Buffer distortion?

[Bonsai]

The attached buffer is designed as the output of a balanced line amp.

If I run the simultation as in the attached cct diagram and look at 20KHz distortion of Vo1, I get around 70ppb (70ppb). If I place a ground between the 2 power supplies the distortion increases to 6.9ppm (6.9ppm) - a deterioration in performance of 2 orders of magnitude.

I only discovered this because I accidently left of the ground between the power supplies. When I put the ground in, the problem arose.

I tried a second experiment. I removed the ground between the output load resistors and with the power supply ground removed, the distortion increased to 1.3ppm (1.3ppm).

The 3rd experiment is with the PSU ground in place and the ground between the load resistors removed. Distortion is 6.2ppm (6.2ppm).

So, the key issue is the PSU ground.

Is this an LTspice or model problem, or am I missing something more fundamental here - e.g. are the distortions cancelling when the ground is removed?

Dazed and confused

[jan.didden]

Originally Posted by Bonsai

The attached buffer is designed as the output of a balanced line amp.

If I run the simultation as in the attached cct diagram and look at 20KHz distortion of Vo1, I get around 70ppb (70ppb). If I place a ground between the 2 power supplies the distortion increases to 6.9ppm (6.9ppm) - a deterioration in performance of 2 orders of magnitude.

I only discovered this because I accidently left of the ground between the power supplies. When I put the ground in, the problem arose.

I tried a second experiment. I removed the ground between the output load resistors and with the power supply ground removed, the distortion increased to 1.3ppm (1.3ppm).

The 3rd experiment is with the PSU ground in place and the ground between the load resistors removed. Distortion is 6.2ppm (6.2ppm).

So, the key issue is the PSU ground.

Is this an LTspice or model problem, or am I missing something more fundamental here - e.g. are the distortions cancelling when the ground is removed?

Dazed and confused

I think the key issue is load current. When you don't have the supply ground, load current is flowing between the two stages and not to ground. That may have some cancelling effect. Try it with just one stage, so you can isolate the cancellation between the stages.
In any case, the situation with supply ground is the 'real' one of course.

jd

[Bonsai]

Thanks Janneman. I get 3.5ppm single ended (which to me seems like reality). I'm concluding that the 70ppb no grounded PSU result is indeed due to distortion cancelling, though I don't think this is a practial connection scheme - the whole thing will be floating around. This result sort of reminds me about balanced symmetrical amps with current mirror loads on the LTP . . .