TDS7054 Power Supply (PSU) repair

[awjennin]

Hello All,

I was up and feeling full of myself, and decided to purchase a non-working TDS7054 from ebay. I figured it would be a breeze to repair, but its been much more challenging than expected!
The unit I got is extremely clean. The dials, exterior, everything looks to have very little use. The insides are mostly dust free and generally look undisturbed. I know this doesn't mean that its low hours, but it looked hopeful. The HDD was missing.

What I have observed is that the unit appears to have no power. When plugging the power cord in with the toggle switch set to on, there's only a faint crackle of power making connection - which tells me theres no big inrush. When plugged in, there are no lights and the front power button is non-responsive (and led is un-lit). After running through the service manual power supply testing section (table 6-5), i measured no voltage on any of the supply rails (exposed at the Front and Rear Power Distribution Cards). I decided at this point, there was either a board-mount fuse, or housekeeping supply that was damaged on the power supply CCA.
I pulled out the power supply and got started. I observe two sections of this card - a high voltage side, and a magnetically isolated low voltage  section. When I apply power, a high pitched whine slowly starts (as large electrolytic  bulk caps charge), and i determined that the noise source was a high voltage film capacitor. I determined this by using paper rolled in to a cone, and placing it over parts, and listening. When I remove power, the whine slowly decreases frequency and gets quieter. Not sure if this is a clue or not..    The high voltage side of this card has 165VDC in a number of locations (i.e. at Drain of switching supply FETs, etc), but no other voltages elsewhere. I've checked the fuses, and capacitors all look to be in great condition, and resistors look to be unburned.
On the high voltage side, I see a number of switch-mode-power-supplies (SMPS) that rely on DC voltage. Based on spec sheets, it appears that this voltage needs to be between 10-20VDC. I've bugged out the board with my meter (fluke 189) and have noted all locations where this "housekeeping" voltage goes, but (and this is the humbling part), i can't figure out where its generated. I sort of suspect it comes off windings of T10, but I am much more versed in digital and analog design - not power electronics. I also don't have an available scope to take time/freq domain measurements (to look for oddities / rails trying to come up) - catch 22!
I'm shining a light through the board and trying to draw out the schematics - its a slow and error-prone exercise.

If anyone has advice / schematics / history with these units, I would greatly appreciate the assistance!!!

Thanks in advance for reading through for for any help!



For explanation of hackery near T10, please see posts 3 and 4

[Hexley]

Check out the area just below T10. There is a brown electrolytic capacitor that looks like it might have bulged a bit, and there seems to be electrolyte damage around the ICs just below.

[shakalnokturn]

If not signs of butchery...

[awjennin]

I will pull those caps and take some measurements - thanks for reviewing!!

The IC located next to those caps is a UC3842B (SMPS controller), and I pulled the spec sheet on it (https://www.onsemi.com/pub/Collateral/UC3842B-D.PDF)
It  accepts a VCC, and provides a 5v Vref - which a number of inputs (comp, fb, etc). There's an under-voltage lockout when VCC is below 10v/above 16v, or when Vref is below 3.4/above 3.6 (not sure if that is an internal measurement, or the measurement of the external pin).
I measured Vref at 1.8vdc (so we fail that condition - likely in UVLO) so was looking in to that. I removed the part and tested external to the CCA. when I powered it external (very simple setup), Vref was 5v as i expected, so i put it back in circuit
I also measured a 15-ohm series resistor at Vout (gate driver) at 45 ohms... and i don't have a 1206 1206 in 15 ohms, so I found a used 2512 and hacked it in to test.
there is an oily substance on the board which i thought was flux at first, potentially from a no-clean production process?... but you can see how nasty my rework was - so now skeptical of what it was...

When i get to root cause, I will order fresh parts, and clean things up.


I will try to get my schematics drawn out and will upload -  I think my next step is to identify why Vref is being held load on this card.


Thanks again!

[awjennin]

Here's a pic of the schematic, as I see it.



Via CR97, this supply does seem to be the origin / VCC housekeeping supply - but I'm still asking myself why she's not firing up. I will re-read through the datasheet again


From what I can tell, this UC3842 is powered by the 2.1mA that gets through the two series 16.6k power resistors. It appears that the diode to GND (near UC3842 pin 7, marked as 15.5v) is a Shottky but it could in fact be a 15.5v Zener making a little shunt supply - would make more sense than a shottky. It seems to be a SOD123, and has a case marking of J5L, but i'm not finding any info on that..   
Not sure the point of the other diode - CR96 - is? maybe it allows the VCC/housekeeping supply (which i surmise to be 18.3v based on the V-feedback voltage divider) to provide power during a removal of prime power to assist with power down sequencing? just a guess...

As I said, the series source resistor at the gate driver / output of the UC3842 was measured to be 40 ohms - which I (poorly, as pointed out) replaced with a 15-ohm.
Now i don't think the slowed slew rate of that fet (when at 40 ohm) would have caused overheating to the point of damage, and i don't see any evidence of thermal degradation near to that fet... but something wasn't right about that. possibly another clue
i used my fluke189 (freq counter mode) but can't see output switching of the UC3842, and in Voltmeter mode i sure don't see any voltage potential, but that could also be damaged..


Some more light reading tonight, and hopefully a revelation in my dreams tonight.
thanks for any help reviewing schematic & pics. Kind regards

[WattsThat]

If not signs of butchery...

Definite signs of the board damage around the 8 pin soic above C804/below T10. The appears to be what maybe is a 1206 resistor that is tombstoned with board damage in that area. It’s directly to the left of the to220 heatsink next to the cap that Hexley mentioned.

That’s not a good sign...

[nctnico]

If not signs of butchery...

Definite signs of the board damage around the 8 pin soic above C804/below T10. The appears to be what maybe is a 1206 resistor that is tombstoned with board damage in that area. It’s directly to the left of the to220 heatsink next to the cap that Hexley mentioned.

That’s not a good sign...

I agree. Someone has worked on the PSU in that area before. The capacitor looks fine to me.

[Hexley]

Here's a pic of the schematic, as I see it.

From what I can tell, this UC3842 is powered by the 2.1mA that gets through the two series 16.6k power resistors. It appears that the diode to GND (near UC3842 pin 7, marked as 15.5v) is a Shottky but it could in fact be a 15.5v Zener making a little shunt supply - would make more sense than a shottky.

1. Thanks for the schematic. Looks pretty straightforward.
2. Maybe I missed it, but have you measured the voltage at pin 7 (Vcc) of the uc3842?
3. While you are in there, could you note down the voltages at all the other pins of the 3842?

Thanks,
Hexley

[awjennin]

Hexley - Thanks for your review!!

These are the node voltages I have measured, per pin (meter on dc+ac, probe leads loosely twisted together for rejection)

1 - 1.12vdc,  0.20vac
2 - 0.01vdc,  0.05vac (sometimes jumps around to 0.2 and a little above, i can't tell if thats the meter picking up 60hz or not)
3 - 0.00vdc,  0.00vac
4 - 1.85vdc,  0.18vac
5 - 0.00vdc (gnd)
6 - 0.00vdc,  0.00vac
7 - 15.6vdc,  1.26vac
8 - 1.85vdc,  0.28vac

I truly hate to shotgun it, but en lieu of an o-scope to probe around, I may order a new IC (and 15-ohm 1206).
from what I can tell, CR97 and CR97 are normal schottky diodes based on the Vfwd i measured on my meter - there are no case marking. I may also try to pull and test the two diodes which i suspect might be zeners (no ref-des, sod123's with J5L case marking) and verify them. Same with the switching FET.
Because shipping cost will dominate the cost of an order, i may add these related diodes/n-fet to the order as well - haven't decided yet.

THANKS again for the help - and Happy Easter!!!

[Hexley]

The data sheet for the UC3842B specifies the threshold voltage for Vcc undervoltage lockout as between 14.5V and 17.5V, with 16V typical.

So your Vcc measurement of 15.6VDC, with 1.26VAC of ripple is concerning. Vcc is too low for reliable startup, it appears.

Some possible reasons, in no particular order:
1. The zener diode might have drifted low over time, and perhaps should be something like 19V.
2. Maybe somebody before you changed out the original switching controller chip -- perhaps originally it was a UC3843B. That part has a lockout range of 7.8V - 9.0V, so it should be happy with 15.6V as a startup supply.  On the other hand, the lower lockout voltage means that a UC3843 tries to start operating much earlier on the charging curve of the bulk (170V) supply than would a UC3842 -- which may or may not be a problem in this particular supply.

Sorry, no definitive answers. Just some thoughts to share.

Another comment in passing: make sure that C807 is in good health. It appears to be the bulk storage cap for Vcc, and needs to hold up Vcc during the period of increased Icc while the chip starts switching and the housekeeping supply comes up to voltage.

- Hexley

[TiN]

Most likely you have shot stand-by power. There is a regulator around 3842 that is used to provide +5VStby rail.
Be sure to replace ALL electrolytic capacitors. I had those leak out nasty electrolyte and ruin the PSU. Even if they look okay, they are already doing their nasty business under the cap  In my case it was dead 3842 and capacitors with gunk shorting stuff on PCB.

[awjennin]

I truly appreciate the input on this!

I removed the suspected zener diode and reverse biased it with 24v (current limited with a 5k resistor), and sure enough it showed about 19.5vdc across it. A little more digging on digikey revealed a SOD123 from OnSemi (seems like this PSU has lots of onSemi parts) which has a J5x case-marking for a 20v nominal zener (MMSZ5250BT1G) - This is beginning to correlate with what Hexley pointed out about the UVLO threshold... which i'm super embarrassed about misreading as a window comparator input vs the VIL/VIH that you pointed out! 15.5+ripple would absolutely be concerning if we aren't meeting the guaranteed VIH threshold! between leaky diodes, caps, pwb damage, and potentially damaged 3842, that could surely explain the voltage drop.

And to both Hexley and TiN's point, the caps do look great - but sure enough, I pulled the 68uF closest to the 3842, and while it measures close to nominal, the terminal finish of the through-holes is degraded, and if it has started leaking - that would explain the residue on the components near by!

The more I type, i'm becoming more and more convinced that the 68uF that you both called out is in fact the root cause. leakage current pulling down the VIN rail, and degraded characteristics not allowing it to keep the rail ripple-free. evidence of the residue and etched through-holes...

I will go through the effort of re-capping this PSU, point well received!!!

Side note, i'm a bit of an audio guy, and a while back I tracked down two sony DA777-ES receivers. I replaced pot's, caps, and retuned it per the service manual. While none of the caps looked bad and generally measured nominal, it made a huge difference to the audio quality!!
I will be just thrilled if i can get this scope going!!


THANKS again! I will report back when I get this SMPS cleaned up, and the PSU as a whole recapped.
Happy Easter all! stay healthy!

[fitch]

Wow, I cannot believe I found this thread.  I really hope I can contribute something here, and maybe fix my power supply as well.  I have a TLA614 logic analyzer that I have had for many years.  It was plugged in and in standby essentially the whole time.  Last time I tried to power it up, nothing.  I found another thread here from another user with the exact LA and his tail of repairing his supply.  Unfortunately, my failure is different.

First, my supply has marked on the big caps:
PS2251 REV. L
119-5806-05 REV. -
12709-5102
The voltages and currents are listed, but these could be configuration dependent.

My supply appears to have the input section working.  The DC on the big caps is there, and the two switching FETs Q3 and Q5 have drive signals on their gates.  Also, my +5VSB supply works.  When I power up the supply with PS_ON# deasserted, I have a single red LED lit.  When I assert PS_ON#, the LED goes out.  I tried to look at the various outputs to see if any of them tried to come up before shutting down, but am unable to see any movement on any of them.

I find it very interesting that on my supply, I had the same cap (C807) leak all over the surrounding circuitry.  I thought that replacing that cap would fix it, but no.  I have since found 2 other of the same type of cap that also leaked and have replaced them as well.

I have been spending time in photoshop with pictures of top and bottom, tracing out the signals to build a schematic of the board so that I can make some sense of what is there.  I too am not a power supply guy, so some of this is a bit foreign to me.  But the engineer in me wants to find the cause and fix it, besides the fact that my LA is in pieces taking up space in my lab:-).

I did get a connector diagram from Tektronix, that was the only bit of information I could get.  Maybe it can help you.  The housekeeping supply you talk about may not come up until the supply is turned on.  The connection diagram shows PS_ON# location as well as POWER_FAIL#.  These might be helpful if they are traced through the circuit.  I have been trying to do that myself, although very slowly.

Since I have my +5VSB working, and at least one other 5V supply, I can take some measurements for you if they help.  Just let me know.

[fitch]

By the way, I never mentioned this, but my power supply looks exactly like yours even though it was from a logic analyzer.  Perhaps it is somewhat standardized in some of the Tek units of the day.

I can confirm that indeed the controller is a UC3842.  Mine is original.

When I enable my supply, I measure slightly more than 18.3V at CR96 and 17.7V on pin 7 of the UC3842.  And, this supply is not there until I enable the supply by shorting PS_ON# to GND (I jumper J1-1C to J1-4C).

With the cap leaking in that area, is it possible some of the connections were corroded to the extent you have little to no connection left to pin 7, and therefore a lot of drop?

[TiN]

You can test PSU by poking PS_ON to ground as see if it chooch. Obviously have some DC loads connected to main power rails, so you can check if all voltages correct. Maybe it's not the power supply you have problems with.

[fitch]

awjennin,
Well, I found my issue.  Underneath one of the 68uF/35V caps that had leaked, the corrosion was so severe it ate all the way through a trace that was feeding the Vcc pin of one of the UC3844 controllers.  Once I found that, I repaired the open and the unit fired up.

Good luck with yours.

[Hexley]

Underneath one of the 68uF/35V caps that had leaked, the corrosion was so severe it ate all the way through a trace...

fitch, be sure that you clean the entire board to neutralize any electrolyte that might be found somewhere else. Isopropyl alcohol and a distilled water rinse might be a good idea.

- Hexley

[awjennin]

Fitch et al,

I'm late to the party! Work has gotten really busy as of late! i didn't even get around to checking this.   I'm thrilled that you found the issue! and I think your observation of power supply being similar is really valuable to the community. It sure makes things easier when a company reuses effort - both for them, and us!

And a huge thank you for posting the pinout of the prime power output connectors!  If that info was in the tech manual, i completely missed it!!
Now i'm really second guessing what I've done, and read TiN's comment about the 5v housekeeping supply in new light. If I can get a few moments I will check that out. What i've learned from this exercise has been great - but if it was the 5v supply all along, then i'm a dummy. ha!

However, I did order all new cap's for the power supply. Your observation (as well as Hexley's and TiN's) about Cap leakage are all supporting evidence that even if the unit is working, some caps may be close to end of life (EOL), and generally new caps every 15-20yrs will likely improve the scope because the cleaner output power rails should reduce the noise floor of the scope. Dirty power, dirty measurements?
So, I went ahead with an order. It included a few other parts which I touched with the iron a few more times  - but that aside, the entire order was ~$90 shipped. I fully intended on grabbing some ferro-flash too - forgot to finish that investigation prior to clicking 'order'!

In detailed design, caps can be picked because of bulk capacitance, equivalent series resistance(ESR), impedance at a specific freq/range, current rating, hours/lifetime, etc...  so I find it really difficult to replace these parts blindly  because i don't know what specifically was optimized/why the original caps were picked, and generally the original's have been obsoleted. Therefore I follow a person rule of thumb to replace caps
1) Replace from a similar company (United ChemiCon & Nichicon, in this case)
2) stay within a similar family, if possible
3) replace with components that are physically similar size, with same capacitance and voltage ratings as the old ones. I figure, its the physics that dictate a lot of the parameters. Capacitance eludes to foil area, voltage rating eludes to dielectric thickness, then the physical construction confines the X&Y dimensions that hopefully result in a component with similar properties in general
so, if you stay with the same manufacture (i.e. same build process), and same physical construct, the replacement part probably has similar electrical characteristics to the old one. I can't say its fool proof, or even accurate in all situations,  but that being said, here's my BOM

Capacitor replacement list
ELXY350ELL152MK40S (30x, CAP ALUM 1500UF 20% 35V RADIAL)
UPM1V680MED1TD (10x,  CAP ALUM 68UF 20% 35V RADIAL)
ELXY630ELL100MEB5D (10x,  CAP ALUM 10UF 20% 63V RADIAL)
EKZE500ELL331MJ25S (10x,  CAP ALUM 330UF 20% 50V RADIAL)
UPV1H6R8MFD1TD (10x,  CAP ALUM 6.8UF 20% 50V RADIAL)
UPV1H220MGD1TD (10x,  CAP ALUM 22UF 20% 50V RADIAL)
EKY-500ELL4R7ME11D (10x,  CAP ALUM 4.7UF 20% 50V RADIAL)
EKMH451VNN471MA50T (2x, AP ALUM 470UF 20% 450V SNAP)

Other parts (casualties to the debugging / introduction to power electronics)
FQPF4N90C (switch n-fet, associated with supply i've been debugging - cut off of CCA to preserve PWB. Old was obsolete, this appears to be very close parameters to original)
MMSZ5250BT1G (20v zener diode)
CRCW120615R0FKEA (15 ohm 1206 250mW resistor - original was bad / measured 40 ohms)
UC3842BVD1G (supply controller)
UC3844BVD1G (supply controller - not debugged, but case markings were almost unreadable - looked overheated though no pwb discoloration, worth the couple dollars just in case)


Fitch, regarding your power supply, do these voltages and currents match?




AND on a total aside...
 a little question that came about when I clipped the leads of that switch fet to remove and test it  -
 I noticed that C1 (C1:pin 2 in my schematic) is bonded to GND plane , but pin 1 is routed to the switch-fet's heatsink - which almost suggests that they expected the fet heatsink to be connected to drain of actual fet. i.e. TO-220 (non-isolated), versus the TO-220FP (isolated) version thats installed on these PSU's.  Odd, because this cap isn't doing an ounce of good, as far as i can tell. oh well!


Thanks again for everyone's input! Hopefully I will get the new caps soon, for a nice little weekend project.

[fitch]

It looks like you and I have the exact same supply right down to the revision.

By the way, I just found out that my supplies would not come up with some load on them. However, they were not 0V, but like .3 or .4V. That was very different from when it did not come up at all, in that case they were all 0V. Once I saw a little movement, I put it back in my LA and it powered up.

I think the UC3842 only supplies the +5VSB.  You might check to see if that is present on J1.

[awjennin]

I got the digikey box the other day, and went through a process of removing all the caps, Removing the parts around T10 where I was troubleshooting, cleaning up the PWB in all areas with 90% rubbing alcohol and a piece of cotton, reinstalling all components, then putting the board through my own version of water wash (hot water + soap, then alcohol wash, then dry) to get the gunk and flux residue off. Then after a quick power test, I put hot-glue drops between the caps to help with stressed solder joints due to vibration, over time. I even moved the stickers over from the large caps. I'm fairly happy with how it turned out.

   

My million dollar observation was this - ALL 68uF CAPACITORS ON THE POWER SUPPLY WERE LEAKING ON THE BOTTOM SIDE!!!  and while all other caps visually looked acceptable, i'm glad i replaced them all. She fired up without hesitation, and started booting!
While the leaking caps were bad, ROOT CAUSE of the failure was this: the 68uF on the input rail (shunt regulator) was bad and caused excessive resistance/ voltage-drop - to the point that the input thresholds were keeping the UC3842 SMPS to stay powered down. The 15-ohm series source resistor between the UC3842 & switching fet measured 40 ohms because the leaked capacitor ate away the resistive element on the 1206 chip.

I now need to track down a hard drive for this! It looks like a standard parallel interface laptop hard drive can be used? I will 3D print a hard-drive caddy, and then plug in the parallel cable directly in to the HDD (and remove the little parallel-to-hotswap adapter CCA inside the scope). Does anyone have experience with that? sound like an acceptable plan?

Mucho Thanks to xDevs.com because i think the SW load will be a breeze with all the resources there!

THANKS AGAIN to all those who read through and helped out with my troubleshooting. I will get a pic of it when she's back together and SW loaded

[awjennin]

I found a IDE HDD at the local PC salvage shop and got it. Was hoping for a 7200rpm, but they only had a 5400. oh well.  While I was loading Windows (with the top chassis removed, of course...) the keyboard fell in to the unit and damaged the Celeron fan. As I removed the board to inspect and repair, I realized there was another Electrolytic cap next to the fan that was blown up like a balloon. I decided to make a final digikey order replacing the two capacitors on the PC board, as well as the fan. On digikey, I also found a much quieter fan that moved about the same amount of air so that was nice. Installed, and power up test, then loaded and booted into Win2k.


Next, I decided to make a little HDD caddy, as my scope came without one. I drew up a quick model and hit print. I have to get my printer printing cleaner, but i will focus on that when winter comes around again. for the time being, this is more than sufficient.








Then got her buttoned up with the covers on!   If anyone has a spare top-plastic, I would be interested in buying that! or a 500MHz probe or two. I like the modern Tek6139's but am open to other good ones.


Now it is time to load software. I have some forums to read through to figure out where and how to get the Tek software & drivers loaded, as well as the ram written to and options loaded.
If anyone has specific threads that they've used, please let me know! otherwise i will dig around.

BTW if anyone's interested, my HDD caddy file is attached to this post. Because my scope came without the caddy, i didn't have the hot-swap to IDE interface board, so I removed the mating board from within my scope - and now the IDE cable plugs directly from the riser board to the HDD. Not perfect, but saved a lot of time and money trying to locate one! I tolerance it to have a nice firm fit into the scope.

 TDS7xx4_HDD_CADDY.STL.zip (9.86 kB - downloaded 38 times.)

Note: Remove IDE cable from inside scope, and attach it to HDD prior to inserting HDD into scope. Then feed cable through opening in back as you slide in HDD caddy. When complete, reattach host-connector of IDE cable into mating connector on riser card.


Thanks again for the help - hope this thread is able to benefit others in a similar spot!

[charlyd]

Hi i know this topic is aged already but my question is related to this psu. So the best place to ask i would say.

halfway "TiN" is talking about planning loads on the output.

but i already did that and my psu startsup in the scope but does not when it is placed apart on my bench. with the switch over J1/P1 and P4.
do i miss something 
i have  load over
J1.  P5,6  +12v.   --- 18 ohm --- gnd
J1.  P11,12.  +5v --- 18 ohm --- gnd
J1.   P19.  +3,3V  --- 18 ohm --- gnd

J1 P1  --- switch ---  100ohm ----- gnd

my red led lit when switch is open , it goes of  when #PS_ON to gnd.
but doesn t lit  when the switch is toggled to power off again.


My T7. J2 supplies are not present  +15v -15 v  and no + and - 5 volts

[charlyd]

i am tracing the secundairy side of these famous Martek PS2251 power supply.  and i am curious if anybody can add info like drawings / selfmade diagrams / hand made  schematics from the  T7  sec.   T10  sec.
i am missing some voltages like. +5v  -5v  +15v and -15v  located on J2.

and my +12V on J1  is +5.94V  so tips are welcome.

[charlyd]

Hello i have this question is it possible to remove fuses for particulair. circuit. to build up the secundairy side part by part.

now my psu is getting in protection every second or so.

so can i remove F50 F51 F52 . so the psu starts in a "small" version. first the basics  primairy and the needed secundairy part.... is that T10 ?