High-frequency power amplifier circuit

[jackmorrison]

I've gone a little outside my area of expertise in a section of a circuit I'm working on, and I'd be grateful for some advice. I need to boost the current drive of a voltage source to power an array of LEDs. Requirements for the circuit are:

Input: range as needed, nominally 1V to 3V, frequency DC to 5MHz; high impedance (direct from DAC or waveform generator)

Power: single-ended supply, prefer 6.5 to 12V

Output: about same voltage range as input (so gain of 0.5 to 1),
at least 500mA, prefer 1A (continuous). Load is 0-64 high-brightness LEDs via analog switches with switch capacitance up to 130pF each, LED capacitance unknown but maybe guess 10pF, so 10nF total

Size: ideally no more than 10x10mm for two channels

My initial design uses OPA2673 in voltage follower configuration: input to +IN, -IN tied to OUT; single-supply (-VS=ground, +VS=10V). This works for a little while, then the ($4) IC heats up and dies - even with no LED load connected. Initially I had the "GND" pin on the IC tied to ground, but now I think that's wrong. Also I notice the datasheet recommends 511ohm feedback resistor even in a voltage follower configuration. This part is only available in a QFN package which is a little tough to experiment with (although there are lower-power relatives available in SOIC).

Now I'm wondering if a simple NPN transistor circuit would be better. I threw together a test with a 2N2222 I had laying around, and it seems to work but a little slow.

So my questions are:
- what's the disadvantage of a single BJT voltage follower compared to an op-amp?
- any idea why the OPA2673 is not happy in this role?
- what's the right way to meet the above requirements?

Thanks for any help.

 

[unclejed613]

a) gotta schematic?

b) did you read the data sheet, especially the part about "Thermal Analysis"? the chip may need to have a heat sink.

c) are the power supply rails properly bypassed at the device? also read the data sheet section on board layout.

d) have you looked at the output with a suitable oscope? sounds like the amp might be self destructing from oscillation.

e) you may want to read up on current feedback amplifiers in general, their behavior is a bit different from voltage feedback amplifiers.

 

[jackmorrison]

a) gotta schematic?

Sure. Couldn't be much simpler... Here's the original design:



This is an updated version, adding resistors in the feedback loop, removing the suspicious "GND" connection, removing the extra output capacitance which slowed the response down to much, and adding additional supply bypass per the datasheet. Haven't built this yet.

b) did you read the data sheet, especially the part about "Thermal Analysis"? the chip may need to have a heat sink.

I'll review that more carefully.

d) have you looked at the output with a suitable oscope? sounds like the amp might be self destructing from oscillation.

There was a small (about 20mV pp, forgot to note frequency but MHz range) oscillation when I removed the capacitor on the output.

e) you may want to read up on current feedback amplifiers in general, their behavior is a bit different from voltage feedback amplifiers.

Thanks, I'll look for clues there.

 

[unclejed613]

https://en.wikipedia.org/wiki/Current-feedback_operational_amplifier

CFA circuits must never include a direct capacitance between the output and inverting input pins as this often leads to oscillation.

the "direct capacitance" could include capacitance of the PCB traces, especially in a multilayer PCB