EEVblog #538 - HP35670A DSA Repair - Part 2

[EEVblog]

Part 2 of the HP 35670A Dynamic Signal Analyser repair.

[vaualbus]

could also not the the ADC it's self but some other circuitary.
I'm repearing an TDS540 that just for only a memory chip, failed in acquisition and proc interface.
So sometime just only a litle thingh can create thousend of errors.

ANd the most important thingh If you not connect the intra board coaxial cable of course some of?  self tests failed....
(and the source is not connect to the board).......

[xenocide702]

At around 20:10 when you were poking around the input jacks, did the LED flash in response to you touching it, or was that just coincidence?

[EEVblog]

ANd the most important thingh If you not connect the intra board coaxial cable of course some of?  self tests failed....

No, that's not it.

[edwardcurrentclamps]

My comment won't help much, but I have to admit that I can't wait for you to fix it. I'm also happy that I actually didn't buy it. Would have gone way beyond my knowledge!!! The 70 pages you refer to in the videos were quite intimidating to me. Being able to see you fixing it though is AMAZING!!!!!!

*Don't stop when you're tired, stop when you're done!*

Thanks

[Rasz]

maybe swap i2c eprom with a fresh blank one
and try to read the old one in the programmer

is it cpld or an fpga?

[edpalmer42]

When the original power supply failed, it must have spiked the voltage fairly high to take out the reverse diode and the regulators.  If that high voltage passed through the shorted diode, it could have taken out lots of things.

You checked for chips that were getting too warm, but what about chips that weren't getting warm at all?  Is the unit's power drain reasonably close to it's spec?

Do those boards use any kind of picofuses (i.e.  soldered in fuses)?

Ed

[c4757p]

I've a bad feeling about the I2C failure.... you're seeing activity on the bus, but are there responses, or just the processor calling out to a dead chip? I'd really worry that something which can take out a digital-only chip like the I2C cal ROM can take out a whole shitload of other stuff as well

It seems a shame that such an expensive piece of equipment wasn't better protected against failure. That power supply should be packed with an army of OVP crowbars...

[EEVblog]

Comparison of Vimeo vs Youtube as I saw it.

[scopeman]

Dave,

Are there any separate analog reference IC chips on the board (+ and - ref)?

Maybe something to look at. I would love to come to Sidney and help you fix it but it would be a long swim from Virginia!

Maybe some HP engineers or techs that worked on this thing will pop in to help or maybe just wait for another one with a different problem to show up on ebay.

Sam

[EEVblog]

Maybe some HP engineers or techs that worked on this thing will pop in to help or maybe just wait for another one with a different problem to show up on ebay.

Well, I haven't even started troubleshooting this yet really. I've only gotten as far as the PSU stuff.

[AlfBaz]

I wouldn't give up hope on the asic just yet. Its most likely the troublesome -15V rail is solely for analogue stuff and if the analogue stuff fails to provide cal signals to the asic the unit's best guess is an asic failure

[EEVblog]

I wouldn't give up hope on the asic just yet.

I'm not even close to giving up on the ASIC yet. I haven't even looked into it, it's more likely than not, not the ASIC chip itself.

[EEVblog]

The example you gave below Dave isn't really sharp enough for me to easily pick a difference.

Look at the Lemo connector on the right hand side outside the chassis. Huge difference IMO.
And it's cleary visible in real time video. The outline of my head for example is very blocky on the Vimeo one, and smooth on the Youtube one.

[bktemp]

I don't think that the power supply created a huge spike when it died:
As it looks like all voltage rails are generated from the same SMPS and there are no post regulators. Those LM3x7 and LM7xxx regulators have a SOA protection and will not die until the input voltage goes above >35V. This is about double the normal voltage. If the power supply had supplied double the nominal output voltage, some digital circuit on the 5V rail would also be dead. My guess is, the negative rail simply failed to output any voltage. The still working positive supplies would then have pulled the negative rail positive and killed the regulators and the diode. Putting a negative voltage (positive for negative regulators) into the output is something voltage regulators don't like at all and tend to fail quiete easily. This missing voltage could have killed some analog parts too.

[HamiltonHipster]

Good video Dave! Should have met you at Electronex, seriously kicking myself now

I have a Marshall B150 Guitar-Bass amplifier sitting on my bench at the moment. I turn it on and all it does is make loud scratchy noises. First thing i did was measure the negative voltage regulator, Lm7915, it was passing straight 20 odd volts to all the analog front end op-amps.

Of course, a big clue leading me to the regulator was the green corrosion spewing out the top of its package. Replacing it didn't fix anything, however...

Looking forward to part 3! :B

[Fryguy]

Did the 19+ volts on the negative rail get through to the analog circuits ? That may have killed some of the analog chips - e.g. many op-amps are rated at +/- 15V max for operation and +/- 18V as an absolute maximum afaik .   

[kizzap]

Did the 19+ volts on the negative rail get through to the analog circuits ? That may have killed some of the analog chips - e.g. many op-amps are rated at +/- 15V max for operation and +/- 18V as an absolute maximum afaik .   

most op-amps are specced as +/-18 Volts, being 36 Volts between rails, -19 to 15 volts is only 34 volts, so there is a chance that the op-amps are still fine. That said, it would pay to check them all out.

[MauriceS]

That is assuming the root cause was not massive overvoltage. The current (kinda working) PSU has 19V, but what was the voltage causing the regulators to pop? They are also speced somewhere in the 30v range usually - so that could have been just -30v or more on the negative bus alone.
But...

Then one would kinda expect one or more tantalums to have died a fiery death. But time to check the opamps i guess.

[Winston]

I would think HP would design a power supply or make sure a subcontractor would provide a power supply that wouldn't damage an entire unit if it failed.  I wonder if the original power supply was taken out by a lightning strike.

[tocsa120ls]

I wonder if the original power supply was taken out by a lightning strike.

A $20k unit plugged into an unprotected socket... seeing that my sub-1k gear is on a UPS, this thought makes me cringe.

Dave, making some progress there! Even less errors than the first time! Can't wait for a little I2C diagnostics

[kaushleshchandel]

Great video again! This is giving be some confidence to buy "Broken" "As is" "For repair" stuff from ebay.

There's a guy who has Datron 9000. And it's not working. And he is ready to sell it to me for $1000.

If you succeed, I may go for buying it and trying to repair it.

[Kirill]

If you succeed, I may go for buying it and trying to repair it.

Dave's success or failure doesn't mean a thing for other cases. Byuing a "for parts or repair" device may end up just repacing a blown fuse as well as literally rebuilding the device from scratch within an existing case. It always is like a state lottery. The only difference is sometimes you can actually win.

[iXod]

My intuitive sense is that the high-V failure of the PS killed not only the regs but also some critical logic components, prob. some unobtainium array stuff.

But that isn't to say that I wouldn't enjoy another vid to confirm this.

Keep up the great troubleshooting. It help give courage to others who might not have thought to tackle such a project.

"The outcome is really secondary."

Cheers!

[holozip]

I'm loving these video's where you get your hands dirty and try and mend stuff - just being able to observe the approach someone takes to troubleshooting is really valuable info; I can't wait to see if you manage to resurrect it.

Good luck!

[Rasz]

Comparison of Vimeo vs Youtube as I saw it.

use this to compare
http://www.freedownloadmanager.org/download/MSU_Video_Quality_Measurement_Tool_19505_p/

comparing on feel is so audiofool

[envisionelec]

I'm going out on a limb and saying that it's not the ASIC and it's probably not unobtanium. Too bad there seems to be no available schematics.  If I was really driven to succeed, I'd build a jumper/jig to get the board out of the chassis for testing.

[vaualbus]

Apparenty it seem that the system use an ADC and the atmel asic (fpga) to controll it. You can see a block diagram of the unit in service manual.
If it so probably the ADC is near by the fpga. If so we are a little bit more lucky given the fact that the asic if is only nasted logic work at 5V at least.
Apparently in the circuit there some dac to introduce a voltage to reduce the adc noise!,a +/- 6.3 reference adc voltage is provided by a dac controlled by the asic (fpga).
So I'll begin to look for that reference voltages.
Second I think that one of the two transistor on the attenuator board is a step up dc - dc converter to generate the test voltages  for the system, try to check also that two transistor.
That as I said on youtube try to see If it it's possible to adjust the -19.2 psu rail as close as possible to -18v the corret voltage.
Check also the back board because aparently also that board use the -18 rail.
The strange think is the i2c error becuase in the manul there no reference on what can cause that error.

[stranger]

Thought on the Atmel FPGA / ASIC, some good progress with the PSU and yes that raises a suspicion re the Atmel part but there is a chance the analogue ‘asic’ error is not related. The memory that finally crawled to the front of my mind was with similar socked parts and is worth a check, easy fix if you are lucky. Dendrites (whiskers) growing under the socket and providing shorts between adjacent pins, fix using a blast of IPA and a small stiff brush poked between pins under the socket being enough to clear some faults in the past.

http://www.purafil.com/PDFs/Technical%20Papers/Industrial/Corrosion%20The%20Unseen%20Enemy.pdf

http://en.wikipedia.org/wiki/Dendrite_(metal)

A real pain is when it happens on an internal PCB layer, 

[Stonent]

Actel. not Atmel.

[os40la]

Dave,

Do you have this Agilent 35670A Service Guide?

http://cp.literature.agilent.com/litweb/pdf/35670-90066.pdf

[NiHaoMike]

It's highly unlikely for an overvoltage to take out the 4148. That would take more than 100V, at which point there would be a lot of burn marks everywhere. Most likely the rail shorted and applied reverse bias for just a moment.

[EEVblog]

Do you have this Agilent 35670A Service Guide?

Yes, of course, but there are no schematics.

[84GKSIG]

silly but I,m hanging on the edge of my seat to see this thing work

[TopGunPk]

Hi Dave. Not sure if this could be anything, but mostly I2C EEPROMS are used to store CAL data and constants. Some would remain same over the years, some might change depending on a lot of factors.

Your 5V (digital) rail is OK, so its unlikely the I2C EEPROM has failed, however have you considered the possibility that when the original PSU went kaput, it might have caused a high spike or kickback, and probably that was enough to either : a) destroy the I2C EEPROM itself, OR b) corrupt / destroy some of the CAL data on it ?

From a software point of view, the software would start by getting the CAL data from the I2C EEPROM and then start reading the ADC, presumably by connecting one or all channels to known reference voltages to perform calibration. Most of the times, you need to have the previous CAL data (and that too within close limits) to get that CAL function to return a proper value. Otherwise, if the return values are not within a specific limit, it MIGHT just assume that the Analog or ADC section has failed !

Just a hunch.

[pickle9000]

My take on this repair is... very entertaining, I love troubleshooting.

- I'd pop out the eeprom and make a copy, mainly so I could have a starting point if needed later. It's good standard practice when possible.
 
- I like the fact that Dave is taking money into account, any repair can be done given enough time but "Is it economical for the customer?". I have no problem doing things purely for enjoyment or education but in a real job always think of the customer. Skipping the manual and checking the other regulators also says he is keeping the customer in mind. A fast repair save the customer money and you get to make more from them in the future.
 
- A very professional way of checking the temp of a chip. Use the back of the finger, it's sensitive and if you get a burn you can keep on working. Very "old school"

[ddavidebor]

Yep that's right!!!  Call data loss usually give adc error! And analog frontend error also!!!

[F4CPY]

According to the 4-15 part and 10-3 section, what code is displayed on the leds during a time of more than 4 sec?

[kaushleshchandel]

If you succeed, I may go for buying it and trying to repair it.

It always is like a state lottery. The only difference is sometimes you can actually win.

I agree its a lottery. Dave's video is a good step by step process on trying a repair. Even getting the Datron 9000 from "no power" to "power up with error" would be an achievement for me. I have never ever tried a repair on a large instrument like that.

[berniwa]

Hey Dave,

generally i think it would be an idea to identify what may have caused the fault.

In my opinion the diode and vreg damage indicates a positive voltage on the -18V supply or a voltage spike. Most IC have internal clamping diode so if the negative supply goes positive this diodes will conduct to gnd and you will get a current flow accros the little input diode you replaced. If I'm correct you may see some of the clamping diodeson the IC pins damage (having different reverse voltages) or dead (be carefull with Analog Pins, which may not have clamping diodes).

In this case the device may be beyond economical repair.

Greets
Berni

[Winston]

I wonder if the original power supply was taken out by a lightning strike.

A $20k unit plugged into an unprotected socket... seeing that my sub-1k gear is on a UPS, this thought makes me cringe.

I've seen far more expensive units than that plugged into unprotected sockets at several different universities and numerable military units, especially the mobile ones. This gets back to the original point of my post - my surprise that HP designs or allows a subcontractor to design $1500 power supplies that take out an entire unit when they fail or allow the power supply to be taken out by the unit after it's replaced due to inadequate output protection.

[LaurenceW]

I'm with BKTemp on this one. Maybe the negative 18V supply failed in the original power supply. The remaining circuitry in the instrument then attempts to put a reverse (i.e. positive) voltage across the negative regulators, and this will take them and any "piss-amp" (must be an ozzy phrase?) diodes with them in a flash. I've done that very thing myself!  Fault intolerant design, in my case.

I'd be VERY surprised if a raw -18V supply reached op amps etc. designed to run at -15V, would cause them any embarressment at all. But -18V onto any rail designed to run at -5V... that could do some damage.

I still need convincing that all of the cards are now carrying the appropriate main power supply voltages, and locally regulated on-card voltages.

[EEVblog]

I still need convincing that all of the cards are now carrying the appropriate main power supply voltages, and locally regulated on-card voltages.

I thought I showed that in the video? All rails on all cards are now good, unless I missed one.
Next video has been shot and uploaded for release. Subscribers have already seen it 

[baljemmett]

this will take them and any "piss-amp" (must be an ozzy phrase?) diodes with them in a flash

"Pissant", meaning insignificant

[vaualbus]

Dave when you release the next video on youtube? By the way you could ask to agilent for the schematics an
saying that you do a video on that. Pheraps agilent give the schematics. I was luck and optained the true schematics for my tektronix tds540 by asking to tek the schematics. Wait for the public release of the video. Best regards Alberto.
Also if you are subscribed there more video or you just see the new video before?

[Galaxyrise]

You showed a chart of measured power supply voltages, and that got me thinking.  When checking a "maybe bad" power supply, when do you check with a scope and not just a DMM?

[Stonent]

Dave when you release the next video on youtube? By the way you could ask to agilent for the schematics an
saying that you do a video on that. Pheraps agilent give the schematics. I was luck and optained the true schematics for my tektronix tds540 by asking to tek the schematics. Wait for the public release of the video. Best regards Alberto.
Also if you are subscribed there more video or you just see the new video before?

Well he did on their forum, but that was towards the time of the first video, they redirected him to the PSU area and he got no responses.

[vaualbus]

I don't mean by ask to the forum, I was in the situation on tek forum, but directly to the agilent help services.

[EEVblog]

Well he did on their forum, but that was towards the time of the first video, they redirected him to the PSU area and he got no responses.

Why they moved it to the PSU area just baffles me.
But in any case, it looks like the schematics are unobtainium.

[vaualbus]

This is bad but I wish you'll able to fix it.
Although have you tried to put a signal in the dsa?

[Stonent]

Well he did on their forum, but that was towards the time of the first video, they redirected him to the PSU area and he got no responses.

Why they moved it to the PSU area just baffles me.
But in any case, it looks like the schematics are unobtainium.

Well I guess because initially you were trying to get it to power up so perhaps that's why. Maybe a new posting about the A5 board will get some more hits?

[k2teknik]

But in any case, it looks like the schematics are unobtainium.

I think that HP / Agilent just not have the copyright to publish the schematic, you get that message now and then when you ask for schematic, drivers and other tech questions, the answer goes like, "Sorry we can't give you that because of copyright".
They buy a finish product, branded in there own name, and a part of the "low" price is the missing copyright to publish the schematic etc.

[k2teknik]

You showed a chart of measured power supply voltages, and that got me thinking.  When checking a "maybe bad" power supply, when do you check with a scope and not just a DMM?

For good measure you can add in a scope and one more DMM, then you can measure AC, DC and see the scope picture, all at the same time, and you should do it every time, but we are just human ain't we?

[EEVblog]

They buy a finish product, branded in there own name, and a part of the "low" price is the missing copyright to publish the schematic etc.

There is no way this is not a 100% designed HP product (excluding PSU)

[alm]

If this product was repaired by board exchange in the service centers and they didn't offer a CLIP, then schematics may very well never have existed outside the product development department.

[Fitzroy]

Hey Dave at around 7 mins into the video. You said there was two 5 V rails one being 5V analog and the other being being 5V Digital. So what is difference between those two(other being called digital and Analog)

Thanks
 

[daddario]

Hey Dave at around 7 mins into the video. You said there was two 5 V rails one being 5V analog and the other being being 5V Digital. So what is difference between those two(other being called digital and Analog)

Thanks

This is usually done to minimise noise ingress to the analog part from the digital part.