SMPS Noise Reducing Tips / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

SMPS Noise Reducing Tips

1 2 next

Move Thread

LAN_403

MRacerxdl

Wed Jun 08 2011, 07:47PM

Registered Member #989 Joined: Sat Sept 08 2007, 02:15AM
Location: SÃ£o Paulo, Brazil
Posts: 476

Hi all,

For a long time I wanted to make an good SMPS. I started with Uzzors2k bench smps schematic:

For testing, I removed the current feedback and protection and put the TL494 pin 4 with a 1k resistor to ground, also a 100nF cap in parallel.

My oscillation frequency is set at 50kHz and my transformer has a 10:1 ratio (not exactly but gave me about max 20V at 320VDC input at bridge).

My target output is 0-12V @ 20A supply. I'm using two IRG4PC50UD IGBT's to switch, its a bit overkill, but its only thing I have to test with 230VAC mains. The output retifier is four (only using 2) 100A 600V Ultrafast Minibrick diodes :

I made a input filter normal as any computer powersupply has. Capacitor -> Indutor -> Capacitor (Actually, I removed a Power supply input filter and put at my power supply.)

The supply is running fine, but I get a nasty noise at output, and the output has a good filter (about 100uH inductor and 10.000uF Cap, and 470nF Cap arranged as Pi Filter), also the noise starts even with no supply hooked up to the bridge, only with gate signals. if I hook up the power supply, the noise amplitude gets bigger (normal).

I measured with the scope, the noise at Bridge supply (exact over the Two IGBT's), here we have:

Thats the noise without the supply (only with gate signais):

And here with the circuit working (gate signals and power supply):

I'm not worried about the supply noise, but with the output noise. I get about Vpp noise at output (I dont think its ripple, because even with no load, the noise is the same with a 10A load (with 10A load I get a little ripple, but the noise is bigger).

I saw that all supply have a RC filter across the Diodes of retification, I tryed too as you can see in the bridge pic. Its a 10R resistor in series with 10nF cap. It reduced the noise (as all other filter done too) but I think the noise level is critical. 12V Supply with 2Vpp noise (swing between 10 and 12V).

Is there any tips to reduce the noise? I'm not too familiar with SMPS's.

Also, changing the voltage output in the potentiometer, doesnt change noise level.

Thanks!

[UPDATED 17:35 GMT-3]

Discovered relationship between Noise and switching wave. It seens to be ring:

2Spoons

Wed Jun 08 2011, 11:56PM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

This may not fix your problem, but is worth checking. Connect the scope probe to the same point as the gnd clip and see how much noise is being picked up directly by the probe - this will at least let you see how much of the noise is 'real'.

MRacerxdl

Thu Jun 09 2011, 12:36AM

Registered Member #989 Joined: Sat Sept 08 2007, 02:15AM
Location: SÃ£o Paulo, Brazil
Posts: 476

Yeah, the noise is real. I already tried that. Also, it only happens when I turn on the logic side (eg: With gate signals).

I tryed to change the gate resistor (doubled) but it didnt change anything :/

Jrz126

Thu Jun 09 2011, 12:57AM

Registered Member #242 Joined: Thu Feb 23 2006, 11:37PM
Location: Erie PA
Posts: 210

what are the specs on those diodes? Have any slower ones to try?

MRacerxdl

Thu Jun 09 2011, 02:15AM

Registered Member #989 Joined: Sat Sept 08 2007, 02:15AM
Location: SÃ£o Paulo, Brazil
Posts: 476

The diodes I'm using is STTH120L06TV1, they are 70ns delay. I only have MUR860 here, and they are 60ns...

I only have a few schottky diodes, but I didnt found any speed info at the datasheets. It worth to try?

Nicko

Thu Jun 09 2011, 03:56AM

Registered Member #1334 Joined: Tue Feb 19 2008, 04:37PM
Location: Nr. London, UK
Posts: 615

May be a bit of a red herring in your case, but I've designed and built a few commercial SMPS PSUs for audio use (mainly tube audio) where noise is a sensitive issue and some for amateurs (e.g.

)

You are doing the right thing for the output side, but you need to look more closely at the source of the noise rather than just trying to cure the symptoms.

I'd start with the layout - breadboards are never good for low-noise SMPS work - they are full of stray capacitances that couple switching noise back into the feedback & control side of the system. With low-noise SMPS design, layout is absolutely critical.

I always use a ground plane and make sure that all high current transients are kept well away from the feedback pin, both on signal & ground return.

Getting an SMPS to "work" is one thing, getting it to work properly is quite another, and breadboards are not good for that. I've always found that the RF engineers trick of "Manhattan-style dead bug" development works pretty well:

Manhattan Building Techniques:

Manhattan Style:

(Swedish - use Google translate)

In addition, SMPS supplies can inject noise back into the mains supply, that then finds its way into other equipment - a good mains filter will help, as will shielding - switching big currents at 50kHz will create a lot of EMI, which again will get everywhere....

HTH

2Spoons

Thu Jun 09 2011, 04:43AM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

You changed *the* gate resistor - do you mean the one on the primary side of the GDT? Is that the only gate resistor? If so, I would put resistors in the gate lines of each igbt - say 4 ohms or so - to prevent ringing between gate capacitance and GDT leakage inductance.

Pinky's Brain

Thu Jun 09 2011, 02:37PM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

Could be the diodes snapping, slower diodes have more reverse current but block it smoother ... so they cause less spiking.

Schottkys and SiC have effectively no reverse current and thus can't snap it off either ... so yeah, if you have appropriate voltage/current Shottky/SiC diodes use those.

Arcstarter

Fri Jun 10 2011, 01:50AM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

I'd start with the layout - breadboards are never good for low-noise SMPS work - they are full of stray capacitances that couple switching noise back into the feedback & control side of the system.

I would like to add emphasis to this. If you take a 74hc14 and make an oscillator using the 'first' gate (pins 1 and 2), you will notice that on a breadboard, all of the inverters on that row of pins will probably oscillate due to capacitance coupling the output of the first inverter to the inputs of the others. The error op amps in a TL494 are very sensitive, so that is definitely something to look in to. A big buck converter i made (not yet finished) had a ridiculous output waveform, until i shortened the wires from the output to the error amp for voltage feedback. It works well now, despite being imperfect... As hell.

ConKbot of Doom

Fri Jun 10 2011, 11:17AM

Registered Member #509 Joined: Sat Feb 10 2007, 07:02AM
Location:
Posts: 329

+1 to trying to lower parasitics. I designed an optocoupler interface at work with a 300v in max current source for the opto coupler. Testing it out on prototyping breakout boards, with ~4 inch leads on the part, + BNC to mini grabber clips made for all sorts of nastyness, enough bounce on turn on to actually have the output go back low again. Between the fast edge from the signal generator, the RG58 test lead(capacitance, maybe transmission line effects since the sig-gen had a 10 ns rise time), and stray inductance in the wire leads, it was doing that.

A tweak in where a peaking cap was in the circuit, reduced it some, and the final version had ~1/2" wire leads from the BNC to the PCB, and I cant get it to misbehave, even if trying.

1 2 next

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