designing the ideal ATX power supply replacement for vintage Macs

[zigzagjoe]

Primary fuse should match the output rating of whatever AC->DC 5v can provide + tiny overhead. I suggest that rather than going for the utter maximum the original can provide, instead design for a practical maximum (all slots populated, a couple drives, accelerator, etc.) as to save costs on requiring a 300 watt 5v PSU. This part of the "PSU" design should be sized to taste anyways - a IIci won't have the same power demands as an IIfx or Quadra 950, for example.

The original PSU fan needs to be preserved or replaced to cover cooling needs. Likely can just provide a JST connector (if the original fans use a connector) and/or a header that the fan leads can be soldered to off the +12v.

Yes, U1 is an abstraction for the DC-DC module(s) or circuits for +- 12 generation.

It is likely that finding a DC-DC that can do enough on the +12 will be well overkill on the -12. So two separate circuits/modules may be required. For cost effectiveness, I'd suggest finding reference DC-DC designs with known parameters (often found in datasheets for switching ICs) and copying that, rather than buying a module. Reputable good quality DC DC modules = expensive, and vendor lock-in.

Fuse on the 12v isn't essential, as AC/DC-DC switching regulators typically have built in over current/short circuit and over temp protection as a minimum. At the amperage we're talking on the 12v rail, regulator will shut down due to OCP on a short before a fuse could blow.

Common negative / ground: Any DC/DC happening inside the PSU needs to share a single common negative output as the main +5V PSU.

Some DC-DC designs provide a new common output that has a voltage differential as compared to the input common. This can cause problems if you join the input common to the secondary DC-DC's output common. Shouldn't be a big problem to deal with, but it is something to be aware of.

As a simple example to think of: you could get a -12v voltage differential by reversing polarities: +12v as -/common, -/common as your +. However, if you try to feed your "-12v" to any IC or other circuit referencing earth ground or the input common, you'd end up shorting +12v, which leads to fireworks.

Chassis ground off the AC input is not connected to PSU negative output inside the casing. This grounding is so there is ultimately the chassis is always grounded and you can't end up with a case where chassis becomes energized with +120vAC or +12vDC etc. Also helps to supress EMI. However, chassis ground doesn't see current from any AC->DC: if current is flowing into chassis ground, this is a Bad Thing and represents a shock hazard. It'll trip GFCI also.

Conclusion: PSU negative output to be kept seperated from chassis ground.

BTW: I am by no means an expert in this stuff, just have done quite a bit of tinkering. If I'm wrong in any of this, folks, please chime in to correct me

 

[cheesestraws]

Some DC-DC designs provide a new common output that has a voltage differential as compared to the input common.

The magic words I used to find converters that do not assume a common ground is 'isolated dc/dc converter', fwiw.

 

[bigmessowires]

Thanks. I spent the morning on this. A few more things:

1. I've struggled to find a reasonably-priced 5V regulated supply with sufficient amps and that will fit inside a Mac IIci PSU enclosure (roughly 150x140x85mm). See here for options: https://www.digikey.com/en/products/filter/ac-dc-converters/133 Most of the higher-power options are long and narrow, which is a bad fit for the Mac PSU enclosure. If you want to match the 110 to 130 watt output power rating of the original PSU and still fit inside the enclosure then you'll need to spend at least $100. You need to dial back to a 5V output rating of 18A (90W) before you can find anything reasonably-priced that fits. Allowing say 30W for the +/-12V converter output, that only leaves 60W or 12A for 5V, which is substantially less than provided by the stock IIci or IIfx power supplies. That's a big enough shortfall that I'm not sure this even makes sense anymore. But something like the 18A Mean Well LRS-100-5 at $22 could work if you can accept the lower amps.

Mean Well RPS-160-5 is 30A and would fit, but it's $50. It has 5V standby and PS_ON functionality already built-in, which is nice. But it's getting closer to the price of a SFX DC-to-DC ATX PSU, a much simpler all-in-one solution.

2. I found no 5V to +/- 12V modules that could supply enough +12V current. I only found one dual-output switching regulator IC that could handle that much current at 12V: the Analog Devices LTC3633, but it's not cheap and that's only for the chip itself. I did enough reading to conclude that designing your own high-power switching regulator circuit is not an easy DIY project even if you have a reference design from a datasheet. It's definitely beyond my skill level, but if somebody with more experience in power circuits wants to take a stab at it, please do. It's something that looks easy enough to put together, but difficult to actually do well enough to achieve a good result. This discussion gives you a taste of all that's involved: component and cap selection, ESR considerations, physical component placement and trace topology to minimize loops, heat dissipation considerations, heat sinking, noise considerations, output ripple, ferrite beads, ??? It's much more than just having a schematic.

It is likely that finding a DC-DC that can do enough on the +12 will be well overkill on the -12. So two separate circuits/modules may be required.

3. Yes. I wouldn't want to try designing my own though, for the same reason I wouldn't try it with a dual-output design. Here is a 5V to 12V 3A boost converter all nicely sealed up in a big heat sink. It's $15.

4. With a single output 12V converter, you'd still need a separate -12V output. This is low-power enough that it's more reasonable DIY territory. Or use a cheap 5V to -12V module like this one. This can only do a few hundred milliamps, so it assumes that the prior analysis was correct that 1A for -12V is not actually needed by the Mac. Beware that the lower power dual-output +/- 12V converters do not support having more load on -12V than +12V, so they can't be used as -12V-only converters.

5. Finally, you probably still need to DIY something for the soft-power on, the main relay or FET, the fuses, etc. similar to @zigzagjoe's example schematic.

 

[bigmessowires]

According to a Kill-a-Watt meter, my "loaded" Mac IIci pulls about 55W from the wall electric outlet when first turned on, quickly dropping to 45W. That means it's using at most 11 amps from the 5V supply, because 5V times 11A is 55W. In reality some of those 55W from the wall are lost to inefficiency or used by the +12V and -12V supplies, so the actual 5V load is probably more like 9 amps. In that case the $22 Mean Well LRS-100-5 would work just fine.

@jmacz I'm curious if you have similar wall power wattage numbers for your loaded IIci and loaded IIfx?

It also occurs to me that with a DC-to-DC design, the less power the Mac pulls from 12V, the more would be available for 5V. So if the Mac isn't using the full 12V rating (which is likely), but is only using something like 15W from 12V, then the LRS-100-5 could still output 15A on 5V assuming perfect efficiency, which is better than the 12A that I estimated earlier. And if the Mac doesn't use any 12V power, then the full 18A would be available for 5V.

 

[bigmessowires]

This whole conversation assumes that you already have a stock PSU that you can use as a donor, for its metal enclosure and the power plugs. But if you don't have any PSU at all, or you don't want to sacrifice your existing PSU, then this project becomes harder. You need to get an enclosure somewhere, or build one, and find a way to secure it inside the Mac so it won't pop loose when the computer is moved. And you need a way to mount the mains power plugs and the fan so that they'll align properly with the openings in the Mac's case.

 

[bigmessowires]

While I agree that going 5V to 12V makes more sense from an engineering standpoint, when you consider the availability and prices of actual components, I think going 12V to 5V will be far easier. There are plenty of 12V-only supplies with sufficient power that are small enough, and a monolithic 12V to 5V 20A module can be purchased fairly inexpensively.

small form factor Mean Well 12V 8.5A regulated supply, $17
small form factor Mornsun 12V 12.5A regulated supply, $20
small form factor Mean Well 12V 10A $21
12V to 5V 20A module with heat sink, $19

 

[jmacz]

@jmacz I'm curious if you have similar wall power wattage numbers for your loaded IIci and loaded IIfx?

I have a kill-a-watt here somewhere... let me see if I can check that as well and also monitor while it's running a benchmark or something. Got my hands full at the moment as I've got three laptops split open for some surgery along with their battery packs. So once I get these closed up, will see if I can run that test... probably tomorrow.

 

[jmacz]

Kill-a-watt results for my loaded IIci

• Right after turning it on: ~59W
• Steady State After Finder is Loaded: ~55W
• During MacBench Tests: ~56W

Will get back to you on my IIfx.

 

[jmacz]

Kill-a-watt results for my IIfx (with one accelerated video card - 24bits 1024x768)

• Right after turning it on: ~64W
• Steady State After Finder is Loaded: ~63W
• During MacBench Tests: ~64W

Results for IIfx with Two Accelerated Videocards (both at 24bits 1024x768)

• Right after turning it on: ~70W
• Steady State After Finder is Loaded: ~73W
• During MacBench Tests: ~74W

 

[bigmessowires]

Thanks! If we call it 75W max power delivered from the electric outlet, then your IIfx is using at most 15 amps from 5V. In truth I would guess it's probably more like 12A, since some of that 75W is lost in PSU inefficiency or used by the 12V and -12V rails.

If you can be satisfied with enough power for a IIfx + two accelerated video cards, then 12A@5V should be an easier number to reach than the 18A@5V printed on the stock PSU. That $22 Mean Well 5V supply would probably be fine. Or if you're using a 12V supply, probably any of those options around $20 that I mentioned before would be OK.

I also discussed this new PSU idea on my blog, and one commenter argued that using a 12V supply down-converted to 5V was superior to a 5V supply up-converted to 12V, in part because it would cause less 5V noise due to switching transients of 12V devices. https://www.bigmessowires.com/2023/...ulation-and-vintage-computers/#comment-266066

But at the end of the day, it's probably still more sensible to use a $90 SFX PSU with DC-to-DC regulation, instead of a $20 industrial PSU plus a $20 12V-to-5V module plus a hypothetical DIY board for -12V and standby power and soft power. And if your Mac power needs are more modest, then just use a cheap group-regulated ATX PSU and it should work OK.

 

[jmacz]

Just to provide context into "loaded" for my two computers tested above. Which might not be as loaded for others.

IIci

• DayStar PowerCache 50Mhz 68030 Accelerator
• SuperMac Spectrum 24 Series III at 24bit 1024x768
• Asante Ethernet Card
• Internal ZuluSCSI (with 5V power cable to motherboard)
• External ZuluSCSI (using SCSI termination power)
• AEKII keyboard and Glide Point Touchpad
• 32MB of Memory

IIfx

• SuperMac Spectrum 24 PDQ at 24bit 1024x768
• SuperMac Spectrum 24 Series IV at 24bit 1024x768
• 3Com Etherlink III Card
• Internal ZuluSCSI (with 5V power cable to motherboard)
• External ZuluSCSI (using SCSI termination power)
• AEKII keyboard and Glide Point Touchpad
• 128MB of Memory

 

[jmacz]

If you can be satisfied with enough power for a IIfx + two accelerated video cards, then 12A@5V should be an easier number to reach than the 18A@5V printed on the stock PSU. That $22 Mean Well 5V supply would probably be fine. Or if you're using a 12V supply, probably any of those options around $20 that I mentioned before would be OK.

Makes sense. Yeah, I don't think I will be adding anything else to this particular IIfx.

 

[superjer2000]

I would love to see data about your Mac's configuration, what ATX PSU you're using, and what your 5V supply voltage is when the system is running. I'm guessing that most people who've done an ATX mod either have lightly-loaded Macs where this isn't much of a problem, or they just never measured the voltage and didn't realize there was a problem.

I was actually a bit surprised at my measurement results. I run a number of Seasonic SSP-xxx-SUB PSUs (xxx - 250 or 300 W) after first finding them as a suitable replacement for SE/30s. In the SE/30 use-case I had typically measured very close to 5V but I may go back here again and check what I'm getting on my SE/30 with a 50mhz accelerator and Ethernet card. Ultimately though, I have absolutely no stability issues with these Seasonics in any of my SE/30s.

I have also adapted these PSUs to a IIci and IIsi and haven't had any issues so never bother measuring the voltages.

I did run some checks on my IIci which has a 68040 40mhz accelerator as well as a Nubus Ethernet card and the 2 card Mac286 board set (so fairly decked out). Again, I have had absolutely no stability issues with this machine (and Bolle had noted previously that the Carerra 040's can be finicky without a good PSU) so I was a bit surprised at the voltages (as measured at the HDD connector):

Case 5V 12V
A. IIci as described above 4.67 12.37
B. As A. with 3.3ohm resistor across 3.3V line 4.78 12.48
C. As B. but with 3.5" HD 4.78 12.48
D. As A. with 10ohm resistor across 3.3V line 4.78 12.48
E. As D. with 3.5" HD 4.79 12.40

I also tried adding a 20 ohm resistor across the 12V line but didn't see much difference in voltages so I settled on the 10ohm resistor across the 3.3V line. I had done this with a previous SE/30 mod at one point as well as the Seasonic PSU I use seems to indicate in the spec sheet that it should have a load on the 3.3V line but when using for SE/30 PSU replacements I hadn't noticed a difference whether the 3.3V line was loaded or not so I have generally just left it. The 10ohm (10 watt) ceramic resistor doesn't really get that warm so I just zip tied it to the metal wall of the PSU to act as a heat sink.

I checked my measurements against the floppy drive port and got similar measurements. At some point I would like to go back and measure directly at the PSU output, although I then pulled out a IIcx with a stock Astec PSU and at the HDD power connector I was getting just over 5.0v.
I will note that what the Seasonic is putting out is within the ATX spec (+/- 5% - so 4.75v is within tolerance). I'm not sure I'll go much further down the rabbit hole here as all of the systems I've retrofitted these PSUs into work fine for extended periods of time.

 

[jmacz]

@bigmessowires and I had an unrelated issue with the ATX PSU involving the sense lines prior to this thread, but once that was resolved, I think I might be the only one who had a real issue with the 5V line. One of my video cards would display garbage artifacts if the voltage dropped below around 4.85V. Above that it’s perfect. I haven’t seen any other issue.

That said, the stock PSUs seem to be putting out around 5.1-ish volts. And my new DC-DC SFX PSU is putting out 4.97-ish volts.