EEVblog Electronics Community Forum

EEVblog => EEVblog Specific => Topic started by: EEVblog on October 18, 2013, 09:45:12 pm

Title: EEVblog #536 - HP35670A DSA Repair
Post by: EEVblog on October 18, 2013, 09:45:12 pm
Dave repairs a HP 35670A Dynamic Signal Analyser.
Well, kinda...

! Private video (https://www.youtube.com/watch?v=YwiZROj3SSI#)
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: JoannaK on October 18, 2013, 10:40:12 pm
Have to say, that 40;Mhz oscillator looks indeed quite a strange..

Well, at least the machine has plenty of self-diagnostics in it now that it manages to boot up. Please keep us informed how this progress.


Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: free_electron on October 18, 2013, 11:15:08 pm
IMMEDIATELY clean off the square rubber feet on that top board that have turned to gunk ! Remove all of em !
The other ones need to come off as well they eventually all turn to gunk. that gunk is conductive. 3M has admitted they had a problem with those.  a plasticiser in the rubber reacts over time and the thing basically 'melts'.

i have sene tha t problem in many a machine that has these square rubber bump feets.

that thing on the crystal is not a resistor but a color coded inductor
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: nitro2k01 on October 18, 2013, 11:36:20 pm
Obvious thing to try next: All other linear regulators you might find. Whatever fault took down the 7912 probably took down some other stuff as well. If another rail is gone, obviously all the analog tests are going to fail.

Edit: The video embed link says private video. You can probably fix this by doing a dummy edit to the post.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: LaurenceW on October 19, 2013, 12:04:27 am
Yeah, the number of listed "fails" is so widespread, that I'd rather expect a single, common cause. My money (not all of it, obviously) is on a dead power supply rail.

At least, if Dave can't fix it, it will make an exciting Dumpster find for somebody else :-DD
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: exp1 on October 19, 2013, 12:09:33 am
I would heavily suspect anything connected to the +/- 18v rails, looks like when the PSU broke the 18v rails soared and took out the oscilator's reg. it could have taken out some analog/mixed parts too.

Btw, did you check the main fuse in the PSU? that could be the broken thing in the AC side.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: EEVblog on October 19, 2013, 12:27:35 am
Obvious thing to try next: All other linear regulators you might find. Whatever fault took down the 7912 probably took down some other stuff as well. If another rail is gone, obviously all the analog tests are going to fail.

Yep, quite likely, that was going to be my next plan of attack. May not be easy to measure some though, will have to see.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: centon1 on October 19, 2013, 03:04:42 am
Noob question.

Does the DSA self-test care if the frequency has not been 'tweaked' again?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: EEVblog on October 19, 2013, 07:25:22 am
Does the DSA self-test care if the frequency has not been 'tweaked' again?

Almost certainly not. The DSA can't know the reference freq is off unless it has another ref freq to compare it to, which it doesn't.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Tothwolf on October 19, 2013, 07:28:02 am
You might also try checking for shorted tantalum capacitors, especially the SMD versions. Tantalum caps tend to fail short when exposed to power spikes or excessive ripple (which punches a hole in the oxide insulating layer). I've seen many tantalum capacitors fail in this manner when a PSU failed (and even when there was just excess ripple due to dried out electrolytic filter capacitors). With enough current, tantalums can even catch fire and/or explode when they short.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: algorath on October 19, 2013, 10:10:06 am
how much did you pay for it? and how can a psu repair possible cost 1600$ ? what the hell did they do with it?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Experimentonomen on October 19, 2013, 11:29:21 am
I noticed a core on one of the transformers looks cracked, the printing on one looks interrupted and shifted up/down as if a crack appeared. Dunno if this could count for anything though as the psu was working at first.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: R_G_B_ on October 19, 2013, 01:25:13 pm
I had something simular happen with a oscilloscope turned out I had a short on a cap. Pulling the volatage down to zero and everything else connected to the same rail. So I would not be to worried by the amount of errors that have been loged if all these compents are running from the same source. Just a thought. Try powering up each board seperatly and test the voltages on each board. 

R_G_B_
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: M0BSW on October 19, 2013, 01:29:50 pm
 I'd love  to find something like this, where if you have the time you can investigate one part then on to the next, satisfaction must be fantastic when completed. maybe a hobbyist wouldn't care about not been viable repair, why would it matter in a hobbyist world, you are a lucky man to find this stuff Dave.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: M0BSW on October 19, 2013, 01:32:39 pm
how much did you pay for it? and how can a psu repair possible cost 1600$ ? what the hell did they do with it?
They probably gave the owner some Grade " A "bull shit I should think
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: sleibson on October 19, 2013, 03:17:07 pm
Another great video Dave. You are a natural!  :-+

I suspect there's one root cause for all the self-test failures. With the ADC clearly not working, the analog section cannot take any measurements. Thus any tests requiring the ADC to work must fail. Since the self-test is complaining about the ADC FPGA, that's the place to start. I don't think that you pulled that board in the video, but if it's another Actel ACT 1 FPGA in a socket, the first step is resocketing that FPGA after using a bit of contact cleaner with the hope that a bit of corrosion is the culprit.

The good news is that the entire digital section seems operational with perhaps the exception of the missing I2C peripheral.

The 39.xxxx MHz crystal oscillator in this instrument is a wonder. It takes up a huge amount of board space! I would have guessed that with all the attention and space given it, this oscillator was supposed to be some sort of ultra-stable clock source for all of the spectral measurements this DSA is supposed to make, but then I would have expected an oven-controlled crystal if stability was that important. I would generally cheat in this part of a design and just use a canned crystal oscillator in a metal can with a 14-pin DIP footprint. It would have been a lot smaller and my designs generally were crimped on board space. Also, it wouldn't need -12V. You could certainly afford one of those in a $20K instrument.

Strange that something killed the primary section of the "new" or "repaired" power supply. A spike that got through and also killed the negative voltage regulator? I can't tell from the video, but it looks like HP bought that supply from a power supply vendor instead of developing one on its own.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Rory on October 19, 2013, 03:23:47 pm
Doesn't it seem like HP/Agilent gear is designed that when they break, they fail catastrophically?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: sleibson on October 19, 2013, 04:17:18 pm
This instrument is 18 years old (based on the date codes) and there are many indications that it's gotten rough treatment. Crud in the BNCs, corrosion on the display, general tarnish and crud inside. It's seen a lot of action.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: bktemp on October 19, 2013, 04:19:56 pm
The datecode on the FPGAs is 94/95. The datecodes on the transformers also say 94/95. My guess would be that they have repaired the power supply. Otherwise I would expect a never datecode.

How does the DC input works? I see no second regulator path. From the placement of components I would guess that both inputs use the main switching transformer (between the L shaped heatsink and the secondary side). Therefore the other transformer must be a 12-28Vdc -> 300Vdc (or something like that) converter driven by a H-bridge (the 4 mosfets mounted on the chassis next to that transformer). The big red toroidal inductor could be the filter inductor. It could also be a PFC inductor but that was uncommon in 1994.
If it is done that way, then it is strange why the AC input is not working. It must something simple like a broken fuse, bad switch contact or connector.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Saulius on October 19, 2013, 05:34:19 pm
For how much did you get this device?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: vaualbus on October 19, 2013, 07:18:54 pm
As I've arleady written on youtube, I've read the schematics diagram on the service manual and apparently,on each board there is voltage regulator to change the +/-18 to other voltages, so if the psu fail and give to much voltage that regulator will die but the baord work.
Dave do a follow up please.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Zbig on October 19, 2013, 08:19:58 pm
These yellow ICs are wicked cool  :-+ I want a yellow microcontroller so badly now. Or a DAC. Or anything. No, yellow tantalum caps won't cut it.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: david77 on October 19, 2013, 08:43:07 pm
The yellow things are probably resistor networks not IC's, sorry to disappoint.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: LoyalServant on October 19, 2013, 09:27:19 pm
Looking forward to the follow-up.

I wonder if the I2C bus is being dragged up or down.... common culprit in a lot of consumer gear at least.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: EEVblog on October 19, 2013, 09:50:29 pm
The 39.xxxx MHz crystal oscillator in this instrument is a wonder. It takes up a huge amount of board space! I would have guessed that with all the attention and space given it, this oscillator was supposed to be some sort of ultra-stable clock source for all of the spectral measurements this DSA is supposed to make, but then I would have expected an oven-controlled crystal if stability was that important.

My guess would be it's a Butler oscillator. How low drift and low phase noise it is I don't know. Probably not hugely special in the phase noise department given the low frequency measurements.
It could simply be a carry-over from an old design, before canned oscillators cut the mustard?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: vaualbus on October 19, 2013, 09:57:24 pm
I think that the reason why the board haven't got the footprint is because this boards for agilent aren't troubleshooting so you have a fault and you replace it by paying thousends of dollars. Dave have you looked at this addition service manual contents?
LINK: h**p://www.home.agilent.com/upload/cmc_upload/All/35670-90067_supplement.pdf, there some more content in there. Probably the schematics will be avaible as soon as the instrument go out of production.
If you really want to repair it try to ask to agilent If they gave to you the schematics, somebody in the world should have those either a repair center or agilent itself.
Have you tried to put a signal in?
Apparently in new revision the instruments use only one cpu board, I've read that on the supplement manual somehere If I recall correct.
Best regards, Alberto
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: N2IXK on October 19, 2013, 11:11:20 pm
Do you have any evidence that the PSU was actually replaced?  Looked like some 1995 date codes on the caps, which dates the PSU to the same age as the rest of the instrument. Perhaps the $1600 was a repair estimate, not an amount actually spent on a failed repair?

Interesting that the DC input still works. Do the DC and AC primary portions of the PSU share the same switching transformer, with all the secondary side rectifiers and regulators in common?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: sleibson on October 20, 2013, 12:00:23 am
The 39.xxxx MHz crystal oscillator in this instrument is a wonder. It takes up a huge amount of board space! I would have guessed that with all the attention and space given it, this oscillator was supposed to be some sort of ultra-stable clock source for all of the spectral measurements this DSA is supposed to make, but then I would have expected an oven-controlled crystal if stability was that important.

My guess would be it's a Butler oscillator. How low drift and low phase noise it is I don't know. Probably not hugely special in the phase noise department given the low frequency measurements.
It could simply be a carry-over from an old design, before canned oscillators cut the mustard?

Well, I was using canned oscillators in the 1970s, predating this design by more than a decade.

--Steve
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Oracle on October 20, 2013, 08:52:09 am
how much did you pay for it? and how can a psu repair possible cost 1600$ ? what the hell did they do with it?

1600$ and they didn't fix anything...  :--
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: GermanMarkus on October 20, 2013, 12:20:32 pm
Dave, I think you missed the black AC fuse at the backside of the PSU!?
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: magnus0re on October 20, 2013, 01:40:58 pm
Nice going. maybe it's the same fault on the adc board.
Please make a followup soon :)
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: ZekeD on October 20, 2013, 03:55:14 pm
I'm hoping the I2 holds the ADC cal data tables, and with corrupt/missing cal data, the self test of the ADC's (and more..) is read as a fail. Now you have me wanting a DSA to muck around with.

Keep us posted. Great stuff.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: hikariuk on October 20, 2013, 04:37:24 pm
how much did you pay for it? and how can a psu repair possible cost 1600$ ? what the hell did they do with it?

1600$ and they didn't fix anything...  :--

Official parts are often hilariously expensive compared to an identical non-official part, going by my experiences with IT related kit.  I don't know if $1,600 is plausible even then though (I've no idea what a PSU of that kind is really even worth; I'm just a curious proto-newbie).
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Stonent on October 20, 2013, 08:10:27 pm
Well if it's complaining about gate arrays, I'd start looking at those Actel chips.

Pull them out one by one  and re-run the diags and see if there's a point where one chip is removed but the diags do not indicate any additional failures, that might indicate that Actel chip has died.  Also maybe check if there's any small GAL chips anywhere on the board.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: G7PSK on October 20, 2013, 09:27:21 pm
Has it been exposed to some form of solvent, that would explain the melted rubber and the screen could also carry plastic into the ic sockets. highly informative on trouble shooting and I for one would love to see Dave carry on with the repair. :-+
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: grayfox386 on October 21, 2013, 03:30:29 am
i would check the i2c eeprom for short it probably has config data for the adc on it if not that then check gate arrays
Title: EEVblog #536 - HP35670A DSA Repair
Post by: ddavidebor on October 21, 2013, 05:22:38 am
Has it been exposed to some form of solvent, that would explain the melted rubber and the screen could also carry plastic into the ic sockets. highly informative on trouble shooting and I for one would love to see Dave carry on with the repair. :-+

That's ugly melted thing melt itself with time. Hp loves it.
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: zinkoxyde on October 21, 2013, 02:28:38 pm
Dave, I recently got a Tektronix 2430a that failed a bunch of self tests with the ADCs and triggers. I ended up clearing the NVRAM and that took care of the problems. It must have gotten corrupted some how. I always blame software (or firmware in this case) before hardware. :)
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: HWgeek on October 21, 2013, 09:29:22 pm
Remember the old axiom - K.I.S.s    ;)

If there was one bad regulator, there may be others, so then the chips that are powered up by other possibly blown regulators will show up as a fail too.   This certainly can be a great debug learning "series" for people who might be interested in what it takes to fix something electronic.

Here's hoping that you follow up with additional video(s) and mayhaps get the darn thing working!
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Napalm2002 on October 23, 2013, 12:48:10 am
Amazing video I think Dave, I would very much like to see you go into further testing. I would say this vid has been one of my favs. Good luck on fixing it Dave! I think you should fix it up kinda no matter what the cost. Maybe some of us could donate parts to you or even start a fund my dsa fix video? Just a thought. I'm a huge Acdc fan by the way
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: Fezder on October 23, 2013, 10:00:58 am
nice video, good that it eventually booted up :)
Title: Re: EEVblog #536 - HP35670A DSA Repair
Post by: iXod on October 23, 2013, 06:00:11 pm
Before you cleaned it up I thought that the goo might have capacitively loaded down the oscillator circuit...

These repairs that end "not fully functional" are great! This is how many of us end up! And its great to see your process and choices you make.

The end? result is really secondary...