[mosiniak]

So it seems that the system clock is on the io4 board? Not sure this is sensible conclusion though as it doesnt feel right... The io4 Dallas is soldered to board, where as on the system controller it is in a socket. There is also what looks like an oscillator chip on the back plane which could be it. Then again maybe it is on the CPU and it was the contact cleaner that fixed it. Did you manage to check where this system clock resides mosiniak?

I got "NO SYSTEM CLOCK ERROR" when I was trying to run Onyx with IPxx board removed. After putting back the IPxx there was no error. I cannot remember if the IO4 board was inserted or removed. I have to check it.

There is little information about he system clock. I found that:
1) System Controller checks for valid system clock sometimes called backplane clock.
2) On the IP19/21 board there is IC from Motorola MC100E11 which is an backplane clock driver.

From 1) and 2) I can assume that system clock is generated by the IP board. Why "NO SYSTEM CLOCK" error shows to You when You removed IO4 board I have no idea.

[Noris]

So it seems that the system clock is on the io4 board? Not sure this is sensible conclusion though as it doesnt feel right... The io4 Dallas is soldered to board, where as on the system controller it is in a socket.

There are two Dallas chips in the Onyx (more if you have multiple IO4s). Both of them store the SSN (system serial number). I believe the one on the system controller keeps the system time and the one on the IO4 keeps the environment variables.

I tried to desolder a Dallas from an IO4 once and gave up on it for fear of destroying the PCB which is very thick. There are probably some ground planes inside which are very effective at absorbing and spreading the heat. I ended up modifying the Dallas while it was installed. You can read about it here: https://www.vdheijden-messerli.net/sgist...io4-nvram/

I did a similar hack on the Dallas on the system controller, but I think the post went down with Nekochan... still have the photos somewhere. For now there's just the attached photo.

Regarding the SSN: in SGI systems, the serial number or MAC address is normally "attached" to the chassis, not the CPU or mainboard. In this case it's stored in the system controller, with the IO4 as a backup. If the serial numbers mismatch, the system controller "wins" and updates the IO4. A blank Dallas in the system controller will be initialized using the IO4.

Only if both are empty will a special procedure work to set the SSN. it is described here: http://www.futuretech.blinkenlights.nl/c...procs.html in sec 6.3.3.5

Hi Jan, thanks for info. I have read about this hack and nor sure I fancy trying it to be honest!!! I thought flat timekeepers are more of an inconvenience rather than being critical i.e you keep loosing date, boot config etc.

Is retaining the ssn critical to booting and basic operation? I had hoped it would just read in as a load of zeros or something if lost but not effect system function.

---

So it seems that the system clock is on the io4 board? Not sure this is sensible conclusion though as it doesnt feel right... The io4 Dallas is soldered to board, where as on the system controller it is in a socket. There is also what looks like an oscillator chip on the back plane which could be it. Then again maybe it is on the CPU and it was the contact cleaner that fixed it. Did you manage to check where this system clock resides mosiniak?

I got "NO SYSTEM CLOCK ERROR" when I was trying to run Onyx with IPxx board removed. After putting back the IPxx there was no error. I cannot remember if the IO4 board was inserted or removed. I have to check it.

There is little information about he system clock. I found that:
1) System Controller checks for valid system clock sometimes called backplane clock.
2) On the IP19/21 board there is IC from Motorola MC100E11 which is an backplane clock driver.

From 1) and 2) I can assume that system clock is generated by the IP board. Why "NO SYSTEM CLOCK" error shows to You when You removed IO4 board I have no idea.

It was robably just cleaning and retinning the vcam earths that resolved then. I think I did this after trying various minimal configurations.

[Noris]

Sorry for such a long post but these poka woes returned...

So it seems this VCAM -12v POKA error was not actually being caused by the VCAM regulators after all, but by one of the IR graphics boards, associated power boards, or perhaps the 3v / 5v power connections at the backplane... this latter option somehow came to me this morning after bizarrely dreaming about arranging some crisps (that is the chips that you eat if you are in the americas) on an IR power board last night!?!?

Basically the re2 setup (ge10, 512x1, and 2x 505 boards) works fine on both backplanes. The IR setup (ge12, 512x2, 303x2 and 305x2) yields poka -12v error, but if I pull the gfx cards and leave these power boards in place then the system boots and the -12v is fine now with no system clock errors being raised and all lights on power boards are ok.

The other symptom that might help trap this error is a red led on the ge12 with 3v written on it lights up at the same time as the -12v error led on the io4 but immediately goes out as the system powers down the vcam. I have tested the voltage on the vcam and can indeed see the negative voltage only reads around -2v with the ir boards but the +12v is fine.

Unfortunately the diagnostic roadmap does not cover the latter IR power subsystem, which seems more complex that re2 with 3v and 5v rails for the graphics, but I assume that the 3v rail for the ir boards is under poka.

The bad -12v points towards an error on the vme bus, but there are no vme cards installed, only the pab2 on the graphics boards. If the pab2 is on vme maybe it is causing error. Can this be simply removed or do jumpers need setting etc?

The light on the ge12 points towards a 3v error and as the 303 and 305 are not indicating errors I guess a 3v error could be caused by poor connection on the relevant pins which aren't used with re2. Is a bad 3v supply to ir boards likely to pull down the -12v for the vcam?

What about power board order? I cant find anything indicating that the onyx is sensitive to order of the power boards so it might be worth shuffling them around some but I'm a bit nervous about doing this! I could also try single 512 from re2 setup to rule out the 512x2 board but am worried this will fry the 512 if too much current is drawn!

Oh yes all board revs are ok for the ir setup... vcam, io4, ols, etc and there is no visible damage to any of the ir boards.

---

Ah ha... well after having head swimming for a couple of days then writing that massive long yarn over lunch, I just went back to machines and almost instantly spotted something on the dg board that looks suspiciously like a blown tantalum capacitor

I doubt that gluing crisps (or potato chips) to it will help but my mate who does actually resemble chewbacca can probably fix it!

There is what looks like some voltage regulators near these caps so it would be interesting to find out what they are doing.

[mosiniak]

79xx series voltage regulators are negative voltage regulators and 7908 produces -8V and must be powered by at least -9.5V. You can perform -12V voltage measurements with oscilloscope or analog meter with and without board with broken cap. Maybe You can see that with board inserted the -12V tail voltage never reaches the value of -12V.

You can also gently remove broken capacitor. There is a chance that this capacitor is making short circuit. In my opinion You can power up the board without this capacitor (for testing and checking) - it's not critical element. It's an input capacitor probably connected directly to -12V rail.

[weblacky]

Possible a forum member with the same board can post a pic to confirm (that ain’t right!). But looking above it the 7908CT is a 8v negative voltage regulator.

It likely feeds off of the -12v rail. The data sheet for this series of regulator claims it should be installed with two capacitors. One .33uf cap on the input -> ground and another 1.0uf cap between output -> ground.

Looking at your pic the damaged cap is connected to pin 2, which is labeled “input” on the data sheet.

So right there you’ve likely found a short between the input -12v and ground. Which would definitely be enough to drag voltage low (as you observed).

My advice would be to get yourself really cheap hot tweezers and remove all of those caps, use normal solder braid and an iron to clean up pads. Then install new caps of the same type and values using the iron (one side at a time, go slowly). Hot tweezers to install often make a mess, parts stick, and move during work and collision with neighboring parts becomes much more likely. Just use hot tweezers to remove caps.

Don’t try to use hot air in this application! You’ll likely roast/damage the board while making no real progress due to the huge/thick board layers that sink away heat. Simple $40 cheap hot tweezers should remove the caps easier and better.

Good spotting! Let us know what happens.

[Noris]

79xx series voltage regulators are negative voltage regulators and 7908 produces -8V and must be powered by at least -9.5V. You can perform -12V voltage measurements with oscilloscope or analog meter with and without board with broken cap. Maybe You can see that with board inserted the -12V tail voltage never reaches the value of -12V.

You can also gently remove broken capacitor. There is a chance that this capacitor is making short circuit. In my opinion You can power up the board without this capacitor (for testing and checking) - it's not critical element. It's an input capacitor probably connected directly to -12V rail.

Yup that's exactly what happens... -12 is fine without the boards inserted but only hits -2 when installed. Lol spot on regarding the input... My electronics friend advised that the cap is likely a non critical component providing smoothing to the input and that tants fail into short condition :) and apparently those regulators are pretty much bomb proof so should be ok.

All the same it seems safer to wait for the replacements to arrive!

---

Possible a forum member with the same board can post a pic to confirm (that ain’t right!). But looking above it the 7908CT is a 8v negative voltage regulator.

It likely feeds off of the -12v rail. The data sheet for this series of regulator claims it should be installed with two capacitors. One .33uf cap on the input -> ground and another 1.0uf cap between output -> ground.

Looking at your pic the damaged cap is connected to pin 2, which is labeled “input” on the data sheet.

So right there you’ve likely found a short between the input -12v and ground. Which would definitely be enough to drag voltage low (as you observed).

My advice would be to get yourself really cheap hot tweezers and remove all of those caps, use normal solder braid and an iron to clean up pads. Then install new caps of the same type and values using the iron (one side at a time, go slowly). Hot tweezers to install often make a mess, parts stick, and move during work and collision with neighboring parts becomes much more likely. Just use hot tweezers to remove caps.

Don’t try to use hot air in this application! You’ll likely roast/damage the board while making no real progress due to the huge/thick board layers that sink away heat. Simple $40 cheap hot tweezers should remove the caps easier and better.

Good spotting! Let us know what happens.

Hi weblacky, great advice thanks! Well I managed to remove it pretty easily with iron... i just gripped it with tweezers and popped each side with only a little flux and heat, but ill let my friend do the repair as he's been doing it for years. I believe he was planning to use hot air so I'll let him know about the heat dissipation issue. It seems most of the power lines are in the middle of the boards and I guess they are fairly substantial!

According to the markings they are 4.7uF 20V so will just go with the same however there are 297 SMD caps on mouser with those ratings, 47 of which are 10% kemet caps with same colour / markings. Once I confirm package size this should reduce to a choice of only 15!!!

I was wondering about replacing other caps too but there are quite a few on the various boards and many are harder to get to than those ones pictured. Is recapping generally advisable for this gear? I'll be ordering them today and they are cheap so maybe it's a good idea to get enough to replace the ones that are located in areas that look like power circuitry?

[mosiniak]

According to the markings they are 4.7uF 20V so will just go with the same however there are 297 SMD caps on mouser with those ratings, 47 of which are 10% kemet caps with same colour / markings. Once I confirm package size this should reduce to a choice of only 15!!!

I was wondering about replacing other caps too but there are quite a few on the various boards and many are harder to get to than those ones pictured. Is recapping generally advisable for this gear? I'll be ordering them today and they are cheap so maybe it's a good idea to get enough to replace the ones that are located in areas that look like power circuitry?

Capacitance is not critical, You can use 4.7uF, 6.8uF or 10uF. Pay attention to the voltage, 12V +50% should be reasonable limit (18V). You can choose any capacitor in the same case/footptint. You don't need to care about "low ESR", "low impedance", "temperature", etc. This capacitor will work in very comfortable environment (non switching operations, and "room" temperatures).

In my opinion there is no need for replacing other capacitors. Tantalum capacitors are not like the electrolyte capacitors.

Package looks like SMD size B (2.8mm x 3.5mm) compared to raster of 7908 (2.54mm). Here You can find dimensions of tantalum capacitor cases: https://botland.com.pl/img/p/424-1088.jpg

[weblacky]

I recommend replacing the other linear regulator caps (not all caps on the board).

You have the board out at this time and all the linear regulators in that region use those caps in the same way. If you had one fail then remember, they’re all from the same batch and used in the same way. So replace that small cluster of caps, just for piece of mind.

Could you just remove the one cap and leave it, maybe. Should you, no.

After all SGI wasn’t being your friend by adding extra parts to the board. They had to spend money and time adding those capacitors. So they obviously serve a purpose, else SGI would have left them out to save money.

Also I’m serious about not using hot air in your application if you’re not going to preheat the board! It’s a bigger risk and takes more time than what you did to remove the bad cap.

Unless your friend has a very large board preheater, it’s likely just going to damage the board surface and any neighboring plastic connectors, trying to overheat the area to get that solder to liquify.

Now if your friend does have a very large board preheating system then by all means use hot air. It can be very convenient. But if he’s just going to turn up the hot air gun, point it at the board, and attempt to desolder, it will fail. These boards aren’t cheap or plentiful, please don’t take the risk, use an iron or hot tweezers to remove caps from these huge boards. Then clean and apply new caps. No need to apply hot air.

I was thinking you'd get something cheap like this: https://www.circuitspecialists.com/Tweez...r-SMD.html
just for desoldering old caps, don't use to install replacement caps, use a nice iron.

Let us know what happens.

[vishnu]

I would recommend using two irons to pop the caps off at each end simultaneously. And as others have pointed out, when it comes to capacitance it's always okay to use more than the original (as long as the cap still fits the footprint of course). The input cap is not necessarily critical but the output cap is. If you read the manual most manufacturers will state in so many words that if there isn't an output cap the regulator might not be able to regulate *at all*.