Thanks for your suggestions!

On 13 sty 2012, at 17:17, John Rehwinkel wrote:
>> I have a supply that kind of works, but I get a large sawtooth-like ripple 
>> on the output. The ripple is 17V peak-to-peak at 170V and the frequency is 
>> 23-80Hz (depending on the load).
> 
> There are four things I would look at: current limiting, feedback 
> oscillation, power supply problems, and thermal issues.
> 
> The MAX1771 has built-in current limiting, which could cause the whole 
> shebang to cycle on and off, which could appear as a sawtooth after the 
> output filtering.  Make sure you have solid, low impedance connections to the 
> current sense resistor and back to the CS input.

The board layout is right here, in case you'd care to take a look: 
https://skitch.com/jrychter/g2tq6/max1771-psu-1.0

This is heavily based on Nick de Smith's layout, although I might have made it 
worse. Of course I can share the full design if I manage to work out the bugs.

> Power supply problems are the easiest - see if there's any similar 
> oscillation at the power supply input pin.  Nick's diagram shows separate 
> power inputs for the inductor and the MAX1771 - are you powering them 
> separately?  Either way, look at the input power, if it's a lab supply that's 
> going into current limit and retrying, you could get behaviour like this.  If 
> you're not powering them separately, it might be worth a try.

Hmm. I'm powering the board from a (cheap) lab power supply integrated into my 
soldering station. It is supposedly capable of providing 1A @ 12V and I'm only 
drawing between 50mA and 200mA, but…

I looked at the supply with a scope and there are dips of around .75V that 
correspond to my ripple frequency. I thought those are to be expected. Nothing 
more serious, though. Would 1771 be that sensitive?

I have a single supply trace. There is a 100uF capacitor, then the inductor, 
and then the 10uF tantalum cap right next to the 1771. I never thought about 
separate traces for power.

> For feedback oscillation, you'd have a phase shift happening somewhere in the 
> voltage regulation loop - or something that pretends to be phase shift.  I'd 
> look closely at the feedback resistors and make sure there aren't any 
> parallel capacitors inadvertently hooked to them.  I'd also carefully check 
> the output capacitors - both the electrolytic and the high frequency one.  If 
> they're not doing their job of absorbing rapid spikes and hash, the 
> controller chip can get confused about what's really going on.  Scoping those 
> points is tricky, due to all the high voltage, high current stuff going on 
> nearby, which will tend to couple into the probe lead, upsetting the 
> regulation further and confusing the scope display.

I thought feedback oscillation would result in higher frequency ripples. I 
don't see any parallel capacitors, just the ground plane below. My feedback 
path is substantially longer than in Nick's design, but it originates straight 
at the 4.7uF output reservoir capacitor and is relatively far away from 
anything else (everything around it is ground), so I thought I would be fine 
here.

The other modifications I made to the feedback path are: 1) I used two 
resistors in series because of high voltage and 2) my divider is really 1360k - 
8k2 (+5k pot) instead of the more usual 1M5 - 10k (+5k pot), because of the 
resistors I had. I did not expect this to make a difference.

Oh, also, the final 100nF capacitor is not mounted yet (I don't have one). I 
assumed this would only make a difference for higher frequency switching noise.

> The last thing that occurs to me is a thermal issue.  The MAX1771 will 
> throttle down its on-time when it gets hot, which could lead to thermal 
> cycling. 

This is definitely not the case, nothing gets hot except for the load 
resistors, I haven't gotten anywhere near peak currents this should be capable 
of.

I will keep looking based on your suggestions.

--J.

-- 
You received this message because you are subscribed to the Google Groups 
"neonixie-l" group.
To post to this group, send an email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/neonixie-l?hl=en-GB.

Reply via email to