[jan-jaap]

Pictures are of:
- Top Layer:

My candidate for anything SGI specific behavior in there: the PIC16C72A with the 'SG' sticker on it. The PIC16C72A is a microcontroller with A/D converter. The NMB PSU was the original PSU for the Fuel. How about this scenario:

* The original NMB design uses the A/D converter to compute 'fan speed' from voltage drop over fan. Easily checked: just block the fan and see what the L1 says.
* The second generation (Sparkle) was a generic PSU adapted to conform to the behavior of the NMB unit, hence the little PCB that fakes the fan speed.

I'm a bit busy with my PowerSeries PSUs right now (knee deep in leaked capacitor fluid, picture will follow), otherwise I'd do it myself.

[jwhat]

Hi Jan-Jap,

I can't confirm the idea on the PIC16C72A microcontroller...

But can confirm the results of "idiot test":

>> # l1cmd env
>> Environmental monitoring is enabled and running.
>>
>> ...
>> ---
>>
>> Description State Warning RPM Current RPM
>> --------------- ---------- ----------- -----------
>> FAN 0 EXHAUST Enabled 920 1188
>> FAN 1 HD Enabled 1560 2410
>> FAN 2 PCI Enabled 1120 1534
>> FAN 3 XIO 1 Enabled 1600 2327
>> FAN 4 XIO 2 Enabled 1600 2177
>> FAN 5 PS Enabled 1349 2205
>>

Now stick soft plastic into fan...

>> # l1cmd env
>> Environmental monitoring is enabled and running.
>>
>> ...
>> ...
>>
>>
>> Description State Warning RPM Current RPM
>> --------------- ---------- ----------- -----------
>> FAN 0 EXHAUST Enabled 920 1196
>> FAN 1 HD Enabled 1560 2428
>> FAN 2 PCI Enabled 1120 1534
>> FAN 3 XIO 1 Enabled 1600 2343
>> FAN 4 XIO 2 Enabled 1600 2177
>> FAN 5 PS Enabled 1349 0

So yes fan is monitored and likely it shows a real speed as it has now gone up to 3409 RPM while I have been typing, likely due to temperature monitoring.

Thanks for the tip and good luck with PowerSeries PSUs.

Cheers from Oz,

jwhat/John

[jan-jaap]

The NMB GM430WTXW01SSV isn't ATX, it's a WTX (Workstation Technology Extended) power supply. Yet another form factor that went nowhere... You can still find the website of the WTX organization on archive.org. From there you can download a WTX Power Supply Design Guide -- which documents the requirements for the fan monitoring signal

So it's likely the SGI Fuel PSU isn't 100% proprietary, it's simply (based on?) a standard that never caught on.

The Sparkle FSP460-60PFN is an EPS12V PSU. EPS12V is another ATX derivative (but more ATX compatible than WTX). If the SGI Fuel mainboard was designed for WTX, than the "SGI" Sparkle PSU would need to (1) re-arrange the pinout from ATX to WTX, and (2) emulate the WTX fan signal.

It never made much sense to me (engineering wise) that SGI would first design a system to require a fan tacho signal, and then modify an off the shelf PSU to fake this signal.

[jwhat]

Hi Jan-Jaap,

yes I did some reading up on WTX specs and appears the Fuel PSU is based on the monitored 460 Watt single fan spec.

I also finally got an original SGI Sparkle and opened it up to have a look at the extra circuit board compared to the PC version.

It has Hitachi HA17393 Controller chip on it:



I will post more once I have more.

Blogging details here for pics hosting etc

Cheers from Oz.


jwhat/John

[weblacky]

The NMB GM430WTXW01SSV isn't ATX, it's a WTX (Workstation Technology Extended) power supply. Yet another form factor that went nowhere... You can still find the website of the WTX organization on archive.org. From there you can download a WTX Power Supply Design Guide -- which documents the requirements for the fan monitoring signal

So it's likely the SGI Fuel PSU isn't 100% proprietary, it's simply (based on?) a standard that never caught on.

The Sparkle FSP460-60PFN is an EPS12V PSU. EPS12V is another ATX derivative (but more ATX compatible than WTX). If the SGI Fuel mainboard was designed for WTX, than the "SGI" Sparkle PSU would need to (1) re-arrange the pinout from ATX to WTX, and (2) emulate the WTX fan signal.

It never made much sense to me (engineering wise) that SGI would first design a system to require a fan tacho signal, and then modify an off the shelf PSU to fake this signal.

Hi Jan-jaap,
According to the little research I did...SGI didn't "design a board that required a tacho signal"... a version of the WTX standard outputs a fan signal by requirement through the 24 pin connector.  So they actually seemed to design for the WTX standard, correct?  https://pinoutguide.com/Power/wtxmain_pinout.shtml

Which then makes one wonder can you buy another WTX power supply of the standard and make it work in a Fuel without modification?

[jwhat]

Hi SGI'er,

I know that this is well worn path, but please bear with me...

In looking at the WTX & Fuel Sparkle wiring is appears that the Fuel is using Intel WTX Wiring Variation that is quite different from the 460 Watt PSU specification.

See this diagram for variations of WTX, Intel Variation, SGI Sparkle and ATX (which PC Sparkle is consistent with):



Could someone please confirm on:
- P3 - What is this ?? Intel Variation has +3.3V DC or based on WTX (brown) it could be +3.3 AUX (which is just another +3.3V supply)
- P12 - Fan-M ??
- P24 - Is this really -5V DC ?? which is not even part of WTX spec

EDIT: Feedback is that P12 is Fan-C (but does nothing), P24 - Fan-M, P3 is +3.3V SENSE

I am asking this so I can just redo the ATX Sparkle wiring -> SGI Fuel WTX without need for adaptor.

Thanks very much for any assistance.

Cheers from Oz,


John

[weblacky]

Hi SGI'er,

I know that this is well worn path, but please bear with me...

In looking at the WTX & Fuel Sparkle wiring is appears that the Fuel is using Intel WTX Wiring Variation that this quite different from the 460 Watt PSU specification.

See this diagram for variations of WTX, Intel Variation, SGI Sparkle and ATX (which PC Sparkle is consistent with):



Could someone please confirm on:
- P3 - What is this ?? Intel Variation has +3.3V DC or based on WTX (brown) it could be +3.3 AUX (which is just another +3.3V supply)
- P12 - Fan-M ??
- P24 - Is this really -5V DC ?? which is not even part of WTX spec

I am asking this so I can just redo the ATX Sparkle wiring -> SGI Fuel WTX without need for adaptor.

Thanks very much for any assistance.

Cheers from Oz,


John

https://forums.sgi.sh/index.php?threads/...rsion.129/

https://wiki.preterhuman.net/Fuel_PSU_repair

I’m pretty sure most of this has been done before. But none of this will solve your ENV fan problem. What you may not realize is that the available fuel ATX adapter is all about emulating the expected fan signal and just swapping a few of the wires to the right voltage signals. There is no voltage adaptation or conversion going on in the adapters. It’s simply pushing a couple voltages to the other lines as indicated by the pinouts in the resources presented. The problem is that the fan signal is not a common tachometer signal from a fan. That’s why the inventor of the adapter had to use a small series of ICs to generate the “expected” signal.

The wiki link above already proves that just shoving a fan signal onto the wires is not appropriate and does not work.

[jwhat]

Hi Weblacky,

thanks for the links I rarely go to SGUG, as IRIX Network give me sufficient SGI hit ;-)

I have updated info based on Peter Human link.

I am aware that you can't simply feed fan to Fan-M, hence why I wanted to look at SGI Sparkle to see details of the extra fan control circuit (to avoid the tacho simulator)

Purpose is just to see if there is DIY approach to convert ATX/PC Sparkle -> SGI Sparkle using the same extra circuit.

The info now means that the wiring is aligned with the Intel 860 WTX variant...

Cheers from Oz,

jwhat/John

[weblacky]

Yes, I’m under the impression it’s the “intel variant” WTX pinout, as you surmised. It’s sad that I cannot find any PDF spec for WTX that describes it.

[jwhat]

Hi Weblacky,

yes especially given that Intel was the author of the WTX specs to begin with...

EDIT: I did find this spec for the original Fuel NMB PSU which has wiring as per WTX Intel Variant (post was courtesy of Jan-Jaap via Nekochan on Way-Back Machine) 

Cheers from Oz,

jwhat/John