Oscillations on output of loaded LM675 for all positive voltages.

Thread Starter

RovingJH

Joined Feb 26, 2016
3
Hello,
I'm trying to use an LM675 power op-amp IC as an output buffer for a signal generator. For stability, I've configured this IC as a non-inverting amplifier with a gain of 11 as per the circuit diagram below. As also recommended in the datasheet, I routed separate ground returns for the signal input and power output. The circuit is powered from a dual PSU at +/-10V. The output is connected to a 10 Ohm power resistor returned to ground (using 4cm long wires).

LM675 buffer.png
However, there are oscillations on all positive output voltages, especially when Vcc exceeds the maximum amplified input signal by around 2V (a little below which there would be clipping). By way of illustration, I have captured a scope trace of the output below for a 200mHz 200mV sine-wave input. This low frequency gave me enough time to increase Vcc between each positive cycle as labeled in the trace. I omitted and interchanged various values for C4 (in cct above) with minimal effect, except to slightly decrease the oscillations near the crossover point. I also added large electrolytic capacitors across the supply rails beside the chip (and C2/C3) to no avail. The oscillations are in the MHz range and are sinusoidal in nature. The negative-going output swings remain clean at all frequencies (up to around 300kHz). There are no oscillations without the load.
LM675 oscillations.png

For a 500mV sine-wave input, the picture is as below.
LM675 oscillations2.png

This behavior persisted when I used batteries instead of the PSU. I'd be grateful for any pointers/suggestions please.
Thanks,
John.
 
Last edited by a moderator:

sghioto

Joined Dec 31, 2017
2,329
Did you build this circuit on a PC board?
According to the data sheet:
1618789055898.png
I see 4 separate ground returns
1618789107243.png
 
Last edited:

Thread Starter

RovingJH

Joined Feb 26, 2016
3
Many thanks sghioto and crutschow for your replies. I soldered this circuit on stripboard and obtained the layout below.

Sig gen stripboard.png

The bridges in my circuit diagram reflect those I used to obtain separate signal return paths. As per sghioto's numbering, and as above, ground #1 is on one stripboard trace directly connected to a ground banana socket. Ground #2 is routed directly to this banana socket via a ~1cm jumper and the capacitors are beside the chip (as are the other passives). Ground #3 and #4 are tied together, and returned to the banana socket via a ~6cm copper strip and a ~3cm jumper. I applied solder along the length of all copper strips for ground return, signal output and +/- rails, to ensure low resistance, and thoroughly removed and cleaned flux between traces etc. The three bridges are heavy-gauge enameled copper wire. I also tried separately routing ground #3 directly to my ground banana socket with no change.

I think I will take your advice and design a PCB layout with a ground plane. It's curious though that only the positive-end cycles are affected - all output excursions less than the ground potential are fine. I also tried a different chip, just in case, but no change.
 

Thread Starter

RovingJH

Joined Feb 26, 2016
3
What value cap are you using for C4?
I have tried various values and none. When I use the recommended value of 220nF (in series with 1R), the oscillations appear on both the upper positive and lower negative excursions. When I use a lower value such as 100nF, or even omit this capacitor, the oscillations appear only on the positive excursions as shown in my screenshot traces at the top of this thread.

I since found this post on the TI website: LM675 high-freq output oscillations, so it appears that trouble can arise even with a ground-plane PCB. The poster resolved their problem by changing the snubber values (maintaining same RC constant) and adding a variable resistor in series with a capacitor from the output to inverting input.

I will try that approach but I'm not confident that such a setup would remain stable across the frequency range in a signal generator buffer application, and indeed the requirement for the snubber (incorporating C4) would add a substantial load at the higher frequencies (up to 150kHz in the intended application).

I might consider other options, such as the OPA541. This does not require an output snubber, but its GBP is much lower at 1.6MHz (versus 5.5MHz for the LM675).
 
Top