[weblacky]

Hi All,
I'm struggling with my last Indy repair and boy I didn't expect this.  I have some "ideas" of what could be happening...but I'm flying blind here.  I need some old timers to chime in and save me here :-)  

Also I hope my records match my memory.  

When I got this system it was assumed to have a bad PSU.  However I've since used another, rebuilt, PSU on this Indy (I'll will try another for chance) and it operates the same as it's original PSU.

The first time I tried to start it (with rebuilt PSU for this point on).  It would bring it's fan on, front LED SOLID GREEN, nothing.  After awhile you get tired of nothing and press the front panel reset button, the front LED momentarily turns RED and then turns GREEN and the graphics appear and everything starts!!!


The first few times I started it I got this (I think this is the unit that gave these...I may be I'm remembering wrong...so grain of salt, sorry):

   

After restarting/unplugging/replugging this error only came up 2-3 times in a row and has never since come up in 20+ boots.  System goes to PROM and says performing diagnostics and no errors occur and you get right to the main PROM menu.

Now when you plug in the Indy, it stays off, you press the front panel power button, it starts the fan and the front LED goes solid green...still goes no farther...you can press the front reset button almost immediately and it will reset and startup fine, with boot chime!!!!  Also RTC is holding MAC address AND datetime correctly between AC disconnects so...I assume it works for now.

BOTH PROM console AND Irix OS CAN control the PSU (e.g. they can issue a hardware turn off).

   

I've tried the following:
1. Totally different RAM...no change (but correctly registers Memory size every time of whatever I put in).
2. Unplugging the internal drive power connector from the mainboard (so drives in the 3.5" bays have no power)....no change, same behavior.
3. I thought I tried a 3rd PSU but I'll try it again later on, but no change from the ones I've tried now.
4. I thought I was able to do one resetenv...no change.

Here's what I'm wondering...what if this system was a frankenstein and is a badly upgraded unit?  I only think this because I do not see a reference online to PROM versions and CPU support.  But some sites claim the CPU (1994) in this thing was made available the year AFTER the PROM copyright of 1993!?!?!  Now that could be just a publishing thing and not a real chonometer of events.  But it's what I have to go on...poor CPU support?


I've seen no other issues, no freezes, no visual artifacts, nothing so far...

I don't own an EPROM programmer so I'd be asking anyone if they have maybe the last R5000 PROM for Indy (I assume I can run the last PROM with an older R4400 CPU?)  Correct me if I'm wrong.

Does this look like a firmware issue?

I know on the surface it looks like a power supply "zombie" issue.  But I've tried other PSUs and they seem to work on other Indys (again I'll make another attempt with another rebuilt PSU, I have just in case).  But this symptom does predate my purchase of the unit (I saw it when I went to buy it, but attributed it to a failing PSU at the time).

Seems, so far, like maybe a mainboard/PROM/CPU issue?

Anyone got any advice/hint for me on this topic?  Also Anyone got a later-model PROM for me?  I don't do this often enough where buying a programmer and ICs is going to be less then just buy a single PROM from someone.

Thanks for any help you can provide.

[robespierre]

I only think this because I do not see a reference online to PROM versions and CPU support.  But some sites claim the CPU (1994) in this thing was made available the year AFTER the PROM copyright of 1993!?!?!  Now that could be just a publishing thing and not a real chonometer of events.  But it's what I have to go on...poor CPU support?

R4400 150 Mhz processor upgrades were manufactured and in customers' hands by at latest December, 1993. So the firmware changes to support them were already written, using engineering test chips, at a prior date.

Your only symptom is that the machine doesn't automatically reset, so the likely cause is (drumroll) the POR circuit (stands for "power-on-reset"). Usually they use a soi dissant "voltage supervisor chip" or just a comparator plus a one-shot multivibrator to generate this signal. It is typically in a corner of the main board where there is extra space because the traces are routed last, after more critical signals. I'll have a look later to see if I can spot anything.

[weblacky]

I only think this because I do not see a reference online to PROM versions and CPU support.  But some sites claim the CPU (1994) in this thing was made available the year AFTER the PROM copyright of 1993!?!?!  Now that could be just a publishing thing and not a real chonometer of events.  But it's what I have to go on...poor CPU support?

R4400 150 Mhz processor upgrades were manufactured and in customers' hands by at latest December, 1993. So the firmware changes to support them were already written, using engineering test chips, at a prior date.

Your only symptom is that the machine doesn't automatically reset, so the likely cause is (drumroll) the POR circuit (stands for "power-on-reset"). Usually they use a soi dissant "voltage supervisor chip" or just a comparator plus a one-shot multivibrator to generate this signal. It is typically in a corner of the main board where there is extra space because the traces are routed last, after more critical signals. I'll have a look later to see if I can spot anything.

 
Yo Robespierre,
Hmmm ... Well, I was really fearing it was the mainboard's ability to control itself versus something I could modularity replace.  However, I have other mainboards for comparison. If you'd be kind enough to point me to the a good geographic region where this circuit might exist on an Indy...I can compare the measurements between the functional board and this one under powered-off testing with my huntron...perhaps narrow down the IC/component that way.  But, since I'm not familiar with these kind of processes/circuits, I'd not know where to begin my search (or even the type of IC I'd be looking for).

Any assistance you can give would be valuable and I appreciate your input.

In the meantime, I'll see about researching the circuit you mentioned to see if it's perhaps made up of a standard recipe of components often enough to visually pick out.

Thanks!

[weblacky]

Am I right that what your talking about it related to the processor ColdStart & Reset feature that comes up in the manuals?

Page 14 errate for Page 216?
https://ftp.zx.net.nz/pub/archive/ftp.sg...1_08_94.ps

User manual for MIPS R4000/R4400 (Resets covered in chapter 9)
https://course.ece.cmu.edu/~ece447/s13/l...manual.pdf

POR timing diagram on printed Page 219...However I'm still at a loss of how the interaction physically takes place, so far. However, now I assume what you may have been trying to tell me was the interval (time) for how long to assert the "reset" is the job of this external POR circuit? If so, okay...the info I have doesn't still give me a better physical description of what's happening to isolate some to trace/pin. But it's good logical info.

While I doubt it's that simple, I'm reading that POR cirucits (general info searching) can vary from being a simple cap and resistor - timed circuit, to something else. But I guess their job is to hold a voltage starting at time for a specific time? Yes?

I still don't yet understand the correlation between the signal and what pins it's using for this signal (e.g. I don't see a "reset" pin on the CPU pins).

So I'll need a little helpful nudge (or better). So at least we now have the diagram of timing of the POR, that's got to be something?

[robespierre]

My guess is that the LM358A on the processor module (low-power opamp) is used to provide power-on-reset.

[weblacky]

Hmm, okay...I can start working with that, thanks for the tip! I have other Indy R4400 modules, I'll compare them!

From this observation, I'd also assume if I swapped CPU cards that my symptom should disappear (confirming an "on CPU card" issue)?

Stay tuned!

[wroteafaq]

My guess is that the LM358A on the processor module (low-power opamp) is used to provide power-on-reset.

I haven't actually pulled an Indy apart to have a look-see at what's there, but the 4K data sheet shows two different reset inputs, and describes fairly complicated conditions for their use - not only the number of clock cycles that they have to be asserted for, but clock synchronization requirements.  So expect it to be a little more involved.

Fwiw, I've long been a fan of supervisory chips like the 1232 rather than just going the old-school simple comparator route (unless it's a uC that has its own more robust reset circuit onboard).

[robespierre]

I'll concede that I only have a R4600 module and it has a '358. There isn't anything on the IP24 that looks like a supervisory chip, so unless it was incorporated in the chipset somehow, I don't think it's on the main board. A lot of SGI's custom chips were done in VLSI gate arrays so they wouldn't normally be able to do analog functions like comparators.

[wroteafaq]

Agreed - it's not going to be in any of the ASICs.

[weblacky]

My guess is that the LM358A on the processor module (low-power opamp) is used to provide power-on-reset.

I haven't actually pulled an Indy apart to have a look-see at what's there, but the 4K data sheet shows two different reset inputs, and describes fairly complicated conditions for their use - not only the number of clock cycles that they have to be asserted for, but clock synchronization requirements.  So expect it to be a little more involved.

Fwiw, I've long been a fan of supervisory chips like the 1232 rather than just going the old-school simple comparator route (unless it's a uC that has its own more robust reset circuit onboard).

robespierre - I'm current involved in a family thing so I've not gotten to this.  But thank you again for lending your time/opinion to this mystery.  I'll certainly try it out.  I'd assume (given the way these older things work, that it's a simpler series of discrete components and it's just as likely some resistor/cap/whatever connected through this op-amp is out of spec/damaged and doesn't cause the op-amp to trigger correctly, obviously could also be said op-amp).  I'll check the measurements to and from and internally (as best I can) through the installed op-amp and see if there's leg that doesn't match a working system's leg.  The POR diagram I found claims that RESET needs to be held for a certain time period to work during POR.  Should we assume (for now) that maybe it's not being held the correct amount of time and so the continued power-on boot doesn't occur?

wroteafaq - I too saw all the timing stuff and (for me) I didn't really understand what physically is being toggled/activated on the CPU (external of the CPU), so I'll of course take any suggestions someone has to start an investigation.  I do understand this POR mechanism has evolved over the years but I'm keeping in mind that the Indy was likely designed between the the extremely late 80's to like 1991 at the most (because you have to manufacture the thing too before you can sell them).  So this was competing with the Intel the 486 like Dx2 and stuff for PCs at the time.  I'd imagine it's much simpler than what we use today.  For all I know it's what's used in simple microcontrollers and it's a pre-tuned timing circuit that's simple gone out of spec of the time is holds the reset?  We are talking like 1992-ish here.

However if you have an Indy and would be so kind as to spare your time and expertise, if you spot a circuit or mechanism that you believe in involved in the power-on process that would cause a symptom (hanging POST, nothing on the screen, reset causes it to proceed correctly) then please point me to it.

I don't have a full lab but I did snag a Huntron 3200S from an unhappy former owner that didn't understand how to use it and wanted to just turn it, so I got it for like 45% of retail.  I can use it to compare the in-circuit and out of circuit regional signatures to try to lead me to "what's different". It's worked very well in the past with my recent fixes posted here on IrixNet.  It's a cheater tool, no question, but it allows me a greater chance of restoring an item to operational. I purposely tracked down a modern Huntron for a ultra-low stimulus test voltages that no other product seems to do.  So I can do ASA against stuff that operates on extremely low voltage and cannot withstand a higher test signal.

There may not be a management chip, but if you see a cluster of known-layout components that classically does stuff like this, please consider pointing me to that to help solve this mystery.  I just need a region to try...I don't have the robotic system that checks every point over 4 hours for me :-(  So I need to get within a few inches or so.

Thanks.