Nearly complete SMPS, not stable / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

Nearly complete SMPS, not stable

1 2 next

Move Thread

LAN_403

cjk2

Mon May 03 2010, 11:57PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

Is there a secret to making these things stable? I believe I designed my circuit correctly. I am currently trying to get my supply stable in constant voltage mode only, Ill fool with the current control only after the voltage mode is working.

I have played with the value of R4 and found that smaller seems better. The value of R40 seems critical to getting the supply to stop hissing and chirping. R40 is currenly a 100K pot set to about 90K ohms. Much increase or decrease at all seems to make it hiss or "crackle" at various loads and voltage settings. The supply seems to regulate fine, the output voltage stays constant regardless of load. The Hissing and crackling sound also changes with the load and I hate this! Can't I have a supply that is stable with any load I connect?

How do you guys manage to build these quiet, stable, SMPS's?

Steve Conner

Tue May 04 2010, 11:55AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I have three variable SMPS on my lab bench at work. Two Xantrex 1kW units, and a third one I modded myself.

They all hiss or whistle at some load setting or other, usually at very low currents, or the boundary between discontinuous and continuous mode. It's annoying, but it doesn't seem to be harmful.

If R40 is affecting the noise, though, maybe you have spikes getting into your current feedback signal. Maybe you need a tiny capacitor from the "Ramp" pin to ground, as shown in the TI UC1825 datasheet, page 5.

Another thing to try is, make C7 larger and C6 smaller. I start with 0.47uF and 100pF. Then replace R6 with a pot, and twiddle for the cleanest load step response. (you know what a load step test is, right?)

You need to tune that compensation network according to the resonant frequency of your output filter, ESR of the output filter caps, and so on.

The current mode feedback helps, if it's working properly it adds another 90 degrees phase margin to your voltage control loop, so the compensation network isn't so critical. But too much current feedback into the Ramp pin can make the thing go unstable as well.

101111

Tue May 04 2010, 02:13PM

Registered Member #575 Joined: Sun Mar 11 2007, 04:00AM
Location: Norway
Posts: 263

cjk2, this looks very interesting.
If or when you get it to work properly, maybe a group buy on PCBs and parts?

cjk2

Tue May 04 2010, 04:39PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

Well it looks like Ive made some progress on this. I added the leading spike suppression cap across the output of my current transformer, and this made my ramp waveforms look a little nicer. I also changed my compensation circuit like you suggested, I used a 150pF cap and 500nF. The slope compensation resistor is still pretty much hair trigger, It is happy where it is, but any change seems to upset it. In board rev.2 I am likely to make it a pot to avoid all of this trouble. My hissing is now greatly reduced and is only seem at low voltages with low loads. I have yet to pull over 100ma from this supply, but I assume that everything will actually improve under large loads.

My turn on overshoot is Terrible, like 5v. fiddling with the compensation pot does not seem to help this much. Perhaps my output filter inductors are much too large?

Binary friend: I would be happy to send you a set of boards once I get this supply worked out and tested for maybe $15, just send me a PM when Im done with this project.

Steve Conner

Tue May 04 2010, 04:57PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

cjk2 wrote ...

The slope compensation resistor is still pretty much hair trigger, It is happy where it is, but any change seems to upset it.

This suggests you maybe still need to work on the compensation network some more.

Also, you need a load step test rig. Richie Burnett convinced me to build one, and it's been really useful. I had to design about a dozen little DC-DC converters at work, and I used it on all of them.

Basically what it is, is a load that switches between 10% and 90% maybe 10, 20 times per second. You use a high value dummy load resistor, and a lower value one switched in parallel by an IGBT, MOSFET or whatever. You synchronize your scope to it and look at how the output voltage bounces around as the load changes. A good SMPS with current-mode control should maybe droop a few percent and recover with a nicely damped exponential. I fettled my DC-DCs at work until they all did that.

One from the bad old days before current-mode control can overshoot and ring like crazy.

wrote ...
I have yet to pull over 100ma from this supply, but I assume that everything will actually improve under large loads.

Probably not.

HV Enthusiast

Tue May 04 2010, 09:57PM

Registered Member #15 Joined: Thu Feb 02 2006, 01:11PM
Location:
Posts: 3068

Well, you probably don't have access to a Venable or Impedance / Network analyzer, but you should probably take a shot at measuring your loop response which can be done "old skool" with an oscilloscope.

I won't describe the exact procedure, but you would simply put a series 50 ohm resistor in your feedback loop, and then use a low-level oscillator to inject a sinewave into that 50 ohm resistor. Then using an oscilloscope, you can measure the input vs. output and get a gain / phase plot. You would just do this incrementally over a variety of frequencies. Then you could see just how stable your loop is.

A search on google will probably yield a ton of resources of how to do this with an oscilloscope and signal generator.

cjk2

Thu May 06 2010, 03:27AM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

Ive been fiddling with this thing all day. Maybe after Ive had some classes on control theory I will know how to stabilize this project

. It seems to work just fine in voltage mode only, that is when feeding CT directly to ramp. As soon as I start to add any current component to the ramp pin however, I get an imbalance in duty cycle, that is, the positive half of the drivers output has a different duty cycle than the negative half!

Both attached pictures are of the output of my current transformer. In the first picture, you see the current waveform as I expect it to look, each ramp is the same as the next (they don't start at 0 tho, perhaps they should?).

In the second shot you see what happens when I add any current component to the ramp pin, consecutive pulses are not the same.

I build a step test rig as you suggested, It uses a mosfet and my signal generator to switch an 8 ohm resistor in series with a 100 ohm resistor.

EDIT: Looks like I discovered for myself why this supply is not stable. According to Fairchild semi, a halfbridge based current mode power supply will not be stable due to the capacitors in series with the primary of the transformer. Looks like it will be voltage mode control for me unless I feel like adding a balence winding to my transformer...

Steve Conner

Thu May 06 2010, 09:12AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

cjk2 wrote ...

EDIT: Looks like I discovered for myself why this supply is not stable. According to Fairchild semi, a halfbridge based current mode power supply will not be stable due to the capacitors in series with the primary of the transformer. Looks like it will be voltage mode control for me unless I feel like adding a balence winding to my transformer...

Oh, worse luck!

I didn't know that halfbridge and current mode don't mix.

cjk2

Thu May 06 2010, 02:07PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

That's alright, I'm actually a bit relieved to find that the problems are likely caused by something I can fix and not some fault that I can't seem to track down, or my inability to correctly compensate the thing. I have decided to try adding a balance winding later today and see if it helps things. I suspect it will as I've found several articles discussing current mode instability in half bridges.

Too bad I didn't know this from the beginning, but this is how you learn!

EDIT: It looks like with a balance winding I am able to get this thing stable! I am seeing very little hissing and all my ramp waveforms are the same! No more oscillating junk

cjk2

Wed Jan 12 2011, 09:38PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

This is a follow-up to this "long time in the making" project.

I got it about as stable as I could which means very little whine at any output setting. I decided to go straight to digital control and skip the pots on the front panel. I tried using the ADC and DAC of an ATXMEGA micro with mixed results. The internal ADC is not very high quality and has probably 20 to 40 LSB of noise, this is pretty unacceptable in my application. I was able to read the output of the powersupply in both voltage and current to 3 significant digits, this is not enough though, I want 4.

I then went on to design the attached circuit with a dedicated 12 bit ADC and DAC. I will be sending off these boards soon and hope to have the 2nd revision of this project done in the next month or so.

Attached is my new schematic and some pictures.

Video is here: http://www.youtube.com/watch?v=YoVEyr2PRYM

Note that actual pictures are of the version currently on my desk with ATXMEGA for the ADC and DAC and older control board. The schematic and board layout images are for my next revision with dedicated DAC and ADC as well as redesigned opamp and power circuitry. Vales of output filter and compensation sections will likely change.

1294868271 51 FT88593 Smps 3 Digital Prelim 1

1294868271 51 FT88593 Smps 3 Digital Prelim 1 Boards

1 2 next

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