Strange SMPS configuration. Can some decipher it for me.

Good idea.

This schematic is hard to read, but you can get a reasonable picture from it.

The SMPS controller appears to be on a floating ground. The mosfet switches power from the DC Bus through an 8.2 ohm resistor to the floating ground. The inductor is connected to this floating ground. Now I understand the purpose of the resistor is to provide current feedback, but thats about where it stops. I can kind of see how it works, and i like it, the whole controller sits between the inductor and the DCBus.

I tried to google for app notes on this but without sucess. I take it if I were to use this same configuration with a different controller, then I would be wise to clamp the voltage across the controller itself?

Regards:

Andy

I emailed IR and got a decent copy. I was trying to find it but I must have deleted the email. I do have a printed copy. If helps any, i will scan the print I have and post it to ABSE or on my website.

Ok, can be found at

or

Be warned, the big one is about 4mb. The smaller is quite visible and is about 1Mb. Sorry I cant get them smaller, i lack the tools on the computer. LEt me know if you want it on ABSE

What is the advantages of doing it this way over say fixing it to ground? The only advantage that I can see so far is by not having to put a resistor between ground and the load for current measuremnt, which I assume would be an advantage in the situation at hand.

Thanks,

Andy

Its a buck converter. What they have done is float the entire peak-current-mode controller with the N-channel MOSFET. The start-up behaviour would be tricky to get exactly right, but nothing very funny here. Thanks for the nice pic :)

Cheers Terry

lack of a level-shifted gatedrive is probably the biggest benefit.

a "conventional" buck would use a P-ch (or PNP) switching element, with a ground-referred voltage-mode controller.

The current sensing resistor senses switch current, not load current (although when the switch is on, the two are related), but yes, this is a real PITA with a "conventional" buck converter.

they would have to be careful to ensure it starts up properly, but the UC3842 has an UVLO, so it ought to survive regardless of how shonky the supply is. DCM would be an issue - D8 keeps the 3842 supply "pumped" every time the FET turns off and inductor current commutates to D13, pulling U5-8,U5-9 to a schottky drop below 0V. During DCM, this can be quite a short interval.

Cheers Terry

In article , The Real Andy wrote: [...]

There are two places that allow you to get at the switch current without having to have a large common mode AC voltage. You can sense at the input side of the switch or at the load side of the inductor.

Input side of switch:

Linear and several others make chips for this purpose. You can also make a do-it-yourself current mirror for the job.

Inductor output side:

If you are doing a roll-your-own switcher controller out of an LM339, this is the way to go. You modify the normal control circuit like this:

Rsense --+-/\\/\\-+-- Vout Vref ! ! ! ! \\ \\ / / /R1 / R2 \\ \\ \\ \\ / / ! ! \\ \\ +-------------!-\\ ! ! ! ! ! ! >-+--- +--/\\/\\/--!!-+ \\ +------!+/ ! ! ! /R3 ! --- ! ! \\ \\ --- +----!-\\ ! ! / R4 ! ! ! >--+-+--------- \\ / --!+/ ! ! \\ ! --- GND / Vref --- \\ ! ! GND GND

This provides a foldback current limit of the switch current. You may get foldback of the load current too depending on other circuit characteristics. Adding a resistor from the "off time one shot" capacitor to the output voltage will give you foldback of the load current.

R1,2,3,4 are unbalanced enough that the switch certain to be held off when there is zero load.

OK, getting this now. I was going to attempt a similar concept with a TL494 (easy to get at local hobby shops) but now I am thinking this may be a bad idea. I can see many areas where the TL494 wont cut it now. I think I am left with using the same controller used in the IR dev kit or going with a different concept with the TL494.

Sooo, the next question is, how best to sense switching current or load current with a conventional buck converter without lifting the ground? Am I going in the wrong direction?

Can you point me to a good app note on the former topology using the

Many thanks for the help, tis appreciated. To have even got a working circuit so far amazes me considering I am a former digital person who now only does software.

Regards:

Andy

In article , The Real Andy wrote: [...]

Pound for pound, a N channel FET has a lower RDS(ON) than a P channel FET of the same voltage rating. To fairly compare prices you need to compare ones with equal RDS(ON) and voltage ratings.

I've run into the problem of there being no 600V P-channels out there and solved it in a very ugly way. I made the whole controller chip etc fly up and down with the source of the N channel FET. This works well but is very nasty to trouble shoot when it doesn't. You have to put a psudo ground plane under the switcher chip and stuff so you end up with a fairly large Cstray to deal with. This is not a circuit you want to try to do at

Thanks Ken. After much reading (On Semi and Philips have a great selection of SMPS app notes) I have decided to follow the current mode path. In the application I am using it (same as the pic provided) i think the current mode controller has significant advantages over voltage mode. I am really warming to the IR design, just got to go find some app notes on that method now :). As an added bonus, the UC3842 is about half the price of the TL494, so the hobbiest who build this will be even happier!

Thanks all, rest assured i will be back with more questions :)

Andy

One more question, why chose the N-Channel Fet? A few quick serches of online suppliers seems to suggest the the P-Channel breed is cheaper. I did notice that there where less P-Channel parts available, perhaps this is the reason? I did notice however, for a search of 600V P devices, no results, 500V, about 3....

Andy

I "printed" Scan20001.tif to a 300dpi PDF via Acrobat... 130kB, very readable, zooms quite nicely.

...Jim Thompson

current mode is the best IMO. A well-designed smps should happily run with a dead short on the output.

Cheers Terry

For US$40 you can download and register PMView for 'doze or OS2-eCS. Excellent viewer as well as image format converter.

Scan20002 .png 101,682 Scan20002 .tif 1,088,064

Scan20001 .png 279,289 Scan20001 .tif 4,352,434

The PNG format is about as good as it gets for b/w images or ones with a few colours. JPG is best for colour photos.

Ted

Cheers guys, will keep that in mind. I didn't have a pdf converter at hand at that time but I think that too would have been my preffered method.