Is Audio Science Review going about it all wrong? Or partly wrong? Or all right?

[Blumlein 88]

snip....

Measuring speakers would be wonderfully good to do but not very practical due to need for specialized room (anechoic chamber). Speakers are also difficult to source due to size, weight, etc. Places that do that (soundstage) rely on manufacturer loans which is not what we are about.

Ultimately ASR Measurements allow you to find the best performing device for least money with features you want. This is something you can't do without the data we provide.

What I'd like to see (and hey no problem for me as long as Amir does the work ), is at least measuring the load of quite a few loudspeakers. As in impedance plots. And put together an adjustable speaker load so amplifiers could be tested with the AP while connected to loads that aren't resistance. I wouldn't think you need an anechoic chamber for that. Some speaker impedance plots are available. Even if not for an actual speaker loads nearly like some common well regarded speakers would be useful. A big ESL, a large Revel, so on and so forth. Test the amp out on that load. JA at Stereophile has one such load he sometimes tested with. Something a little more advanced in that direction. He only showed changes in FR. I'd like to see the full suite of measurements with an analog of a speaker load or two. It would of course mean running all the tests three times once with a resistive load, say once for a box speaker load, and once for an ESL type load.

Given that so far many of the amps you've tested are surprisingly sub-par I think bigger differences would emerge with a load like a genuine speaker presents.

 

[restorer-john]

And put together an adjustable speaker load so amplifiers could be tested with the AP while connected to loads that aren't resistance.

SIY had offered to build such a load box for Amir IIRC. It'd be good to put up a basic schematic and recommended parts list so anyone can knock one up. Two channels, 8/4/2 ohm switchable loads and two each switchable C and L values would be a good start. As long as the load resistors are 100W ceramic non-inductive (<10%), they can take 3 or 4 times that short term.

 

[SIY]

SIY had offered to build such a load box for Amir IIRC. It'd be good to put up a basic schematic and recommended parts list so anyone can knock one up. Two channels, 8/4/2 ohm switchable loads and two each switchable C and L values would be a good start. As long as the load resistors are 100W ceramic non-inductive (<10%), they can take 3 or 4 times that short term.

The reality is that even regular wirewounds have low enough inductance to be pretty negligible. A couple microhenries is typical. Having the APx1701 is useful, but with some jiggering, you can use a "normal" power amp.

 

[DonH56]

I suggested a Power Cube some time ago but they are not cheap... https://www.audiograph.se/ I think a few loads covering some of the bounding conditions would be good. Something with a big impedance spike in the bass then a dip to emulate a ported driver, maybe a hefty capacitor to emulate an ESL...

One concern Amir probably has is that he does not want to overstress (e.g. destroy) an amplifier loaned for him to test.

 

[amirm]

One concern Amir probably has is that he does not want to overstress (e.g. destroy) an amplifier loaned for him to test.

This is definitely an issue. I damaged one already and seen some others get pretty unhappy at the limit.

 

[restorer-john]

The reality is that even regular wirewounds have low enough inductance to be pretty negligible.

I have found the same. Even cheap Chinese wirewounds are remarkably accurate and low inductance. They are however rather 'optimistic' in their dissipation ratings...

 

[restorer-john]

This is definitely an issue. I damaged one already and seen some others get pretty unhappy at the limit.

The trouble is, the performance at the limit (overload recovery etc) and the fragility, or lack thereof are important tests and deserve exposure where they fall down. I do however understand you'd rather not be repairing blown-up loaner gear full of bespoke SMD parts and explaining to the owners you killed it.

It's a bit like those dyno fails on youtube. Hotted up cars pushed to the limit and then they go 'boom'. Funny but sad.

 

[Blumlein 88]

The trouble is, the performance at the limit (overload recovery etc) and the fragility, or lack thereof are important tests and deserve exposure where they fall down. I do however understand you'd rather not be repairing blown-up loaner gear full of bespoke SMD parts and explaining to the owners you killed it.

It's a bit like those dyno fails on youtube. Hotted up cars pushed to the limit and then they go 'boom'. Funny but sad.

So would you suggest Amir always video the testing he does with such loads. Then have a youtube channel called, "letting the smoke out"?

Okay different suggestion.

@pkane has gotten his Deltawave software pretty well developed. Testing with tones at max high power and such are potentially destructive. If you have a track or two of music you always use, maybe recording the amp output terminals when hooked to speaker-like loads, and do a null test against the original would tell you something. This way the amp is being stressed as it normally would with music and any high power moments are fleeting and short-lived. I know it can tell you frequency response without having to do a frequency sweep.

 

[solderdude]

Edit: Ray showed the errors of my ways.. peaks in that FR part can be just below 0dB.

 

[cjfrbw]

I don't know. A You Tube video of Amir surrounded by flaming amplifiers and dac's might be pretty cool.

 

[Sir Sanders Zingmore]

RULE NUMBER ONE in ANY listening comparison sighted or blind is :

YOU MUST MATCH LEVELS!!!!!!!!!!!!!!!!!!!!!!

Even if you prefer the listen and describe subjective impressions methodology you still must match levels. Without that no conclusions are worth the time to think of them or express them.

Is matching level by ear good enough? NOOOOOO!
Levels must be matched to within .1 db and matching by ear is likely only good to 1 db or perhaps a half db. .2 db level difference is enough to make one bit of kit sound better/different vs another.

A good simple method is to measure voltage with a test tone (maybe -20 db 1 khz) at the speaker terminals. You want to get it to within 1 percent. So if one amp measures 10 volts, you want the other to be within 9.9 to 10.1 volts. Figure out how to switch gear and adjust volume between gear to maintain this. Then listen to music.

This is step one. This is the step without which none of your impressions are valid.

What does level matching mean if the frequency response is different between the components being compared?
You might level match to within 0.1v at 1kHz and still be un-matched at some other frequency.

 

[RayDunzl]

no music signal ever has 18kHz at the same level as say 100Hz.

Quick look:

Daft Punk Random Access Memories Track 4 decibel display:

Ok, came close...

 

[RayDunzl]

What does level matching mean if the frequency response is different between the components being compared?

I'd think it means you step back and ask "Why is the frequency response different?"

 

[Dimitrov]

Can anyone recommend an inexpensive switch box to compare, let's say an AVR to a stereo amp? I want to be able to demonstrate that golden-ear audiophiles are just as stone deaf and tin eared as the objectivsts (per the usual claim).

 

[Blumlein 88]

What does level matching mean if the frequency response is different between the components being compared?
You might level match to within 0.1v at 1kHz and still be un-matched at some other frequency.

Usually matching at 1 khz will match over the majority of the response (excepting speakers). One could match over more than one frequency, or one could fire pink noise thru and match levels. Just as a for instance, if you had pink noise flat with one device and the other begins to droop at 10 khz and is 6 db down at 20 khz, the RMS level for the those two files differs by less than .2 db and it'll be the response droop you hear. Up until 10 khz they are the same and 6 db droop would be rather large for anything other than speakers.

I think for speakers the suggestion is to use pink noise which has been filtered leaving the noise between 500 hz and 2000 hz for level matching. With any non-speaker gear that would result in the same setting as using a 1 khz tone for all, but gear with a very strange response.

 

[Theo]

What does level matching mean if the frequency response is different between the components being compared?
You might level match to within 0.1v at 1kHz and still be un-matched at some other frequency.

I guess the answer is in the question. Differences in the spectrum mean that there are actual differences between the devices, so if you're trying to prove that, it's done. OTOH, stating that there are differences if levels are not matched is wrong. So, you should follow Blumlein88 suggestion in any case.

 

[DDF]

Can I suggest another test that would be easy to construct....I propose a multitone tone test exactly similar to the one he uses now when it comes to frequencies but sloping in amplitudes in a similar way as music content is.

This is how I conduct my multitone distortion testing. I weight the tones according to the spectral density defined in the IEC spec
https://www.diyaudio.com/forums/mul...ion-loudspeakers-multitone-2.html#post4544215

 

[solderdude]

This is how I conduct my multitone distortion testing. I weight the tones according to the spectral density defined in the IEC spec
https://www.diyaudio.com/forums/mul...ion-loudspeakers-multitone-2.html#post4544215

Makes sense for speakers...

 

[LTig]

Usually matching at 1 khz will match over the majority of the response (excepting speakers). One could match over more than one frequency, or one could fire pink noise thru and match levels. Just as a for instance, if you had pink noise flat with one device and the other begins to droop at 10 khz and is 6 db down at 20 khz, the RMS level for the those two files differs by less than .2 db and it'll be the response droop you hear. Up until 10 khz they are the same and 6 db droop would be rather large for anything other than speakers.

I think for speakers the suggestion is to use pink noise which has been filtered leaving the noise between 500 hz and 2000 hz for level matching. With any non-speaker gear that would result in the same setting as using a 1 khz tone for all, but gear with a very strange response.

The problem with pink noise signal is that its amplitude varies quite a lot which means that a simple volt meter will not do - you have to average the readings over 10 sec or more.

I would suggest a multi tone signal where the amplitudes are A weighted (or some other real world weighting).

 

[JohnYang1997]

Multitone is definitely useful. But recently I measured smsl su-8 which the sloping of the multitone will drastically change the distortion level. In rew there is an option to select white or pink.

About audible transparency, I don't care all other people say. The dacs sound different and they measure different. I am still learning the correlation of things. Especially the ess ones.
Also I'm currently designing an amplifier. Different opamps certainly sound different despite that they all achieved 0.00007% thd for at 100mw into 16ohn load, and 0.00017% 1.5W into 4.7ohm. They just do. Fet input (i use parallel design so reduced noise to 0.7uV and below) just has fuller bass and more solid imaging than bjt input say opa1612. I even only have 1kohm input impedance/250ohm source impedance. Also opa2156 and opa1642 sound different. Opa1642 just has better sound stage. I use one opa1642 for each channel so it must not be the separation of the chip itself. I match volume to 0.1db at 1khz. there are difference. They measure a bit differently but no where to correlate to actually explain why.
I chose opa2140 in the end for lower offset version of opa1642.
Things just don't always work as we think they do. We just don't know everything yet. If you can duplicate 100% sure it will sound the same but that's different from having some similar numbers in measurements.