trouble with CCPS / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

trouble with CCPS

Move Thread

LAN_403

Turkey9

Thu Sept 03 2009, 02:15AM

Registered Member #1451 Joined: Wed Apr 23 2008, 03:48AM
Location: Boulder, Co
Posts: 661

Hi.
I've built a CCPS to charge my 320 uf 1200v caps. It has great power and with a small capacitive load the current waveform has a good characteristic shape. The problem is that when charging the larger cap, the wave form seems like it's getting a huge amount of interference/noise/distortion. No discernible pattern can be seen in the trace, just basically white noise. The circuit itself also makes a loud hissing/buzzing sound similar to what happens when corona forms but there is no corona breakout. The noise increases in intensity slightly as the voltage goes up.

Also, if I turn the supply voltage down after it has charged to a good voltage, the current waveform can be seen and is pristine, but the cap voltage doesn't go up at all due to too low of transformer output voltage.

The cap voltage is monitored by a comparator that controls the enable pins on the gate drive circuitry, just fyi.

GeordieBoy

Thu Sept 03 2009, 01:29PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

What is the topology? This will help the power electronics people here speculate on what might be going wrong.

Posting a circuit diagram and waveforms is always a good move too.

If I had to make a total "plucked out of the air" guess, I'd say switching noise getting into the control circuit and causing instability, but it could be anything...

-Richie,

Turkey9

Thu Sept 03 2009, 11:09PM

Registered Member #1451 Joined: Wed Apr 23 2008, 03:48AM
Location: Boulder, Co
Posts: 661

Its a SLR inverter, sorry forgot to mention that. It's a mosfet h bridge driven through a gdt with 5ohm primary side resistance. UCC27322 or whatever chips drive the gdt. A pwm chip sends the pulse train to the driver.

1252018721 1451 FT1630 0903091613a 322391

waveform on the base of one MOSFET without and bridge voltage.

1252018721 1451 FT1630 0903091617a 281700

This one is with about 80V on the bridge and cap voltage of almost 900v

1252018721 1451 FT1630 0903091612a 446327

The Entire Setup.

1252018721 1451 FT1630 0903091605a 301850

Current waveform normal....

1252018721 1451 FT1630 0903091609b 332256

Current wareform at 800v cap charge, single shot storage on oscilloscope

1252018721 1451 FT1630 0903091606b 315563

Real time normal trigger oscilloscope shot of current waveform at 800v cap charge.

GeordieBoy

Fri Sept 04 2009, 12:06AM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Some observations:

The first waveform shows an almost ideal gate drive waveform, except for quite noticeable droop. This is likely either due to too little magnetising inductance (not enough turns on the GDT) or due to using a DC blocking capacitor that is too small on the primary side of the GDT. Also check that you have a 100nF ceramic and 10uF electrolytic across the supply rails to decouple the supply right at the gate drive ICs.

The second waveform clearly shows that the gate drive pulses are getting messed up by noise getting into either the PWM driver chip or the gate-drive chip's enable line. Since there are little pulses where the drive is meant to be off, i'd suggest that the noise is getting into the PWM controller chip that sets the duty ratio, rather than onto the enable line at the drive IC which could only disable the existing PWM signal. You are right not to turn up the supply voltage until all of the switching waveforms are clean and rock solid, otherwise a device failure is almost guaranteed!

What are you using as your PWM controller IC? Is it a proper SMPSU controller like the UCxxxx parts, or is it something you have fashioned together from 555's, latches and logic gates?

The layout in the picture could be improved. Particularly the power section to reduce stray inductance that tends to cause ringing. I might also suggest changing the MOSFETs for modern IGBT's with fast co-pack diodes. The SLR inverter is much more suited to IGBTs because of it's high peak currents, zero current turn-off and need for fast recovery free-wheel diodes.

The current waveform looks the right shape, at least during this test, albeit a bit "ringy". See earlier comments about layout and changing to fast IGBTs.

To be totally honest I'm not sure what is happening that is causing the PWM signal to breakup in the next picture! It looks like the controller IC is freaking out due to some interfering signal. I don't know what controller you are using so can't really tell. All I can suggest is three courses of action:

Firstly, decouple everything thoroughly with ceramics and electrolytics. Decouple the power side DC bus. Decouple the low-voltage controller power supply rails. Decouple the supply right at the gate drive IC's. And finally decouple any reference voltages and thresholds for comparators. In power electronics when devices switch current quickly they produce electrical noise, who's aim in life is to get on every signal in your circuit! These decoupling suggestions prevent it from being able to mess up critical voltages that need to be clean.

Secondly, you said that the problem goes away if you turn the DC bus voltage down? Are you sure the diodes on the secondary side of the transformer are adequately rated in terms of voltage and are fast recovery types.

The SLR inverter is intended to model a constant current source. So the first thing I would do is bolt a short across the output and see if I can get the DC bus voltage up to the full design voltage without things going haywire. This at least proves the inverter can source a constant current into a short-circuit with full DC bus voltage without going unstable. If it does go unstable you can wind the DC bus voltage up and down around the threshold whilst probing around on the controller side to investigate the cause.

Finally, I would disconnect or disable any over-voltage comparator, voltage feedback, or "charging complete comparator". I suspect this might be where the problem lies. It is often when you try to measure something on this power output side and feed it back to the controller when noise gets into the feedback loop from the power side. Therefore, disable this loop and see if you can charge a capacitor up to the required voltage whilst monitoring with a meter. You can switch off the power manually when it gets to the required voltage, or just connect a load across the capacitor so that it reaches equilibrium at the required voltage and doesn't charge any further.

Lots of ideas for you to try there.

-Richie,

PS. If you want to get more replies to this thread, posting some design details like switching frequency, power throughput, power device part numbers, rectifier part numbers, controller IC type, and details of the transformer, etc might help others to come up with suggestions.

Turkey9

Fri Sept 04 2009, 04:14AM

Registered Member #1451 Joined: Wed Apr 23 2008, 03:48AM
Location: Boulder, Co
Posts: 661

Thanks Richie! I'll start working on all that as soon as possible.

Here are some specs for you all....

Goal:
Lowish power system with supply preferably below 100v. 500 watt would be good but the devices need to have as little heating as possible but still with decent power. Just need to charge a 290J cap fairly quickly.

High side power:
Variac rectified by 600v Hyperfast diodes from a coilgun project. 1640 uf 200v electrolytics decoupled by two .1 uf film caps... not ceramic.

I have an additional 470 uf on each side of the h bridge at the devices with another couple .1 uf decoupling caps.

Bridge:
4 W25NM60N MOSFETs
15 turn gdt with 1uf film and 5 ohm on primary side. Nothing on secondary side.

Driver:
UC38086 smps controller from TI. 1 uf tantalum and .1 uf ceramic on power pins as close as possible to chip. Two signal lines from UC38086 fed to two UCC27322 MOSFET drivers. Both driver chips have a 1uf tant and a .1uf ceramic across the power pins. The voltage supply is 12v from a computer psu.

driver frequency is at 18kHz driving the transformer at 36kHz.

Charging transformer:
8 turn primary with .56 uf in series... tv flyback core.
GBPC3512A 1200v 35amp bridge rectifier feeding the cap. Cheapest thing that was in one package and had right ratings on digikey... should probably switch to ultrafast discrete components.

Comparator:
lm339 with a 20 turn 10k pot to set reference voltage. Divider across cap ran to comparator past the base of the transformer.... could be it right there.
No decoupling at all on comparator.

I just received some 10uf tantalums that I'm going to replace the 1 uf ones on the driver. When I had the driver circuit on the breadboard the sound would occur if I brought my hand close to it in order to adjust frequency.

If a comparator is too much trouble to keep noise out, could I monitor the cap voltage with a microcontroller and use that to regulate? I'd imagine that that would be even more susceptible to noise.

NOTE: just noticed that the sound is picked up on my computer speakers so I'm sure it's interference. Also just checked some stuff... Disconnected the comparator feed and that helped already. Now at low voltages there is a clean wave form that changed as voltage on the cap increases. It looks similar to what i would get while tuning the circuit so I but it is my SMPS controller.

Thanks! Jesse

GeordieBoy

Sun Sept 06 2009, 05:16PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

You don't need Hyperfast diodes for the mains side input bridge-rectifier. A decent chunky mains bridge rectifier will work fine here and be much more robust than fast recovery diodes.

100nF polypropylene film caps across the bridge are fine.

15 turns on the GDT sounds like far too little. I don't know what the core material is, but I'd suggest about 30 or 40 would give better waveforms given the low switching frequency (18kHz!)

Charging transformer primary sounds like it has WAAAAAY to few turns! I'd shoot for more turns, and a higher resonant frequency like 100kHz, with a 50kHz switching frequency for the power level you are operating at. Otherwise you may saturate the transformer or at the very least overheat it.

Output rectifier bridge GBPC3512A is totally unsuitable for a switch-mode power supply output rectifier. The datasheet shows it is characterised for 50/60Hz and doesn't even show forward and reverse recovery times!!!! You'ld be better swapping your rectifiers round!

Definitely decouple the reference to the comparator too.

...did I mention changing the MOSFETs for IGBTs too?

-Richie,

Turkey9

Mon Sept 07 2009, 05:26AM

Registered Member #1451 Joined: Wed Apr 23 2008, 03:48AM
Location: Boulder, Co
Posts: 661

Thanks for all the suggestions!

First, how exactly would i decouple the reference for the comparator? Would i need to just put some 100nf caps across the divider resistors or from the reference to ground?

I don't know much about IGBTs, so will they have low switching losses? Like I said before, i need to have the least heating possible. Would these devices be suitable? FGA180N33ATTU

I don't quite remember why I used such a low frequency, I think I was attempting to get as much current through as possible because I originally limited the bridge voltage to 32 volts.

Thanks for all your help, i'll post my results soon!

Thanks again!

-Jesse

Turkey9

Sat Sept 19 2009, 10:41PM

Registered Member #1451 Joined: Wed Apr 23 2008, 03:48AM
Location: Boulder, Co
Posts: 661

RESULTS:

I decoupled everything like crazy and that seems to have fixed most of the problems. The current trace is now very clean but the frequency is still changing as the charge on the cap goes up.

As for the GDT, I was wrong about the turns. I built a new one before the first post with 32 turns. I increased the primary side capacitance and it didn't change anything. I went all the way to 6 uf.

I haven't changed to IGBTs yet, I'm currently doing research on the best ones to get.

I'll work on changing the drive frequency soon and the other suggestions to improve performance.

-Jesse

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob