18 replies to this topic

[Whatmore]

When I compare two components, I always try to ensure I'm listening to both at the same volume. My clumsy attempt to do this involves playing the same bit of music with both components and making sure the SPL on my iphone app is the same.

I'm sure there is a more "correct" method and I'm guessing it involves more than just having a more accurate SPL meter...

Does it involve matching white noise SPL?

If so, what happens if the frequency response of the two components is different?

I guess, in a roundabout way I'm asking if it means anything to say that two components are "volume matched" if they have different frequency responses?

[GFuNK]

When I've ABX'd DACs in the past I run a 0 dBFS test tone through my squeezebox and then measure the voltage output from the DAC with a voltmeter. Your aiming for < 0.1 dB difference in voltage.

This however assumes a constant frequency response. You could repeat with several frequencies if you were concerned.

Also, I always use the same input to the amplifier to avoid another possible bias. I do this with an ABX unit.

Not sure if you can get a < 0.1 dB tolerance on the iPhone, compare it with using a voltmeter and let us know how close it gets.

[Whatmore]

I think the crux of my question revolves around what happens if the frequency response is not constant ... Using your method, voltage matching across several frequencies will lead to different volume settings for each frequency

[JA]

Hello W,

Good question. Well if the components under test have different freq responses and you try to volume/level match using noise it will provide some sort of broadband average and it's probably more fair than either using a spot frequency match (say at 1 kHz typically used for various refereneces) or not volume matching at all. There's probably also a good argument that could be made for level matching using a test signal that is weighted in the region where the ear is most sensitive and/or level matching using a broadband noise and using a weighted measurement, for example dB(A) for acoustic measurements. You also don't need to level match acoustically, you can also level match using electrical measurements, for example perhaps with preamp testing..

In any event, if the level matched components have different enough frequency responses you will hear this anyway, for the right reason, the frequency response difference, rather than it being due to any outright level mismatch on the test, which was the original intent of the test. Is that circular enough ????

EDIT: Taking it to the nth degree, perhaps the fairest (and most time consuming) might be to use a broadband test signal or sweep and a freq response measurement, with the levels in both devices adjusted until the tester was satisfied that the freq response curves were the best approximation of one another, counting all the dips and peaks, sort of like curve fitting.

Best
JA

Edited by JA, 12 January 2012 - 11:14 PM.

[LogicprObe]

Maybe white noise would be the best?

[GFuNK]

I think the crux of my question revolves around what happens if the frequency response is not constant ... Using your method, voltage matching across several frequencies will lead to different volume settings for each frequency

If the frequency response varies wildly it's not hi-fi

Any method, in which case, will be an approximation. You could try to capture the frequency response of a broadband source and analyse that.

There isn't any need to use a SPL meter if: you use the same input Chanel and level with the amp, the same speakers and if you use the same listening position.

Assuming a linear system, the same voltage to the amp will give the same SPL.

[monaro8]

Good morning Trev,

I've tried this before to make sure I am comparing components at the same volume level.

In order to avoid too much mucking around, repeatedly getting up and down off my listening position, you can try this method...

a. if the volume sounds a bit louder than before...partially block both ears with your fingers (make sure you wash your hands first)(cotton wool may also be used)

b. move your head back or forth being careful not to strain your neck muscles

c. if you have a remote control, then of course, use that

d. when you get tired of it...get a glass of wine or a beer, sit back, relax...and enjoy the music.

e. completely ignore the above and go right on with your test....

Vanch...

(yes, I'm still here mate...)
(and...No, I haven't blown up another valve amp lately...touch wood)

Edited by monaro8, 13 January 2012 - 07:28 AM.

[Whatmore]

Aaah there you are Vanch
we've missed you. How are those speakers going?

[Whatmore]

If the frequency response varies wildly it's not hi-fi

Any method, in which case, will be an approximation. You could try to capture the frequency response of a broadband source and analyse that.

There isn't any need to use a SPL meter if: you use the same input Chanel and level with the amp, the same speakers and if you use the same listening position.

Assuming a linear system, the same voltage to the amp will give the same SPL.

I'm interested in this question for a few reasons. One of these is - if there is a subtle frequency difference between components then it may still be enough to cause a preference for one component even when you think you've volume-matched

[Drizt]

I'm interested in this question for a few reasons. One of these is - if there is a subtle frequency difference between components then it may still be enough to cause a preference for one component even when you think you've volume-matched

That is true. No real way around it though. All you could do would be to measure the devices and then be aware of their frequency response anomalies.

[Ando]

I believe that current thinking is that for comparing things like loudspeakers that can have large frequency response differences the best way of matching perceived loudness is using a sound level measurement with a weighting that rolls off lows between the standard B and C weightings ( starts around 100Hz, down about 12db at 20 Hz ) and has flat high frequency response.
I think JA made a really good point above in that frequency response differences are real differences , if possibly the most easily fixed. I suspect that ( subtle ) frequency response difference may account for many of the differences between components attributed by some people to all sorts of other mechanisms.
Cheers Mike

[Arg]

I think the standard is B weighted pink noise.

[Ando]

I think the standard is B weighted pink noise.

I think this is outdated by the work of Soulodre and later others as the standard.

http://www.sfxmachin...nd_dynamics.pdf

Edited by Ando, 16 January 2012 - 08:16 PM.

[Whatmore]

OK this has gotten a wee bit too technical for me. Can someone explain what weightings are (as in B and C weightings) and also what is pink noise (as opposed to white noise) ?

[bhobba]

The big boys at double blind tests do it with an occiliscope to get them as close as possible. The claim is perceived amp differences is due to volume differences. When I compare amps etc and want to be careful I don't go to that extreme but rather use a SPL mater. However IMHO its nothing to really worry about because all you do is vary the volume of both up and down and you will notice with actual differences they do not change as you do that.

Thanks
Bill

[Ando]

W, White noise and pink noise have different energy distributions. White noise has equal energy by frequency bandwidth. From 40 to 45 Hz has the same energy as 1200 t0 1250 Hz etc. Pink noise has equal energy by octave so 40 to 80 Hz has same energy as 1200 to 2400 Hz. Pink noise which has a flat frequency response is more "natural " and is commonly used for loudspeaker/acoustics testing.
The noise weighting curves A, B,C etc ( http://www.cassafe.c...-responses.html ) are used to try to convert sound pressure level readings into measures of perceived loudness. They are needed because of the ears different sensitivity to different frequencies which varies with loudness levels. ( equal loudness curves : http://en.wikipedia....oudness_contour ). The different weightings have different uses. The A weighting which discounts low frequencies the most was developed for use at lower volumes.
The RLB weighting that I described above was developed as an improvement over using the B weighting in the type of situation you asked about. I was wrong to call this a standard as it seems no real standard exists and people use the method/weighting they think will work best in their situation.
As JA pointed out in his post no type of weighting will make two speakers, or whatever, with grossly different frequency responses have the same perceived loudness.
Cheers Mike

[Arg]

I think this is outdated by the work of Soulodre and later others as the standard.

http://www.sfxmachin...nd_dynamics.pdf

Hi, that's an interesting paper but I still think audio researchers use B weighted pink noise to equalise the level of different loudspeakers before conducting listening tests.

[Ando]

I still think audio researchers use B weighted pink noise to equalise the level of different loudspeakers before conducting listening tests.

Hi Arg, No doubt you are correct that many researchers do. I think Canadian NRC and Harman use the RLB and they seem to be amongst the leaders in home loudspeaker listening tests .
Cheers Mike

[davewantsmoore]

If so, what happens if the frequency response of the two components is different?

Gross generalisations:

For two equal responses, slightly louder sounds better
High Q deviations can be either added detail, or missing detail
Low Q deviations, are more noticeable, and affect tonal balance