Dear forum members,

I am interested in knowing your opinions on audio performance (THD+N) when it comes supply voltage ripple.
I have the responsibility to write datasheets, and characterize audio performance of audio amplifiers. Under the direction of the senior apps the decision was to place a THD+N vs. Supply Ripple curve in the datasheet it has been difficult to implement a typical PSRR (Ksvr) test circuit with the Audio Precision to capture the PSRR vs. Frequency curve. I am currently looking at a low frequency VNA to implement this test.
I have never seend a THD+N vs. Supply Ripple curve in any other audio amp datasheet. Perhaps no one has thought of it before, do not see the added value, or it is bogus.
The test setup is rather simple: a power amp is used to create the DC supply while superimposing a sine wave for ripple. The amplitude is known (200mVpp) and is swept across frequency while THD+N is meaursed with an audio analyzer.
I would feel more comfortable seeing a mathematical relationship much like how (DC) PSRR is defined: (change in Supply/change in Vout) and manifests itself as DC offset at the output.

Any thoughts on this?

 

dis, i don't see any use in this parameter either. could you go ask your senior apps again for what purpose is this parameter there?

if they insist, maybe you can just do this.
the thd+n is there,and you've got the parameter.
the psrr's parameter is there too.
superimpose the two parameters.
well, do you think it too simple?
but you've already got the two parameters needed right?

 

Just winging it, but it's hard to see how PS ripple would affect THD. Showing up as noise might be another thing. That should show up with no applied signal - just see what happens to the output noise figure when you add/subtract capacitance from the supply rails.

Amps have to be fairly robust in the power supply section for normal operation. You would think the lack of filtering would show up under heavy loads as part of the THD figure, but this is not too hard to demonstrate by adjusting filter caps.

I feel that nothing succeeds like excess, and have about 50,000 uF capacitance on my homebrew power amp rails. I have no idea about it's THD, but I've never been able to drive the amp into clipping.

 

When I worked for a pro-sound manufacturer, I turned up the sub-woofer's amp in the demo room and it blew a 15A electrical mains breaker. My eardrums were at their max p-p displacement but the amp wasn't clipping.

Then I substituted the main amp for my home-made little portable amp for the beach and nobody noticed.

 

dis, i don't see any use in this parameter either. could you go ask your senior apps again for what purpose is this parameter there?

The reason is that the traditional PSRR curve has been a pain point to setup with the Audio Precision. I have tried several routes:

1) Using a TI based application circuit that uses the AP itself as the generator/analyzer.
2) Using a power amp as the generator and the AP as the analyzer.
3) I tried it on a competitor evaluation board just to take our board layout out of the quesiton.

Something is wrong (consistantly) in the setup I suppose.

So this led to the decision to use a THD+N vs. Supply Ripple. I have to look at the data again, but I feel that it is not something that you can easily correlate. Niether of us are audio gurus, but I am more up to speed in terms of understanding audio specs and researching this particular issue.

The THD+N vs. Supply Ripple is almost like an intermodulation distortion test. But the difference is that we are sweeping the ripple frequency.

 

sorry dis,
i'm not familiar with the abbreviations, could you describe them for me plz? (TI and AP). thx

 

TI - Texas Instruments
AP - Audio Precision

 

i see.
i think ripple in supply voltage doesn't correlate or has little to do with thd+n, but it at the same time directly affects the measurement of thd+n.
don't you think that while measuring the thd+n, the ripple (occuring according to the psrr ratio) will be measured as additional noise and harmonics too?
i thought it'd be like measuring the power of an engine versus the displacement of the pedal.

 

i see.
i think ripple in supply voltage doesn't correlate or has little to do with thd+n, but it at the same time directly affects the measurement of thd+n.
don't you think that while measuring the thd+n, the ripple (occuring according to the psrr ratio) will be measured as additional noise and harmonics too?

I have no doubts that ripple will show up as noise, and cause additional distortion on the output. The problem is if you have ripple voltage of 200mV @ 1kHz on the supply can you drop that into an equation to predict the %THD+N?
Similar to % overshoot in the time domain correlates to phase margin in the frequency domain.

 

I have no doubts that ripple will show up as noise, and cause additional distortion on the output. The problem is if you have ripple voltage of 200mV @ 1kHz on the supply can you drop that into an equation to predict the %THD+N?
Similar to % overshoot in the time domain correlates to phase margin in the frequency domain.

well you've already had the psrr ratio. that means you got the % 'noise' as additional to the existing noise. although simply adding them ( adding noise a and b is like this right? sqrt(a^2 + b^2) ) maybe would not be as accurate as measuring on the fly.
what do you think?