Electronics > Beginners

Filtering 19kHz switching noise from UPS AC output

(1/3) > >>

Novgorod:
Hi!
We have some critical lab equipment, which we want to power through a UPS in case of mains power interruption. Our UPS is an Eaton 9PX series with 1.5kW max. but typically much less of protected load (100-200W). Everything works fine except for one device, which really doesn't like the AC power provided by the UPS but works fine directly on mains power. This device is basically a high-voltage supply (<10kV), which takes only 70W of mains power during normal operation. When it's powered by the UPS, the high-voltage supply still works but it can't stabilize the HV output, i.e. it fluctuates heavily. When switched back directly to mains (e.g. UPS set to bypass mode), it stabilizes immediately (behavior is the same with or without power cycling).

I assume the HV is directly derived from the AC mains input without an intermediate DC conversion step and is thus very sensitive to the mains waveform. Before getting the company involved and losing sanity to tech support, I want to try to find a simple solution first. The UPS is more or less decent and is supposed to output a clean AC sine waveform, so I measured it with a scope (under light load, i.e. with the HV device connected) and compared it to the mains from the wall (we have 230V/60Hz here):

Red: mains AC from the wall
Blue: AC output from the UPS





The sine output from the UPS is actually cleaner with less harmonic distortions than the mains from the wall, but it clearly has some small but significant switching ripple at 19.2kHz, see the zoomed bottom image, with an amplitude of a few volts (roughly 1-2% or so). I don't have much experience with UPS but I suppose it's nothing out of the ordinary, as UPS are primarily meant for IT stuff with their switching power supplies, which don't care much about the input waveform anyway. I can only assume that the small high-frequency ripple is the reason why our high-voltage supply has problems with it.

My question is: can this 19.2kHz ripple be sufficiently filtered out in a simple way? I guess a low-pass filter with a high enough power rating should do but I really wouldn't want to cobble together some custom solution because I have very little experience in playing around with mains power. Maybe that's a common industry problem that has been solved many times already but I don't know where to start looking. Could it even be as simple as putting some ferrites around the cable and what specs would I need?

Thanks for the help!

T3sl4co1l:
Probably the output has a big capacitor on it, so the output impedance at those frequencies will be low.  So a series "trap" would be called for.  This is a parallel resonant tank (L || C), connected in series with the source, and acts to increase the output impedance at its resonant frequency.  Probably design it for 5 or 10 ohms at 19kHz, with some give-or-take for final tuning (adjust the capacitor or inductor by selecting values or making parallel/series combinations, or adjust inductor gap or turns/taps).  So, that'll be L = (10Ω) / (2*pi*(19kHz)) and C = 1 / (2*pi*(19kHz)*(10Ω)).

If that's not good enough, an R+C in parallel with the load may be helpful (R ~ 1-5Ω, C a few times (>= 2.5) bigger than the trap C), or a trap (series L+C, calculate the values similarly).

Tim

ogden:
If you have 1:1 mains transormer around- try it first. Then there are power conditioning filters usually combined with surge protection. APC manufacture such for example.

MagicSmoker:
Tim's suggestion of a parallel resonant trap is not a bad one except for assuming the output impedance of a HV supply will still be capacitive at 19kHz - there tends to be a lot more stray inductance with the large clearances and transformer winding arrangements required above 3-5kV or so.

Consequently, I would recommend the boring old LC low pass filter on the output of the UPS (series L, shunt capacitor) with a corner frequency around 1/10th the ripple frequency (e.g. - 1.9kHz fc for 19kHz fripple) and a characteristic impedance 1/10th or less than the effective load impedance. In this case, the load is 70W with 230VAC mains so it has an effective impedance of 756R; a filter with less than 75R impedance should do the job. For examples using standard E12 values: 4.7mH + 1.5uF (Zout = 56R); 3.3mH + 2.2uF (Zo = 39R); etc.


Novgorod:
Thanks for the advice. A custom-built filter box would be my last resort if I can't find an "off-the-shelf" solution. The proper design would be probably more complicated, since the power draw of the HV supply device ist not constantly 70W but can vary by a lot during normal operation (up to ~300W but typically below 150W and most of the time around 70W). However, it's a kind of composite device with additional functions (other than the HV supply) and it has a rated total power draw of up to 1kW (those high-power functions are used rarely and only for a few minutes at a time). I could probably integrate a custom mains filter just for the HV supply portion of the device, but that would require messing around inside the device itself.


--- Quote from: ogden on June 29, 2019, 08:26:38 pm ---If you have 1:1 mains transormer around- try it first. Then there are power conditioning filters usually combined with surge protection. APC manufacture such for example.

--- End quote ---

This appears to be the component I'm looking for. I don't have anything like that "at hand" but I found plenty of power line EMC/EMI filter modules from various vendors, which might be suitable. These EMC filters are typically designed for high rejection at high frequencies (>1MHz), as I couldn't find any specifically for the few-kHz range. However, some of the "high-performance" type of filters seem to be good enough down to the 10kHz range, for example this one:

https://hken.rs-online.com/web/p/power-line-filters/7024031/

According to the data sheet, the 10A version has the best performance at low frequencies with ~20dB insertion loss for both differential and common mode noise around 20kHz, although I don't know whether this is load-dependent. This should reduce my UPS ripples by a factor of ~100, i.e. to some ten millivolts amplitude. Is this worth a try or am I completely mistaken?

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod