Measurements vs Listening

[av_phile]

While I strongly believe that accurate well made measurements can give you an idea of how a gear can sound, there are some measurements that don't reveal the entire picture, expecially when they are truncated or summarized for conciseness as often the case with published technical specs in brochures. Especially when comparisons are made. In amplifiers for instance, two identically powered amps with the same 20hz-20khz +-1db frequency response can still sound different from each other, all other specs equal. That's because one's FR can have some frequencies boosted 1db in much of the same frequencies were the other is attenuated by the same amount. That accounts for 2 db difference which is audible. In which case, only a complete frequency-amplitude plot across the bandwidth will tell the entire story, not just a descriptive technical phrase. But it is rare that amps and speakers sell with a complete FR graph.

 

[WmAx]

INdeed, you are correct that the standard manufac. spec is not very useful. Ironically, the one good thing Stereophile is good for is the very good and complete set of amplifier meaurements they provide, at various loads and amplitudes. I tend to only read the measurements in stereophile.

I rarely trust measurements as published by the manufacturer.

-Chris

While I strongly believe that accurate well made measurements can give you an idea of how a gear can sound, there are some measurements that don't reveal the entire picture, expecially when they are truncated or summarized for conciseness as often the case with published technical specs in brochures. Especially when comparisons are made. In amplifiers for instance, two identically powered amps with the same 20hz-20khz +-1db frequency response can still sound different from each other, all other specs equal. That's because one's FR can have some frequencies boosted 1db in much of the same frequencies were the other is attenuated by the same amount. That accounts for 2 db difference which is audible. In which case, only a complete frequency-amplitude plot across the bandwidth will tell the entire story, not just a descriptive technical phrase. But it is rare that amps and speakers sell with a complete FR graph.

 

[djoxygen]

But it is rare that amps and speakers sell with a complete FR graph.

For those interested in an amp and speaker that does, every Mackie HR series active monitor (of which I have 2 pair) comes with its own personal FR graph measured on that exact monitor after it's assembled and before it's packed up and shipped from the factory.

 

[Rip Van Woofer]

In amplifiers for instance, two identically powered amps with the same 20hz-20khz +-1db frequency response can still sound different from each other, all other specs equal. That's because one's FR can have some frequencies boosted 1db in much of the same frequencies were the other is attenuated by the same amount. That accounts for 2 db difference which is audible.

Is this more common than I have thought? I always assumed that aside from the more, uh, eccentric high-end makers that one could safely assume that electronics from mainstream mfrs. were flat, or at least avoided purposeful FR diddling.

[If it is true, please break it to me gently. I would have a hard time performing my duties as a moderator if I were to spend the next several days or weeks curled up in a fetal position next to my equipment rack, whimpering and clutching a tearstained copy of The Audio Critic...]

 

[Rob Babcock]

On a related note, I really don't think that even highly detailed speaker measurements are of much use to the end user in determining how a speaker will sound (at least not the measurements normally used). A good example would be B(l)ose 901's. Measuring them on-axis anechoically won't give you any meanful info since the little "gem" is designed to reflect the majority of sound. This means the true in-room response will be dictated mostly by interaction with the room. The radiation patters of Dipoles or line-arrays also will engage the room in unique and interesting ways.

In the spirit of philosophical question I originally asked, then one would have to permit a thorough & accurate measurement conducted in-room, but naturally the results would be valid only for that room, no other (speaking for the B[l]ose example- although I doubt the results would be very impressive given the particular speaker in question! ).

 

[djoxygen]

On a related note, I really don't think that even highly detailed speaker measurements are of much use to the end user in determining how a speaker will sound (at least not the measurements normally used).

For those who haven't built up a library of experience relating measurements to sound, measurements (and especially FR graphs) can be used as much to confuse and mislead as they can be used to inform and evaluate.

 

[WmAx]

Is this more common than I have thought? I always assumed that aside from the more, uh, eccentric high-end makers that one could safely assume that electronics from mainstream mfrs. were flat, or at least avoided purposeful FR diddling.

I am not awmare of any reason for modern electronics to have deviations in response of that magnitude in the audible passband. Ironically, it seems like some of the most poorly performing products are at the top price shelf(hi-end SET, etc.).

-Chris

 

[djoxygen]

I am not awmare of any reason for modern electronics to have deviations in response of that magnitude in the audible passband. Ironically, it seems like some of the most poorly performing products are at the top price shelf(hi-end SET, etc.).

I think FR is really only of concern for speakers. Any reasonable solid-state receiver, amp (and cable!) should be pretty darn flat from 20 to 20k. Tube gear *might* be a little more variable under some circumstances, but still reasonably flat. With things other than speakers, I'd think the specs we'd be more concerned with are THD, S/N, and such.

 

[WmAx]

SET tube amps usually have very high output impedances(in the single digits!), causing severe frequency response devation at varying loads(manifested in the impedance variatons at the resonant peaks of the bass alignment and the load variaton around the crossover regiion(s). Here is the frequency response(beyond pathetic) of one of the world's most expensive amplifiers(Wavac SH-833):

http://www.stereophile.com/images/archivesart/WAVacFIG01.jpg

Stereophile text: Fig.1 Wavac SH-833, 8 ohm tap, frequency response at 2.83V into (from top to bottom at 80Hz): simulated loudspeaker load, 16 ohms, 8 ohms, 4 ohms, 2 ohms (1dB/vertical div.).

A 'little' more variable?

-Chris

I think FR is really only of concern for speakers. Any reasonable solid-state receiver, amp (and cable!) should be pretty darn flat from 20 to 20k. Tube gear *might* be a little more variable under some circumstances, but still reasonably flat. With things other than speakers, I'd think the specs we'd be more concerned with are THD, S/N, and such.

 

[djoxygen]

SET tube amps... severe frequency response devation at varying loads

A 'little' more variable?

I'd think one could do better than that with a tube if one really tried. Maybe the "warmth" doesn't just come from THD. That +5 at about 80 Hz might beef up the sound a little, eh? >;-)

 

[mtrycrafts]

In amplifiers for instance, two identically powered amps with the same 20hz-20khz +-1db frequency response can still sound different from each other, all other specs equal. That's because one's FR can have some frequencies boosted 1db in much of the same frequencies were the other is attenuated by the same amount. That accounts for 2 db difference which is audible. In which case, only a complete frequency-amplitude plot across the bandwidth will tell the entire story, not just a descriptive technical phrase. But it is rare that amps and speakers sell with a complete FR graph.

I have yet to see an SS amp that has the +/- 1 dB variation at midband or some other frequency other than extremes around 20 Hz and 20kHz. At those locations 2 dB overall variation is well below known JND thresholds of detection. But hey, I have been know to be wrong, from time to time

In the publication of the 1990, 8th International AES Conference, there is a review of 23 DBT amp tests, 13523 trials, one positive outcome when a 10 watt amp was compared with a 400watt amp. 4 others were mechanical problems, broken amps
But, as I was informed elsewhere, we don't know everything in audio and what we can really hear, especially when non hi end audio components are used
But hey, I have been know to be wrong, from time to time

 

[av_phile]

I was talking hypothetically to illustrate that 2 amps offering the same measured specs as indicated can still sound different if for the reasons cited. And unless a more thorough FR plot is made available, will not yield an accurate picture based on a single range + variation spec. My point is, if measured specs would be any basis at all, they should be complete and thorough.

And yes, while it may be difficult to hear 2db variations at either end of the audible spectrum at normal listening levels, they can be heard at loud levels. (3db is supposed to be the threshold of audibility, but you'll be surprised there are people sensitive to as low as 0.5 db volume increments. And that is why some HT receivers have 0.5db increments in their level settings.) I know I did in my younger years building and testing various amp kits. I don't know about those DBT tests you mentioned, but I had heard differences between 3-4 amplifers with close THDs and FRs at different output levels. The differences can be significant especially at close to clipping levels - the point at which most published specs are made. Either that or the published specs were not accurate at all.

 

[WmAx]

I was talking hypothetically to illustrate that 2 amps offering the same measured specs as indicated can still sound different if for the reasons cited. And unless a more thorough FR plot is made available, will not yield an accurate picture based on a single range + variation spec. My point is, if measured specs would be any basis at all, they should be complete and thorough.

I don't think anyone is against thorough measurments. But, audible variatons among modern, proplery engineered equipment due to FR variations should be very rare today.

And yes, while it may be difficult to hear 2db variations at either end of the audible spectrum at normal listening levels, they can be heard at loud levels.

Please clarify.

(3db is supposed to be the threshold of audibility, but you'll be surprised there are people sensitive to as low as 0.5 db volume increments.

Where does this 3dB come from? In careful controlled comparisons[1], 0.5dB is average detectable level of broadband changes with low dynamic range material(such as levels used in standard tv ads) for overall level differences, among subjects. The detectable level is lower for specific broadband test sounds(white noise), averaging 0.2dB. For music with larger dynamic range, 0.5dB JND is less likely to occur.

[1]The Subjective Loudness of Typical Program Material
Soulodre, Lavoie, Norcross
AES Preprint: 5892

-Chris

 

[Unregistered]

+3dB is given as the minimum audible level increase but the audibility of level changes varies quite a bit with frequency. 3dB is thus just a ballbark figure when considering the full frequency range for music.

The ears are less sensitive to level changes in the very low and very high frequencies. While .5dB increases may be perceptible in the midrange, most people will not be able to detect a change that small in lower or higher frequencies. The studies that have shown that increments as small as .2dB were audible use pure tones - definitely not what we listen to when we listen to real music.