EEVblog #540 - HP35670A DSA Repair - Part 4

[EEVblog]

Dave continues the investigation into repairing the HP 35670A Dynamic Signal Analyser. This time using the Flir E8 themrla camera to check for any overheating components on the suspect A5 analog board.

Part 1:
Part 2:
Part 3:

[jesuscf]

One quick test is to check the output of the op-amps.  If the output is stuck to +-15V the op-amp may be bad.

[EEVblog]

One quick test is to check the output of the op-amps.  If the output is stuck to +-15V the op-amp may be bad.

Yep, can do. I might finally be getting the schematic for this board though, that would be nice.

[synapsis]

Thanks for explaining the quiescent current relationship to heat. I have a Flir (damn you Mike!), but I've so far been lucky to find things either hot enough to cook an egg, or dead cold.

So far the best use I've found for the Flir (besides tuning engines and looking for house insulation leaks) is aiming IR lasers.

[pickle9000]

EEVblog #4769854 - HP35670A DSA Repair - Part 764.5

For those of you still following.......

[EEVblog]

EEVblog #4769854 - HP35670A DSA Repair - Part 764.5

yes, it may come to that, because I suspect this one is BER. But a lot of people do want me to persist with it.

[ludzinc]

I had a weird fault with an almost_but_not_quite working Space Invaders clone (IREM).

No schematic to hand, it was time to check everything and reverse engineer.  Only took me a decade, but I got there and the kids love it. 

[han]

simple 5 minuite DMM/Oscilloscope on pin - , + and Out will tell everything..


Out = positive if in(-) < in(+)
Out = negative if in(+) < in(-)

if not put X mark on the opamp

[Rutger]

Maybe a silly suggestion, but isn't is quicker to just replace all the opamps with your new desoldering station. I know this a kind of a brute attack to the problem, but with the low the price of these opamps and the time it takes with a desoldering gun you might have fixed the board in 1/2 hour. Onces you have put the ic sockets back you can always put the old opamps back and try to debug the problem one opamp at a time.

[c4757p]

Maybe a silly suggestion, but isn't is quicker to just replace all the opamps with your new desoldering station. I know this a kind of a brute attack to the problem, but with the low the price of these opamps and the time it takes with a desoldering gun you might have fixed the board in 1/2 hour. Onces you have put the ic sockets back you can always put the old opamps back and try to debug the problem one opamp at a time.

AD845 is $7 each! And I think they're all AD845, or most at least.

Besides, it's not necessarily the op amps (and if you replace them all, Murphy'll make sure it's not... ), and op amps are pretty easy to check individually. Make sure the output isn't pegged, and make sure the inputs are equal.

[Rutger]

AD845 is $7 each! And I think they're all AD845, or most at least.

Wow, I didn't realize that this opamp was so expensive. So that was a silly suggestion, oh well back to troubleshooting.

[echen1024]

Just check a few op amps randomly, and Bob's your uncle!

[pickle9000]

Maybe a silly suggestion, but isn't is quicker to just replace all the opamps with your new desoldering station. I know this a kind of a brute attack to the problem, but with the low the price of these opamps and the time it takes with a desoldering gun you might have fixed the board in 1/2 hour. Onces you have put the ic sockets back you can always put the old opamps back and try to debug the problem one opamp at a time.

In a business situation it's a valid method. You may have a group of 5 transistors that always cause issues part of the service procedure could be to pull and replace those. Ideally you would test them as well (on or off board) and then proceed with a your regular tests.

In this case you are looking for education so that changes the equation. There are things you can do to help but basically a schematic is best. Because of the number of op amps Dave could take an 8 pin test clip and solder on some jumpers the hook those up to a couple scopes. That would save time and help ensure proper connections. In reality though that only reduces test time and it's just so much easier with a schematic.

 

[vaualbus]

First like a said last time there is another chip like the one that you replaced so replece them as well. Then with this expansion board we've to figured out a way to connect the coax cables but this is another story.
If i were you would try see the jocking voltage reference problem you 've had last time (vref+ is vref- and vice versa).
It securly be an opamp problem. Than asked again on the same guy that send you the schematics for this board schematics.
Also i think that there also a -18v regulator on the back board.
Wish you keep on withe the series.
To be sure take also a look at the signal path from the coax through the board and see if at the adc input the signal are present. For me the transistor on near the coax connector are bad.

[daddario]

Well, it could be nothing but why is the -15V rail reg dissipating so much more power? On the thermal camera it looks to be 20C warmer.
The heat sink looks roughly in the 20K/W ballpark to me, thus dissipating around 1 watt more.
I don't really know what's the input voltage, but I'd assume it's dropping ca. 5V.
I=P/U ;I=1/5; I=200mA
200mA is quite a substantial current.
Of course it might just have some extra circuitry on that rail or it is designed to have some op amps driving a load with its output against the neg rail, while not in use etc.

[EEVblog]

What's the story behind getting a schematic for the CPU board but not the Opamp board?

Someone said they'd send me the schematics and that's all I got. Can hopefully get the rest as well.

[eV1Te]

Maybe a silly suggestion, but isn't is quicker to just replace all the opamps with your new desoldering station. I know this a kind of a brute attack to the problem, but with the low the price of these opamps and the time it takes with a desoldering gun you might have fixed the board in 1/2 hour. Onces you have put the ic sockets back you can always put the old opamps back and try to debug the problem one opamp at a time.

AD845 is $7 each! And I think they're all AD845, or most at least.

Besides, it's not necessarily the op amps (and if you replace them all, Murphy'll make sure it's not... ), and op amps are pretty easy to check individually. Make sure the output isn't pegged, and make sure the inputs are equal.

A quick glance on the datasheet and I'm puzzled why they are so expensive? 25 nV/sqrt Hz, 250 uV offset and 16 MHz bandwidth (1.75 MHz under load)

There must be both cheaper and better op amps today?

[digital]

Well done on having another attempt even though it is frustrating, I hope you have another attempt especially if you can get the schematic

[G7PSK]

I noticed that one of the voltage control chips was far hotter than the other 2 is that just down to the extra current on the -15 volt rail or is it faulty itself.

[DL8RI]

Hi,

I'd check TP4 on A5 (see Page 323 in the PDF).
With this TP you can check the whole Analog stuff of the source. If I remember correctly the source crashed against it high border of the DC-Offset.
Maybe the Ref-Voltage from the Signal-DAC to the Offeset-DAC is faulty.

If that's not helping maybe you should feed in a signal (1k Sine) into TP6 and look on the outside.

Same thing with TP402 for the ADC. If a signal is fed in, and the controlling works. You should see you analog signal there.

[jancumps]

Maybe a silly suggestion, but isn't is quicker to just replace all the opamps with your new desoldering station. I know this a kind of a brute attack to the problem, but with the low the price of these opamps and the time it takes with a desoldering gun you might have fixed the board in 1/2 hour. Onces you have put the ic sockets back you can always put the old opamps back and try to debug the problem one opamp at a time.

...

In this case you are looking for education so that changes the equation. ...

And let's not forget for entertainment, and the 'play-along-at-home' factor.

[BMac]

I think I would replace one of the OP amps and see if the thermal signature changes.
Dave I really like this continuing saga, I love a good fight!

BMac

[RupertGo]

Power supply overvolt failures are so often BER - even if you fix all the faults, you don't know whether the surviving components are still within spec or have been pushed close to an early death. Sometimes it's worth just not bothering to fault-find to any precision, just renewing all the affected components - which is SOP for vintage restoration, where you often re-cap (and replace known problematic resistors) before even applying power for the first time.

Dave's done this thermal imaging trick for other problematic boards in the past (was it a Lecroy scope?), but I don't remember it ever working! It's a Macbeth move: when theatre companies were in trouble, they put on Macbeth as a last-gasp, can't fail trick to get bums on seats, but of course by the time you go there it's usually too late. Which is why Macbeth is seen as a cursed play in thespian circles and cannot be mentioned by name...

[squeezebox86]

Thanks for continuing on with this one! I was just yesterday thinking yet again that I would love to see more work on this one! The three regs that you show early on with the Flir, the one appears to be much hotter than the other two. Is this the expected thermal signature for that part?

[c4757p]

AD845 is $7 each! And I think they're all AD845, or most at least.

A quick glance on the datasheet and I'm puzzled why they are so expensive? 25 nV/sqrt Hz, 250 uV offset and 16 MHz bandwidth (1.75 MHz under load)

The "AD" logo is made of pure platinum.

[olafnew]

I'd try changing that buffer, the LM6321, it does look suspicious - and it surely may be one of the key rpoblems for the output not working. Especially given, that you do have the nifty desoldering gun - desoldering dip 8 is surely quick and easy

[mpep]

Just my opinion, but seeing as both the AC section of the power supply is damaged, and you've found voltage regulators damaged, I suspect that maybe Lightning Strike is the primary cause.

Good luck with trying to fault find, but as you've stated, most likely BER.

[Elandril]

I was wondering if I missed something in the other repair videos, but I distinctly remember that there was some I2C communication problem previously. Did this problem clear up, after fixing the voltage rails? If not, then there is still a problem on the digital side of things...

I also concur with what was previously stated for the -15v voltage regulator, which seemed much hotter than it's two companions.

[EEVblog]

Dave's done this thermal imaging trick for other problematic boards in the past (was it a Lecroy scope?), but I don't remember it ever working!

Maybe not in my videos, but I've used it countless times to find faulty parts.
Checking chip temps is a very standard troubleshooting technique.

[EEVblog]

So much for the schematic. My request has been turned down at the highest levels of the Agilent mothership. 

[BMac]

Is this still a supported product?

[pickle9000]

Dave's done this thermal imaging trick for other problematic boards in the past (was it a Lecroy scope?), but I don't remember it ever working!

Maybe not in my videos, but I've used it countless times to find faulty parts.
Checking chip temps is a very standard troubleshooting technique.

As is freeze spray.

[vaualbus]

So much for the schematic. My request has been turned down at the highest levels of the Agilent mothership. 

Really that guy on the forum seem to be favorable for the schematics.
Despite that next you could check the resistence beetweeen the -15 to gnd. Of course somethingh is wronf otherwise why the -15 is hotter?
So eventually some op amp could be shorted .
After I will try to inject a signal in the board and follow the signal path till you found where, If, it stop.

[Len]

So much for the schematic. My request has been turned down at the highest levels of the Agilent mothership. 

"And by the way, our name is Keysight!"

[RupertGo]

I was being facetious about using thermal testing to find broken components - have singed my fingers many times doing just that...

The ability of manufacturers to withhold servicing information is iniquitous, and illegal in the automotive market in the UK and US. Efforts to create a similar "right to repair" directive for all electronic goods haven't come to anything in the EU (at least, as far as I know), but the fight goes on.

[vlad777]

You should suck all them op amps. They suck really nice with a desoldering gun.
Make a schmitt trigger arrangement on a breadboard and test them one by one.
Make it a batch process.

I really want to see it fixed.

[han]

Thermal camera is great when troubleshooting high power device, like in industrial environment
but when used in low power board is useless
The only time the thermal camera is useful is when detecting intermittent problem.
It is difficult detecting functional fault without have comparison..
the only thing can detected by thermal camera without comparison is shorted fault component in VCC and GND , and it can easily detected using ohmmeter

[Kilroy]

Given that this is a power supply failure I would tend to suspect any op amps on power rails to be highly susceptable candidates for failure. Usually, I would just go hard core....suck them out with a Hakko 808, stick them on an unpowered bread board, and compare them relative to a known good component. You feel like a complete hack when trouble shooting in this way, but it is often the quickest way to completely factor out a large number of possible and plausible failure modes.

With a Hakko 808...roughly 15 seconds to remove an 8 pin chip...about 4 chips in a minute...around 5 minutes or so for 20 chips. Pretty efficient use of time, really.

[januszb]

There was an very interesting lead in this post:
  https://www.eevblog.com/forum/blog/eevblog-540-hp35670a-dsa-repair-part-3/msg318593/#msg318593
a followup to rs20's find:
  https://www.eevblog.com/forum/blog/eevblog-540-hp35670a-dsa-repair-part-3/msg318114/#msg318114
which gave an explanation for signal observed.
Could be an interesting starting point for investigation, maybe by probing the pins on the DAC as the digital input is probably fine and working the way from there.

Keep up these videos Dave, they're great and very educational (and with this particular one there is a certain element of suspense!)

[max666]

Please continue trying to fix this puppy, Dave.
Even if it turns out irreparable in the end, like the Lecroy scope. I just love this failure hunting videos. 

How do they say? "The journey is the reward!"

[Agent24]

With a Hakko 808...roughly 15 seconds to remove an 8 pin chip...about 4 chips in a minute...around 5 minutes or so for 20 chips. Pretty efficient use of time, really.

Couldn't agree more, when I got a desoldering gun it was night and day.. the old manual sucker gets hardly any use any more, besides, with PTH you basically *need* a proper desoldering tool unless you want to f**k around with braid all day.

[ckm]

So much for the schematic. My request has been turned down at the highest levels of the Agilent mothership. 

The only conclusion to draw from this is that, if you are in the market for a DSA, you are far better off buying an older used one where the schematics are available unless there is a specific use-case for the capabilities of a new unit.  Or even keeping older units in working order....

That's what I'll be telling my clients when they ask.  Given their current cost-cutting affinity, I'm sure they will be very happy to hear this.

[TheRuler8510]

Hey--whatever happen to this project? Is it just on the back burner?

Will we see another update?

I'd love to see a resolution...

[ziggyfish]

Looking around, you have tested LP4, but you haven't tested TP6 LPF2. If this is ok, then you can be 100% sure that the FPGA is working as it should. This is because LP6 is actually on one of the FPGA outputs (page 8-27 on service manual).

If its not ok, then then you know its the programmable low pass filter.

To narrow it down even test TP7 (Atenuator DAC) and TP3(DC offset DAC), and if that is ok, then test TP9, this should be the final output of the source. And if that looks ok, you know its between TP8 and TP9.

The other thing I found was the A5 board is actually part 35670-66505, so maybe that is useful for someone to find the schematic for the board.

The other thing that may help find the schematic is to look at the 35660 service manual (maybe they use the same design in both models).

[vaualbus]

Hi dave you will continue with the series?
Can you post here the op amp names on the forum so some could send you the op amp as mailbag!
Wish to have another video, they are so interesting.

[Quarlo Klobrigney]

Not sure if it had been cross referenced yet, but that Raytheon part is a TRW TDC1001, ADC, 18 pins and all. Use your favourite datasheet search engine. At least some data will help including the parallel interface that the ADC chip apparently uses.

[Towger]

Come on Dave, time to send the patent to Dr. Shahriar...

[EEVblog]

Shahriar would just waste his time like I have

[c4757p]

It's pining for the fjords, needs to be let go. Nobody's going to have half a chance at repairing this without schematics, and even with schematics it's likely that it has a lot of dead bits.

Thing just doesn't have any magic smoke left, and I doubt Dave has one of these.

[Quarlo Klobrigney]

After reading this "35670A schematics are proprietary - we cannot share them" Well kiss my ass Keyshite.

Dave, I have done similar to this board and worse without schematics. Can't you just remove and replace all the op-amps and try to see if it works? Surely they can't be that expensive. A vacuum desolder station helps rather than solder-wick. As far as the ASIC/CPLD well you're of course SOL. The ADC-DAC IC's I don't know about them as far as price.
I have a Spectrum Analyser from HP. I can't upgrade the firmware (fixes & features) for the same reason.

 :--Screw them that would offer no support and then orphan the user. No one wants to replicate their ancient technology to make money from it anyway. Shame indeed. Goodbye HP-Agilent-Keyshite or whoever you are today, and hello disposable, hackable China. And you wonder why your going out of business.

[mikerj]

. Can't you just remove and replace all the op-amps and try to see if it works? Surely they can't be that expensive.

Already suggested on page 1 of this thread.  The AD845's are $7 each, and there are a lot of them.

[c4757p]

As well as proprietary chips on the same rails, I believe.

[EEVblog]

Dave, I have done similar to this board and worse without schematics. Can't you just remove and replace all the op-amps and try to see if it works? Surely they can't be that expensive. A vacuum desolder station helps rather than solder-wick. As far as the ASIC/CPLD well you're of course SOL.

They are expensive, but I could socket them and put in cheaper one to test. And yes, if it's not them and it and ASIC or whatever them I'm screwed, the odds of that are high. There is a ton of stuff on that rail that I'm pretty sure went sky high, it could be anything. I've already spent enough time on it, I couldn't be bothered any more.

[Quarlo Klobrigney]

Sorry, I was trying to offer encouragement, not to foster anger or frustration. Maybe at a later date when things cool off you'll try again, or maybe one of the kind intertoob people will provide the unobtainium parts it needs. Either way keep on keepin' on.

[EEVblog]

Sorry, I was trying to offer encouragement, not to foster anger or frustration. Maybe at a later date when things cool off you'll try again, or maybe one of the kind intertoob people will provide the unobtainium parts it needs. Either way keep on keepin' on.

[TheDefpom]

Dave, if you have given up on this, perhaps you could send it my way to have a crack it it?