OPA552 as power regulator

I need some programmable power supply rails, maybe -5 to +40 volts or so. We have some OPA552s in stock, a nice 60 volt power opamp. It's a decomp, and it will be driving a big capacitive load, so there is an oscillation hazard. One of my guys got the model from TI and kluged it into LT Spice. The symbol is ugly but it seems to work. We don't know how to make it look like a real opamp.

formatting link

This looks fine, even without C1. Maybe I'll include C1 on the pcb layout just to have another knob to turn.

I'd like to use the OPA552 thermal cutoff as a secondary, slowish current limit, but I'd have to tune the pcb thermal resistance somehow to control that. Maybe a bunch of thermal-zero-ohm jumpers? Maybe a tweakable amount of gap-pad under the board, or above the amp? I suppose I could selectively drill out vias on the rev A board, but that's really ugly.

I want a thermal conductivity trimpot. Or DAC.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

What's wrong with using FLAG?

Is that 1p for C1 right? You have stray that's 10x that.

Reply to
bloggs.fredbloggs.fred

Copper area fanning out round the chip, with the area split into lots of se ctions like a fanblade. Each fanblade narrows where it gets close to the ch ip: you can cut through those to make a fair difference to heat loss. I do that if desoldering something from an unwanted board.

NT

Reply to
tabbypurr

There's a simple way to safely use decompensated opamps, often useful to get faster slewing, etc. I looked in AoE III, to give the page, but discovered I didn't get it in there!

First, you can add a resistor between the opamp's two inputs (for an inverting amp this would be to ground), so it thinks it's working at a higher gain, while not interfering with the actual in-out function. But this adds extra output offset, multiplying the input offset by the additional DC gain. However, since you only care about high-frequency loop-compensation, just add a capacitor in series with the resistor. So that's it, add a series R + C to insure meeting the opamp's minimum gain spec. To combine this with the standard capacitive-load isolation trick, Figure 4.76, page 264, add a resistor in series with the new feedback cap.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

For what? We'll probably not connect it in real life. I guess it could light up a red LED or something.

That's a placeholder for "maybe we don't need a cap there." I said that.

Tweaking a bit more, it looks like C1 is a good idea: 1 uF.

The opamp also (according to Spice) oscillates when it current limits. That will happen when the dac voltage steps and a lot of current is briefly forced into the load cap. That's probably not a big deal, but we can slew-rate limit the dac output, in hardware or software.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Production wouldn't like doing that. And we don't prototype.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I need gain from DAC to Vhi, so the gain divider, where it is, takes care of the decomp issue.

This is a little better:

formatting link

I've got to go clean some windows.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

opa547 ?

Reply to
Lasse Langwadt Christensen

That's nice, with the programmable current limit. Probably worth adding to stock. It swings reasonably close to the rails, so with a

48v supply we could make 44 or 45 volt output pulses.
--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

For programmable pulse generators, how about using programmable rail voltages, with MOSFET switches?

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Put a resistor right next to the OPA552 (or two, one on each side for better uniformity) as a heater and vary the power to control the "ambient" temperature the 552 is dumping it's heat to.

--
Regards, 
Carl Ijames
Reply to
Carl

sections like a fanblade. Each fanblade narrows where it gets close to the chip: you can cut through those to make a fair difference to heat loss. I do that if desoldering something from an unwanted board.

I thought you wanted a proto were you could cut off the islands 1 by 1 to s ee how much heatsinking got you to the right place.

If cutting islands in production is out, which I assumed it would be, obvio us other ways are a) solder in links in production, eg 0 ohm Rs to connect thermal islands b) solder in (or not) additional heatsinking items, eg bits of copper, bars for heatspread etc.

If you really want it to be fully variable rather than preset, I guess a bo lt in a nut (soldered or crimped) could provide variable conduction. Not a very likeable arrangement though.

NT

Reply to
tabbypurr

That's what I'm trying to do, generate programmable 4-quadrant Vh and Vl rails. Then I'm thinking about using GaN switches.

I've done some nice fast pulse gen output stages with mesfets and then phemts, but the parts keep going EOL. I've never had much luck using mosfets; high Cg-d makes the pulses ugly.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

I'm surprised at that, I've had excellent results. You need to avoid using a too powerful part, keep capacitances down. If it's sub 15ns tr tf you're seeking, go ahead and use SiC. They're fine at low voltages too. For good pulse waveforms into coax, you have to include 50-ohm back termination.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

The problems happen when Vh gets close to Vl, 1 volt say. The volts of gate drive leak into the output and the pulse gets ugly, like a big precursor in the wrong direction.

I'd like to get below 1 ns. SiC needs tens of volts of gate drive, which is awkward. GaN needs roughly 4, which is bad enough. I really miss phemts; 0.7 volt swing would switch them, and Cgd was close to zero.

Our little time-domain biz is whiplashed by the cell phone/telecom market. If they can't sell a billion parts a year, they shut down the fab. Broadcom bought Avago and killed the good stuff.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Maybe there's room for a biz that buys untraced parts & tests & certifies them.

NT

Reply to
tabbypurr

I've had surprising good luck finding hidden stock of un-obtainium parts in Asia, searching Alibaba sellers. Put up a part number and ask for quotes on 250 pieces.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Yes, lots is gettable that way. Could test & certify replace manufacturer traceability sometimes? Not for milspec normally, but I remember a big military ask for any old 486 PCUs folk had lying around, so even mil can accept that in some circumstances.

NT

Reply to
tabbypurr

I'll give an example, I needed a certain 600V MOSFET that had been long been discontinued, placed a request for 250 pieces on AliExpress. Got quote at $0.38 each, ordered with DHL shipping. The parts quickly arrived. Right packaging, right labeling. Sets of capacitance measurements showed the part's unique values. Would a counterfeiter go to the relabeling trouble for a $95 order? Did he have the time? How could he spoof the physical measurements? One has to ask, how are the parts not going to be exactly what they appear to be?

I'm amazed these parts hid for years, without Google revealing any sign that they existed, but they indeed had hidden, and the Alibaba RFQ system turned them up.

As a followup, the instruments made with these parts worked perfectly.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I had similar good results with an unobtainium 1.2kV MOSFET, off the market for years, only a few pieces available for about $10 each, got 250 at ~ $0.60 each.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.