt. I've attached a marked up of amplifier schematic, and a sample waveform of the output in comparison to what is driving Q709/Q710, where you can see some distortion in the signal when the output begins to misbehave. I'm at work now, but I can capture better waveforms later this evening if they would be helpful.

[capt bullshot]

For testing purposes, you could replace Q718 with some other general purpose PNP (similar Uce / Ic / P rating), if Q718 is the culprit you'd see the ouput signal restored at lower frequencies, but maybe distorted at high frequencies (if your replacement transistor isn't fast enough).

[FuzzyOtter]

That sounds like a reasonable idea. I have nothing suitable on hand, but I did some sleuthing on Digikey for a readily available device I could pop in for testing. The selection of PNP BJT devices with even remotely similar specifications is super slim... did this technology suddenly vanish with the advent of FETs? I think the closest I can get is the 2N4033: https://www.digikey.ca/product-detail/en/central-semiconductor-corp/2N4033/2N4033CS-ND/4806882 which matches up very closely spec-wise, except for the lower fT and higher DC current gain.

The only 2N5583's I could find are on eBay, and are either "unmarked" devices from the USA, or marked but questionable parts from China.

[pigrew]

A while ago I saw a similar issue which was due to cold-solder joints on the output transistors. I suggest you try reflowing them. on the other hand, my unit didn't have the other circuit damage you reported, so you likely are suffering from a different issue.

[FuzzyOtter]

The 33120A signal generator has been repaired. I wanted to give a debrief for anyone following in my footsteps.

SYMPTOM:
The signal generator's negative half of it's output signal clips as you increase the amplitude of the signal. The clipping is independent of frequency and applies to all waveforms (sine, square, ramp, arb). The clipping "resets" as you increase the amplitude and the attenuator switches steps, and then appears again.

VISUAL INSPECTION:
Overall the mainboard looked fine with no immediate damage apparent. Close inspection of resistor R771 revealed that it was burned so badly that the resistor had turned from it's normal blue colour to flat black, but it did not scorch the board or surrounding components, so it almost looked normal. Resistance should be 26.1 0.1%, but the resistor measured 172 ohms.



ELECTRICAL INSPECTION:
I suspected that a transistor in the amplifier stage was the cause of the fault, so I measured each transistor in situ to check for obvious failures. The meter I used for this was my GW Instek GDM-8251A. Here are my measurements:





Obviously something looks fishy with Q718 not providing a measurement similar to the others. I had expected to see a faulty transistor fail short instead of open.

Q718 is a Motorola 2N5583 RF PNP BJT. The key specification for this part is it's fast delay/rise/fall times, and it's transition frequency spec of 1.3GHz. Tracking one down seems to be hit or miss depending on the day. Before putting effort into sourcing one, I decided to replace the transistor with another "close enough" transistor from a readily available source, like Digikey. In this case, I picked a 2N4033 based on having close current gain, and being decently fast. I also picked up a replacement resistor for R711 (part number RQ73C2A26R1BTD).

I replaced R711 and sucked out Q718, temporarily replacing it with the 2N4033. Thankfully, testing the generator showed that the output signal has been restored. I cycled the amplitude up and down it's entire range and the clipping was eliminated. The transistor performed fairly well up until around the 10MHz mark, after which it began to rapidly distort and heavily attenuate the signal.

I located a replacement 2N5583 from an eBay seller in the UK selling various NOS parts. There appear to be many "unmarked" or "Motorola" (in suspicious quotations) from eBay sellers in China for fairly cheap, but they seem super sketchy, so I decided not to risk it. When I received the part, I popped it in and the generator seems to work properly again. Here is what the sinewave output looks like at 2Vpp into a 50 ohm termination at 1MHz and 15MHz (signal on top, sync output on bottom).





The only outstanding problem, if it even is one (as I can't find captures of what a known good 33120A generator 15MHz square wave output looks like) is that the rise/fall times of the square wave output are relatively poor. It looks OKish up until 10MHz or so, and then the poor edge performance becomes dominant. Even the sync signal performs better. Here are some captures of what the square wave output looks like into a 50 ohm termination:

[nfmax]

I am currently working on a 33120A with a similar fault, exhibiting itself as negative clipping, and also 'crossover' distortion at higher frequencies. Luckily, I received two of these units as 'dead', and the other one was suffering only from a blown output fuse. In the faulty unit, R771 measures OK, but Q718 is apparently open-circuit base to emitter. This puzzled me, because the generator will still provide a negative output up to the clipping point. I found that the output was actually being driven negative by the positive side driver, Q714, through the base-emitter junction of the positive output device Q713, in reverse (Zener) breakdown mode. In my unit it dropped just under 5.4V in breakdown. This explains why the clipping occurs - the maximum swing is limited by the voltage drop across the Zener-ing junction. The crossover distortion comes from the limited slew rate at the output of the gain stage, now being asked to move about 6V between positive & negative half-cycles.
This may explain why your R771 failed, although mine is still OK, as the full output current now passes through this resistor.
However, operating a transistor in reverse BE breakdown is not, generally speaking, a Good Thing - it may result in reduced gain as well as increased noise. I plan to replace both output devices as a precautionary measure: you might consider doing the same.