[weblacky]

Okay....small problem...that I'm unsure is a real problem...but problem!!!!

I rechecked all caps (100% agreed with my discovery doc), soldered in the large daughtercard (with FAN ENV info plug on it) to the LV board.  Tried to start it on my current limiter, but limiter bulb blinked (repeatedly) and couldn't start. I've seen this before when ~2A isn't enough for inrush of a 500W PSU to start when I fixed an older Dell 500W PSU on the limiter. So I chanced it and placed it on an unlimited line then the PSU started (with FAN).  I used my old standard ATX PSU tester (plastic thing that shows voltages you plug into 24-pin ATX header to start and give live values for standard PC PSUs).  Since Tezro is ATX with extra plugs, I bet on a NORMAL ATX startup...I was right!  So go me!

Now the problem...maybe, I couldn't hook up the CPU 12V power plug (non-standard, so 12V2 was listed as gone).

Values the meter displayed were:
5v=4.9v,
3.3v=3.3v
12v1=11.2v...ahhhhhh, what?

I cannot access the 3.3V lines without major soldering (no connector) and I don't want to do if I don't have to.  This COULD be caused by not enough load on the 3.3v line for great regulation (there is currently no real load on 3.3v...my old tester doesn't load anything).  I modified a molex adapter and set my Load tester to 2A @ 5V  & 2A @ 12v... and hooked it to one of the two molex connectors on the PSU and it did pull 2A on each line and the 11.2v voltage didn't go down (drop) at the tester's ATX connector (but did at the wire to the tester to 10.9v, somewhat expected voltage drop, locally).

So here is the "problem" (I'll explain why I placed it in quotes in a moment).  It's supposed to be a modified ATX PSU, by ATX standard the 12V nominal voltage is supposed to be with + or - 5%, well that's 11.4v...and I'm 11.2v (~7%) at the ATX connector with 4A load (doesn't go higher with NO load).  So I'm just sitting outside of the ATX spec.


Now the SGI doesn't use 12V that much (drives?) but used 3.3v heavily.  It's POSSIBLE that a good 3.3v load MAY increase these values to just inside the correct standard with a correctly loaded PSU...unknown.

Here's why I placed that phrase in quotes. According to the Tezro manual, the environmental monitoring only shuts down the system when voltages are +/- 20% of nominal voltage.  So for 12V that's like 9.6v....so does this mean SGI did in fact know about this and compensated with allowing a pretty far out of standard PSU to be A OKAY as long as it was stable?

I never started this PSU (got it two years from eBay), I didn't start it because it's old enough that it could have failed right then, I saved it to rebuild, then started it.  So other than being told, it worked, I have ZERO history.


The fact that the other values are extremely close (but exact or low), leads me to wonder is it's a regulation issue and I need to just install it in a Tezro and see what happens (won't explode) or if these values are in fact OK and were accepted by SGI due to their own design and settings tolerances.


Can other Tezro (desk and rack) & O350 users/owners do me a favor and post their voltage output table from your L1s to see what some of you are running at these levels, please?  

I know a few tenths won't matter.  But if you guys are also running a little low, it makes me feel a heck of lot better moving forward.

This is just too odd to be a failure, I'm about 45% sure it's a regulation issue due to underloading...but that's not hugely confident.  If you guys are all running under 12v or even under 11v, I'd feel better that this is SGI specific and therefore doesn't meet the ATX specs (specifically) :-)


Thanks for your help on this...we're close...

[Raion]

I don't currently have access to my Tezro or its L1 because it's at my old house in Virginia. But I'll let you know.

I do know that the 5-pin signal connector signals to the system to increase 3.3v or decrease using a triangle wave apparently?

[weblacky]

Right, that connector COULD make a subtle difference, unknown at this time.

If you could spread the word to people you know, that would be good to get the ball rolling! My hope is, it's fine. But I really need more evidence before I assemble it back up and clear the space for getting my Tezro off the shelf and swap PSUs. If the system only shuts down at 9.6v on the 12V line, then I'm sure it can "handle this" but I'd like to see others first.

In regards to fixing this, man, working but a "little off" isn't something I can do anything about (given the adjustment on the daughtercard aren't labelled!). Not starting at all at least gives you something more severe to look into. I assume some person at the factory does a load test while on a meter or something and adjusts all those little pots for something (likely not JUST voltage). If they'd labelled them I might try that, they didn't label them at all (that I can see). So a little off but it works would be fine by me...if that's the case.

I cannot believe that all the other values are that close...and yet 12v is that far off, really doesn't make much sense unless this miminial load thing has teeth to it.

---

***UPDATE***, I just found a screenshot of an ENV from a Tezro on SGUG,

https://forums.sgi.sh/index.php?threads/...75&slide=0

The table claims it will start warning you at 10.80v and will shutdown at 9.6v. So, it looks like if the voltage doesn't drop below 10.98v (my end of wire, Molex, voltage drop value), it won't warn you. The value at the ATX connector never dropped below 11.1v in my tests on bench, so far. Still trying to decide if I should screw it all back together and plan to just try it and see if the values are better when inside the system....

[weblacky]

***UPDATE 2***

Okay, still cannot sleep...so I'm thinking about what I can do without poking the bear.

The PSU has several (in a row) variable resistors on the LV daughtercard. They were very poorly glued, actually there is only glue affecting the last (largest) POT. Now I have no idea what adjusts what. But I'm kind of assuming that these control the primary voltage feedback values only and I sort of want to believe the biggest POT may adjust them all (like a global offset). I cannot be sure of course.

So right now I'll plan to do Ohm checks on all POTs, then mark positions on all of them after recording current settings. Then I'm going to check if each POT has an electrical connection to a specific voltage line (via some kind of continuity check). It seems a lot of the ways this is done is one side of the POT is connected to the output voltage through a coil or some other resistor. I'm hope this is the case, then I'd know if I found a 12v specific offset to play with or at least where a setting MIGHT go.

If that proves fruitful, I'll then play with that adjustment (for the 12V line). If I find this is a digital thing and there IS no direct connection with any voltage outputs, then I'll see if the larger POT is connected differently than the other smaller ones and see what confidence that gives me.

This may not be needed as I said, as long as the 12V line NEVER dips below 11V the system will never see it as a problem...but that is too close for comfort right now, I'd love to give myself a margin into 12, so I can dip by 0.5v during sudden loads and not worry about momentary warnings from the Tezro...more to come.

[weblacky]

OK,
After stewing last night and suddenly growing a pair...I decided to try the variable resistors and I got a very good result...NOT GREAT...but VERY GOOD.  I've attached the pic below.  There are 4 total POTs in the LV daughtercard.  The glue job was so poor that the glue they used didn't even touch the first 3 POTs (I just snapped off the glue because it was barely on them)

   

But the last one is firmly in place.  I tried my best to probe all these and basically I came up with very strange results. The top three POTs aren't directly connected to the output lines...this mean everything is an abstract digital control...sucks. The top 3 POTs are 105 (1 Mohm) and the bottom is 100 Ohms (101)

The last (bottom) one is glued so tightly, that I don't dare fiddle with it.  But it's connected to about everything! It literally has low (under a Kohm) to ALL DC output lines and to ground.  In fact I cannot really make heads or tails of WHAT it does.  But since it's very glued...I wasn't able to try it.  

I'm considering purchasing brand new trimpots (POTs) and desoldering these and inserting new ones so they are free of the glue....in the meantime I KNOW that two of them are 12V and 5V adjustment (but not MUCH adjustment) and the top one is LIKELY 3.3v (but I didn't fiddle).  The bottom MAY still be some sort of adjustment offset, because the trimpots don't allow a huge amount of adjustment.

My previous values were:

3.3v = 3.3v
5v    = 4.9v
12v  =11.2v

I gently tried each POT and found they don't just start jumping, you have to move a fair ways before the 0.1v values starts to move...that's not a great sign...it means the adjustment only allows so much.

So MAXING out the 5V POT (I did back off to the point it really changed...it's just that further movement made no additional gains on this one) and the 12V POT.  My new values (rebooted PSU 3 times and did WITH and WITHOUT 4A of load on 5V and 12V lines to test settings).

3.3v  = 3.3v
5v     = 5v
12v  = 11.8v (MAXed out...couldn't get any more).

So I'm now inside the safe zone and only missing 0.2V, the voltage drop on the loaded 12V molex connector was 11.65V.  So still JUST within the ATX standard.  I'd like to gain those 0.2V offset on the 12V.  My guess is the bottom POT either controls HZ or it moves the entire range of adjustments as a global offset...but I don't have the experience to tell me, nor the data via probing to tell me.

So it really depends on how hard the PSU will voltage drop under load on the Tezro.  It still may drop out of the ATX standard but it shouldn't drop to the first L1 reporting threshold of 10.9v!  So at this point, I'm going to count my chickens (mixing metaphors) and go ahead and start glueing and assembling the PSU, if the entire range shift UPWARDS when the Tezro is loaded it will really only affect the 3.3v into maybe a 3.4v-3.5v and that might be an issue.  But the 5V only cam up 0.1V so the 12v will be the one to launch.  Seeing as how this was a change of only 0.6v...if it swings that high it likely at least won't reach like 13.5v in all likelihood (still inside the Tezro L1 limits!)

So this is a big deal discovery!  Adjustments for at least 2/3 are tested and likey the top one is 3.3v...don't know what the bottom does but that one apparently really mattered at the factory.  Given they are all mixed with resistors everywhere they aren't actually directly measurable (I tried limit testing so ensure they were functional...got back values that only make sense with resistors in parallel).

So unless I unsolder (or replace) I cannot be sure if I've had a failure on a POT but I'm extremely close to perfect.  I don't think I'll press it.

More to come.

[jwhat]

Hi Weblacky,

great to see that you have completed the rebuild.

I have only just seen your callout:

"Can other Tezro (desk and rack) & O350 users/owners do me a favor and post their voltage output table from your L1s to see what some of you are running at these levels, please?"

As you are well aware O350 have completely different swappable power supply to the Deskside Tezro. But FYI, I have powered up my 6 x O350 rack and done L2 env, all the 12V reports are pretty much bang on 12V.

See bottom of this set of hinv reports for the L2 env report from 6 x O350 (a nice statistical sample ;-) ): https://tips.graphica.com.au/what-is-ins...-sgi-hinv/

You may recall that one of these chassis had voltage issues, which you provided feedback on: https://forums.irixnet.org/thread-2563.html
Which I had to address by removing the Adaptec "DM10" card.

Cheers from Oz,

jwhat/John

[weblacky]

Yo Jwhat,
Yeah, I’m hoping to get to the 1.8v issue soon. Thanks for the ENV post, your at 12v yourself, so my suspicion is still gnawing at me, this 11.8v business has me all hot and bothered. Why can’t it be 12V!

After I finish all these Nidec Indy supplies, I’m going back to Trezo PSUs and I’m going to attempt two more rebuilds. One of these Tezro PSUs, last one actually, can be bench started and won’t break immediately. I can then see a before and after voltage reading.

If that first Tezro PSU is still an outlier at 11.8v, I’m going to find a replacement trimpot and replace the glued one (after recording measurements) and see if that guy can bump my 12v past the 11.8v I adjusted it too, into like 12.1v! I’d feel a lot better.

You’re correct in that I don’t care about the PSUs, I wanted Tezro and Tezro-like systems (not fuel) to get running voltage examples to see if anyone’s running a lower voltage like this is. So far, it looks like this isn’t a common issue. So I’ll need to open the patient up again for surgery and replace that Trimpot and start fiddling in the hopes I can bump (adjust) the running voltage to perfect!

It’s still odd though, since the adjustment is digital to a management IC, the adjustments don’t kick in until after Power Good is asserted, which I don’t like.

So when I see the voltages on my ATX tester they are the original voltages (before I fiddled with the line trimpots), the after about 1/2-1sec the values all change to what I adjusted them to and stay there (even under a test load). By they are applied late where even a human can even notice them from startup. That’s really frustrating!

While I know the L1 will accept it as it takes that time for it to start and 11.2v is still within the 10% before 1st warning threshold, if it jumps up the full 12v the system shouldn’t complain. But that last thing I need is some stupid warning during startup due to the PSU lagging the adjustments I set.

We’ll see if that happens (I don’t think so), but it’s so close to the thread hold that it bugs me (no margin).

Thanks for posting!

[weblacky]

I've ordered several 101 (100 Ohm) side-adjusted trim pots from mouser that I "think" will work on the Tezro PSU as the "fourth" adjustment, that was glued down.

I cannot let that adjustment go...though I might be tempting fate. I guess we'll see!

Cheers.

[weblacky]

Okay I took the time today and pried the PSU apart just a little to get to that lower (lowest) trimpot. I tried to buy replacements...but they ended up being bigger (correct footprint) then the original and other parts blocked its installation. So I desoldered the original, measured all the legs, then picked all the glue off it until it was clean, remeasured it, then reinstalled it (tight and a pain but did it).

I ran the PSU (unloaded) with an ATX tester showing ALL the line voltages (nice tool!). I slowly turned the trimpot...nothing changed on the voltages after 1/4 of a turn, tried the other way...nothing changed. So I then set it back and said...done. I guess it doesn't shift all the trimpot windows or something...sad.

Oh well, I'm in spec so I guess I'll keep going and I'll need to get things open to get my broken Tezro PSU open and on the bench for rebuild. Once I do that I can at least have a comparison after rebuild to see if the 12V1 rail is low as well or it's just this unit.

So since ~10.9v is the 10% warning threshold and I'm 11.8v, which is only 1.6% off nominal, as long as it doesn't drop too much, it will work. I'll just stop and work on new things.

So I guess I don't know what that last trimpot, that's heavily glued, is for.

So that finalizes the Tezro PSU rebuild process. While not 100% perfect, it's within spec and will work for many years (compared to before) so we'll just keep moving forward.

[vvuk]

It may be well after you need this, but just in case here's voltages from my two O350s (I've removed the XIO/PCI voltages):

Code:

O350 w/ODYSSEY:
Description    State       Warning Limits     Fault Limits       Current
-------------- ----------  -----------------  -----------------  -------
          1.8V    Enabled  10%   1.62/  1.98  20%   1.44/  2.16    1.777
           12V    Enabled  10%  10.80/ 13.20  20%   9.60/ 14.40   12.000
        12V #2    Enabled  10%  10.80/ 13.20  20%   9.60/ 14.40   12.000
          3.3V    Enabled  10%   2.97/  3.63  20%   2.64/  3.96    3.337
        12V IO    Enabled  10%  10.80/ 13.20  20%   9.60/ 14.40   12.000
        5V AUX    Enabled  10%   4.50/  5.50  20%   4.00/  6.00    5.070
      3.3V AUX    Enabled  10%   2.97/  3.63  20%   2.64/  3.96    3.285

O350 without: (NUMAlinked to previous; previous one is primary)
Description    State       Warning Limits     Fault Limits       Current
-------------- ----------  -----------------  -----------------  -------
          1.8V    Enabled  10%   1.62/  1.98  20%   1.44/  2.16    1.777
           12V <not present>
        12V #2    Enabled  10%  10.80/ 13.20  20%   9.60/ 14.40   12.063
          3.3V    Enabled  10%   2.97/  3.63  20%   2.64/  3.96    3.320
        12V IO    Enabled  10%  10.80/ 13.20  20%   9.60/ 14.40   12.063
        5V AUX    Enabled  10%   4.50/  5.50  20%   4.00/  6.00    5.044
      3.3V AUX    Enabled  10%   2.97/  3.63  20%   2.64/  3.96    3.285

You're right about the fault/warning limits, so your 11.2 isn't even in the warning range... but all of my PSUs sit right at 12.0V.  But, these are the rack PSUs and not the tower ones.  (I'm.. surprised by the 12V not present in the second output but uh.. everything seems fine?)