Failing power supplies and the "usefulness" of the machines in question - Printable Version

+- IRIX Network Forums (//forums.irixnet.org)
+-- Forum: SGI/MIPS (//forums.irixnet.org/forum-3.html)
+--- Forum: Hardware/Triage/Repair (//forums.irixnet.org/forum-11.html)
+--- Thread: Failing power supplies and the "usefulness" of the machines in question (/thread-2503.html)

---

Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

I really wonder if switch mode power supplies are more prone to failing when running off a 120V source? I have a 220V source on my side, and I have lots of machines spread across a wide age spectrum and I've never seen a power supply fail, touch wood! (It's maybe worth researching!) I am aware that Trippynet does have an Indigo2 Max IMPACT with a flaky power supply and in the UK they do have a 220V power supply, but there's always the exception to the rule.

As a collector, in my opinion, excellent condition retired "known to be working" SGI's are worth just as much as ones that are in daily use.

In my opinion, the only SGI's worth keeping running well into the future are the likes of the Onyx, Onyx2, Tezro and maybe the Octane2! (The rest don't offer good enough performance to warrant the effort, in my opinion, especially not machines such as Indigo2's.) [So I'm simply going to retire them as working machines!]

Here's my reasoning: (This is just my opinion, so don't let my reasoning detour you, as your performance needs may differ!!!)

According to the graphs, running anything less than a 195MHz R10K is going to be nothing more than frustrating!!!! (Even the 195MHz R10K is frustrating to use for general tasks, so R5000's, R4000's and R3000's are utterly useless in my opinion!)

So if you're looking to get into SGI, I'd suggest that you rather save and purchase one of the more powerful machines, than have a whole bunch of slower machines! (Unless you want them for collection purposes of course!)





I've made good use of all of my machines and I've thoroughly enjoyed them in the process. (I suggest others to do the same, while they still can, as that nothing lasts forever except hopefully your experience!)

The Onyx dual R8000 that I'm purchasing from Ian early next year will more than likely be my last SGI purchase. (The R8000 really interests me, although I'm not expecting it to have good performance, not for general tasks anyway!)

In my case, it's now time to move on to a complete Apple 1 replica, ZX Spectrum, NEXT cube and the like!

---

RE: Failing power supplies and the "usefulness" of the machines in question - Trippynet - 10-10-2020

Well, I have two flaky PSUs for my Indigo2 to be honest! That's part of the reason it hasn't been used all that much this year.

Personally though, the performance isn't what interests me about my SGI machines. If it was, I'd use my Fuel more regularly as it is a much quicker and more capable system than the Indigo2, yet I found that my Indigo2 was the system I found myself powering up far more regularly (when it worked properly). Simply I guess because I have much more of a soft-spot for it. I might look at getting another PSU for it, but I'd definitely be interested if we ever manage to get "new" built PSUs for them because anything I buy will be 20-25 years old now.

Still, it depends what you use it for. I don't really do much 3D rendering work on my SGI machines, so the poorer performance of the Indigo2 isn't an issue here. On the occasions I do fancy something requiring more grunt, that's what my Fuel is for.

---

RE: Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

@Trippynet I respect your opinion!

With regards to your power supplies, do you know where they came from, as they may have been running from a 120V source for most of their lives.

My Indigo2 Max IMPACT came from Germany, where they also have a 220V source.

In my reasoning, it's possibly easier (or less strenuous) for a switch mode power supply to develop a high output current, when its input voltage is higher. (This is just a thought, so please don't crucify me!!!)

---

RE: Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

When operating from a 120V source, the input rectifier has to operate as a voltage doubler, adding extra caps into the circuit.

---

RE: Failing power supplies and the "usefulness" of the machines in question - Trippynet - 10-10-2020

Both PSUs came from Ian - one was the original one when I bought the Indigo2 from him, the second was bought later when the original started acting up. Unfortunately I don't know what supply they originally used. It's likely a 220v source as they probably originated either in the UK or Europe, but not 100% certain of course.

Edit: Should add that I've also had a NIDEC PSU in an Indy fail on me many years back. This was definitely a UK system that had only been used on 220v, I acquired it from the University that originally owned them when they were being replaced. Of course, the NIDEC supplies don't have the most solid of reputations...

---

RE: Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

Input rectifier stage

AC, half-wave and full-wave rectified signals
If the SMPS has an AC input, then the first stage is to convert the input to DC. This is called rectification. An SMPS with a DC input does not require this stage. In some power supplies (mostly computer ATX power supplies), the rectifier circuit can be configured as a voltage doubler by the addition of a switch operated either manually or automatically. This feature permits operation from power sources that are normally at 115 V or at 230 V. The rectifier produces an unregulated DC voltage which is then sent to a large filter capacitor. The current drawn from the mains supply by this rectifier circuit occurs in short pulses around the AC voltage peaks. These pulses have significant high frequency energy which reduces the power factor. To correct for this, many newer SMPS will use a special PFC circuit to make the input current follow the sinusoidal shape of the AC input voltage, correcting the power factor. Power supplies that use active PFC usually are auto-ranging, supporting input voltages from ~100 VAC – 250 VAC, with no input voltage selector switch.

Here's some info on voltage doublers, and how they are affected by a heavy load:

https://ec.kemet.com/blog/voltage-doublers/

Im not proposing that this is the main cause of failures, but it could be a contributing factor.

More advanced SMPS feature protection circuitry which monitors the output voltage with reference to a reference voltage by means of a circuit containing an optocoupler for isolation purposes. @Trippynet, if your power supply was recapped and the FET's were also replaced, it could be the protection circuitry that's not calibrated properly.

---

RE: Failing power supplies and the "usefulness" of the machines in question - weblacky - 10-10-2020

While I’m not a professional on this subject, I have been studying for nearly two years on PSU basics to prepare to repair these things. 

I think you have a misunderstanding on about how a PSU works in relation to its main failing components- capacitors.  As you know aging caps fail in both the ability to store the same amount of energy and the ability to tolerate/handle their rated voltage ripple (they let it pass through more than absorb it as time goes on).  Along with chemical changes of course. 

Now based on ohm’s law, you are right about higher voltage drawing less current during normal applications.  However, that’s not really the exact observation with switching PSUs.  It’s really based on the fact that high-frequency AC is easier to switch effectively (less on-time, longer switch intervals between on), that’s what the main FET does before the transformer.  It’s chopping up the rectified AC , which are still DC humps that go from 0v to the running filter cap voltage (less as cap charges).  On 120v systems the filter caps run about 160v.  On 220v they run about 300v. 

Everything after the transformer is really the same values on both electrical systems.  At the end of the transformer are much lower DC values and the main management IC will change frequency of the FET based on voltage drop from current load (ohms law) on feedback (lines, optoisolator, small transformer, etc).

So while I’ve just explained a few benefits, I hope you noticed the drawbacks - higher filter cap voltage.

Yeah, the filter caps are huge and some PSUs that run just 220v have only a single cap instead of two (unneeded during normal lifespan due to increased efficiency). 

But you’re running your primary side filter caps near their intended rating (say 400v rating).  As the cap degrades you all on 220v will hit the cap’s limits much faster then us on 120v.  That cap effectively bakes in much more margin for us on 120v because it has to be rated so high for you on 220v. 

As the cap ages and stops being such a great filter, that’s noise then is transmitted down the line and stressing all your filter caps, FET, and your secondary output diodes (because less primary filtering and AC back feed protection).  Then either short or PSU protect notices the out of tolerance situation and keeps kicking in.  PSU lifespan reached. 


In short, 220v works the PSU components at a higher voltage , as they age they fall out of tolerance on 220v long before they’d have issues on a 120v system. Exceptions do happen of course. 


Also the biggest issues is actually lack of usage (I’m guilty of this as well).


The military has a specification that says electronics in storage longer than 2 years accelerate capacitor degradation.  If the PSU is used regularly the caps can do some amount of reforming from slight damage.  An unused cap (after a few years in storage) may actually be in worse shape then a used cap for the same period.  The first couple years don’t really matter, but it accelerates after a few years.  So electronics that are not regularly powered on a for a little while, have chemical cap degradation exceed that of normal use.  It rots in storage!


As part of a wide initiative, after PSU rebuilds I plan to hook up my collection with a schedule they they are booted once a year for a few hours each.  I’ve kept them off for so long that this was the real issue. And many of the people recently coming into the forum had the same thing - worked 5 years ago - just now tried it - problems.  A 10 year old PSU, put in storage for 5 additional years, is worse off than a 15 year old PSU being used regularly.

That’s likely more the observation you’ve seen, casual  use of your systems has actually preserved them.  I didn’t know this fact until about last year, horrified to learn you’ve done something wrong you thought was helping.  I wasn’t, I should have been turning my stations on every now and then to slow degradation. 

Old PSUs from storage much more likely to fail on 220v because their age doesn’t allow them tolerate the higher working primary side voltage.  In the land of old PSUs, the 120v fares a better chance of operating a NOS PSU.

---

RE: Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

What I took from what I read, is that because these power supplies are able to handle both 120V and 220V, they make use of a voltage doubler in the input rectifier stage when operating at 120V to get the voltage supplied to the filter cap, feeding the oscillator on the input side of the transformer to equal what it would if a 220V was being used, requiring the introduction of additional capacitors into the circuit.

A fixed 120V power supply would obviously have a lower voltage (160V) across the filter cap.

Please correct me if I'm wrong.

---

RE: Failing power supplies and the "usefulness" of the machines in question - weblacky - 10-10-2020

I can only find paid references so sorry, no links.

No, that’s an incorrect assumption of a design decision. You are likely correct on PSUs that have a manual switch on the back to switch between the two standards. The switch engages a doubling circuit on 120v side so the PWM frequency is tighter in range, as if you were on 220v for simplicity.

However that’s not what happens on modern auto ranging power supplies (most SGI PSUs don’t have a voltage selector switch) The primary filter caps don’t operate at the same voltage between the two systems. You see this a lot in like TV PSU repair videos on YouTube where the PSUs are autoranging. 120v system the DC voltage across the legs of a primary filter caps is roughly 160v, the same measurement on, 220v system are at least 300v. I’ve seen 356v before.

It’s actually easier (cheaper) now to make autoranging PSUs now, as they just vary the FET frequency in a greater range. Higher voltage means slower switching, low voltage means faster FET switching.

I only have paid literature that told me this info. But if you watch enough YouTube repair videos you’ll often see people detect 165v on US systems and 350v on foreign channels across primary cap leads. That’s the only reenforcing info I’ve found.

---

RE: Failing power supplies and the "usefulness" of the machines in question - Irinikus - 10-10-2020

Cool, thank for the info!

---