formulas for LC circuits on outputs of switch PSUs?

[Svart]

looking for formulas for choke/cap ripple circuits on the outputs of various powersupplies. any help greatly appreciated!


:thumb:

 

[CJ]

Your going to use a switcher> What's the app?

 

[Svart]

building one.. :grin:

really it's for work, I designed a PWM PSU(for motor speed), variable via touchpads with a reversible Hbridge current amp utilizing SCRs as high side sources and IGBTs as lowside switching elements. I posted part of the reset circuitry a while back on the forum if you remember. It all works well but I need to implement ripple filtering on the outputs due to the switching noise and other things. the switching is not making the brushes happy at all, nor is the screaming from the 1khz switching frequency making my ears happy.

so far it looks like a max of 150VDC @ 3 amps at 100% duty will be the max output needed.

I'd like a low value inductor but that will cause me to use a much higher value capacitor if I'm not mistaken(higher ripple). I've got ideas on what I'll need but am having trouble finding info on formulas to get exact numbers.

thanks! :thumb:

 

[Svart]

also I'm finding a lot of engineers saying that inductors are not needed and can cause problems with variable switching.... and just as many saying it's needed..

 

[Svart]

anyone?

please?

:guinness:

 

[bcarso]

Svart, I think we really need more details of the circuitry. When you mix brushes with inductors the results can be very messy indeed. Without knowing the details it sounds to me more as if you need some sort of R-C snubbers.

L-C filters will need some damping resistances in any case. At really high frequencies much of the damping comes from the core characteristic itself, as used in ferrite beads. Methinks your problem is quite a bit lower frequency-based.

 

[Svart]

yeah it's switching at 1khz. the motor is a 4 brush standard DC motor. the problem I am seeing is only at the current threshold needed for rotation. the pulses are not meeting that threshold so the motor is not turning but they are causing the unit to sing at 1khz. moving lightly above that threshold, the motor turns fine and the noise diminishes. So in layman's terms, at such large deadtimes the pulses don't supply enough current when ON to do anything but make the motor vibrate(and burn brushes). I am looking to smooth the pulsetrains out to an average DC level, the less deadtime/the more ON time, the higher the average DC level. It seems that inductors won't work properly with such high variations in pulses from what i can find. I'll have to experiment with caps on the outputs but I can't have too large a discharge time after the bridge turns off due to the need to reverse the motor. I could limit the lowest duty cycle to something over the current threshold but I really need it to go from 0-100% A snubber might work, that's in my Try This list.

also for more info, when switching into a direction, the PWM controller softstarts, or ramps up to the set speed over a couple of seconds to limit inrush current and the like.

:thumb:

 

[bcarso]

Are the pulses basically coming from voltage sources? If so an L input filter shold be o.k. Don't just throw caps in there with nothing else or you will get the dual behavior of trying to interrupt currents to inductors.

I gather there is no closed-loop feedback to allow a tachometer-based system for example to goose up the drive a bit to get past stall?

 

[Svart]

it's a multi purpose brute force circuit to drive halogen lights, small reversable motors and large single direction motors. the drive for the lights and large unidirection motor is only one lowside IGBT. This specific setup doesn't have current or voltage monitoring although I am also working on another that will but not for this arrangement.

the current amp bridge is essentially a SCR sourcing the current from a fixed rail(highside), and an IGBT(or MOSFET) actually doing the switching to GND(lowside). this switching is the actual speed control. so yes, in a way it's coming from voltage sources but the pulses are to the GND rail..

one thing that I tend to forget in the midst of all the design is that this drive system will be driving motors at the end of 1000 FT of cable, so add some more capacitance and inductance to the equation and it might not be *that* bad.. but i need to also plan for much shorter runs too.

thanks! :thumb:

 

[bcarso]

I see it now, creeping along the mine shaft or ocean floor.

To see if the smoothing concept will really help your stall torque, I would do an experiment: when it stalls with the existing circuitry, quickly switch over to a constant voltage source and see if it moves at that point.

You might be able to do something that senses stall just due to the extra current drain, and move your duty cycle to ~100% until freed up.

 

[Svart]

through sewer pipes usually!

the problem with the PWM is that the torque is related directly to the pulse-widths. let's say at 10% duty, the pulses are on 10% of the time so you are only allowing the motor to see an average of 10% of the current available from the sources even though the pulse itself is at full rail to rail amplitude.

When testing the stall torque, I assume you mean to use a constant voltage source of the same average potential as the PWM output. I have done this and the motor responds as one would expect, speed related to voltage level. the motor responds just as well above the point where it stalls and complains(what I called the vibration.. :green: ) but is slightly inefficient due to some pulses arriving at the motor between commutator pads contacting the brushes simply doing nothing. another reason I would like to see some kind of smoothing..

As for current sense, I could implement it if nothing else works.

thanks Bcarso! I'm beginning to feel that you might be the only one who cares about my plight! ha ha

:guinness: :thumb:

(edit) oh by the way: http://cobratec.com/products.html