mV ripple on a HIGH voltage power supply

[AlienCircuits]

Hello,

I wrote a long explanation of what I'm doing and trying to figure out, but I'll just get the question out:


If I'm regulating a high voltage at about 800V and need to sense the output for my feedback loop and it needs to be divided down to ~1.25V (factor of .00156), how do I hope to minimize ripple to mV range when the feedback loop will probably put this signal below the noise floor (the ripple is about 20 µVpp at the feedback input)?

My spice simulation with my chosen opamp spice model seems to have no problem responding to this very low µV signal, but I am using high value resistors and I doubt this opamp will really see the µV ripple when I go to build it.


I almost think that I need to make a separate feedback path with an AC coupled feedback that only is used to eliminate ripple.

 

[erikl]

... I need to make a separate feedback path with an AC coupled feedback that only is used to eliminate ripple.

Right. I think you only need to bridge the upper resistor of your voltage divider by an appropriate capacitor. Try and analyze it!

 

[FvM]

I presume you are talking about a linear voltage regulator, although that's not said in your post. Apart from noise consideration aspects, which are ignored in a SPICE AC or transient analysis, you'll need sufficient loop gain at the ripple frequency to achieve a regulation effect, as well as DC loop gain to get DC regulation. Capacitive bypassing of the voltage divider is a usual means to increase the loop gain at higher frequencies. It's the somehow trivial implementation of the mentioned "separate AC feedback path".

 

[DarrellTH]

This may be far off-topic but, I had a HV circuit design once to drive a TWT (traveling wave tube). The output had to have extremely small ripple. The switching power supply could not achieve this due to the small HV cap value so, I put a small linear regulator on top (reference to rtn via a capacitor) to "soak up" the last few volts of ripple.

 

[AlienCircuits]

The problem with putting my bypass cap to get more AC ripple in my feedback is that it made my PID controller very unstable. This is because its proportional gain is tuned to the sensitivity of my regulating MOSFET circuit. I turned proportional gain down and tried to just get my derivative gain up and it works now, but still a little unstable at start up (oscillates to saturation at startup) and I have less performance than I did before, but at least I know that the ripple signal has a better chance of being picked up by my control circuit in a real circuit.

I think the reason my PID controller was so unstable with the AC bypass cap added is because the ripple (5mVpp) is 160,000 smaller than my DC signal, and so the AC signal is getting a lot more gain, and I don't think my PID can be designed to control both. I am thinking of 2 feedback paths, one sensitive that is PID for the DC feedback, and another less sensitive separate feedback for the small mV ripple.

 

[chuckey]

Its a bit crude but I would hang a string of zener diodes from the live output to earth (with a suitable pot to maintain the zener current). Then you will have an error voltage you can do something with. If its a mains PSU then your amplifier will have enough loop gain at 1 X or 2 X the mains frequency to clean up both the DC and the ripple.
Frank

 

[bepobalote]

maybe this can help you: Linear Technology APP NOTE 118
Written by Jim Williams it teaches how to measure ripple from real HV power supplies...