problems with compensation in boost regulator

[chimi]

Hi everyone I am new at this forum!!!
I am designing a boost regulator to get 15Vdc-4.4A from the +12Vdc from the PC PSU. Thing is that I seem to fail somewhere in the design of the error amp compensation network, first of all I want you to know that everything is free of noise and well snubbed with RCD snubbers so the are apparently no current transcients when switching on
The circuit uses a current mode IC: the famous UC3845 and the MOSFET is an IRFZ44N which has a RDSon=17.85 mOHM which is close to ideal

the configuration of the compensation network is the typical 1 diode and 1 resistor paralleled that some authors call "single pole with in-band gain", no optical isolation (obviously).

What happens: some of the switching pulses (normally 1 out of 2) are stopped just at turn on time, thus having almost no switching for the entire cycle, at first glance I thougth It was the current sense, but the transient in the on periods that happen with no problem are really bigger than those truncated, so, I don't think so....

I attached a picture of this seen on my oscilloscope so you can get the whole picture
Thanks!!!

[jtkraj]

I will appreciate if you can post the schematics as well.

[chimi]

Sorry didn't send it before, but I had to make it "understandable" and put it up to date...
also you must know that the operating freq. is approximately 125KHz
Cheers!

[jtkraj]

In this case. i am not sure of using a snubber. I will not use a snubber while designing a buck converter.

[chimi]

But it is a boost converter, and before actually mounting the snubber in the PCB, at a 3A load, the noise was enough to make the inductor sound due to fast-changing duty cycle, the inductor current sense was crazy

After I single-point grounded the control logic and all of its external components, the supply works fine in light loads (0.5 to 2A) but at 3A and above I experience this problems again, not as severe as before, but it doesn't do any good anyway

Thanx!

[chimi]

As crazy as it sounds (at the beggining) I tried removing the snubbers and the supply worked just fine, I posted a pic of the stable switching after the removal of the snubbers.

I reached the conclusion that there was no need of a snubber and the single point ground for the logic was crucial, so before I implemented the single point ground, the snubber was "needed" because all the noise ended up in the logic, but afterwards only increased the voltage transcients over the sense resistor

You were right, thanks, you do not know how much I needed this to work

[cerrem]

You still need to examine the closed loop response of the system..
You can measure it on the bench...
But first it is advisable to get out the BODE plot semi-LOG paper.
Derive and draw out the OPEN-LOOP response, then your ERROR-AMP response....then you can see your CLOSED-LOOP response... Check your PHASE MARGIN and GAIN MARGIN....if they are not within spec, then re-adjust ERROR-AMP values....
Also...watch out for the variation of LOAD, which may put you in DISCONTINOUS or CONTINOUS MODE transisitions, this can introduce a RHPZ (Right Hand Plane Zero) which may put the loop unstable....


Chris

[Eva]

Snubbers are intended to damp ringing (reduce the amount of parasitic oscillations) and they may be required or not depending on the quality of the layout and input/output filtering.

Are you including input filter capacitors? They are not shown on the schematic.

[chimi]

Well...
I already tested abroad with a very fast-changing load: a 600Hz PWM switching over a 5A load all the range from 10% duty cycle to 100%, and works just fine, never goes in continuous mode.

Regarding closed loop response there are a lot of this I've got to learn, how to make a bode plot of phase and gain margin, this always was difficult for me, in high-school and in college... so I only do what I can with bibliography about power supplies on how to calculate just the components of the compensation network.

Regarding input filtering, i decided to try first without, see what happens and then mount the caps or not, and, seeing this works just fine, didn't put it in the circuit though, what I did not include in the schematic was a filter at the track just in the supply Vcc of the 3845, which is a series 12 Ohm resistor from the +12V followed by a 100uF aluminum electrolythic cap, which made really better the waveform of the out pin and the gate signal.

And finally the snubbers: I designed before no other thing than off-line supplies. generally flyback and push-pull and only one half-bridge, always isolated by the well known TL431 and the 4N25/35 opto, in this designs, a snubber was always needed because of the high RDSon and output capacitance of the 500V/600V transistors, in fact I tried all kinds of normal and lossless snubbers to see which was better for each of these topologies, even implemented a lossless snubber for the push-pull similar to that used in a flyback or a single switch forward converter, so, when I had instability in this particular circuit I thought "well, let's mount the snubbers" didn't know this would be a problem....

Thanks to you all !!
I'm still worried about closed loop response, I'll get deep into that

Regards!

[chimi]

By the way, here's the schematic of the circuit just as it's working, with input filtering for Eva and everything, including the modification I "narrated" before

Cheers!!!