EEVblog #540 - HP35670A DSA Repair - Part 3

[adcurtin]

I gotta recommend the Hakko 808 desoldering gun. If you chose to go the route of replacing all the op amps, I'd bet you could have all 15 on that board replaced in an hour. Seriously, this thing is absolutely awesome for anything through hole.

[wbooth.clearscene]

It may be worth thinking about this from a different angle. Why is it the 15 grand replacement power supply failed? Figure this out and you could be on the way to unravel the cause of all the other symptoms your seeing. Eg a simple shorting problem on the main board takes out the supply, but that problem could screw with reference generation, etc... Worth a shot before calling it a day.

[Zad]

Given the silkscreen claims on the board, I would do a sanity check on those +/- voltage levels. Is there continuity on the pin with a positive voltage, to the positive supply pins on the op-amps? It may be that the silkscreen really is mis-printed, but there may be a reason further down the chain that the voltages have swapped.

Consider removing the panel and applying a +/- supply to the op-amp power feeds, inject a small signal to the inputs and measure the outputs. That should give you an indication of which (if any) have failed, without having to power up the digital side etc.

Tack a ribbon cable to the input pins of the DAC, and bring them out to the scope. We aren't talking about 100MHz signals so you should be able to get some indication of what is being generated while the board is active.

[tri-be]

The DAC at U2 (AD565AJD,  which I presume is for the front panel Source output) is fed by the two TTL serial to parallel ICs (74HC595) at U1 and U17. U1 provides the most significant 8 bits to the DAC. The 'scoped sine wave is repetitive which at least suggests the serial data is OK. It's possible that one of U1 or U17 is damaged such that the output enable is not working, so the "steppy" output might be due to that (U17 handles the least significant bits of the DAC,  so it would most likely be that IC in that situation). I would somewhat expect U17 to be the cascaded output of the U1 device, the serial data being sent least significant bit first from the digital board.

However, the DAC datasheet says it is powered up to +/- 18V, so I would guess it's powered from the same +/- 15V supply as the dead op-amp, and I would suggest that it is actually fried.

You could probe/scope the digital inputs of the DAC to see if they all (or at least the lower bits) change on sine wave generation, and directly scope the output of the dac on pin 9.

I think if the DAC is dead, there would be no dishonour on giving up on this.

[Psi]

At this point i would probably replace all of the opamps on the board at once.

[azi]

The extender to allow more testing would be a good next move, but lacking that, I would desolder the opamps one by one and test them on a breadboard to see if they work. If many have failed, then I would replace all of the semiconductors on the +/- 15 volt rails. Hopefully, the +5 volt chips are all ok.

[Rasz]

AD565AJD
I think if the DAC is dead, there would be no dishonour on giving up on this.

AD565AJD is $10 with free shipping

[EEVblog]

If many have failed, then I would replace all of the semiconductors on the +/- 15 volt rails

That's about 25 opamps, and a handful of chips, minimum, on that one board.
And that doesn't take into account almost certainly failures on the analog input board as well.

[Jonny]

Make your own card bus extender

http://nz.mouser.com/ProductDetail/AVX-Connectors/108457096002025/?qs=sGAEpiMZZMud62t8luTOJtxE%252bQ%2fcV%2fAVEzdlj0dzMv4%3d

Cheap as chips! And I'm sure you'll have contacts that would make a one-off pcb for it 

I can sympathize with the groans associated with getting so far and wanting to do more but not knowing if it'll be worth it.... been there.

Top effort so far 

[Fagear]

Because it's stuck inside the box!
I really need an extender card or cable before I can properly work on this any further.

So why not just solder three wires by each opamp (or clip some hooks on it) and slide the board in? Just as you did with test points and alligator clips? You can see everything on each of them by scope. Its something about 5-10 minutes to sold wires, slide board in, power it up and probe with your scope!

Given the silkscreen claims on the board, I would do a sanity check on those +/- voltage levels. Is there continuity on the pin with a positive voltage, to the positive supply pins on the op-amps? It may be that the silkscreen really is mis-printed, but there may be a reason further down the chain that the voltages have swapped.

Its a good thing to check this. But this voltage is not feeding opamps, they are on +-15 volts. Not so easy to check by opamps pinout. As service manual says - it is used as reference voltage by ADC. If ADC is in that ASIC you don't know its pinout and cannot correctly probe some pins.

Consider removing the panel and applying a +/- supply to the op-amp power feeds, inject a small signal to the inputs and measure the outputs. That should give you an indication of which (if any) have failed, without having to power up the digital side etc.

I don't think it's a good idea. Some mixed chips (that are digital+analog and use separated power supplies) don't like situation when one supply is on and other is off. For example - PGA2311 audio volume control IC. It starts to overheat if you apply +-analog voltage and do not apply digital power to it. I think soldering some wires and put the board back is better solution.

One thing I noticed in service manual (I have three versions of them now): in the table "Functional Tests All Self-Test Group" it says that A5 board is doing something with IIC bus. And there was error message on the screen about it. I suspect digital control board is sending data through A5 board to the DAC via this bus. If there is something wrong with it it can corrupt some messages causing such weird output from the DAC.

[Flump]

I cant offer any words of wisdom as fixing this is
way out of my league but i hope you dont give up on it,
maybe have a break from it, and attack it at a later date
armed with some new ideas and hopefully a schematic
will turn up.

[daddario]

So why not just solder three wires by each opamp (or clip some hooks on it) and slide the board in? Just as you did with test points and alligator clips? You can see everything on each of them by scope. Its something about 5-10 minutes to sold wires, slide board in, power it up and probe with your scope!

I've used this method and with such a large number of required test points the wire soldering takes up far more time. Plus having just a bunch of wires on your bench instead of a board to measure is absolutely unintuitive, especially if you don't have a schematic and the board layout is the only thing to rely on. The only recommendation I can give in this kind of a situation is to use a ribbon cable with one end terminated with an IDC connector and the other end(s) soldered to the needed test points. This way you'll have this monolithic multipin from where to measure and this helps to keep things a little bit less messy and confusing.

[rs20]

Working atop nitro2k01's great suggestion that some of the DAC's input pins are stuck high or low, I tried simulating this in MATLAB. Forcing certain bits on or off didn't quit produce the right output (in particular, the signal would show frequent excursion to fullscale high and low, which you don't see here). But, long story short, if you simulate that the DAC has failed in such a way that it just sums all the input bits and just outputs that sum (outputting b1 + b2 + b3... instead of 1 * b1 + 2 * b2 + 4 * b3...), you get a very familiar looking graph (attached).

This doesn't feel like good news for the state of health of the DAC, although it is circumstantial evidence at best.

Source:

% Random tweaking value
dacbits = 13;

% Ideal sine wave
y=sin(0:0.0003:2*pi);

% Values to DAC
vals= repmat(uint16(y*(2^(dacbits-1)-1)+(2^(dacbits-1))), 16, 1);

% Convert into individual bits (from LSB to MSB)
bits = 0 ~= bitand(vals, repmat(uint16(2.^((0:15)')), 1, numel(y)));

% Simply sum all the bits, instead of do a sum weighted by 1 (LSB) ... 2^N
% (MSB)
plot(sum(bits))

As a side question, I'm wondering if this idea makes any sense: I'd go around and test the voltage across the op-amp inputs (input- vs input+). According to the whole virtual earth principle thing, they should be equal (unless the output is saturated). If you found just one or two op-amps that failed VEP, maybe it would be worth a shot at socketing+replacing those. This obviously requires you to be able to probe the live board, though.

[vaualbus]

It not make a lot of sense because the point where dave measure the strange signal is not the dac output but is a point mark 1 pass filter. On the output of the dac a signal that seem a sine wave apper. I think is made in that wave and than pass through an active filter to clean and make the sine wave so If the ic is break no signal pass or some noise.

[Anks]

I think that the bits are probably not reaching the required level to switch them in and maybe cause by the reference voltages or even line contention.

[RupertGo]

I'd so walk away from this one.

[EEVblog]

So why not just solder three wires by each opamp (or clip some hooks on it) and slide the board in? Just as you did with test points and alligator clips? You can see everything on each of them by scope. Its something about 5-10 minutes to sold wires, slide board in, power it up and probe with your scope!

Because it would be a ridiculous mess. There are at least 25 opamps on this board alone, not to mention the DAC, ASIC, MUXes and other points.

[EEVblog]

I'd so walk away from this one.

At this point normally I likely would too, but it will make for some more interesting troubleshooting videos I think.
Just don't expect it to work at the end of it...

[c4757p]

Rather than wire up every single op amp, why not take the time to wire up an extender?

[EEVblog]

Working atop nitro2k01's great suggestion that some of the DAC's input pins are stuck high or low, I tried simulating this in MATLAB. Forcing certain bits on or off didn't quit produce the right output (in particular, the signal would show frequent excursion to fullscale high and low, which you don't see here). But, long story short, if you simulate that the DAC has failed in such a way that it just sums all the input bits and just outputs that sum (outputting b1 + b2 + b3... instead of 1 * b1 + 2 * b2 + 4 * b3...), you get a very familiar looking graph (attached).
This doesn't feel like good news for the state of health of the DAC, although it is circumstantial evidence at best.

Nice work!

[Stonent]

Given the silkscreen claims on the board, I would do a sanity check on those +/- voltage levels. Is there continuity on the pin with a positive voltage, to the positive supply pins on the op-amps? It may be that the silkscreen really is mis-printed, but there may be a reason further down the chain that the voltages have swapped.

Consider removing the panel and applying a +/- supply to the op-amp power feeds, inject a small signal to the inputs and measure the outputs. That should give you an indication of which (if any) have failed, without having to power up the digital side etc.

Tack a ribbon cable to the input pins of the DAC, and bring them out to the scope. We aren't talking about 100MHz signals so you should be able to get some indication of what is being generated while the board is active.

That's a good point, see if the +/- test points on this board show continuity with another point.

And still I'd like to see the board diagnostics with the Actel chip removed. If it shows additional failures with the Actel chip removed, we could assume that it isn't dead (or not fully dead)
If the failures stay the same then, well I guess we're still at 50/50.

[c4757p]

Working atop nitro2k01's great suggestion that some of the DAC's input pins are stuck high or low, I tried simulating this in MATLAB. Forcing certain bits on or off didn't quit produce the right output (in particular, the signal would show frequent excursion to fullscale high and low, which you don't see here). But, long story short, if you simulate that the DAC has failed in such a way that it just sums all the input bits and just outputs that sum (outputting b1 + b2 + b3... instead of 1 * b1 + 2 * b2 + 4 * b3...), you get a very familiar looking graph (attached).
...
As a side question, I'm wondering if this idea makes any sense: I'd go around and test the voltage across the op-amp inputs (input- vs input+). According to the whole virtual earth principle thing, they should be equal (unless the output is saturated). If you found just one or two op-amps that failed VEP, maybe it would be worth a shot at socketing+replacing those. This obviously requires you to be able to probe the live board, though.

If the DAC is something like an R-2R, you could get this effect if all of the low-side drivers were disabled. Check the negative rail to the DAC!

Edit - or the high side, of course, you'd just have the inverse... but we've already been checking negatives.

[Stonent]

Rather than wire up every single op amp, why not take the time to wire up an extender?

An idea I had thought of earlier in the other thread is some kind of low profile 8 Pin DIP test clamp that goes on the chip where you can take all the INs and OUTs and visualize them on one or more scopes.

[parabuzzle]

so how much did Dave pay for this thing? I know in the first video he said it would be an expensive paper weight if he couldn't fix it...

[TMM]

A board extender would be very nice to help for testing (the op amps).

A board extender is pretty much essential at this point.

If you do get/make board extender maybe you can use that AIM TTI Current probe that you reviewed to quickly find out what op-amps etc are buggered?
I assume you still have that probe.

The currents for the signal stuff would be way too small to pick up anything useful. It may find a short on the powersupply but imho an IR camera is even better for pinpointing those faults.