dirty smps

[Simon]

So attached is the diagram and scope view of the circuit. I can't work out why I'm getting all the "ringing". perhaps because it's on a breadboard ?

I actually got a very good output when I coupled the mosfet straight to the comparator output. I'm using a load of 120 ohms. I'm not after a perfect output just a rough stepdown

[cybergibbons]

Quite likely - it's hard to breadboard SMPS and not have issues like this.

[Simon]

yes that is what I thought, there are wires all over the place, I was going to try vero boarding it to see if it improves

[muvideo]

Hello, perhaps only an error in the drawing, but  shouldn't D1 and C? be connected to GND?

[Simon]

yes drawing error, will correct

[Harvs]

This is kind of an interesting circuit...  What's setting the frequency of oscillation?  From what I can see, it's the feedback loop combined with the low pass RC filter on the voltage divider input to the comparator.

Maybe why your seeing less oscillation when you directly couple the mosfet gate to the comparator is because the comparator doesn't have the drive capacity to switch the gate as fast (only 16mA from the datasheet), and therefore it's an extra time constant in the circuit stopping it going into high frequency oscillation.

Either way, I personally would much prefer to use something like a nice and cheap TL494 controller IC that'll give you a fixed frequency output with PWM, soft start etc.  It also has a bandgap 5V ref on chip so you could get away from the regulator as a reference.  If you haven't seen them, have a look at the extensive application notes on the web.  They're an IC that's been around for ages and is in a huge number of computer power supplies. ~$0.30 from digikey last time I checked.

[SeanB]

Put some hysteresis on the comparator, use 1M between output and pin 5 ( or 4, depending) with a 1k resistor to the rest of the circuit. Should stop the oscillation almost totally. Try also adding a 1n capacitor across the output to ground.

[slateraptor]

If I'm reading that Rigol screen correctly as 5V/div, then all I have to say is that's looking damn nasty.

[Simon]

If I'm reading that Rigol screen correctly as 5V/div, then all I have to say is that's looking damn nasty.

quite, the actual Vrms is what is should be but there are elements of it at the input voltage

[SeanB]

Output I was referring to was the comparator, it should slow down the rise to high a lot, and this will stop the ringing which is probably the comparator ringing merrily as it crosses the switch threshold.

[Simon]

This is kind of an interesting circuit...  What's setting the frequency of oscillation?  From what I can see, it's the feedback loop combined with the low pass RC filter on the voltage divider input to the comparator.

Maybe why your seeing less oscillation when you directly couple the mosfet gate to the comparator is because the comparator doesn't have the drive capacity to switch the gate as fast (only 16mA from the datasheet), and therefore it's an extra time constant in the circuit stopping it going into high frequency oscillation.

Either way, I personally would much prefer to use something like a nice and cheap TL494 controller IC that'll give you a fixed frequency output with PWM, soft start etc.  It also has a bandgap 5V ref on chip so you could get away from the regulator as a reference.  If you haven't seen them, have a look at the extensive application notes on the web.  They're an IC that's been around for ages and is in a huge number of computer power supplies. ~$0.30 from digikey last time I checked.

Yes I'm relying on the natural delays in the circuit and the feedback loop to put the breaks on the frequency that otherwise will be a bit load dependant. I gave the project the tile or "dirty smps" before i even started because my aim is to just hash something reasonable together with parts I already have and that are quite generic without expecting a perfect flatline output. Ripple is ok as this is at the moment to power my laptop and as that will have it's own built in regulation I'm not fussed of a few hundred mV ripple as I'm not fussed about extrme output voltage acuracy which is why i bunged a cheap 78L05 in there. This is also a good learning exercise for me

Indeed driving the switch mosfet straight from the comparator gave it a delay of it's own, you could definitely see the slop in the gate voltage and the mosfet (with no heat sink) was getting hot with not a lot of load. The mosfet gets less hot with the buffer/driver in the circuit despite the amount of switching. The .1 uF capacitor in the feedback loop seems to make no difference at all, even a 10 uF capacitor did not change anything.

I'll try the capacitor on the output next.

I know I can get a cheap reg chip to do this but the idea was the learning curve and an excuse to put hand to breadboard and scope.

[Simon]

The 1M feedback resistor and output capacitor made some difference but now the output is every rippled and there is still a lost of ringing. Pushing the wires about causes random reactions so I'm guessing that it maybe time to move away from the breadboard.

[Simon]

The capacitor on the comparator output actually causes the buffer BJT's to heat up, Perhaps to slow a signal now and causing shoot through ? Maybe I should increase the values of the resistors between the BJT's and the mosfet switch

[Simon]

I have to keep the gate drive resistors about equal as the mosfets gate cannot go to -23V (and later in the final version -28V).

I could increase their values. C4 has no effect on the timing or chatter. Perhaps I should look at a fixed frequency design. For the mosfet this needs to provide up to 5A with I guess 10+A switching bursts and the mosfet is a 23A one so about right margin ? As i said the mosfet seems to switch quite fast with the driver, the problem is the feedback loop