Complex Load for Power Amplifier torture testing

[Blumlein 88]

Do we need a new standard for rating speaker impedance? One called musical impedance.

Maybe play a specified level of pink noise (more or less mimics frequency balance in music) through a speaker and see what the 20 khz band of power, voltage and current is to arrive at the effective musical impedance of a speaker. I'd bet most ESL's would get something like a 16 ohm rating this way. Some of those large ported box speakers will probably get a rating below 4 ohms.

 

[restorer-john]

I owned some ESL63s for a decade. You really should get some. I still think in some ways nothing is as good.

Does anyone have an up to date opinion on the more modern Quad ESL incarnations that came out of IAG?

 

[March Audio]

Just one other point which may not have been clear/obvious from the previous set of plots.

In the plot below and the Kenwood in the OP, we are plotting each harmonics distortion against the fundamental frequency. That is to say if the input signal is 1kHz then harmonic 3 is at 3kHz even though its distortion level is plotted at 1kHz.

This is why the lines stop short of the graphs end. The measurement was to 48kHz bandwidth, so for example harmonic 5 stops at 48/5 = 9.6kHz. Although looking at the plot REW seems to truncate a little before that.

If we plot the harmonics at the actual harmonic frequency we see an interesting thing with the Purifi. All the (very low) harmonic distortion is above 20kHz.

i.e. inaudible.

As it progresses in frequency any speakers output will be reducing, reducing it even further.

So it clearly doesnt care about capacitive loads or pathologically low impedance values. In band, where it matters, its completely "unphased" by it.

With the Kenwood from the OP, although its not presented in this format we can see the distortion ramping from about 300Hz, so all the harmonics are having an impact within the audible band below 20kHz.

 

[restorer-john]

@March Audio Just remind me again, that's just 5W isn't it? Those loads of yours can take a whole lot more than that, let's see the amplifier at rated power.

 

[ajawamnet]

@ajawamnnet The Futterman schematic doesn’t look right. If memory serves, in the Futterman circuit, the cathode resistor of the split load inverter was returned to the amp output rather than to ground.

The 4CX and 3CX transmitting tubes are not good for audio. Besides the high voltages, noise from forced air cooling, and grid current, they need a high impedance load, which makes output transformer design very challenging.

Sounds kinda like the issues with a turntable and vinyl... yea, so impractical. Might mean it would sell even more.... super exclusive audiophile club where you need a separate room just for your amplifiers...

Makes me think of a general I was talking with at one of our sites - right after the Columbia "disassembled" on re-entry. I mention how awful that was...

"Yea - not as bad as that Challenger... at least with Columbia, we got our stuff up there. There's always stuff in that payload bay that ain't on no manifest..."

"But what about the crew?" I asked...

"Ehhh .. the more it blows up the more whacko's are gonna wanna fly on it..."
I just wanted to puke at that point... my father was in the military for 39 years, and he didn't think that way...

And yea - there'

 

[March Audio]

@March Audio Just remind me again, that's just 5W isn't it? Those loads of yours can take a whole lot more than that, let's see the amplifier at rated power.

Yes, same as Pavels test. I cant go further with that reactive load as its only a 3 watt resistor. It gets very warm as it is and we will release magic smoke

 

[restorer-john]

I cant with the reactive load as its only a 3 watt resistor. It gets warm as it is and we will release magic smoke

Fair enough. If you've got 4 of the big aluminium ones at 4ohms each, you could a high power 1 ohm resistor and the cap.

 

[restorer-john]

It gets warm as it is and we will release magic smoke

Ah, gotta love the smell of burning nichrome in the morning.

 

[March Audio]

Actually I have just noticed an error in the test. It wasnt at 4.5 V rms output (5 watts into 4 ohms) it was actually 6 v rms which is 9 watts.

So if we ar getting down to 0.5 ohms at high frequencies thats 72 watts. Very pleased I chose a short sweep!

 

[March Audio]

I just ran 8 sweeps which REW will average to reduce noise increase accuracy. They are only 5.5 seconds duration each but that poor little resistor was begging for mercy!

Purifi reactive load 9 watts 6 volts rms

 

[Blumlein 88]

Does anyone have an up to date opinion on the more modern Quad ESL incarnations that came out of IAG?

Haven't listened to any really. One model is just a copy made less expensively in the Far East. One is something of an oversized version. So I don't see why it wouldn't work okay. I'd like to see them just double up the size in every dimension, add more segments and adjust as needed for it to work.

Of course I'd like for someone to make one with many more segments and feed them separately with DSP.

 

[March Audio]

For reference this is the Purifi 4.45 V rms 4 5 watts into 4 ohms 8 averages

 

[amirm]

Do we need a new standard for rating speaker impedance? One called musical impedance.

Maybe play a specified level of pink noise (more or less mimics frequency balance in music) through a speaker and see what the 20 khz band of power, voltage and current is to arrive at the effective musical impedance of a speaker. I'd bet most ESL's would get something like a 16 ohm rating this way. Some of those large ported box speakers will probably get a rating below 4 ohms.

Paul Miller of Hifi News has done something like this. He has some statistical data on music spectrum which he then applies to impedance curve and angle. The result is then an effective resistive only impedance. I tried to research the details but did not get far.

 

[March Audio]

For further information this is my P701 amp, Hypex NC1200

4 ohm purely resistive load, 5 watts 4.45 V rms

reactive load, 5 watts 4.45 V RMS

So you can see the older Bruno design is not quite as good as his new Purifi design, but again no issues and almost all the harmonic distortion resides in the inaudible ultrasonic region above 20kHz.

 

[DDF]

Paul Miller of Hifi News has done something like this. He has some statistical data on music spectrum which he then applies to impedance curve and angle. The result is then an effective resistive only impedance. I tried to research the details but did not get far.

IEC 268-1, 1985 is a standard signal whose mean power spectral density is meant to be representative of a wide variety of music types. It’s basically flat from 100 Hz to 2 kHz, approx 6 dB down at 5 kHz and 12 dB down at 10kHz

Peter Chapman at B&O published a paper (AES preprint 4277) comparing a large number of CDs to the standard, while also looking at peak data. The average powers of most music followed the spec closely, except for hip-hop, which was almost flat from 300 Hz to 10 kHz, with 6 dB prominence at 60 Hz. There's also useful data in "The Spectral Amplitude Distribution of Selected Compact Discs", JAES Vol 37 No 4 plots power spectrum distributions with 1 to 90% interval graphs, for various instruments or ensembles.

I remember when working on a compressor spec years ago, I found that absolute peak levels were fairly flat across frequency (up to 7kHz, the cut off of my analysis), and it was duration that reduced as frequency increased, leading to the low pass nature of the power spectral density.

 

[amirm]

The data on music spectrum is not an issue. The issue is the mapping he uses to then convert that, plus full impedance and phase into a representative resistive load for the speaker. I can easily reason how that would work but would be nice to have the spreadsheet/matlab used to produce the single value.

 

[pma]

@pma @restorer-john

OK, I went rummaging in my box of crazy crap and it turns out I had the parts to try this test so threw it together on the bench. Slight variation, 30uF and 0.56 ohms.

Purifi amp running 4 ohms resistive at 5 watts

Purifi reactive load. HF 0.5 ohms

No issues at all.

Thanks for doing the tests, @March Audio . There is a small difference, as mentioned by @restorer-john , though your modules are about 400W rated power, my KA-5010 clone with only 2x25V power supply is rated at 40W/4ohm. Speaking in power terms, talking about 5W makes sense only for the resistive load. With the R//R-C load according to post #1 and constant output voltage of 4.47V, 5W is true for 50Hz frequency, but at 20kHz the power into 0.5ohm is 40W and output current is 9A. This is just at the limit of the small KA-5010 clone. For your amplifiers, this is deep below the limit and to make a comparable test, you should run it at 14.14V, this would be 50W/50Hz/4ohm and 400W/20kHz/0.5ohm. Output current would be 28A/20kHz. Then the tests would be comparable as of the stress created to the amplifiers.

Just one more comment - testing in REW and STEPS is not equal.
REW sends a sine sweep with duration of several seconds for the whole frequency band. STEPS sends triggered sine wave at each frequency, frequency increment in my case is 1/24 octave and at each frequency the wave is of 200ms duration continued by 200ms pause. The whole measurement takes several minutes, not seconds.

 

[restorer-john]

Pavel, to be fair, @March Audio wanted to knock up some quick tests with parts at hand. He's probably not got the parts repositories you and I have. I'm glad he presented what he did. Of course, more power is needed, but he may not have a big reserve of modules just to play with at the moment. After all, the one he really takes to the limit may have to end up being a test mule and not sold. They are expensive.

I don't think we need to worry too much, the light has been shone on reactive load testing, difficult phase angles and stability in power amplifiers. Amir is slowly coming around. He even gave you a square wave...

I think what we can do, is talk about what complex loads are important going forward, so we can perhaps standardize on a bank of loads that can be built inexpensively and can offer comparative test results.

 

[March Audio]

Thanks for doing the tests, @March Audio . There is a small difference, as mentioned by @restorer-john , though your modules are about 400W rated power, my KA-5010 clone with only 2x25V power supply is rated at 40W/4ohm. Speaking in power terms, talking about 5W makes sense only for the resistive load. With the R//R-C load according to post #1 and constant output voltage of 4.47V, 5W is true for 50Hz frequency, but at 20kHz the power into 0.5ohm is 40W and output current is 9A. This is just at the limit of the small KA-5010 clone. For your amplifiers, this is deep below the limit and to make a comparable test, you should run it at 14.14V, this would be 50W/50Hz/4ohm and 400W/20kHz/0.5ohm. Output current would be 28A/20kHz. Then the tests would be comparable as of the stress created to the amplifiers.

Just one more comment - testing in REW and STEPS is not equal.
REW sends a sine sweep with duration of several seconds for the whole frequency band. STEPS sends triggered sine wave at each frequency, frequency increment in my case is 1/24 octave and at each frequency the wave is of 200ms duration continued by 200ms pause. The whole measurement takes several minutes, not seconds.

Nahh.

Come on, earlier you said this is a test any decent amp should pass, now you want to make excuses for the Kenwood.

Plus I would have to purchase additional high power resistors.

How about you find a 400 watt A/B amp, or any you can lay your hands on, and and try the 5/9 watt test

What I can do is try my P122 which is closer to your Kenwood.

Well, REW is free, can you repeat your test with it? It would also allow us to compare clearly the in band / out of band distortion levels as you can plot the harmonics at the harmonic frequencies.

 

[Blumlein 88]

IEC 268-1, 1985 is a standard signal whose mean power spectral density is meant to be representative of a wide variety of music types. It’s basically flat from 100 Hz to 2 kHz, approx 6 dB down at 5 kHz and 12 dB down at 10kHz

Peter Chapman at B&O published a paper (AES preprint 4277) comparing a large number of CDs to the standard, while also looking at peak data. The average powers of most music followed the spec closely, except for hip-hop, which was almost flat from 300 Hz to 10 kHz, with 6 dB prominence at 60 Hz. There's also useful data in "The Spectral Amplitude Distribution of Selected Compact Discs", JAES Vol 37 No 4 plots power spectrum distributions with 1 to 90% interval graphs, for various instruments or ensembles.

I remember when working on a compressor spec years ago, I found that absolute peak levels were fairly flat across frequency (up to 7kHz, the cut off of my analysis), and it was duration that reduced as frequency increased, leading to the low pass nature of the power spectral density.

So the 1985 standard is basically flat to 2 khz and then a pink-noise like roll off.