Asistance required on troubleshooting Corsair CX500

[KidD01]

Hello,

I just got a Corsair CX500 which got a blown primary capacitor (a Chemicon SMQ 220uF/400V), judging from the capacitor appearance it seems there are something wrong besides the capacitor as the plastic on capacitor body had some traces as if it was overheat.

Would be grateful if anyone can provide information which parts to check.

Thank you.

[goodpsusearch]

Re: Asistance required on troubleshooting Corsair CX500

Is the fuse blown?

[sam_sam_sam]

Re: Asistance required on troubleshooting Corsair CX500

If the fuse is blown check the bridge rectifier for a short circuit

[Agent24]

Re: Asistance required on troubleshooting Corsair CX500

If the bridge caused the capacitor to fail, it could have killed the PFC MOSFETs and controller too.

Or the PFC circuit could have a fault which caused an over-voltage condition on the capacitor.

It's still possible for primary capacitors to fail by themselves, too. Just not as often as the secondaries.

[KidD01]

Re: Asistance required on troubleshooting Corsair CX500

Thank you for the response, I'll try to check based on the advise given.

I'll try to update if there are things come up.

[momaka]

Re: Asistance required on troubleshooting Corsair CX500

Probably an APFC failure, causing the primary cap to blow (and perhaps the fuse + other components too?). Time to heat up your iron and start pulling parts.

[MHTSOS]

Re: Asistance required on troubleshooting Corsair CX500

Remove the heatsinks with the mosfets on the primary side and check them. Most probably some of them, if not all, are shorted.

[linuxguru]

Re: Asistance required on troubleshooting Corsair CX500

Same issue - Corsair CX500 with the primary Panasonic 220uF/400v blown. I pulled it, removed the APFC heatsink and checked the 2 MOSFETs (13N50P - 13A, 500v) and the rectifier (STTH8S06D - 8A, 600V). One MOSFET and the rectifier were shorted. I removed the dead MOSFET, and replaced the shorted rectifier with a 5A. 600V ultra-fast rectifier. Also replaced a shorted 1N5406 3A rectifier in series in the same circuit. Fuse was blown, replaced with the same rating (T6.3). Primary bridge rectifier was fine. Replaced the primary cap with a Hitano 220uF/400v/85c, which fit in the space available - my 105c caps are a bit too large.

Powered up, and +5Vsb came up and was stable. I then shorted PSON to ground, and obtained rail voltages for about 1 minute, and then the primary cap vented again with a lot of smoke.

Back to the bench, so I removed the APFC switching FET and the ultra-fast rectifier from the heatsink. Checked the remaining power devices, bridge rectifier and fuse - all were intact. Checked the charge paths to the primary cap, and all was fine. Replaced the primary cap with a pulled 220uF/400v (just for testing with APFC bypassed).

Powered back on, and +5Vsb is present and stable. The primary did not vent for as long as I left it powered on - it appears that it vented earlier due to overvoltage from an APFC fault, so with APFC bypassed, it did not vent.

However, the main switcher did not turn on with PSON shorted to ground. It just sits there with a stable +5Vsb. It's possible that I cooked the main switching devices in the first phase, when it worked for about a minute before the primary cap vented - the primary overvoltage may have avalanched the main devices. I'll check that next and replace if needed.

Before I start pulling heatsinks and FETs again, I just need some tips on whether I need to bypass some other protection/monitoring circuit to get it to start with APFC bypassed. As noted above, +5Vsb is fine and the primary cap is not distressed or venting now with APFC bypassed, hence primary B+ is present and presumably within spec.

Secondary caps are the OEM SamXon, and not yet replaced. I wanted to have stable rails before replacing all the secondary caps, which look OK - not bulged or vented. How should I proceed next?

[momaka]

Re: Asistance required on troubleshooting Corsair CX500

*sigh*

Another of these POS APFC circuits causing trouble again.

My #1 tip would be to use a 450V-rated primary cap instead of 400V (or maybe 420V if you don't have any 450V ones or they are too big to fit.) This is really a band-aid to the problem, though... which so far I haven't actually determined if it's due to a fault with these CM6800/6802 PWM+PFC driver chips overshooting the APFC voltage too far or something else, as I've seen this problem across various different makes and models of PSUs. In any case, the problem also seems to be related to manufacturers using caps that are too small for the application - both in terms of voltage rating and rated ripple current... hence the suggestion with going with a 450V cap and of largest capacity possible that will fit.

That aside, what's your line voltage? 110-120V or 220-240V AC? If it's the former, you probably won't be able to run the main PS, even if you did bypass the APFC voltage feedback going to the CM6800/6802 - not without a 115<-->230 V AC transformer. Now, with 220-240V AC mains, you'll have about 320-340V DC roughly across the main cap. So with a bypassed APFC voltage feedback, you should be able to get the main PS to run... though you may not be able to reach full power output on the PSU.

I just need some tips on whether I need to bypass some other protection/monitoring circuit to get it to start with APFC bypassed.

I actually did try to convert an APFC-enahbled Enermax PSU to a non-APFC PSU, as it had a blown primary cap just like your PSU (and blown NTC thermistor, but no other bad parts, luckily.) You can read more about how I did that here (and this hould answer your question how to bypass the APFC voltage feedback circuit):
https://www.badcaps.net/forum/showpo...9&postcount=10
https://www.badcaps.net/forum/showpo...4&postcount=11
https://www.badcaps.net/forum/showpo...8&postcount=14

One note I have here is that I never really went back to fully update that thread again... but I tried using the above Enermax PSU in a more power-hungry PC with an OC'ed C2Q and GTX560Ti, and the PSU would shut down when I tried to fully load both the CPU and GPU to the maximum. IIRC, ~200-220 Watts is all I could pull (measured at the wall) before it would shut down.

Thus, you should be able to disable the APFC - at least if you have a CM6800 or CM68002 chip or similar in your PSU. Just keep in mind that the power output of the PSU may be reduced. Actually, depending on the primary side design of your PSU, it is still possible that the PSU may not run. But at least it should try, once the APFC voltage feedback is disabled on the CM6800/6802.

[linuxguru]

Re: Asistance required on troubleshooting Corsair CX500

Thanks - I'm in a 220-240 VAC region. As noted earlier, +5Vsb is working now, just the main converter doesn't start. I'm OK even if it delivers 50% power reliably, so will take a look at your adventures with the Enermax.

Edit: I have the SCK054 thermistor as well, but just perfunctorily checked that it had continuity and reinstalled it - perhaps it had gone high-Z or open as well, will check.

[Agent24]

Re: Asistance required on troubleshooting Corsair CX500

*sigh*

Another of these POS APFC circuits causing trouble again.

My #1 tip would be to use a 450V-rated primary cap instead of 400V (or maybe 420V if you don't have any 450V ones or they are too big to fit.) This is really a band-aid to the problem, though... which so far I haven't actually determined if it's due to a fault with these CM6800/6802 PWM+PFC driver chips overshooting the APFC voltage too far or something else, as I've seen this problem across various different makes and models of PSUs. In any case, the problem also seems to be related to manufacturers using caps that are too small for the application - both in terms of voltage rating and rated ripple current... hence the suggestion with going with a 450V cap and of largest capacity possible that will fit.

Has anyone tried scoping the APFC circuitry to see if there are voltage spikes etc?
Even comparing between a working and faulty version of the same model?
I guess you'd need a pretty fast, high-voltage differential probe though... sounds expensive.

[momaka]

Re: Asistance required on troubleshooting Corsair CX500

Has anyone tried scoping the APFC circuitry to see if there are voltage spikes etc?
Even comparing between a working and faulty version of the same model?
I guess you'd need a pretty fast, high-voltage differential probe though... sounds expensive.

I don't own a scope, unfortunately. It's one of those tools that I kind of want and needed a few times, but really nothing that I see myself using that often, at least given the stuff I repair. Thus I can't really seem to justify the cost of getting one (yet) - even the cheap ones. For the few audio applications that I wanted to look at the signal, I made it by with Visual Analyzer and an old laptop with Line Input.

Yeah, it would be interesting to scope the PSU to see what is going on. However, I still think the issue has something to do with undersized primary caps for the application. In particular, all the expensive PSU's I've seen in the past that come with 10 year warranty always employed much bigger primary caps, and often at least 2 of them. For example, one Silverstone 600W unit I checked in the past had 2x 390 uF caps at either 420 or 450V, I can't remember - but they were huge. In contrast, cheaper units like this Thermaltake M850W that is rated for 850 Watts, only had a single 470 uF 400V cap on the primary. And guess what - it was blown! (Then again, that is also a CWT-built unit, so maybe not a coincidence after all?) Either way, I "sort-of" fixed that one by putting a big 500 uF 450V cap in it (sort of, because that cap is too big and I have the fan bolted on the outside of the case, lol.) I've been using it for a few years now as a bench test PSU for high power GPUs, and so far it has held up OK. So whatever caused the original Panasonic cap in it to blow isn't causing an issue now with a higher-voltage cap.

But IDK, I think it's also still possible that something could be going out-of-spec with these CM6800/6802 controllers, causing the APFC to go haywire after a number of years. What's your thoughts on that?

[Agent24]

Re: Asistance required on troubleshooting Corsair CX500

I don't own a scope, unfortunately. It's one of those tools that I kind of want and needed a few times, but really nothing that I see myself using that often, at least given the stuff I repair. Thus I can't really seem to justify the cost of getting one (yet) - even the cheap ones. For the few audio applications that I wanted to look at the signal, I made it by with Visual Analyzer and an old laptop with Line Input.

Second hand CROs are pretty cheap, but who can say it will be useful in this scenario, especially if there were random spikes or such.

Yeah, it would be interesting to scope the PSU to see what is going on. However, I still think the issue has something to do with undersized primary caps for the application. In particular, all the expensive PSU's I've seen in the past that come with 10 year warranty always employed much bigger primary caps, and often at least 2 of them. For example, one Silverstone 600W unit I checked in the past had 2x 390 uF caps at either 420 or 450V, I can't remember - but they were huge. In contrast, cheaper units like this Thermaltake M850W that is rated for 850 Watts, only had a single 470 uF 400V cap on the primary. And guess what - it was blown! (Then again, that is also a CWT-built unit, so maybe not a coincidence after all?) Either way, I "sort-of" fixed that one by putting a big 500 uF 450V cap in it (sort of, because that cap is too big and I have the fan bolted on the outside of the case, lol.) I've been using it for a few years now as a bench test PSU for high power GPUs, and so far it has held up OK. So whatever caused the original Panasonic cap in it to blow isn't causing an issue now with a higher-voltage cap.

You are quite possibly right. The primaries are always fairly expensive parts so I can imagine them rating them right down to minimum spec for the job, if they were tight on a budget. Why care if the cap blows early, as long as it lasts just over the warranty.

Another thing, are these capacitors Low ESR? APFC runs at higher frequency than just rectified mains, I'm sure. That should also shorten the life, in addition to running close to max voltage etc.

But IDK, I think it's also still possible that something could be going out-of-spec with these CM6800/6802 controllers, causing the APFC to go haywire after a number of years. What's your thoughts on that?

Not a lot, I don't think. I am not an APFC expert or anything. At the heart of it, it's a boost converter, but I know the real operation is more complicated, and varied.

Taking a more general approach, I think it would be odd for different implementations from different manufacturers to always fail the same way. It is possible I assume, for a circuit designed around the same chip to fail similarly if some other common failure items always cause a certain issue. The design of the CM6800 etc may be such that it always does X when support component Y goes bad.

This could be exacerbated if component Y was incorrectly specified in the reference design, or is often substituted with low quality options, for cost saving..


It *might* be a low-level flaw in the chip (Remember those Intel Sandy Bridge Southbridges where the SATA ports failed due to a silicon-level design defect?) though that seems quite unlikely, but who knows. I expect that would be hard to prove even if true.
(I guess you'd need to profile various different PSUs with that same IC, both working and faulty, try to understand what exactly is happening, and then de-cap many chips and look for a repeated failure in the same location?!?)


This is all speculation on my part, of course. My #1 suspect would be bare-minimum rated primary capacitors. Penny-pinching is a well known root cause of faults, as we all know.

Also, is the CM6800 a low end IC? Likely be used in budget designs with budget capacitors? That could make it seem like the chip is associated with failures when it is not really the problem.

[linuxguru]

Re: Asistance required on troubleshooting Corsair CX500

Another thing, are these capacitors Low ESR? APFC runs at higher frequency than just rectified mains, I'm sure. That should also shorten the life, in addition to running close to max voltage etc.
...
Also, is the CM6800 a low end IC? Likely be used in budget designs with budget capacitors? That could make it seem like the chip is associated with failures when it is not really the problem.

I'm seeing 85c GP caps in a lot of these PSUs, including Corsair, Coolermaster and so on. In some instances (Cooler Master MWE-series v2) there is space to install a larger diameter, higher-spec'ed cap, but in some (Corsair CX) there isn't space - one has to stay with 85c, and the same modest values like 220uF/400v.

The CM6800 is widely used because it is a combo PFC + PWM controller - it does everything required on the primary side, so it saves space as well as costs. However, it is used on higher-end designs also. It seems to survive these cap failures, so it's probably a robust design, just marginally implemented in-circuit in the lower-end designs.

[momaka]

Re: Asistance required on troubleshooting Corsair CX500

Another thing, are these capacitors Low ESR? APFC runs at higher frequency than just rectified mains, I'm sure.

Most of the time, no. Like linuxguru, I find 85C -rated parts quite frequently - especially on cheaper/budget units. 85C caps for sure means they are GP and not low ESR. And even when I see PSU with 105C -rated primaries, they are still usually GP series... though it does depend on who made the PSU and what its price point is. The better-made PSUs always use at least one very large or two still significantly large caps rated for 105C. As for the cheaper PSUs, it kind of varies a lot - sometimes it's 85C caps, other times it's 105C caps... and the capacity and voltage rating are all over the place too. Then there's the factor of Japanese vs. non-Japanese. Seems that the PSUs with non-Japanese primary caps typically have those bigger and rated for a higher voltage... but not always, again. As to whether it's better or not to have a bigger non-Japanese cap with higher voltage and capacitance rating than a smaller Japanese one, I can't tell.

In any case, like you said, seems that manufacturers just typically go with whatever part is the minimum needed to get the job done and last past the warranty, especially on cheaper PSUs. And it also seems that manufacturers really use whatever brand was best available at the time at the lowest price.

That should also shorten the life, in addition to running close to max voltage etc.

Yeah, caps running near their max voltage is never a good thing, so that definitely contributes too.

I am not an APFC expert or anything. At the heart of it, it's a boost converter, but I know the real operation is more complicated, and varied.

It is indeed a lot more complicated than just a simple boost converter, because if you think about it, to get near unity PFC, the current draw from the line should increase and decrease directly with the line voltage, like a resistive load would behave if connected across it. For example, if the line is 10V AC at one instant and you have a 1k resistor connected across it, then the current will be 10 mA. The next instant, when the line is at 50V, the current will be 50 mA. And when it's 120V, the current will be 120 mA. In other words, the input current drawn by the APFC circuit needs to vary with the line voltage. But the complicated part comes from the fact that the output of the APFC is connected to another power converter circuit, whose power draw varies as well. So the APFC not only has to mimic a resistive load on its input, but also constantly adjust it so that it meets the voltage and power draw requirements of the PS connected after it. I suspect this is what causes so much stress on caps. From what I remember from reading briefly reading on APFC circuits, the voltage on the input electrolytic cap tends to vary a lot, and its also under a lot of high-frequency ripple current. So I think those primary caps just cook themselves internally when they are not rated adequately for the task, and that's what eventually does them in. Either that, or the constantly-varying voltage spikes from the APFC perhaps eventually make them short-circuit internally.

In any case, I bet if the APFC circuit is replace with just a simple boost-converter circuit that completely ignored Power Factor and just keeps the voltage steady on the primary caps, I think they will last a lot longer.

Taking a more general approach, I think it would be odd for different implementations from different manufacturers to always fail the same way.

Good point!
Which I think concurs again with my suspicion that it's just down to cap choice (in terms of capacity, voltage, type, and brand.)

It *might* be a low-level flaw in the chip (Remember those Intel Sandy Bridge Southbridges where the SATA ports failed due to a silicon-level design defect?) though that seems quite unlikely, but who knows. I expect that would be hard to prove even if true.
(I guess you'd need to profile various different PSUs with that same IC, both working and faulty, try to understand what exactly is happening, and then de-cap many chips and look for a repeated failure in the same location?!?)

Yup, that's what I was thinking too - it would be a lot of work to determine how many PSUs with CM6800 have failed APFC caps vs. those with other chip... and also how much more or less popular is CM6800 compared to other chips (i.e. the population size of chip X vs chip Y vs chip Z.)

Also, is the CM6800 a low end IC?

Don't think so. I've seen it used in both budget and high-end PSUs. Don't think there are any cheap clones of the chip floating around either... and even if there are, it doesn't seem like PSUs that use them to have failures more often than ones with a genuine Champion-Micro one.

[goodpsusearch]

Re: Asistance required on troubleshooting Corsair CX500

At the company where I work as IT, one computer suddenly wouldn't post.

The fans would spin when the power button was pressed but the motherboard would not "beep" and load the OS.

I checked voltages with multi-meter and found them unstable (can't remember details cause it's an old incident). I replaced the psu with a spare and the computer was back up and running.

Then I took the faulty psu to check voltages again without load. Voltages appeared better without load but no equipment would post with it connected.

Then I opened the case and the visual examination showed nothing wrong. But when I touched the primary big capacitor with my finger it felt slightly bulging.

The cap tested with esr micro and uF scale of multimeter AND Chinese transistor/C/diode tester, all of them found it was totally open!

The psu is Corsair VS450, the CWT version, so CWT is the OEM again.

Several weeks later, a friend gave me for free another failed PSU, this time Corsair VS350, again the same CWT platform and once more the same failure mode. Fan spins, voltages barely on spec and primary capacitor totally failed open, capacity is 0uF!


The failed cap is Aishi 400V 150uF 85C on the VS350 and Aishi 400V 180uF 85C on the VS450.
I think the cap brand is crap. Please correct me if I am wrong.
I also think 400V is too low for APFC.
I also find ridiculous the 150uF/180uF ratings for units that actually do 350W/450W accordingly in reviews (and more in overload tests without blowing). The cap must run at 395V-400V to be able to deliver those watts, because, as we know Q=C.V so when the C is too low you have to compensate increasing the Voltage.
Finally, 85C is just wrong, given the stress that the APFC puts on those caps.

I already ordered and received some replacement primary caps for these psus and I plan to do a full recap and enhancement on them and then use them on my pc builds but unfortunately I postpone it al the time due to lack of time.

[linuxguru]

Re: Asistance required on troubleshooting Corsair CX500

There may not be enough space in the Corsair CX... units for a 450v/105c cap - choices are limited, but there's a Samxon HP(M) 220uF/450v/105c cap that will fit - it's in the tall and thin form-factor:

https://cdn.badcaps-static.com/pdfs/...b9a5d6b9ef.pdf

No idea about reliability, though. Where there's more space available, I use Yageo 220uF/450v/105c. Yageo is the parent company of Teapo, and tends to make higher-quality stuff.

[momaka]

Re: Asistance required on troubleshooting Corsair CX500

At the company where I work as IT, one computer suddenly wouldn't post.

The fans would spin when the power button was pressed but the motherboard would not "beep" and load the OS.

I checked voltages with multi-meter and found them unstable (can't remember details cause it's an old incident). I replaced the psu with a spare and the computer was back up and running.

Then I took the faulty psu to check voltages again without load. Voltages appeared better without load but no equipment would post with it connected.

Then I opened the case and the visual examination showed nothing wrong. But when I touched the primary big capacitor with my finger it felt slightly bulging.

The cap tested with esr micro and uF scale of multimeter AND Chinese transistor/C/diode tester, all of them found it was totally open!

Wow, I'm surprised the APFC circuit didn't explode or take out the primary switching transistors for the main PS. Perhaps the fact that you live in a country with 220/230/240V mains AC might have helped it a bit. With 110/115/120V AC main, the current pumped through the APFC coil should be 2x... which means the back-EMF could get really nasty high with an open primary cap.

The psu is Corsair VS450, the CWT version, so CWT is the OEM again.

Several weeks later, a friend gave me for free another failed PSU, this time Corsair VS350, again the same CWT platform and once more the same failure mode. Fan spins, voltages barely on spec and primary capacitor totally failed open, capacity is 0uF!


The failed cap is Aishi 400V 150uF 85C on the VS350 and Aishi 400V 180uF 85C on the VS450.

Thanks for sharing this info!

So looks like a lot of Corsair PSUs are likely expected to meet the same kind of doom someday - probably sooner than later, now that many of these are rather old.

I think the cap brand is crap. Please correct me if I am wrong.

Nope, not wrong at all ... though it looks like APFC can be hard even on the good Japanese brands if they are not sized up to the task.

I also find ridiculous the 150uF/180uF ratings for units that actually do 350W/450W accordingly in reviews (and more in overload tests without blowing). The cap must run at 395V-400V to be able to deliver those watts, because, as we know Q=C.V so when the C is too low you have to compensate increasing the Voltage.

Although I also think it has to do with the energy stored in the cap too... i.e. E = 0.5 * C * V^2
Since the energy increases with the square of the voltage, a slight increase in the voltage could offset a much greater drop in the capacitance... so that's how I think they are getting away with much smaller caps. IME, most APFC circuits with 400V caps typically run at about 380-385V... which is still pretty damn close to the max. voltage rating, especially since this is just the DC value and not considering the peaks and ripple current, which could well be closer to 390-395V (if not 400V) when added together.

And while electrolytic caps can withstand a decent over-voltage, it is not advisable to always run the cap near its maximum rated voltage and ripple rating... which is what APFC seems to be doing, hence why even good cap brands tend to fail.

There may not be enough space in the Corsair CX... units for a 450v/105c cap

In that case, perhaps drop down to 420V and see if you can keep the same (or get slightly increased) capacitance.

I don't know if 85C vs. 105C will matter much either... but perhaps it does.

In the end, I think it all comes down to finding the biggest capacitor that the PSU can fit dimension-wise and going with at least 420V. Doing these two should give hopefully a little longer usable life than what the original cap(s) did.

choices are limited, but there's a Samxon HP(M) 220uF/450v/105c cap that will fit - it's in the tall and thin form-factor:

https://cdn.badcaps-static.com/pdfs/...b9a5d6b9ef.pdf

No idea about reliability, though. Where there's more space available, I use Yageo 220uF/450v/105c. Yageo is the parent company of Teapo, and tends to make higher-quality stuff.

I probably still wouldn't trust it as much as the Japanese brands, but that's my personal opinion / skepticism towards non-Japanese caps. (Apologies to all non-Japanese caps that might find these words offensive/racist, lol )

[goodpsusearch]

Re: Asistance required on troubleshooting Corsair CX500

At the company where I work as IT, one computer suddenly wouldn't post.

The fans would spin when the power button was pressed but the motherboard would not "beep" and load the OS.

I checked voltages with multi-meter and found them unstable (can't remember details cause it's an old incident). I replaced the psu with a spare and the computer was back up and running.

Then I took the faulty psu to check voltages again without load. Voltages appeared better without load but no equipment would post with it connected.

Then I opened the case and the visual examination showed nothing wrong. But when I touched the primary big capacitor with my finger it felt slightly bulging.

The cap tested with esr micro and uF scale of multimeter AND Chinese transistor/C/diode tester, all of them found it was totally open!

The psu is Corsair VS450, the CWT version, so CWT is the OEM again.

Several weeks later, a friend gave me for free another failed PSU, this time Corsair VS350, again the same CWT platform and once more the same failure mode. Fan spins, voltages barely on spec and primary capacitor totally failed open, capacity is 0uF!


The failed cap is Aishi 400V 150uF 85C on the VS350 and Aishi 400V 180uF 85C on the VS450.
I think the cap brand is crap. Please correct me if I am wrong.
I also think 400V is too low for APFC.
I also find ridiculous the 150uF/180uF ratings for units that actually do 350W/450W accordingly in reviews (and more in overload tests without blowing). The cap must run at 395V-400V to be able to deliver those watts, because, as we know Q=C.V so when the C is too low you have to compensate increasing the Voltage.
Finally, 85C is just wrong, given the stress that the APFC puts on those caps.

I already ordered and received some replacement primary caps for these psus and I plan to do a full recap and enhancement on them and then use them on my pc builds but unfortunately I postpone it al the time due to lack of time.

Happened again , same platform, same failure. Notice, that all the other caps were good.

What I like a lot is that the psu just fails to power pc any more nothing blows up

[goodpsusearch]

Re: Asistance required on troubleshooting Corsair CX500

Wow, I'm surprised the APFC circuit didn't explode or take out the primary switching transistors for the main PS. Perhaps the fact that you live in a country with 220/230/240V mains AC might have helped it a bit. With 110/115/120V AC main, the current pumped through the APFC coil should be 2x... which means the back-EMF could get really nasty high with an open primary cap.

Highly possible. But I have also encountered HEC 450W APFC psu with primary cap totally dried and fuse / mosfets blown. So can't tell for sure.

I don't know if 85C vs. 105C will matter much either... but perhaps it does.

I think 105C helps a lot and also the max ripple current specs of the cap and of course choosing made in Japan caps when possible. I hope 420V caps are enough to help too because I find it very difficult to fit 450V caps in there