Designing buck converter power stage

[alltheway]

Hello to all,

I am trying to design a step down converter and i have issues with the ripple in power stage.
The converter is powered from a 24v bus battery and i have a higher ripple at input and may affect other devices connected to the battery.
How can i reduce the ripple at the input?
I already tried to add capacitors at input but the result is the same 

Below you find the requirements and a few waveforms measured at different points.

Input voltage: 22v - 28v
Output voltage: 14v
Output current: 10 A (continuesly)
Switching frequency: 200 Khz

Input ripple: <= 20 mVpp
Output ripple: <= 20 mVpp

Have a nice day 

[Siwastaja]

As you already have RC snubber over the diode, give it a shot and add another one over the FET. It won't help with actual ripple current, but if you are measuring switching noise, it can help. (Even though your switch node looks quite clean.)

[alltheway]

I'm going to do that because i want to improve the switching performance.
At 10A the mosfet temperature goes up to 65 degrees celsius on a 45x90 mm aluminum plate (tickness is 2mm) and in my point of view is to high.

But for the moment let's go back to the ripple issue.
I changed the load to 5A and is worst, now the voltage goes up to 200 mVpp.
Does not make any sense...why at 5A load the ripple is higher than the ripple at 10A load.

[TheUnnamedNewbie]

I think at higher currents di/dt will be higher and thus the inductor will filter harder.

[MagicSmoker]

Worst case ripple current through the input capacitor of a buck converter occurs at full load and 50% duty cycle (Vout = Vin/2, neglecting losses) and as the buck draws pulses of current from said capacitor it can be a bit of a challenge to minimize the noise/ripple reflected back to the source.

You can use some inductance before the input capacitor to assist with filtering/ripple reduction, but be aware that said inductance can cause instability/oscillation and may require either a damping resistor (either across the inductor, or in series with a DC blocking capacitor then in parallel with the input capacitor).

[Niklas]

In a 1kW frequency inverter for a water pump that I worked on a couple of years ago most of the filtering an buffering was made with metallized polypropylene capacitors. 230 Vac was converted to 450V dc by a PFC and stored in a 470uF electrolytic capacitor. Apart from that, most of the capacitors were X2-rated film 220nF, 470nF and 3.3uF with some addition of SMD ceramic 1nF 3kV. The switching frequency was about the same as in this case.
The electrolytics at the input might not be fast enough for the switching frequency and also too small in capacitance for the load current. Look at the input voltage and either the output or switch node voltage. Switch on => dip in intput voltage, switch off => recovery of the input voltage

[T3sl4co1l]

What's the controller and compensation circuit?

[mrpackethead]

What's the controller and compensation circuit?

That was my question as well. 

[Phoenix]

Can you run the circuit open loop from a function generator? I'm not sure which ripple you're worried about - the low frequency input ripple which is suspiciously ~10ms (100Hz/120Hz), or the 200kHz switching ripple?

I'd first recomment a ceramic capacitor on the input and output side parallel with the electro's. Say, 100nF, it'll help with any high freq. spiking.

(Also, that's one hell of a snubber, how did you design it?)

[diyaudio]

1) 10A peak current?. PCB Layout is mandatory.
2) Type of Load? Large di/dt.. become a human network analyser  , check ALL switching node especially the high current return paths.
3) Controller/PID? Type of compensation network I, II or III ?
4) Wrong Catch Diode, too much Qrr ?
6) What coil/type of material are you using? you could be saturating the core. don't randomly select switching frequency, random turns? study the resonance of the whole system.
7) Loop simulation may help.. 

Did you just randomly select values ? if true, start with the math its not that tedious.. a sea of app notes available on the net.

 
   

[diyaudio]

1) 10A peak current?. PCB Layout is mandatory.
2) Type of Load? Large di/dt.. become a human network analyser  , check ALL switching node especially the high current return paths.
3) Controller/PID? Type of compensation network I, II or III ?
4) Wrong Catch Diode, too much Qrr ?
6) What coil are you using? you could be saturating the core. don't randomly select switching frequency study the resonance of the system.
7) Loop simulation may help.. 

Did you just randomly select values ? if true, start with the math its not that tedious.. a sea of app notes available on the net.

 
   

Plecs will help, if the OP has access to it.

Even simple Ltspice will do. Gosh back in the day it was primitive brute force paper and pencil. "non of this simulation rubbish"   

[alltheway]

Thank you all for the replies.

Bellow is what i did to try reducing the input ripple.

- I added polypropylene caps up to 10 uf and doesen't make any difference.
- I adjusted the switching frequency from 150 Khz to 400 Khz but with no luck.
- I boosted the input capacitance up to 3300 uF and the input ripple had the same value.

With all this changes the ripple amplitude and frequency remained the same, something is terrible wrong with my converter.
At 5A output current nothing heats up, converter works just fine only the input ripple problem.
I'm going to do the math again and redesign the power stage.

One thing, for powering the converter i use a AC-DC SMPS 24V 16A (i don't have a 24v battery at home), to be this the problem?
I ask because i put a 3A load on this power supply and i checked the ripple and it looked pretty much the same (170 mVpp and 800 Hz)

# diyaudio

1 - For tests i don't have a dedicate pcb (the power stage is build on a prototype pcb)
2 - I'm using resistive load
3 - Controller is TL5001 and i'm using type III compensation network
4 - I need to study this
5 - it's an arnold ring core MS-090075-2 with 12 turns (wire: 10x0.25 mm)

I used this document to calculate the values: http://www.irf.com/technical-info/appnotes/an-1162.pdf

# Phoenix

I'm worried about low frequency ripple, attached you find how to design a snubber circuit.

Thank you all for helping me 

[diyaudio]

Bellow is what i did to try reducing the input ripple.

- I added polypropylene caps up to 10 uf and doesen't make any difference.
- I adjusted the switching frequency from 150 Khz to 400 Khz but with no luck.
- I boosted the input capacitance up to 3300 uF and the input ripple had the same value.

The above  mentioned are all typical misconceptions, throwing more capacitance or increasing fsw.. they will only make the problem worse.

Before you do anything what does your PCB layout look like? 
   

[diyaudio]

I used this document to calculate the values: http://www.irf.com/technical-info/appnotes/an-1162.pdf


Thank you all for helping me 

Did you read the datsheet at least 30 times ? Im not joking. Its says.

output transistor
The output of the TL5001/A is an open-collector transistor with a maximum collector current rating of 21 mA and
a voltage rating of 51 V. The output is turned on under the following conditions: the oscillator triangle wave is lower
than both the DTC voltage and the error-amplifier output voltage, the UVLO circuit is inactive, and the short-circuit
protection circuit is inactive.

21mA is not enough to saturate most power FET`s used as a switch,especially at frequencies and current levels you working with. .. you need to calculate gate drive requirements.

A simple totem-pole arrangement can work as its cheap and reasonable buffer drive system, no need for expensive gate driver ics..but if you have one at hand then use it. your mosfet may be operating in ohmic mode. 
       

[alltheway]

AcHmed99

I will change the power supply with a linear one.

diyaudio

The mosfet is not directly connected to TL5001.
I use a gate driver (IR2117)

I made the math again and i get pretty much the same values for the power stage components.
I didn't tough is such a pain to design a buck converter   

Bellow are a few screenshots with the waveform after the input low pass filter (measured across the mosfet D-S)

 

[T3sl4co1l]

Still waiting on a schematic and pictures/layout.

Tim

[johansen]

3 - Controller is TL5001 and i'm using type III compensation network
[...]

I used this document to calculate the values: http://www.irf.com/technical-info/appnotes/an-1162.pdf

try using a type II

[mrpackethead]

You can always cheat a little and use Texas Instruments workbench system. Its actually pretty good.

[Richard Head]

It looks like you have a control loop instability possibly caused by an excessively high output impedance of the power source. I would also be inclined to reduce the value of the input filter inductor.

[diyaudio]

Still waiting on a schematic and pictures/layout.

Tim

+1

[Siwastaja]

Really, get a car battery or something to test it with. Your converter's input ripple is the PSU's output ripple, so you can't easily see which one is causing the ripple, or it might be a problem that only exists with the combination of the two.

[eneuro]

... get a car battery or something to test it with.

Do not forget about fuse, alse if something goes wrong your PCB (if you have one) tracks will explode in few hundreds amps rush current with "magic smoke"

[alltheway]

I'm in a bussiness trip these days and i can't work at this project.
When i will be back i will remake the schematic and i will post it, so we can continue.
I will try to get 2 car batteries to be sure the power supply for my converter is clean, i don't trust this chinese smps that i have right now.
For the moment i want to thank you all.