Here's the writeup for the repair I did on a stanford research systems DS345 arbitrary waveform generator a while back. It was listed on eBay for cheap as failing the self test, but sometimes not failing. As received, it consistently failed the self test with a doubler error. (I think it worked maybe a total of twice without failing the self test with a doubler error.) It was also missing two of the case screws.
The triangle, sawtooth, noise, and arbitrary waveforms all worked fine. The sine and square waves both saturated to a flat line after a relay click. This is consistent with a doubler error. There are two paths the signal from the main output DAC can take: one for the arb functions through a 10 MHz, 7th degree bessel filter which is supposed to give good step response for the arb functions (including the built-in triangle, sawtooth, and noise waveforms), and a second path through a 9th degree 16.5MHz Cauer low-pass filter which goes to a doubler to give the full 30MHz range for the sine and square waves. The signal isn't filtered again after going through the frequency doubler, but I guess it works ok and gives acceptable performance (when working).
The DS345 can sample the output signal, and switch in various relays during the self test, which is how it knew that there's a doubler error. The symptoms (arb waveforms working, but not the regular sine wave) pointed to a problem with either the 9th degree low-pass filter, the doubler itself (an AD834 multiplier with a +/- 1V differential voltage input and a +/- 4mA differential current output), or the circuit (labeled as "level shifter and differential output amplifier" in the schematic) which converts this differential current output back to a differential voltage to feed another AD834 multiplier used for gain adjust.
Now...time to open it up! Inside was a little dusty, and I cleaned it out a little with compressed air. The unit is layed out as two main boards connected by a ribbon cable. The top board hinges open, and has rubber bumpers so it can rest on the case when opened without shorting anything to the case. The unit is all through hole, but pretty densely laid out. It's a multilayer board with thin annular rings on a lot of the pads, and also thin traces so that multiple signal lines can be run between pins.
Below is a zoom in of the doubler section. The 9th degree low-pass filter is in the very bottom right (lots of toroids, all signals differential), with the 7th degree bessel filter for the arb functions right above it. The AD834 multiplier U600 used as the frequency doubler is in the bottom left. The MPQ3906 quad-transistor array U604 (used to convert the differential current output of the doubler back to a differential voltage input for the next AD834 multiplier used for gain adjust) is in a DIP package just left of the center of the image.
Probing around a bit, the signal (at half the frequency) did make it to the input of the doubler U600, so that at least ruled out the 9th degree low-pass filter. Probing the output of the AD834 multiplier is pretty difficult since it's a differential current output. On my unit, op-amp U406B, which is used to maintain the common mode of the output of the differential current to voltage converter (quad 3906 transistor array U604 and associated components) at 0V, was saturated with its output at -11.26V, and was completely failing to maintain the common mode at 0V. Here's a zoom in of the differential current to voltage converter:
The service manual says the following about this section:
"U600 is a multiplier configured as a frequency doubler which extends the DAC's 15 MHz output up to the 30 MHz final output range. The network and quad transistor array U604 converts the current output of the multiplier to a level-shifted output to drive the output amplitude control multiplier. Op-amp U111B [sic] serves to correct the output DC levels, and the voltage summed at U604B's emitter via R614 nulls the mixer's DC offset (which is proportional to the square of the RF amplitude). Peaking inductors in series with R607 and R608 provide some gain boost to account for output roll-off above 20 MHz."
I think U111B here should actually read U406B, and might be a reference designator left over from an older schematic. The circuit works as follows (looking at just one side of the differential current to voltage converter): U604B (or U604A) acts as a common-base amplifier configured as a current buffer to buffer the varying current output of the AD834 multiplier. This current develops a voltage across R607/L603 (or R608/L602) of 250mV/mA. U604C (or U604D) acts as an emitter follower to buffer the voltage developed across R607 (and R608).
Since this is all differential, the two halves together give a total voltage swing of 500mV/mA, which along with the +/- 4mA full-scale output of the AD834 multiplier gives a differential voltage swing of 2V. Op-amp U406B maintains the common mode of this output at 0V so that the swing is +/-1V as needed to feed the next AD834 multiplier U702 used for gain adjust.
On my unit, op-amp U406B was saturated at -11.26V, and unable to maintain the common mode of the output at 0V. Input to the inverting input pin 6 of op-amp U406B was 1.17V, emitter of U604D was 1.69V, and emitter of U604C was 0.66V. These should all be close to 0V during normal operation. (The output of op-amp U406B was about -5.7V after I fixed the unit, although the exact value probably depends a lot on settings, and it's part of a feedback control loop to maintain the common mode of the output of the differential current-to-voltage converter at 0V.)
The control loop works as follows: if the common mode (of the output of the differential current-to-voltage converter) is above ground at the inverting input, since the non-inverting input is at ground, the op-amp's output swings lower, causing more current to be drawn through R612 and R613, and this steals current from current buffers U604A and U604B, which leads to less voltage developed across R608 and R607. Emitter followers U604D and U604C follow this lower voltage, which lowers the voltage at the inverting input to op-amp U406B as expected. And vice versa if the common mode is below ground.
The schematic says "emitters are at +7 VDC", and I did in fact measure 7.04V at the emitter of U604A, and 7.03V at the emitter of U604B as expected. This implies that there is in fact 34mA being sourced by R630/617/618, and 32mA being sourced by R620/619 as expected. The 7V at the emitters is set by 8.7V zener diode D600 along with a 0.7Vbe drop for U604A/B (very little current flows into the base of any of the transistors, and thus there is negligible voltage across base resistors R631/632/633/634).
Next, the schematic says Ic=18.2mA, which should put the base of U604C/D at -0.65V. In fact though, I actually measured -0.1V at the base of U604C, which implies a little more than 20mA, and 0.9V at the base of U604D, which implies a current of about 25mA.
These currents are higher than expected (although not by that much), and also more current flows through R612 and R613 than normal since the output of U406B is slammed negative...but where does this extra current come from? I lifted a leg of D701, and also removed R635 (probably unnecessary...the output of U406B is probably usually negative, and R635 to the -15V supply helps to sink some current so the op-amp doesn't have to sink as much). Lifting these components didn't help things.
There is another current path, which is through R614. According to the schematic, this is used to null DC offset (mixer DC output is proportional to square of RF amplitude). I measured -2.58V on the other side of R614, which together with 7.03V at the emitter of U604B on the other side of the resistor gives a current of 2.4mA across the 4.02k resistor.
Now, adding up all these currents (implied through voltage measurements, not directly measured), there's 2.4mA through R614, 20.4mA through R607 and U604B, 25mA through R608 and U604A, and 18mA total through R612 and R613. In total this comes out to about 66mA. This accounts for all of the current sourced through R630/617/618 (34mA) and through R620/619 (32mA).
The only unfortunate conclusion is that the doubler (AD834 multiplier U600) isn't sinking nearly any current at all (both outputs are supposed to sink about 8.5mA each according to the datasheet...10mA and 6mA according to the schematic). Unfortunately, both supplies to the frequency doubler also measured good at 5V and -5.18V.