Author Topic: A niche chip (LT4320) for mains powered AC -> low volt DC "linear" based psu ?  (Read 38592 times)

0 Members and 1 Guest are viewing this topic.

Offline tszaboo

  • Super Contributor
  • ***
  • Posts: 8217
  • Country: nl
  • Current job: ATEX product design
Datasheet sais exposed pad must be connected to OUTN. That pad also needs more copper to cool the device down. I dont see the reason to use bent TO220 if you can just use D2PACK mosfets. and more vias for high current. Just increase the copper size, it doesn't cost anything.
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
@digsys, thanks for the update.

Btw, are you planning to use gate resistor ? I thought we don't need that according to the datasheet.  ???

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: BravoV
Btw, are you planning to use gate resistor ? I thought we don't need that according to the datasheet.  ???   
Technically, this is a "prototype" layout for me, so I leave the capability of monitor points to test stability etc
If everything is fine, I'll probably leave them in anyway, and just use 1R resistors. Checking gate drive capability is
important to me, as it will hopefully go in a commercial product. Vin-Vop switching characteristics I can do externally.

Quote from: NANDBlog
Datasheet sais exposed pad must be connected to OUTN. That pad also needs more copper to cool the device down.
I dont see the reason to use bent TO220 if you can just use D2PACK mosfets. and more vias for high current.
Just increase the copper size, it doesn't cost anything.   
The posted artwork isn't 100%. The PCB is based on an existing 30A H-Bridge that I made years ago, and works
extremely well !! The PCB is 8oz, exposed, with a copper blocks under the TO220s, screwed down hard, which is why
I use them. Artic silver cement thermally bonds the crap out of it all. Thanks for your comments.
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Quote from: BravoV
Btw, are you planning to use gate resistor ? I thought we don't need that according to the datasheet.  ???   
Technically, this is a "prototype" layout for me, so I leave the capability of monitor points to test stability etc
If everything is fine, I'll probably leave them in anyway, and just use 1R resistors. Checking gate drive capability is
important to me, as it will hopefully go in a commercial product. Vin-Vop switching characteristics I can do externally.

Ahh .. I see, never thought of that, learn something new and good knowledge today, thank you.  :-+

Offline resistor

  • Regular Contributor
  • *
  • Posts: 74
So... how did the prototype turn out?
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: resistor
  So... how did the prototype turn out? 
Funny you ask .. All PCBs assembled and wired up, was about to Solder in the FETs I chose, but just found out the parts
are on hold because the account wasn't paid .. GRRRRRR
I'll probably just use the 25mR FETs I have, instead of the 0.9mR - 1.9mR that are on hold, at least just to prove it and
get a few readings (with my FlirE48+ :-) )  Hopefully answers in a few hrs
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
BTW, just noticed there is an update in the datasheet, and apparently its now available in DIP 8 package  :clap:, great news for hobbyist who want to build a traditional AC to DC step-down transformer based linear power supply.

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: BravoV
BTW, just noticed there is an update in the datasheet, and apparently its now available in DIP 8 package 
Just when I perfected my thermal pad artwork and soldering method :-)
Preliminary results - as per PCB linked before - Using CSD18502KCS NFET, TO220, RDSon=2.4mR, GateCap=3900pF
FETs bolted to a double sided PCB x3oz using Arctic Silver and no other heatsink
Ran at 10A continuous for several minutes - 4,700uF and 16,700uF - FETs barely warm.
- as above but at 15A continuous - FETs quite warm, but not too hot to touch.
The waveforms were the most impressive, absolutely clean with no detectable switching steps, in all conditions.
I'll run harder tests tomorrow, along with Flir images. So far - VERY impressed. The test PCB should go to 30A.
Hello <tap> <tap> .. is this thing on?
 
The following users thanked this post: jram

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Quote from: BravoV
BTW, just noticed there is an update in the datasheet, and apparently its now available in DIP 8 package 
Just when I perfected my thermal pad artwork and soldering method :-)

LOL ...  ;D , also curious why LT made such package, I thought thru hole is dead ?  :-//

Anyway, its so new, when I tried the LT's online shop, the DIP version still not available.


Preliminary results - as per PCB linked before - Using CSD18502KCS NFET, TO220, RDSon=2.4mR, GateCap=3900pF
FETs bolted to a double sided PCB x3oz using Arctic Silver and no other heatsink
Ran at 10A continuous for several minutes - 4,700uF and 16,700uF - FETs barely warm.
- as above but at 15A continuous - FETs quite warm, but not too hot to touch.
The waveforms were the most impressive, absolutely clean with no detectable switching steps, in all conditions.
I'll run harder tests tomorrow, along with Flir images. So far - VERY impressed. The test PCB should go to 30A.

Thanks for sharing this.  :-+

Sounds quite impressive, and 15A for my personal project should be more than enough.  :clap:

Please, if it's not troubling you too much, share few scope's shots the waveforms you mentioned above, also could you please do a startup capture, just curious to how it performs when turned on for the 1st time.

Btw, those tests were done at what voltage ? and whats your expectation or target on the current at your circuit ? 30 Amp ?  :o

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: BravoV
Btw, those tests were done at what voltage ? and whats your expectation or target on the current at your circuit ? 30 Amp ?
Ran them at 5VAC and 25VAC. Next I'll check lowest AC possible and a max of 35VAC (These FETs are only 40V)
I need to see that they run perfectly fine at 30A, in our 20A units. Still waiting for other FET types, including the 0.9mR ones.
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Next I'll check lowest AC possible ...

Great, so eager to see if the UVLO kicks in, and how those mosfet drivers perform at quite low voltage. I guess they designed this chip with the target circuit for linear regulator with output minimum at 5V, isn't it ?

Quoting the datasheet :
"OUTP Undervoltage Lockout (UVLO) Threshold, Conditions : (INn = OUTP, Other IN = 0V) , Min 6.2V, Max at 7.0V"
« Last Edit: November 28, 2013, 03:31:28 pm by BravoV »
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Only had time to run a few more tests, some pix later.
- With small load, min AC is app 6V. Literally just kicks in / drops out.
- Increased Caps to 60,000uF and 18A Load. Ran several mains on/off sequences rapidly - didn't bother it or confuse it. annoyed ...
- As above into a dead SHORT for 5 secs. FETs started heating up damn fast, IC still kept trying to run them, and instantly recovered
once short cct was removed. Transformer was reading AC 83A RMS !!!! at 24VAC ... it's a big transformer :-)
So far, failed to blow the sucker up. I'll try huge Inductor loads next week.
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
So far it looks pretty good eh ?  :clap:

About the upcoming destructive tests  >:D with inductor, do you put any TVS as recommended by the datasheet ?

With the inductor kick back voltage, I guess it can easily surpassed the absolute max. rating at 80V for the OUTP pin, so exciting to hear the results of this test.

Thanks for keeping us updated.  :-+
« Last Edit: November 29, 2013, 09:04:18 am by BravoV »
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
In Version 2 I used - IRFS3004-7P : D2PAK, 0.9mR, 280A, 9200pF, 40V. The LT4320 had no problems at all driving the higher Gate capacitance,
BUT once again there was a sudden LARGE temp rise at around 16A. Turns out that the IC will operate at 6VAC, but at ~16A+ you need a
min of 12VAC to give the charge pumps sufficient drive. (even though the waveforms looked perfectly OK).
With the new layout 3oz both sides, the PCB starts to get warm after 15-20 secs at 21A. Max dropout is ~100mV for a few mS during the
capacitor charge time. Once I upgrade the E-Load (once again) to 30A+, I'll try for a destructive test. All in all - I LOVE IT
More pix at - www.pbase.com/digsys/electronics_stuffz
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Thanks for the update !  :-+

I guess at 6VAC is barely enough power to feed the gate driver isn't it ?

Just curious, at +30A you mentioned that you are going for destructive test, at what voltage ?

PS : Seeing your photo gallery there, to be honest, I sort of lost the interest of this chip for a while, too much distractions, there are tons of cool T&M equipments you have there, just wow !!!!!!  :o
« Last Edit: December 25, 2013, 11:59:56 am by BravoV »
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Up to ~15A, 6VAC is fine, and it also depends on the level of ripple on the DC caps ! If you look at the Gate drive waveform,
it is heavily "influenced" by it. I don't have an exact correlation because I'd be guessing on how the charge pumps work.
Now that I know there is a 3 way relationship, I'll want to find the min AC at max I (at lowest reasonable ripple for now).
That sudden temp rise is most likely a result of "leakage" current, but now I have to set up an accurate current sense and
plot all the waveforms. At least I know it's not an issue over 12-15VAC for 20-30A respectively.

All that (old) gear is going to be sold, once I find time to work out the heck some of it is ! I'm also getting rid of several boxes
of various circuit boards from a 15yr old installation that's just been de-commissioned. So much for holidays :-)
Hello <tap> <tap> .. is this thing on?
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Now that I know there is a 3 way relationship, I'll want to find the min AC at max I (at lowest reasonable ripple for now).
That sudden temp rise is most likely a result of "leakage" current, but now I have to set up an accurate current sense and
plot all the waveforms. At least I know it's not an issue over 12-15VAC for 20-30A respectively.

That is almost half a kilowatt there.  :o

Btw, what is the 3 way relationships you're talking about ?

Looking forward to see the result on min AC while at max current, should be interesting and will be useful info.  :-+


All that (old) gear is going to be sold, once I find time to work out the heck some of it is ! I'm also getting rid of several boxes
of various circuit boards from a 15yr old installation that's just been de-commissioned. So much for holidays :-)

That are butt load of gears you have there, I guess that can not be done in one or two days job.  ::)

Offline tszaboo

  • Super Contributor
  • ***
  • Posts: 8217
  • Country: nl
  • Current job: ATEX product design
In Version 2 I used
There are two minor problems I saw with your layout. One is that you put the fuse (in parallel? dont do that) on the output of the bridge, and it should be on the input. If your something fails on the bridge because of the destructive test, nothing is protecting your input. The other problem I see is the lack of high frequency bypassing. Otherwise it is interesting.
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: NANDBlog
   ... One is that you put the fuse (in parallel? dont do that) ...
Thanks, but they're actually Polyfuses, and they work extremely well in parallel. We even heatshrink them together for
tighter thermal "coupling".

Quote from: NANDBlog
  on the output of the bridge, and it should be on the input ...
I don't intend to place them there, I just needed a "test point" and that artwork looked sort of relevant to an observer :-)
It's a current "pick-off" point and a thermal break (for my FLIR E8+ .. oops I meant E4) to test the 3+3oz Cu temperature.
Usually I ALWAYS use 4+4oz, but had a short cut-off time for delivery before xmas. I usually have all protection somewhere
where it's user-serviceable, NEVER on the main PCBs anyway.

Quote from: NANDBlog
... lack of high frequency bypassing. ...
There's enough HF filtering, just not shown on the pix. All done as per design notes, but I do like to see what goes "wrong"
under those circumstances. It's all part of proper design / testing.
Hello <tap> <tap> .. is this thing on?
 

Offline Phaedrus

  • Frequent Contributor
  • **
  • Posts: 714
  • Country: us
What's your efficiency looking like?
"More quotes have been misattributed to Albert Einstein than to any other famous person."
- Albert Einstein
 

Offline digsys

  • Supporter
  • ****
  • Posts: 2209
  • Country: au
    • DIGSYS
Quote from: Phaedrus
  What's your efficiency looking like? 
DAMN GOOD, even at 20A! If you check the waveforms, V2 has a drop of ~100mV for 1-2 mS. Even with that small PCB,
the FETs didn't get too hot, well under their rating. V1 are good to 15A with very little heat.
In V3, I'll go back to 4oz and add small Cu wings, just for extra headroom.
Hello <tap> <tap> .. is this thing on?
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 20357
  • Country: gb
  • 0999
Just thinking. Another use for this IC could be a brushless motor driver designed to maintain compatibility with a permanent magnet DC motor h-bridge. The driver circuit would always be powered from the same polarity but an input could be connected to the bridge rectifier input to reverse the motor when the input polarity is reversed.
 

Offline BravoVTopic starter

  • Super Contributor
  • ***
  • Posts: 7549
  • Country: 00
  • +++ ATH1
Just want to thank the "Santa Claus" in this thread here, attached photo below the thing I found under my chimney.  :-DD

Offline Phaedrus

  • Frequent Contributor
  • **
  • Posts: 714
  • Country: us
Hm, would be nice if they had a higher-voltage version that can work directly off mains voltage (up to 264VAC). The bridge rectifier is one of the last remaining major loss zones in an ATX PSU. Being able to get rid of those diode losses is a major goal. If a high-voltage version of this chip could be implemented more cheaply than a digital controlled bridgeless PFC, then this could be a great option...
"More quotes have been misattributed to Albert Einstein than to any other famous person."
- Albert Einstein
 

Offline tszaboo

  • Super Contributor
  • ***
  • Posts: 8217
  • Country: nl
  • Current job: ATEX product design
Hm, would be nice if they had a higher-voltage version that can work directly off mains voltage (up to 264VAC). The bridge rectifier is one of the last remaining major loss zones in an ATX PSU. Being able to get rid of those diode losses is a major goal. If a high-voltage version of this chip could be implemented more cheaply than a digital controlled bridgeless PFC, then this could be a great option...

The problem really wouldn't be the driver it's the switches. The pickings for single digit or even sub 20moHm fets (500V- 800V) are slim to non-existent, what does exist is expensive. They do make some IGBT modules but again these aren't cheap. When it gets to the point bridge losses are excessive building a small 1W isolated power-supply for driving some fets would just be a drop in the bucket.

They do have some decent power rectifiers you can always parrallel them to help with the losses a bit. I know you don't get perfect current sharing but thats about it. VISHAY makes some reasonablly priced ones VS SERIES in T0247 so you can bolt them to a basic heatsink and spread the dissapation out. Even a bridgless PFC really isn't all that much better you still need at least one or two addtional switches can't recall exactly off the top of my head.
Not to mention, at 6V the diode drop is 10%, at 230 it is only 0,3%. Well, just close numbers.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf