Limiting Distortion? PA3s

[haskellcurrying]

Hi All,

After reading the reviews, I picked up Topping E50 and PA3s (they showed up today!!) to go with my PSB Imagine Minis. Everything seems pretty great at this point, but I had a question about the review and a comment Amir made:

Allowing for more distortion at 4 ohm gives a bit more power

Beyond changing the input or output volume, how does one limit (or increase) the distortion tolerated by the amp? Currently, I just have the PA3s turned all the way up to max and use the E50 in pre- mode to adjust output. It would be nice to limit the range in such a way as to keep the distortion under control through the full input voltage range(I hope that statement made sense :/ ).

I realize I could just bring the PA3s amplification down from max and hope that I've limited my output to the "sweet spot" range on the graph in the review, but that doesn't sound all that scientific!

Thanks.

 

[somebodyelse]

One doesn't limit the distortion tolerated by the amp. It's just down to where you decide to draw the line for maximum power. In the power vs. distortion graphs it's at the point the distortion starts to rise steeply, not a fixed distortion level. If you draw the line at a fixed distortion level then for most amps you've gone a little way up that steep curve and get a higher power number. Draw it at a higher distortion level and you get a higher number - chip amp datasheets often specify power output at 10% distortion, and amps using those chips sold cheaply on eBay, Ali etc. tend to quote the datasheet number not what the amp being sold can actually achieve. There have been long discussions about different ways to specify power output and which are more or less representative or fair, without much consensus but with plenty of strong opinions.

 

[haskellcurrying]

That is a very solid answer!

I'm newb guessing here, but a more reasonable version of my question is... I get this is decisive but I am working from nothing... I've got a multimeter, a power clamp, and a graph that characterizes the impedance as a function of frequency for my speakers.... how can I use those tools / that information to set the max output volume of my amp to ensure that my peak output is within that flat bit of the power graphs before that inflection to the steep monotonic increase?

Or is that question nonsense?

 

[NTK]

Assuming the impedance of your speakers is 4 ohm, the max power before the rapid distortion rise per Amir's measurement was 46 W. That means the max output voltage is sqrt(4*46) = 13.6 Vrms.

Disconnect speakers from the PA3s, ** Edit: See next post ** use a tone generator (I tried this one from the web below) to generate a sine tone of 100 Hz, maximize all volume settings (computer mixer, player, etc.) except the one in the PA3s. Set the PA3s volume to minimum. Measure the PA3s output voltage, increase the PA3s volume knob until the output voltage just reaches 13.6 Vrms (or a little bit below). That will be your "safe" volume setting where the PA3s should not clip under most circumstances.

For an extra 3 dB margin to avoid inter-sample overs clipping, reduce the PA3s volume setting for an output of 9.6 Vrms.

Online Tone Generator - generate pure tones of any frequency

www.szynalski.com

 

[NTK]

Thought over this a little bit more and realize that class-D amps may not work as intended without a load. Class D amps such as the PA3s output pulses and the output filter will need a load to function.

Therefore the test as described in the above post should be modified to include a load. Test instead with the speakers connected and wear hearing protection during the test, or use a dummy load. To lessen the stress on the speakers, you may use two speakers connected in series. This will halve the total sound output and reduce the power dissipation of each speaker to 1/4 of the single speaker loading. The change to the load impedance will have insignificant effect on the results.

 

[abdo123]

The only thing you would be doing is limiting your system’s dynamic ability.

 

[NTK]

The only thing you would be doing is limiting your system’s dynamic ability.

No, it doesn't. Not when one avoids listening to clipped sound.

What this does is to determine the optimal gain staging of the signal chain. It sets the power amp (last piece in the chain) to its lowest gain while still able to reach full (unclipped) output. This maximizes the output signal levels of the DAC/pre-amp, and will therefore provide the best signal-to-noise ratio the system can give.

 

[abdo123]

No, it doesn't. Not when one avoids listening to clipped sound.

What this does is to determine the optimal gain staging of the signal chain. It sets the power amp (last piece in the chain) to its lowest gain while still able to reach full (unclipped) output. This maximizes the output signal levels of the DAC/pre-amp, and will therefore provide the best signal-to-noise ratio the system can give.

how is this related in Distortion to anyway? the only way to limit the distortion of the amp is not to push it to its limits.

 

[NTK]

how is this related in Distortion in anyway? the only way to limit the distortion of the amp is not to push it to its limits.

This is to find the maximum volume knob setting so that power amp never clips.

 

[abdo123]

This is to find the maximum volume knob setting so that power amp never clips.

For that they need an oscilloscope and I agree that’s the best way to set the gain in the system but again you can’t limit distortion without also limiting the amplifier itself.

Also many would argue that it’s best to have the amplifier clipping than the DAC or pre-amp.

 

[haskellcurrying]

Thank y'all so much for the info and the back and forth discussion. So I've going to try the following workflow, let me know if I'm off...

(1) Plug the banana leads into my mm with a 100hz sine wave playing and tune the max output the the appropriate voltage.
(2) If the above doesn't work try the above again with the speakers attached. I can try just attaching the MM in-line and hope it doesn't explode under load or I can try using the the power clamp around the speaker wire and hope that the measurements are not too far off.
(3) If that volume is too low at let say -10db out from the dac/pre, I just bring it up because I rather have the amp distort than the dac clip (could we go into a bit of detail about this?)

Does that seem okay?

Also I've got a follow up question about frequency response vs impedance:
Some review was nice enough to plot the impedance vs frequency for my speakers. And for a large window(i think like 500hz) below the 2200hz crossover point the total impedance is below 4 ohms. Above the crossover the impedance can clear 12 ohms... how do I use that information decide the peak voltage? It seems like a tweeter doesn't take a lot to power, so I could just consider the woofer range when selecting a voltage target(and maybe just bump down a bit?). That said, I don't really know, that could also be a rather bad assumption. On the other hand, if I use the 12 ohm max impedance to decided the functional peak output is going to be rather low...

Audio is hard :/

 

[NTK]

For that they need an oscilloscope and I agree that’s the best way to set the gain in the system but again you can’t limit distortion without also limiting the amplifier itself.

Also many would argue that it’s best to have the amplifier clipping than the DAC or pre-amp.

Amir did the "oscilloscope" testing and determined the clipping threshold for a 4 ohm load is 13.6 Vrms. That is where the power amp output level is to limit to.

A properly designed DAC never clips as the highest level of any digital signal is known in advance (i.e. digital full scale). A properly designed pre-amp shouldn't clip either since it isn't required to output anything with substantial voltage or power.

 

[abdo123]

Amir did the "oscilloscope" testing and determined the clipping threshold for a 4 ohm load is 13.6 Vrms. That is where the power amp output level is to limit to.

A properly designed DAC never clips as the highest level of any digital signal is known in advance (i.e. digital full scale). A properly designed pre-amp shouldn't clip either since it isn't required to output anything with substantial voltage or power.

in this context clipping is the equivalent of running out of voltage output imo.

 

[mdsimon2]

Thank you all, so much for the info and the back and forth discussion. So I've going to try the following workflow, let me know if I'm off...

(1) Plug the banana leads into my mm with a 100hz sine wave playing and tune the max output the the appropriate voltage.
(2) If the above doesn't work try the above again with the speakers attached. I can try just attaching the MM in-line and hope it doesn't explode under load or I can try using the the power clamp around the speaker wire and hope that the measurements are not too far off.
(3) If that volume is too low at let say -10db out from the dac/pre, I just bring it up because I rather have the amp distort than the dac clip (could we go into a bit of detail about this?)

Does that seem okay?

Also I've got a follow up question about frequency response vs impedance:
Some review was nice enough to plot the impedance vs frequency for my speakers. And for a large window(i think like 500hz) below the 2200hz crossover point the total impedance is below 4 ohms. Above the crossover the impedance can clear 12 ohms... how do I use that information decide the peak voltage? It seems like a tweeter doesn't take a lot to power, so I could just consider the woofer range when selecting a voltage target(and maybe just bump down a bit?). That said, I don't really know, that could also be a rather bad assumption. On the other hand, if I use the 12 ohm max impedance to decided the functional peak output is going to be rather low...

Audio is hard :/

I think the suggestion to measure voltage with a multimeter at full power output while connected to speakers is a bad one.

There really is no reason to be super scientific about this. I would adjust the PA3s gain such that speaker output was about as loud as I would ever want with the E50 volume control at max. I might bump up the PA3s gain a bit to account for recordings that were mastered at a lower level. The only way that you can set the PA3s gain "wrong" would be if you were not getting anywhere near to maximum output on the E50 and were listening at really low levels (say -50 dB) which would be detrimental from a noise perspective.

Michael

 

[haskellcurrying]

I think the suggestion to measure voltage with a multimeter at full power output while connected to speakers is a bad one.

There really is no reason to be super scientific about this. I would adjust the PA3s gain such that speaker output was about as loud as I would ever want with the E50 volume control at max. I might bump up the PA3s gain a bit to account for recordings that were mastered at a lower level. The only way that you can set the PA3s gain "wrong" would be if you were not getting anywhere near to maximum output on the E50 and were listening at really low levels (say -50 dB) which would be detrimental from a noise perspective.

Michael

So currently with the amp all the way up, my comfortable listening output from the dac/pre is between -30 and -10 with the usual value being about ~ -25. Sometimes for really quite recordings, I drop 8, but I don't have my subwoofer plugged in yet (because I'm waiting on a couple cables to show up) and I don't think that happens very often.

I guess I am worried that I am creeping up to the max output of the dac. The previous amp I had was noticeably more powerful so I just set it to like 65% and never thought about it again...

 

[mdsimon2]

So currently with the amp all the way up, my comfortable listening output from the dac is between -30s and -10s with the usual value being about ~ -25. Sometimes for really quite recordings, I drop 8, but I don't have my subwoofer plugged in yet (because I'm waiting on a couple cables to show up) and I don't think that happens very often.

I guess I am worried that I am creeping up to the max output of the dac. The previous amp I had was noticeably more powerful so I just set it to like 65% and never thought about it again...

That is an important point, this amp does not have tons of power output. To me if you limit the amplifier output such that it will never clip at 0 dBFS input you will probably not be satisfied with the volume on low level recordings.

From Amir's review and the Topping specs it seems like gain with balanced input is 19.5 dB, I assume this is at maximum volume position on the PA3s. If you did want to make some measurements I would measure gain and use that to help guide you on the PA3s volume setting. From Amir's review of the E50 you know that at 0 dB volume the output is 4.25 V. You can then calculate the DAC output voltage at a volume setting of say -30 dB, 4.25 x 10^(-30/20) = 0.135 V. If you play a 0 dB test tone (I would use REW's tone generator) you can measure the amplifier output (with a load connected) and use that to determine gain. At 19.5 dB gain you would expect to measure about 1.27 V at the amplifier output, 20 x log (1.27 / 0.135) = 19.5 dB. With the gain measured you could then understand at what volume levels on the DAC you have the potential for clipping the amplifier and determine how much turning down the volume knob on the PA3s reduces the gain.

But I still stand by my first point that all of this is a fun academic exercise but if you do not have any issues with noise then optimizing the gain structure really isn't doing anything for you. The amplifier will always distort at the same volume levels, there is no way around this.

Michael

 

[NTK]

Thank y'all so much for the info and the back and forth discussion. So I've going to try the following workflow, let me know if I'm off...

(1) Plug the banana leads into my mm with a 100hz sine wave playing and tune the max output the the appropriate voltage.
(2) If the above doesn't work try the above again with the speakers attached. I can try just attaching the MM in-line and hope it doesn't explode under load or I can try using the the power clamp around the speaker wire and hope that the measurements are not too far off.
(3) If that volume is too low at let say -10db out from the dac/pre, I just bring it up because I rather have the amp distort than the dac clip (could we go into a bit of detail about this?)

1/2. I wouldn't run this test on a class-D amp without a load. Without a load, the amp output filter may not function properly and the electrical signals your MM measuring will probably not be a sine wave. The voltage readings may therefore be bogus.

I would also stay with voltage measurements (clamps measure currents). Amplifiers amplifies voltages, so voltage measurement are easier to interpret (and voltage measurements are also more precise). Current depends on output voltage and load impedance. If impedance is complex (i.e. have both real and imaginary components), there is also a phase angle that needs to be considered.

3. It depends on how/where you want to adjust your volume. If you adjust volume through your computer software, the optimal is to set the PA3s volume to the "safe" setting as describe in the procedure above. If you adjust volume using the volume knob on the PA3s, then set the computer volume to max, and you know where the not-to-exceed position of the PA3s is. This is the option that will give the best performance.

Also I've got a follow up question about frequency response vs impedance:
Some review was nice enough to plot the impedance vs frequency for my speakers. And for a large window(i think like 500hz) below the 2200hz crossover point the total impedance is below 4 ohms. Above the crossover the impedance can clear 12 ohms... how do I use that information decide the peak voltage? It seems like a tweeter doesn't take a lot to power, so I could just consider the woofer range when selecting a voltage target(and maybe just bump down a bit?). That said, I don't really know, that could also be a rather bad assumption. On the other hand, if I use the 12 ohm max impedance to decided the functional peak output is going to be rather low...

Since music is broad spectrum, picking the "representative" impedance value of the speaker will be an inexact estimate. The only values in Amir's measurements are for 4 & 8 ohm, with the 4 ohm output number giving us a more conservative result. IMHO the 4 ohm power output is the appropriate choice.

Audio is hard :/

"Hi-rez" music is only 192 kHz sampling rate. The SOTA oscilloscopes have ADC of 200 GHz (well, they only have a few bits resolution). Which means for each sample of "hi-rez" music, a SOTA scope can take 1 million samples. Now that's hard

 

[mdsimon2]

As a reference point a volume position of -10 dB on the E50 is a DAC output voltage of 1.34 V. At a gain of 19.5 dB this is 40 W at 4 ohm and 20 W at 8 ohm so definitely within the bounds of the amplifier.

Michael