Standardized Low to Medium Power Testing

[amirm]

I am developing a set of tests for power amplifiers. In there, I am graphic five power levels in between 1 and 50 watts. I picked 1 watt because that is sort of an industry standard. I selected 50 watts so that it encompasses vast majority of amplifiers and is pretty decent level for continuous music listening. I also want to avoid cooking the amplifier too much as this is a lengthy test.

Here is an example with ncore400 amp:

Note that I will be running with different types of loads (e.g. realistic speaker load and resistive) so having more than 5 while easy, can be hard to interpret on screen.

I have picked a bandwidth of 45 kHz for this as to make this "high-resolution audio" ready.

Note that the steps are picked by the analyzer. I will be picking the two extremes and then let the analyzer find the steps between them.

What say you?

 

[levimax]

I am developing a set of tests for power amplifiers. In there, I am graphic five power levels in between 1 and 50 watts. I picked 1 watt because that is sort of an industry standard. I selected 50 watts so that it encompasses vast majority of amplifiers and is pretty decent level for continuous music listening. I also want to avoid cooking the amplifier too much as this is a lengthy test.

Here is an example with ncore400 amp:

View attachment 20382

Note that I will be running with different types of loads (e.g. realistic speaker load and resistive) so having more than 5 while easy, can be hard to interpret on screen.

I have picked a bandwidth of 45 kHz for this as to make this "high-resolution audio" ready.

Note that the steps are picked by the analyzer. I will be picking the two extremes and then let the analyzer find the steps between them.

What say you?

Extending the bandwidth out to 100 KHz would better show what the 20 KHz distortion actually is, in the graph shown it appears THD is dropping at 20KHz when it is actually going up rapidly. The importance of 20 KHz distortion is unknown to me but it tells us something about the amplifier. Is there any disadvantage to extending bandwidth to 100 KHz?

 

[restorer-john]

1 watt at 8 ohms (2.83V) (the defacto sort of standard)
1 watt at 4 ohms is only 2V however.

By the time you do 5 power levels for just 4 and 8 ohms, plus a simulated speaker load, that's 15 plot lines, or 3 different plot screens. You'll run out of colors too. Letalone if you do single channel and both channels driven situations, it will double again.

I think do 1 watt, maybe your 5 watt (that you love) half and full rated power. Lots of amps are giving rated power/2 THD ratings I notice these days- Hypex is one of them. Are you building an automated group of tests and don't want to set the half/full rated power each time?

What happens with a lower than 50 w rated amp?

Extending the bandwidth out to 100 KHz would better show what the 20 KHz distortion actually is, in the graph shown it appears THD is dropping at 20KHz when it is actually going up rapidly. The importance of 20 KHz distortion is unknown to me but it tells us something about the amplifier. Is there any disadvantage to extending bandwidth to 100 KHz?

Class D is the problem. Must use a filter in front of the AP or it will throw a tantrum.

 

[amirm]

Extending the bandwidth out to 100 KHz would better show what the 20 KHz distortion actually is, in the graph shown it appears THD is dropping at 20KHz when it is actually going up rapidly. The importance of 20 KHz distortion is unknown to me but it tells us something about the amplifier. Is there any disadvantage to extending bandwidth to 100 KHz?

Yes. The issue is that music content has far lower levels as frequencies go up. The sweeps on the other hand, have constant level from 20 Hz to max frequency measured. In that sense, the distortion seen at high frequencies is hugely exaggerated relative to actual level.

Let me show this as a real example since I am in the lab. Here is the spectrum and hence distortion and noise on Hypex NC400 at 10 Khz, first at 10 watts:

We have a second harmonic at 20 kHz and third at 30 kHz. Here is what happens when I go down to 1 watt:

Notice that the harmonic distortions are now buried in the noise.

As we know, the tweeters receive far less power than woofers/subs for this reason.

Edit: plotted the wrong graph for 1 watt. Corrected.

 

[amirm]

Here is a quick spectrum of a high-res file (Reference Recordings):

By the time we get to 20 khz, the level is down about 35 dB which is a massive drop in power and hence distortion.

 

[amirm]

By the time you do 5 power levels for just 4 and 8 ohms, plus a simulated speaker load, that's 15 plot lines, or 3 different plot screens.

Right now I am hoping for a lot of these more in-depth measurements that the load can be just 4 ohm. 10 graphs seems to be manageable:

It is hell though if I want to compare two amps with all the variations. So I would probably show them grouped and side by side in two graphs.

 

[amirm]

Here is the same graph as post 1 but with 90 kHz bandwidth:

It is much less cluttered and easier to understand even though I am showing all 10 graphs now. So maybe you are talking me into running it this way.

 

[amirm]

Class D is the problem. Must use a filter in front of the AP or it will throw a tantrum.

The AES-17 filter that I use has flat response to 40 kHz. It keeps going almost at that same amplitude to 90 kHz so it is not a concern here.

Its job is to cut down switching frequencies above 200 kHz or so.

 

[restorer-john]

So you'll give all the data descriptions (in the legend box) like

NC400 1W 4R+2.2uF S
NC400 1W 4R S
NC400 50W 4R S
NC400 5W 4R S

Maybe use S for single channel, B for both driven? (or 1, 2 etc)

I presume you are saving all the individual sweep data sets permanently, because you seem to be able to retrieve them at will for comparisons.

 

[amirm]

I have not tried to label them correctly for these shots. Wish the software would auto-label them but it does not.

But yes, when I save an AP project, it will save the graphs. Sadly this is not true of all measurements but most are that way.

 

[Blumlein 88]

You aren't requiring much current. If you wish to use 4 ohms, I'd like to see the test to 100 watts. 20 volts. Few if any amps will be worse at 8 ohms so testing 4 ohms only is fine. Maybe 50 watts for any amp capable of 100 watts or less, and 100 watts for those that are capable of more. Plus this is enough power with most speakers to squeak out over 100 db SPL which is probably around the headroom you might want over average listening levels.

 

[miero]

Who can hear a distortion of the 20kHz tone?
- https://www.dpamicrophones.com/mic-university/the-basics-about-distortion-in-mics

 

[miero]

How about displaying a distortion number in dB instead of %?

 

[restorer-john]

...Few if any amps will be worse at 8 ohms so testing 4 ohms only is fine...

The 8 and 4 ohm testing is incredibly useful for determining the quality of the power supply and yes, distortion varies considerably at high powers into 4 and 8 ohms.

 

[March Audio]

Yes. The issue is that music content has far lower levels as frequencies go up. The sweeps on the other hand, have constant level from 20 Hz to max frequency measured. In that sense, the distortion seen at high frequencies is hugely exaggerated relative to actual level.

Let me show this as a real example since I am in the lab. Here is the spectrum and hence distortion and noise on Hypex NC400 at 10 Khz, first at 10 watts:

View attachment 20383

We have a second harmonic at 20 kHz and third at 30 kHz. Here is what happens when I go down to 1 watt:
View attachment 20387


Notice that the harmonic distortions are now buried in the noise.

As we know, the tweeters receive far less power than woofers/subs for this reason.

Edit: plotted the wrong graph for 1 watt. Corrected.

Plus to add to this most tweeters are significantly dropping their output above 30kHz so the ultrasonic harmonics that are there are reduced even further than the plots above show. (not forgetting that you cant hear them anyway )

 

[March Audio]

So you'll give all the data descriptions (in the legend box) like

NC400 1W 4R+2.2uF S
NC400 1W 4R S
NC400 50W 4R S
NC400 5W 4R S

Maybe use S for single channel, B for both driven? (or 1, 2 etc)

I presume you are saving all the individual sweep data sets permanently, because you seem to be able to retrieve them at will for comparisons.

Yes, the graphs are confusing without this sort of labeling.

 

[restorer-john]

How about displaying a distortion number in dB instead of %

Take the THD percentage, divide by 100, take the inverse (1/x), log it and multiply by 20 (as it is voltage we are talking about).

Easy. Keeps your brain active and your scientific calculator on your desk where it should be.

 

[March Audio]

Who can hear a distortion of the 20kHz tone?
- https://www.dpamicrophones.com/mic-university/the-basics-about-distortion-in-mics

Interesting point, I havent looked at the link yet, but I might try an experiment. generate a track with a pure10kHz tone (ie something we should all be able to hear) and then additional tracks with added harmonic distortion at increasing levels. Youll have 20, 30, 40 etc harmonics. Lets see if we can hear a difference.

 

[restorer-john]

Yes, the graphs are confusing without this sort of labeling.

It would get mighty tedious keying that in 10-15 times, manually, for each sweep test though. Late at night, errors creep in. Maybe a standard set of colour codes would be easier, with a cut and paste legend?

 

[March Audio]

Take the THD percentage, divide by 100, take the inverse (1/x), log it and multiply by 20 (as it is voltage we are talking about).

Easy. Keeps your brain active and your scientific calculator on your desk where it should be.

or if you are really lazy like me go here
http://www.sengpielaudio.com/calculator-thd.htm