[dexter1]

You da Man, Jan-Jaap! Well done

Wicked to see a Professional Series alive and kicking...

[CB_HK]

Outstanding! Im so glad to see it’s back up and running. Really good job, Jan-Jaap!

Enjoy that beer

[Jacques]

Yay! Very good work! Well done.

Could you grab a video of it powering up, posting and booting? Not seen one of these in action before.

[jan-jaap]

Yay! Very good work! Well done.

Could you grab a video of it powering up, posting and booting? Not seen one of these in action before.

Maybe eventually, but right now it's rather complicated, I have to (cold) boot it with serial attached, then key in all the NVRAM values because the battery is dead and 'resetenv' doesn't work on these, re-initialize and boot it.

I'll take a skin panel to a paint shop today. The plastics are white, painted brown. It's chipped left and right and I want to know how to restore that without spray painting the whole thing because I would loose the logos. The pink top hat is very ugly so I may simply paint it teal, the colour matching the GT graphics.

[CB_HK]

Yay! Very good work! Well done.

Could you grab a video of it powering up, posting and booting? Not seen one of these in action before.

Maybe eventually, but right now it's rather complicated, I have to (cold) boot it with serial attached, then key in all the NVRAM values because the battery is dead and 'resetenv' doesn't work on these, re-initialize and boot it.

I'll take a skin panel to a paint shop today. The plastics are white, painted brown. It's chipped left and right and I want to know how to restore that without spray painting the whole thing because I would loose the logos. The pink top hat is very ugly so I may simply paint it teal, the colour matching the GT graphics.

Oh, that sounds pretty amazing. If it's not too much trouble, would you mind grabbing a few photos as you go? I'd love to see a bit of that restoration. I'm already imagining a super clean IRIS and I have to say it's pretty exciting!

[jan-jaap]

More surgery, with a happy end! Sorry, this is a lengthy post with a lot of photos.

A couple of years ago I replaced the original L&H Research "Mighty Mite" with a PowerOne SPM5 series because the original PSU made horrible coil whine noise under load. I decided to revisit the original PSU to see if this could be repaired.

First of all, the PSU "mostly" works. It puts out the correct voltages, and while the oscilloscope shows some noise on the output it wasn't all that bad. But the noise made it impossible to be in the same room as the computer which kind of defeats the purpose of having it.


L&H Research "Mighty Mite" PSU, open, light load. Voltages correct, no noise.

The noise was unmistakably coil whine. Now, there aren't all that many coils in the PSU. There are some in the input section to filter out crud from the mains:


L&H Research "Mighty Mite" PSU, input filter section and rectifier (under heat sink).

For a moment I thought I had found the problem when I realized the nylon screws holding the filter coils were pretty loose. But no luck. Should have known better, with clean mains power these things don't do much.

The only other "coils" are the transformers in the output sections. The big one is the main +5 VDC output, rated at a screwdriver-melting 200A. What you're looking at here: output filter capacitors, 3x 18000uF, transformer (2x), filter coils (2x), rectifier diodes (4x) and a snubber to suppress semiconductor switching noise. I know the popular opinion here is to simply replace all capacitors in case of trouble, but these still tested good. Blindly replacing the main capacitors in this PSU would have cost me over €200 as well.


L&H Research "Mighty Mite" PSU, 5VDC output section.

With the input section ruled out, the output transformer was next on my shortlist of suspects. In short, a switch mode PSU (SMPS) works like this: rectify AC mains, chop at much higher frequency, run this through a transformer and rectify again. The much higher frequency makes the transformer more efficient (read: much less bulky) and requires less filter capacity in the output.


L&H Research "Mighty Mite" PSU, -5VDC, +/-12VDC sections, 5VDC control and chopper.

Some more of the guts of the PSU. Like I said, the PSU chops the rectified mains voltage. For the main 5VDC output, this is done by the 4 large TO-3 transistors in the right PCB. Since we couldn't find much wrong with the output components we turned our attention to these 2N6678 NPN power transistors.


L&H Research "Mighty Mite" PSU, chopper.

These transistors come out easy (no soldering required).


L&H Research "Mighty Mite" PSU, chopper (continued).

A better look at the way the 2N6678 transistors are mounted. We did some measurements outside the PSU on these transistors, but they require (expect) a lot of current under normal operation and one was destroyed in the process. They did have a much higher than expected base-emitter voltage drop though. That would limit the amount of current going into the transformer, causing the field inside the transformer to 'collapse' continuously, causing the coil noise. Or so was the theory at least, now I had no choice but to replace them.


L&H Research "Mighty Mite" PSU, new chopper transistors installed.

Unfortunately, the original Motorola parts are no longer available, so I settled for some no-name parts. NOS from 15 years ago.

With the new parts installed, the PSU didn't explode, and still put out correct voltages. So I spent several hours to re-install the PSU in the Professional IRIS "Twin Tower" chassis, but when I powered up the system ... it was still making noise. This time not a coil whine, but more like a buzzer. A single frequency. The oscilloscope showed ~ 0.6V TT AC ripple on the +5 VDC rail at a frequency of 2500Hz, probably also the frequency of the beep I was hearing. That was a new one, before the PSU was noisy but at least it put out the stable, correct voltages.


4D/70 AC ripple on +5 VDC rail

That's when I put the whole thing aside for a week, disappointed. However, at some point I realized the 2500Hz frequency wasn't the chopper frequency and I was looking at an oscillation. Now, there's a feedback loop in this system: there are very thick wires going from the PSU to the backplane of the system, and 'sense' wires going back. The sense wires allow the PSU to compensate for voltage drop over the wires to the backplane. On a whim, I removed the sense feedback from the PSU. Success!


4D/70 AC ripple on +5 VDC rail, sense feedback eliminated

Much better! And no more noise! Now the only noise is the remaining chopper noise. The period is roughly 12.5us -- 80KHz. I don't know why it's oscillating. Maybe I damaged something on the control circuitry? Maybe the replacement 2N6678 parts are not close enough to the Motorola parts (but then why does it work without sense feedback?)


4D/70 +5VDC as measured on the PSU terminals.


4D/70 +5VDC as measured on the backplane.

As you can see, despite the massive cables running from PSU to backplane, there's a drop of 50mV between PSU and backplane.

I thought a bit about it, and decided to leave the sense disconnected for the moment. The PSU is officially specified for +5VDC, 200A, so I adjusted the output voltage to 5.05V which makes exactly 5.00 VDC on the backplane. Theoretically, if I would add a lot of hardware to the system, the extra load would cause additional voltage drop over the PSU cables etc, but this just isn't going to happen.

So, there we are. The patient lives! This is one of the very few surviving Professional IRISes. It was in very rough shape when I first got it. Normally I would have passed on it, but you don't have that luxury when it comes to these. So far, I've had to deal with these issues:

1. Chassis blower dead. I replaced the motors of both. They are centrifugal blowers.

2. Probably related: GT graphics raster managers dead. I replaced them with RM1 boards from a PowerSeries GTX.

3. IRIX kernel died with a bus error when starting graphics. Replaced GM board no result. Took a long time before I realized I was looking at a VME bus error and the backplane was jumpered incorrectly. Which moron did that?

4. The PSU. First replaced, now repaired.

Still need to:

1. Reassemble the thing. It has been in pieces for years, hope I didn't loose any of the screws etc.

2. Mod the IP4 CPU board battery backed RAM with a CR2032. This is slightly more complicated than your average Indy because there are two batteries inside and nobody did this I think so I may have to sacrifice one before I know how to do this best. Currently, I have to boot the system with a serial console attached and re-enter correct NVRAM variables every time the system is powered on.

3. Fix the skins. The Professional IRIS skins are white plastic painted brown, not solid brown plastic. It's chipped left and right. The hinges of the doors could use some TLC as well.


To be continued...

Finally, a big shout out to my buddy Casper! I probably wouldn't have been able to fix this one without him.

[weblacky]

Yo, So I'm no expert and you've had a great deal of success with this, but the coil whine does bother me, this is something that should be addressable and you shouldn't have to live with it. So let me ask the basics...you have a lot off neat meters and instruments. So I'm going to make assumptions that you have...well...everything at your disposal to test with.

1. Have you tested the PSU coils in circuit with either an LCR or a curve tracer to compare them to both each other and to another sister/duplicate PSU you may have? How do they compare? Do they read/compare similar?

2. Have you tried low-voltage Curve tracing (200mV-300mV range) on the Filter Cap groups between "like" PSUs to confirm similarity? I feel like the whine could be a shift in the physical of the caps (even as a collective). Perhaps you measure them as a group first, then if you see a big difference, you can separate one of the pair and measure outside to see if one is REALLY worse than the other.

3. The feedback circuit also bothers me, it's like it's overshooting and undershooting via correcting. Doing a simple online search (my mind goes to Buck COnverters on regulators here) it's this indicative of a failure of a capacitor on the feedback line? I know I've been told that when ESR changes substantially on the feedback circuit it causes this kind of wide under-over shooting. So could you check the feedback circuit you disconnected with a low-voltage curve tracer (again 200mV-300mV) to isolate the cap outside of semi-conductance to check between like-PSUs. I think you'll find you have a change in resistance on the feedback circuit. Assuming there is a management chip (and not a TON of passives doing this) it shouldn't be too hard to trace back from Management IC to find capacitance on the feedback line. This is the only thing I know about output oscillation because of the feedback control on this kind of stuff (limited knowledge).

Great posts and great work, I hope you make more positive progress on this and can continue to dedicate your valuable time to writing these posts up for us!

[jan-jaap]

Couple of clarifications:

1. I only have a single L&H Research "Mighty Mite" PSU. It belongs to the Professional IRIS. So I don't have a working reference.

2. When I got this system 10 years ago, the PSU already made the coil whine noises. So I replaced the original PSU with a PowerOne SPM5 model. I did not adapted the wiring harness to supply 'sense' feedback to the PowrOne PSU either.

3. The original L&H PSU would put out correct voltages, under load, even though it would make unbearable coil whine noises.

4. With the new 2N6678 power transistors, the coil whine was gone, but an oscillation on the +5V rail had appeared (and corresponding 'beep' sound from the PSU)

5. With 'sense' feedback disconnected, the oscillation (and the beep) stops. The L&H puts out correct voltages, no noise on the rails beyond what can be expected and basically works. I understand the need for sense feedback in the design here, but since the system configuration is static I think I can get away with leaving it disconnected.

I do have some tools, like an oscilloscope (with isolated probes), a multimeter, a capacitor meter with ESR meter, and have access to quite a few tools from the lab at work. But this is not straightforward to troubleshoot. In the 4th picture of my previous post you can see the -5, +/-12V modules and the 5V control logic with the faulty 2N6678 chopper transistors. Where more modern designs use a single chip like a UC3844 and an optocoupler in the feedback loop, this uses a whole lot of analogue components to do the same. There's half a dozen adjustments (potmeters) on the side. The original 2N6678 transistors were probably matched -- somebody wrote '329' on all of them. But matched for what?

As it is right now, I have a functional PSU, for as long as I leave the sense disconnected. Sure, it would be nice to have everything work 100%, but without schematics or some sort of tech manual I'm not sure it's realistic. Also, it takes a lot of time to remove this PSU from the system and wire everything back up. I do have a set of Motorola 2N6678 transistors on the way -- I did find some in the end. Will they be genuine? Who knows. It was a lot of 4 and they were cheap, if they were counterfeit it didn't bring much. Hopefully it's just old stock. When I get back in I'll take this module to the lab to inspect it for bad soldering joints and can eventually replace any small caps in there. But I'm hesitant to mess with all this analogue circuitry without a way to calibrate it when I'm done.

Next, I'm going to turn my attention to the PowerOne PSUs of the PowerSeries. I have several working and several dead specimen. A much more structured approach will be possible here.

[Geoman]

What an interesting read and a fine work done!

As for the "Dallas-sandwich-chips", 2 days ago I have done the deed and excavated the expoxy whith both coin cells in it with a dremel and had the bare PCB. The quarz oscillator got damaged in the process, so I soldered a new one in. Then I attached 2 cables with a coin cell holder. In the end the thing looks very clean, much cleaner than in the image. I dared to do this, because I had 2 times success with my Indigo2s (see picture).

So If my Crimson works fine with this hack, I can offer to repeat the process on your Dallas chip, now that I know whats inside the layer of epoxy.

[jan-jaap]

What an interesting read and a fine work done!

As for the "Dallas-sandwich-chips", 2 days ago I have done the deed and excavated the expoxy whith both coin cells in it with a dremel and had the bare PCB. The quarz oscillator got damaged in the process, so I soldered a new one in. Then I attached 2 cables with a coin cell holder. In the end the thing looks very clean, much cleaner than in the image. I dared to do this, because I had 2 times success with my Indigo2s (see picture).

So If my Crimson works fine with this hack, I can offer to repeat the process on your Dallas chip, now that I know whats inside the layer of epoxy.

The NVRAM has two parts: the RTC with embedded batteries and an SRAM on top of it (with the SONY logo).

Where does the SRAM go now?