RFC: Switching mode power supply schematics

[jharvey]

This is an app that jharvey has found, and you can see the actual values I'm using

For some clarity, that was MoveX that found that ST page. I simply keep bringing it up because I think it's cool

[KW1252]

I did some searching on the subject and came up with two articles and they were quite toward using switched power supply with ADC's.

http://www.analog.com/static/imported-f ... d_ADCs.pdf
http://www.analog.com/static/imported-f ... ingADC.pdf

It seems the greatest worry is about powering the ADC rather than the signal itself, and at the moment the only ADC's in use are in the MCU, which is already powered with a linear unit... Signals should be easy to filter as they don't need to deliver much power.

[nitrousnrg]

Good to know, thanks Karry.

Checking the datasheet of that IC (Analog's) I see that the maximum input voltage is 5.5v, which leads me to think that if you want ripple readings in the microVolt order you can't go wild stepping down the voltage.

When I realized about that and checked my circuit, I saw that the SMPS present is in Puma's schematic right now has 14mVpp with a 14v input , but 6uVpp (!!) with a 6.5v input.

So, ADC+SMPS is totally cool if the input voltage isn't too high :-)

[KW1252]

I think I found a pretty good candidate. The LT3694 from Linear Technologies has adjustable output voltage and switching frequency (from 250kHz to 2.5 MHz), 2.6A Iout, input range of 4-36V and even two LDO regulator controllers (requires external NPN BJT's). It also has soft start, so it should behave pretty well. How does that sound?

[nitrousnrg]

Yep, sounds good. Willing to simulate the SMPS output ripple with a 14v input?

In a quick reading I didn't see performance values for the LDOs, I don't expect better values than a proper 5v LDO, but it would be nice to know.

[KW1252]

I'm building the circuit in LTSpice, I have the LTpowerCAD too but the damn thing won't work properly! Anyway, I'm working on it. I'll try some filter systems and we'll see how it turns out.

Too bad the switching systems are finicky about their layout so you can't just build one on a breadboard or even a vero board, you'll need to go for a proper PCB right away. Fortunately those data sheets often contain a preferred PCB layout

[nitrousnrg]

Cool, keep us posted. I'll share my tests results here, in 20 days or so, when I get my components :-)

[KW1252]

Here's the reference design from the datasheet:



It looks pretty neat, the only problem is it's designed for 4-layer PCB and the bottom layer is supposed to be untouched ground plane. However, I think we're going to need only one LDO output, so the traces on bottom layer will remain very small and long traces across the zone are not needed.

As for the simulation, well that doesn't look too pretty yet. The reference design yielded this v/t graph:



The 0.15mV ripple doesn't look bad, but that 1mV sawtooth does. I'll test a bit more and see if some caps and inductors will make it behave better.

[KW1252]

This was after putting an ample slab of ferrite on the board... at least it's not sawtooth. That was a typical stabilized condition, the maximum voltage peak-to-peak is still 1mV, but the RMS voltage much less.

[KW1252]

well that's odd and borderline embarrassing, those both graphs are screwed up pretty bad.

What I have now is a pretty funky looking modified 200µV sine wave at 1,8MHz modulated to a stable 2.4mV triangle wave at 7.5kHz that resists all filtering efforts. I guess learning the software better is the key.