A niche chip (LT4320) for mains powered AC -> low volt DC "linear" based psu ?

[necessaryevil]

The LT4320 DIP version is available!

I asked them about the DIP-version, it turned out that they simply forgot to set to flag the part as active! It's now on their website and it will be available on digikey soon.

I made a 2 layer PCB design with  the help of some people on a dutch forum. Should be able to handle 15 amps. I'll order a small batch on itead.

[digsys]

   The LT4320 DIP version is available!  ...

Well, that'll help with heat dissipation in the IC, and that slightly tricky thermal pad.

... Should be able to handle 15 amps. ...

With the FETs I've tested, max 2.5mR RDSon, 15A is a piece of cake. 16-20A and you start to need to think about some heatsinking.
They should be good to 30A. I've only done short bursts to that, and all looked OK.

[peter.mitchell]

hey digsys, have you tried it above 600hz?

[digsys]

   hey digsys, have you tried it above 600hz?   

Sorry no. I didn't buy the higher freq version, as I mainly work in 50-60Hz. But we do have some systems running 400Hz,
so next test run.

[peter.mitchell]

   hey digsys, have you tried it above 600hz?   

Sorry no. I didn't buy the higher freq version, as I mainly work in 50-60Hz. But we do have some systems running 400Hz,
so next test run.

Ok, doesn't particularly matter if you buy the high spec or not, I am more interested in this things limits, i already know mosfets are good high current switches.

[necessaryevil]

With the FETs I've tested, max 2.5mR RDSon, 15A is a piece of cake. 16-20A and you start to need to think about some heatsinking.

Well, I meant the PCBs could handle 15A (calculated, conservative settings)

I use similar mosfets in my design. I did some hand written calculations and I came up with the same conclusion about heat sinking. Nice to have this conclusion confirmed.

[digsys]

  Well, I meant the PCBs could handle 15A (calculated, conservative settings) 

Did you use 4oz Copper? Any other "tricks"? Exposed section/over-solder/silver-solder/wire over tracks?

[digsys]

Ok, doesn't particularly matter if you buy the high spec or not, I am more interested in this things limits, i already know mosfets are good high current switches.    

OK seems there's a mis-communication here :-).   LT4320 is only rated to 60Hz   LT4320-1 is rated to 600Hz.  I didn't buy the xxx-1

From the datasheet - "Higher frequencies of operation are possible depending on MOSFET size and operating load current"
From my waveforms, I only saw very little degradation on gate drive when I pushed the FETs to 20A+, so I'd say if you stay
under ~10A, the max freq could be QUITE high.

[BravoV]

hey digsys, have you tried it above 600hz?

Just curious what are you looking for in using this chip way above it's specification ? 

[peter.mitchell]

hey digsys, have you tried it above 600hz?

Just curious what are you looking for in using this chip way above it's specification ? 

Magic, rainbows, world peace, ect...

But seriously, a high current, high speed, low loss rectifier would be amazing for lots of applications.

[tszaboo]

I'm guessing that gate driving is what is the limit here. There are no external capacitors, so the internal charge pump only can supply so much current into the mosfet gate. So at higher frequency, it cannot close/open fast enough (especially at voltages like 72V) and efficiency drops. Now if only we would use MOSFETs with small gate capacitance and lower voltages...

[danifel]

This is a chip built for PoE, to replace the 70V diode bridge present at every PoE Powered Device. But it is really not the best alternative around for this purpose. There is an Ideal Diode Bridge by Microsemi, PD70224, which is a lot more integrated.

[tszaboo]

This is a chip built for PoE, to replace the 70V diode bridge present at every PoE Powered Device. But it is really not the best alternative around for this purpose. There is an Ideal Diode Bridge by Microsemi, PD70224, which is a lot more integrated.

And also a future product. From a company which is not one of the few companies I would buy stuff from.

[digsys]

  This is a chip built for PoE, to replace the 70V diode bridge present at every PoE Powered Device.... 

I disagree with you here .. I don't see many PoEs needing to supply 20A without needing a heatsink.
It's exactly what it says it is - a POWER Supply bridge rectifier replacement, and they are working out perfect in our products.
Your PD70244 is NOT in the same ballpark. It's 1A (not 20A+), has 160mR inbuilt FETs (I'm using 0.9mR !!) AND it drives these
10,000pF NFETs absolutely fine to 600Hz (and beyond IF I limit the current under 20A !!).. which is ALL I (we) need for DC rectification.
I have no idea why you think it's for PoE

[BravoV]

This is a chip built for PoE, to replace the 70V diode bridge present at every PoE Powered Device. But it is really not the best alternative around for this purpose.

Curious why you claim you know better than Linear Technology that designed it ?  Reading the wrong datasheet maybe ?

Fyi, Linear Technology made a lot of varieties of PoE ics, try visit their website once in a while. Not an expert, but I'm pretty sure they know what they're doing.

[M. András]

i would love to see carry more then a few amperes by those ethernet cables

[tszaboo]

i would love to see carry more then a few amperes by those ethernet cables

It is rated for 0.6A
Voltage about 60-70V though.

[M. András]

i would love to see carry more then a few amperes by those ethernet cables

It is rated for 0.6A
Voltage about 60-70V though.

sorry i meant to be sarcastic. i know their ratings flimsy little signal cables thats what they are for me

[tszaboo]

i would love to see carry more then a few amperes by those ethernet cables

It is rated for 0.6A
Voltage about 60-70V though.

sorry i meant to be sarcastic. i know their ratings flimsy little signal cables thats what they are for me

OK, I should go to sleep. Jóéjt.

[necessaryevil]

  Well, I meant the PCBs could handle 15A (calculated, conservative settings) 

Did you use 4oz Copper? Any other "tricks"? Exposed section/over-solder/silver-solder/wire over tracks?

The only trick is that I use copper planes on both sides of the pcb. For the DC path I use about 2x300mil. The copper layers are only 1oz/ft^2. I'm not really going to use them for 15 amp, just a design goal.

I used this calculator: http://circuitcalculator.com/wordpress/2006/01/31/pcb-trace-width-calculator/

Some images:



I just have sent the second version of my layout to Itead.
I've received the v1 pcbs and got them working, but I realized that I made some misstakes. The holes for the input and output wires are too small and I forgot to connect one plane (resulting in current flowing through only one side of the pcb).

//edit
legenda:
Green = bottom layer
Red = top layer
Yellow = both layers

the white squares are for the wires. They have a 3.2 mm hole in them (m3 screw), which is not visible in the pictures.

size: 50x50 mm (about 2"x2")

[PerBuch]

Is I possible to buy a couple off your V2 boards ?

[digsys]

A blast from the past :-)  If you don't get any answer from mr evil, I should have heaps of several versions lying around. FOC .. was in the process of throwing
out heaps of stuff anyway. Just let me know. Regular postage here is quite cheap.

[PerBuch]

Is't it expensive to send from Australia to Denmark ? :-)

[exe]

It's probably cheaper to order new boards from China.

[digsys]

Is't it expensive to send from Australia to Denmark ? :-)

If I can stuff it into a 260 x 360 x 20mm envelope (padded) ~ OZ$8-18 (125-500gms
Else package ~1Kg OZ $30 ish
Not sure about your end taxes etc. Gift?