High efficiency digital power supply w/ATTiny25V controller.

Go To Last Post
17 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello freaks!

I have been working on a little project with an AVR32 and TFT lately, and I have just completed a draft of the power supply unit... This is intended to run off a 7.4v 3200mAH 1C 2S lipo pack, I have not included the charger or protection circuitry here because well, I can't give it all away... ;)

This takes the lipo pack 7.4v and generates the following supplies:
- 1.8v 600mA max for avr32 core
- 3.3v 600mA max for most of the ics on board
- 5v 1.2A max for TFT module AVDD, also cascading 3.3v and 1.8v (should not draw more than 900mA together, to test...)
- 20v 800mA for 6x leds TFT backlight

This is all done through high efficiency buck and boost converters. Because of board needing a sequenced startup, I included a Tiny25V to control the power up, power down and reset sequence. The program for it is done and simply creates proper power sequencing for the hardware, as well as pauses on reset and shutdown for linux to terminate properly, via the converters' enable pins, and a power N-channel MOSFET on 5v to the board. The gate voltage is elevated to 6v from 5v avr output through the lm321 opamp, to satisfy the Si1050X's gate saturation treshold. The Tiny25V is always in deep sleep and woken up from int0 low level, the pushbutton... If the unit is already running, a momentary push on delays 10 seconds (while the avr32 receives the pushbutton signal and shutsdown, unmount flashes, etc..), then issues a board reset. Holding the pushbutton for a while (4 secs) delays 10 secs, and then sequences the power down. If the unit is off, a momentary press sequences the power up.

I would like your comments, suggestions, and corrections on this before I get a couple of boards done... Most of the chips are very small QFN or SOT23, so a board is crucial, and costs quite a bit for a prototype... :)

This has not been yet tested, and is open for improvements.

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My first thought is that the use of an op amp for gate level shifting does not seem to fit with my idea of "high efficiency". Further, you really want a scheme that can provide LOTS of gate current to switch the transistor fast as that is one of the efficiency factors; an op-amp really does not excel at that.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Do you have any suggestion as to how to drive that gate more efficiently?

I selected the lm321 because of it's small footprint, low quiescent current (about 400uA), and ease of use in this application, but I am open to suggestions here... In the case of gate current, what do you define as lots? The opamp outputs 40mA, and large signal test ramps up to 6v in about 6uS. Turn on time for the MOSFET is 6.2ns, with minimum gate treshold voltage of 0.35v (0.9 fully on).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sorry if this freaks you out.

I think that efficiency will be / should be, the least of your worries. You've got 4 switchers, and of those you have 3 different chips which probably means at least 3 if not 4 different switching frequencies and none of them are synchronised. You are looking at an EMI/EMC nightmare.

That thing is going to be screaming like a Banshee.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well as far as EMI goes, the power supply will be on a completely separate board, by itself, from the rest of the project... Adequate filtering at the mainboard's power input will get rid of any stray EMI... There is really no way I can go around this, unfortunately... I tried looking at dual-triple-quad chips, but either they were outrageously pricey, or couldn't drive my power requirements... Going with linear regulators is out of the question because it would reduce efficiency (alogn with it battery life) by half, and produce way too much heat inside the slim casing...

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

FET gates in switchers can take peak currents exceeding an Amp, with 100ma peaks being the norm for even low-current switches. Normal (ie, CMOS or processor output pin) logic usually does a poor job of driving switcher gates.

I'll take a look around and see what I can find.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Some points I wonder:

LM27313 is operated is Up-Converter. I think the
feedback connection is wrong (must be behind diode).

5V, 3V3 and 1V8 regulator each have a diode
behind. That reduces voltage, regulation and
efficiency.

There is the MOSFET that you use to switch on/off
the 5V. I think its body-diode may prevent
switch-off (If the symbol you used is right).

Thats what I found in a first inspection. There may
be some more issues !

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thank you ossi!

Indeed I made a mistake, the diode on the LM27313 should be before the feedback line... Thanks!

The diode on the outputs was just as a simple means of protection for the prototypes, to prevent blowing the whole PS in case I short something upstream... They will most likely be removed on the final design...

As was mentionned before I am opened to suggestions regarding a better/cleaner way to switch that 5v output. My only pre-requisites are small footprint (SMD), and relatively low cost.

Cheers!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

OK, looking again at the circuit more closely, I now see that the FET and LM321 are really used to turn the external 5V on and off. That does not need a special driver and you really don't even need the LM321. Get rid of the Schottkey diode in the power supply for the micro and you should be able to drive the FET directly with the port pin.

The body diode in the 5V switch FET will prevent the switch from really turning off. A solution I have often used is a pair of P-chan FETs (in an SO-8 package) back-back with the gates tied together and driven by the single control line. Yes, the current flows "backward" through one FET but its happy with that.

Feedback IS correct on the LM27313. Its all those series output diodes that are confusing. Do you really need them?

I also had a hard time figuring out what "PRK22J5BBBNN" is. The symbol certainly gives no hint that it is a switch.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jim,

You are correct that the FET and LM321 are only used to switch the external 5v on and off. The reason for the schottky on the Tiny VCC is to prevent the ISP-supplied power from powering the rest of the board while programming. Obviously this board will need to be off while programming.

I am interested in your dual p-channel setup, care to elaborate? :)

The feedback circuit itself on the 27313 is ok, but the diode was indeed misplaced. Diode's anode should have been connected directly after the inductor, and feedback circuit supplied from after the diode..

The output diodes, read above, no I don't really NEED them... :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

With the dual P-chan scheme, connect the two gates together and the two drains together. One source is input, the other is output. Series on resistance is clearly higher than a single FET but it turns off!

Be careful and add some good bypass cap on the Vcc of the micro. I would use at least 0.1uf ceramic. The series-R of the Schottkey diode will give you more noise than you might expect. At the low currents needed by the micro, the forward drop across this diode is likely to be small enough that you still don't have to worry about driving the 5V switch FET. I would not bother with the LM321.

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:

also had a hard time figuring out what "PRK22J5BBBNN" is. The symbol certainly gives no hint that it is a switch.

Hehe yeah I used a DIP header symbol and footprint because the switch is panel mounted with wires coming out... :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hmmm a little bit of searching got me a single IC SOT-23-6 0.52$ solution... ;)

http://www.fairchildsemi.com/ds/...

So goodbye lm321, goodbye si1050X... :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thought I would let you folks know how it turned out, so here are some pics of the prototype board... ;)

Attachment(s): 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

For those wondering, this is the power supply described in this thread + lipo 2-cells charger with balancer... :)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I haven't looked up the specs on the chips involved, but could you not power the 1V8 & 3V3 switchers from the battery volts rather than via the 5V converter? That would give the 5V converter an easier time.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Unfortunately no I can't, Vin is between 1.8v and 5.5v...

Finding switching (or even linear) regulators that output 1.8v or 3.3v with a wide input range is tricky... I spent hours looking for the ones I have here... Hehehe..

I am aware that cascading the regulators isn't the most efficient way to do this, but I couldn't find any alternative. I really do not expect the 5v and 1.8v to draw much current under full load anyways, so it should be alright... I hope... Hehe.