Catch 22- regulating V to a regulator

[dknguyen]

I'm looking into building a voltage regulator module, but I've run into a bit of a catch 22. How do you normally regulate the voltage for components to a voltage regulator? In block diagrams for regulator ICs, they seem to just include a linear regulator. BUt I'm not sure how they even pull this off because this seems to be what they do for 40V and 75V supplies, and even a small amount of current would generate a tremendous amount of heat.

My situtation is a power supply for a large FPGA. It has multiple voltage requirements and draws too much current for linear regulators. However, I also need proper simultaneous ramp up between the various supplies and synchronized clocks which seems a custom solution might be needed.

I thought it might be good practice to use an FPGA to build the power supply. A much smaller FPGA with much lower current (though it still requires the same voltages as the large FPGA) so maybe it could be started up with linear regulators. However, my desired max input voltage is at least 10V, preferably even higher to 20V or beyond. But too much heat is generated for 1.2V linear regulators, even at 100mA.

I'm having a hard time figuring out how to bootstrap the regulator FPGA so it can actually start up.

 

[Sceadwian]

What do you mean it draws too much power for linear regulators? I've seen linear regulators up to a few hundred watts, they just use current bypass transistors and big heatsinks and fans.

 

[Mikebits]

Not sure if this will help but have a look here.
https://www.electro-tech-online.com/custompdfs/2010/05/AN1133.pdf

 

[mneary]

There are switchers for half a dollar that would almost do the job. A few months ago someone mentioned the MC34063A in another thread. At a glance I don't see any problems with it. When I checked the price it was $0.39 in 10+ quantity. You would also need a 220uH inductor, a few resistors, and a diode. Its reference is 1.25V, so you might have to shop for a similar part.

I would expect you also need 3.3V and/or 1.8V. A common way to do this is a switcher to your highest logic supply (3.3V?) and then a linear down to 2.5, 1.8, and 1.2. Anyways, laptops handle this all the time.

 

[Sceadwian]

In general switchers are relatively high in ripple and variation, so it'd more sensible to regulate down to 5-6V with a switcher and then use linear regulators to handle the ultimate regulation. You're a little vague on what you need and what you have.

 

[dknguyen]

Sorry, I suppose I was vague.

Here's the deal:
-FPGA needs 1.2V and 3.3V
-PLL needs 1.8V and 3.3V
-input voltage max of at least 9V with current of 2A for each source
-the regulators need soft startup, tracking during ramp-up, and pre-bias startup capability, and synchroznized clocks

THis last part has a couple of catch 22s:
- since I want the regulators to be running off of one of the clock outputs on the PLL, but at the same time the regulators need to power the PLL. I had reasoned that I could get around this catch 22 by powering the PLL off of a linear regulator directly since it draws so little current on it's own.

-and heatsinkinks and the like won't really work because there is a board height requirement due to stacking the FPGA module (nothing specific, but mainly the heatsink can't be something that protrudes obviously above the other components on the board which kind of cuts out heatsinks in general).

@MNeary, those pre-bias, synchronization, and tracking capability rule out almost every buck IC out there. THe only ones I've found that have even 2/3 of those only have maximum input voltages of 5.5V. I've not actually found any that have have all 3. It's hard enough finding one with tracking and synchronization, but finding ones that can support pre-bias startup is even rarer.

 

[Mikebits]

since I want the regulators to be running off of one of the clock outputs on the PLL

Never heard of such a thing. Maybe I am confused. 2 amps sounds like a lot.

 

[Gary B]

I used to fix high energy, multiple voltage power supplies in my youth. Things took 120V input and provided: 5V @ 50A, +12V @ 2A, -12V @ 2A, +36V @ 5A, and 48V @ 10A. They used an SCR phase control pre-regulator to get close and then a linear regulator to hold the output exactly on voltage.

The pre-regulator got close enough that it could provide the power for the post-regulator’s components.

 

[dougy83]

As for soft startup and tracking supplies, the attached (or similar) circuit will do it. For the high current (2A?) the transistors marked NPN could be a Sziklai pair (complementary darlington) or similar.

Granted the attach circuit is a lump of junk, you can get multiple output converters (surely some should have a sync input - if not, then it shouldn't be hard to inject a sync pulse into the timing circuit). Obviously the FPGA cannot control the converter before it's powered up, so don't worry about it; just get it to control the converter once it is powered up! (run the converter clock asynchronously while the FPGA powers up).

 

[Hero999]

I would thought you'd use a push-pull converter to reduce the 75V to 5V, similar to the type of PSU use in a PC except the isolation might not be important.

Then use a multiphase buck to reduce the 5V down to the really low voltage i.e 1.2V.

Where are you getting the 75V from?

You'll probably need overvoltage protection on the output, at least for safety reasons, although 75V DC is unlikely to be fatal, it'll give you a shock.

 

[indulis]

You might want to take a look at the MAX8537/MAX8538/MAX8539 or a SC416.

 

[Ubergeek63]

In general switchers are relatively high in ripple and variation, so it'd more sensible to regulate down to 5-6V with a switcher and then use linear regulators to handle the ultimate regulation. You're a little vague on what you need and what you have.

depends on the switcher ... they are generally all that is used these days unless you are getting into critical analog in which case the last 0.25V is linear

 

[dknguyen]

I would thought you'd use a push-pull converter to reduce the 75V to 5V, similar to the type of PSU use in a PC except the isolation might not be important.

Then use a multiphase buck to reduce the 5V down to the really low voltage i.e 1.2V.

Where are you getting the 75V from?

You'll probably need overvoltage protection on the output, at least for safety reasons, although 75V DC is unlikely to be fatal, it'll give you a shock.

75V is just a possible future proofing thing because RC airplane batteries can get that high. 15V is the max for most things (separate batteries for the electronics from the power system). You do raise a good point though about isolation. In a 75V (really, >40V) power system I'd probably want some kind of isolation, though that's likely to be unneeded complexity in lower voltages.

Never heard of such a thing. Maybe I am confused. 2 amps sounds like a lot.

2A is a lot (mainly for the 1.2V and 3.3V. 1.8V not so much but it never hurts if you have lots of ADCs running), about half of that is sucked up from the FPGA alone, the radio sucks up the other half so I might actually need more than 2A. The LM2745 seems to fit the bill. Lots of parts though and I'll need 3 of them. I'll see what I can do.

Any buck regulator with a sync input will do. I'm just trying to sync all clocks on the board- FPGA, and power supply.

I'll also look into the MAX8537/MAX8538/MAX8539 and SC416 as well.