RFC: Switching mode power supply schematics

[KW1252]

That LT1086 was for reference only

LT3694 is an expensive bugger, but on the other hand it contains two linear/LDO regulators, so if for example there is a need for regulated +12v rail (EGT circuitry etc) it's available with minimum component additions. If it's felt the linears should be discrete controllers, then a cheaper chip should be used.

[jharvey]

I see the LT3694EFE#PBF from Newark/Farnell/Element14 and Digikey, I know the first three are basically the same vendor, but that is technically multi vendor. Also it's around $9 to $8 in qty 1, and almost $5 at qty 25. I don't see that as expensive, although I guess it is compared to the LDO's

I think it's a good find, and I hope to see it as an option on future boards. Not required, but optional.

[jbelanger]

3.3V out of one of the LDO regulators might be interesting for an SD card.

[nitrousnrg]

I'm having a though time trying to get those components from a single supplier, I think its going to fall from my testing BOM...

I started to like that chip, although I see it a bit expensive if you count the parts it needs (I didn't search for similar and cheaper passives). Also, I see it valuable to have the option to remove the whole SMPS and just jumper VBAT to the 5v linear regs, it saves $10 or more for those who won't care about temperature or efficiency.

If I can't get the components I'd like to see some real testing data, in case someone does that work :-)

[nitrousnrg]

Ok, i did some testing. Fortunately I didn't post it, because it would have been a waste of time.

I measured 60mV of ripple at high loads, but it was because of the oscilloscope probe.

Right now, the other probe gives me 20mV of noise coming from somewhere, with the board connected but not powered to an AC power supply. Its just a matter of suppying it with a battery. That will happen in some hours.

Anyway, the difference in noise between "powered up" and "just plugged and powered off" is barely different. I only need to tidy up the testbench.

In the meanwhile, pics of the setup will be here:
https://picasaweb.google.com/lh/photo/I ... directlink
The wrong measurements are going to be deleted (If I get why I can't delete any picture, or change captions)

[jharvey]

Nice work. I'm curious did you get a scope capture picture you can post? I suspect your noise is caused by a soggy supply, that huge cap isn't working the way you expect. I'd bet if you put in some smaller caps on it, you would drop the ripple quite a bit.

[Fred]

If the huge cap is an electro, Jared is almost certainly right. Their high frequency performance is very bad, usually.

[nitrousnrg]

If the huge cap is an electro, Jared is almost certainly right. Their high frequency performance is very bad, usually.

Indeed, I was trying out the extremes. With and without the huge cap is pretty much the same, but I can't see accurately inside that much noise. Keen to get the car battery out for testing.

The setup is:
220uF cap -> SMPS (With LC filtering) -> 10uF cap near the reg ->0.1uF ->reg -> 0.1uF -> 22uF

0.1uF are ceramic, 22uF and 10uF tantalum, 220uF electroytic, and the LC filter has a 330uF tantalum

The complete album is here, check the bottom.
https://picasaweb.google.com/1065946256 ... directlink
10mV/Div in all the images

Sorry, I thought the first link would make the other pics visible, and it doesn't. Also, I moved to picasa since imgur hides pictures when you exceed the 220 pictures limit :-/

[jharvey]

Do you recall the time scale?

[jharvey]

Oh, also scope make model, and probes used. Did you let the scope warm up?

If cranked to the scopes limits, it might be the scope that's causing problems. I just attended a Tektronics show, and part of a lab I did included using a calibration step. With out it, the latest and greatest scope had an error of 5.6%. Basically the scope probe lead capacitance varied that much from day to day, that it would wander a bit. Letting an older scope like that warm up, can increase your accuracy but a fair bit, especially if you push the scope limits.