The name brand parts actually tend to be fab'd in SE Asia, nowadays, and their prices are actually lower, into the local market. So don't defame other mfrs, who are using basically the same technology (and even the same machinery), bought fair and square.
RL
The current production is with 100 nF, which is running too hot when sustained 50% duty cycle occurs. I'm getting some C0G 10 nF caps and will see how they work.
The signal is darn close to DC for the brush motors, and under 200 Hz, normally, for the brushless motors. The highest supply voltage is
120 V DC.Well, my voltage is 10 X higher. The current design with 100 nF caps gets a theoretical 1.5 A RMS in the capacitors, which is why they are getting hot! Geez, no surprise. This is a sign/magnitude amp, not a synchronous antiphase design like the typical class-D amp, so at least this 50% duty cycle condition is not the idling state. If I ran this at
120 V in synchronous antiphase, the inductors would smoke in a minute.Thanks,
Jon
it.
used,
I have done enough capacitor testing that i usually can guess right the first time. Point me at some data sheets and i will say what i think it is. 1 kHz is by far the most common though.
?-)
Not much use as data, then.
I suspect the engineers included the test frequency in the data they passed to whatever publicist produces the datasheets, but it got left off, 'cos it doesn't have "impact".
I wouldn't guess, I'd measure. No samples, no business.
Me, too.
Electrolytics are commonly done at 100/120Hz.
Or 100 KHz
Unless they're low-Z types, in which case 100kHz.
True, but they ain't "common" ;-)
The dissipation factor is for the dielectric--at low frequencies the admittance of the capacitor goes like omega*C*(j+delta). Apart from the actual resistance of the wire, it's the ESR that's poorly defined, not the dissipation factor.
Cheers
Phil Hobbs
I think I see what you mean. Treating dielectric losses and series resistance separately. Kemet's models do that. Not the easiest thing to measure.
t
tor,
sic
,be
You need to stick with NPO/ COG ceramics for low dissipation. I've not looked at 250 Volts, but at low voltage there as good as or better than polypropylene (sp).
George H.
n
Check with John Larkin. IIRC, he has stated that ceramics can take many times their rated voltage. Should be in this group somewhere.
Like John, I've put hundreds of volts on capacitors rated far lower. Leakage current levels stayed extremely low. I haven't checked how much
*capacitance* they have at these bias levels. NPO/C0G might not do too badly, but most other dielectrics would probably lose much of their relative permittivity.Reliability is another issue - MTBF generally declines rapidly in the vicinity of maximum voltage specs on most parts.
YMMV.
I've had one HV application requiring extremely low distortion. I tested some 100p/150V NP0 caps at the kV level with loawer (the cap meter couldn't resolve better) than 0.01% capacitance variation. Didn't check any film capacitors because they weren't suited to the app, but I'm not sure they'd be as good.
ght
ductor,
.en
ious
sonic
ies,
ut
ld be
y ots.
an
it
Neat, digikey has 10nF / 250 V NPO's, So that's where I'd start.
George H.
No, the 5X was an X7R, I think. Well, I got these 10 nF 250 V C0G caps in, and they are QUITE good, it fact I can't get a D reading on them. The meter nulls with the D knob all the way against the stop. With the Polyester caps, I can easily get a D reading, although it is pretty low. Well, this should solve the heating problem, now I just need to see if the 10 nF works well in the filter application.
Thanks for the great idea, I didn't know I could get a C0G in that large a capacitance * voltage.
Jon
I had a similar disagreement with someone else here last month on the subje= ct of needing higher and higher Q inductors when the application was intrin= sically low Q. Don't let specmanship get ahead of functionality like he did= .
Simply designing a low pass filter might be optimal from a textbook sense b= ut the issue really is, you have high frequency energy going into the filte= r but not going out to the load. Therefore it either has to be reflected ba= ck to the source, or dissipated as losses in the filter. Which is why the L= 's and C's get hot or break down, or high peak voltages/currents reflected = back to the source damage it.
When the filter network is designed to absorb energy outside the L and C co= mponents, it is often called a "snubber". In the RF world they might have a= n explicit diplexer with a load resistor on the "HF" output port, but in mo= tor drive filters the intentional R is usually put in series with C's (outs= ide the C so it doesn't result in your problem!), and the intentional R cou= ld even be nonlinear (varistors or diodes or zeners).
The source won't mind a puny few hundred mA when it is designed to run 20 A continuous. The inductor current gets "circulated" back to the DC supply. So, this filter is not designed to absorb energy, as an LC filter can't do that very well. It is just supposed to keep the huge Dv/Dt of the FETs switching from getting to the output terminals and then being conducted out to cause mayhem.
But, of course, a high-Q filter resonant at the PWM frequency would be a bad idea. It would look like a short to the output stage and would also develop high voltages at the L-C junction. So, my idea was always to have the resonant frequency above the PWM freq. so that resonance could not be excited. And, a low Q would be fine, again so it wouldn't "ring" the LC excessively. I'm using a big power toroid with a solid wire winding, so it can't be high Q.
There is a snubber in the servo amp to reduce the Dv/Dt to protect the FET driver chip. But, it doesn't reduce the slew rate very much.
Jon
Have something to add? Share your thoughts — no account required.
Ask the community — no account required