Replacing the fan of an Eaton Ellipse Pro 1200 UPS

[0xdeadbeef]

In the past weeks, I have been tinkering with replacing the fan on my Eaton Ellipse Pro 1200 UPS which I bought unaware that the fan is always on and pretty much annoying.
The 1st problem was that the fan is a Sunon MF60152V2-1000C-G99 which is a 60mm 24V 3pin MagLev fan and companies like Noctua and Papst only offer 60mm fans with 12V supply.
In some Reddit group, I found the suggestion to use the Noctua NA-VC1 to convert the 24V to 12V and then use a Noctua NF-A6x15 FLX as replacement.

This seemed to work at first, but after some hours or sometimes already after some minutes, the fan stopped and the UPS displayed a fan error.
Now, admittedly, I jumped to the (wrong) conclusion that the Eaton detected the lower RPM of the Noctua fan and went so far to implement a full fledged µC based RPM simulator that exactly reproduced the speed signal of the original Sunon fan and checked the Noctua's speed signal (with UART output and all).

While I was at it, I also ordered a Papst 612F/2 together with some parts of my circuit. Surprisingly, the Papst fan worked without fan errors (using the NA-VC1). As I was still developing the µC approach at this point, I actually interpreted this as confirmation that the RPM monitoring was the problem. The only other difference I noticed between the Papst and Noctua fans was that the Noctua fan stopped slightly earlier (~4.6V vs. ~4.1V).

Anyway, after sinking a lot of time and money in my µC approach and testing thoroughly that it worked exactly as intended at the bench, it didn't change the situation at all when used inside the Eaton (with the NA-VC1 and the Noctua fan behind it). I.e. there were still fan errors reported with the Noctua fan after some time. For the sake of it, I also tested it with the Papst fan and the Papst fan still worked.
I did a lot of other checks including using the UART output of my µC to see what's going on, but the measurements seemed weird and inconclusive and in the end, the Noctua fan just stopped at some point.

So, finally I did what I should have done first: I actually measured the frequency and voltage at the fan connector on the Eaton PCB.
The measurements were done with an JST XH3P extender cable cut in half to add measurement hooks. So all measurements were done on the Eaton side.

First of all, this is how the speed signal and voltage of the original Sunon fan looks in standby or normal operation:

The speed input shows a frequency of  ~82.5Hz (2475rpm). The voltage signal differs wildly with a mean value of ~12.5V.
The voltage drops are in sync with the PWM edges, so they are most certainly caused by the inrush current into the fan.

Now to the NA-VC1 + Papst 612F/2

Surprisingly, the voltage (measured before the NA-VC1) is now more sinusoidal with a mean value around 4.7V. I guess the NA-VC1 filters out the spikes and creates a permanent voltage drop for its own supply. Still, the massive drop of the mean voltage is surprising to say the least.
Note that the voltage is never really much above 6V where the NA-VC1 is specified for an input voltage range of 12-24V. At voltages much below 12V, the NA-VC1 will more or less output the input voltage minus 0.3V or so.
Also, due to the lower voltage the frequency is only ~46Hz (1380rpm). Much lower than the 82.5Hz of the original fan.
Note that the Papst fan stops slightly above 4V, but the mean value of 4.7V seems to be sufficient to let it run continuously even though the voltage drops significantly below 4V all the time.
So, this works, but the fan speed is much lower than that of the original fan. This also shows that there is no strict rpm monitoring. The Eaton doesn't care if the fan rotates at 1000rpm or 2500rpm. It just checks that the speed signal is there at all.

Finally, the NA-VC1 + NF-A6x15 FLX

Now, this looks totally different. The fan seems to stop shortly every 1.2s - at this point, the voltage jumps up to ~22V.
Actually, I'm somewhat puzzled about the voltage ramping down. I'm not sure where this comes from but most certainly, the fan stops when the voltage reached its lowest point.
You would think that the fan stopping causes the voltage jump, but the timing is not always consistent:


I guess the stopped fan reduces the current to almost 0 and the logic creating the speed signal sometimes needs a bit time to catch up.

Also, it doesn't really looks like the frequency depends on the voltage at all even though the voltage varies between 4V and 23V.
The ~35Hz (1050rpm) measured between the stops is what I measured when supplying the NA-VC1 with 5V at the bench.
Anyway, this behavior is strange and concerning. It seems that the Noctua fan is never really running continuously, Instead it stops shortly all the time. Just not long enough to trigger the fan error ... until it does:

Now, the voltage stays permanently high but the fan stays permanently off. So this is obviously not caused by the Eaton. I guess this must be some kind of emergency shutdown inside either the Noctua fan or the NA-VC1. Not sure what would be more concerning.

Anyway, even though I still don't really understand what's going on there, this shows that using the Noctua fan is not a viable approach, at least not when supplied by the Eaton's fan voltage. I guess it would be possible to supply it from somewhere else and set it to some sensible speed.

The Papst fan is working but the fan speed is significantly lower that that of the Sunon fan (about 56% of the Sunon). This also partly explains why the Papst is so much quieter compared to the Sunon.
For the moment, I guess I will use this approach even though I'm a bit concerned now about temperatures inside the Eaton. Then again, the design of the air flow is total rubbish anyway.