For the case that John brought up, you're actually measuring at -15.5dB. But if you've made up your mind in advance (not you specifically, Dick, "one") that you don't like modern technology, then there's always a reason, always a story, always an uninformed excuse not to use it. Fashion and preconception must be served, and mere evidence is not necessary.
I really hope I never get so old that I instinctively reject the best technologies available to me.
I really hope I never get so old that I instinctively reject the best technologies available to me.
The study that everyone is referring to (Fig. 7 is the essence of the “Blauert criterion” adopted by some loudspeaker manufacturers) :
http://forums.klipsch.com/forums/storage/3/1027021/7805blauert.pdf
The official publication of the J. Blauert- P. Laws work:
Group delay distortions in electroacoustical systems | Browse - Journal of the Acoustical Society of America
A good one
http://www.radiolab.com.au/DesignFile/DN004.pdf
Pages 111-124
http://www.ee.iitb.ac.in/student/~bhagwan/communication%20theory/Radio%20Frequency%20and%20Radar%20Systems/Standard%20Handbook%20of%20Audio%20and%20Radio%20Engineering.pdf
There is a lot of paper on internet.
My opinion on GD is that it is one more way of representing phase issues.
As it is (mathematically) the negative of the rate of change of phase
with frequency, it follows that when there is a steep change in phase within a frequency band, one should expect an increased group delay in that band.
For electronic gear testing, I don’t think it is of much help. It is a great visualization help when testing electromechanical transducers.
George
Attachments
Hi Guys
For loudspeakers anyway, I have not found GD to be particularly useful as a tool.
The problem is that GD IS useful for narrow bandwidths, the problem for wide band anything is that for a given operation, filter or whatever, normal operation of it takes say 1000 times longer at 20Hz than at 20Khz.
This leads people to be horrified by a subwoofer with large GD caused by a low corner Frequency, when that same GD is what should be happening.
On the other hand, as the late Dick Heyser correctly identified, one can imagine a full range loudspeaker as having a frequency dependant front to back position which is indicated by the loudspeakers acoustic phase.
The acoustic phase and amplitude response both governing how the loudspeaker reproduces an input wave shape and how it rearranges the harmonics in the signal in time when they radiate.. Only when a loudspeaker can unilaterally (anywhere in front) reproduce the input wave shape is it retaining the signals harmonic structures in time in that frequency range .
The acoustic phase here as Dick saw it is the measured phase with the rotation cause by ALL of the fixed time delays removed. For his system (the first to do this) TDS, one first measured the time delay with an “ETC” measurement.
That is a VERY clever rick I always thought, here is how it works. You sweep with a linear sweep of X Hz per second. That signal feeds the speaker but also one side of a multiplier (modulator or mixer in RF). The mic signal is delayed because of the path lengths and delays so what arrives at the mic is X Hz behind the sweep F. The mic signal feeds the other port on the mixer and the mixer output is the Sum and Difference frequency. The sum is filters out and the Difference fed into an FFT who’s horizontal scale was changed from frequency to distance or time. Now, what appears is the energy envelope VS time, one picks the highest point (which is a high frequency being linearly weighted) which is the “time = Zero” reference point for the TDS measurement.
While TDS (the TEF machine) is near an antique process now, it actually has a couple VERY strong features like having much more noise immunity and immunity to non-linearity than the sequence systems , to get a robust measure that way, requires a number of averages and the low frequency limit is often noise contaminated at only -20 or -30 dB down (free lf extension it looks like). Back then, like the Beta tape format, TDS was better but narrowly licensed while VHS was widely licensed.
A TEF class I went to in the ancient days of the TEF 10, they had 12 TEF machines, the outputs mixed together to an amp, driving one loudspeaker and one mic split into 12 outputs with multiple measurements running simultaneously w/o interference..
Best,
Tom Danley
Danley Sound Labs
For loudspeakers anyway, I have not found GD to be particularly useful as a tool.
The problem is that GD IS useful for narrow bandwidths, the problem for wide band anything is that for a given operation, filter or whatever, normal operation of it takes say 1000 times longer at 20Hz than at 20Khz.
This leads people to be horrified by a subwoofer with large GD caused by a low corner Frequency, when that same GD is what should be happening.
On the other hand, as the late Dick Heyser correctly identified, one can imagine a full range loudspeaker as having a frequency dependant front to back position which is indicated by the loudspeakers acoustic phase.
The acoustic phase and amplitude response both governing how the loudspeaker reproduces an input wave shape and how it rearranges the harmonics in the signal in time when they radiate.. Only when a loudspeaker can unilaterally (anywhere in front) reproduce the input wave shape is it retaining the signals harmonic structures in time in that frequency range .
The acoustic phase here as Dick saw it is the measured phase with the rotation cause by ALL of the fixed time delays removed. For his system (the first to do this) TDS, one first measured the time delay with an “ETC” measurement.
That is a VERY clever rick I always thought, here is how it works. You sweep with a linear sweep of X Hz per second. That signal feeds the speaker but also one side of a multiplier (modulator or mixer in RF). The mic signal is delayed because of the path lengths and delays so what arrives at the mic is X Hz behind the sweep F. The mic signal feeds the other port on the mixer and the mixer output is the Sum and Difference frequency. The sum is filters out and the Difference fed into an FFT who’s horizontal scale was changed from frequency to distance or time. Now, what appears is the energy envelope VS time, one picks the highest point (which is a high frequency being linearly weighted) which is the “time = Zero” reference point for the TDS measurement.
While TDS (the TEF machine) is near an antique process now, it actually has a couple VERY strong features like having much more noise immunity and immunity to non-linearity than the sequence systems , to get a robust measure that way, requires a number of averages and the low frequency limit is often noise contaminated at only -20 or -30 dB down (free lf extension it looks like). Back then, like the Beta tape format, TDS was better but narrowly licensed while VHS was widely licensed.
A TEF class I went to in the ancient days of the TEF 10, they had 12 TEF machines, the outputs mixed together to an amp, driving one loudspeaker and one mic split into 12 outputs with multiple measurements running simultaneously w/o interference..
Best,
Tom Danley
Danley Sound Labs
Or, as a guy working for a (very successful) PC game company once said: Innovation often comes from perfect execution.
Thanks for this George. I'm surprised to find that I was carrying out similar experiments around the same time Is Linear Phase Worthwhile? though with a rather different slant. My conclusions were very similar.
However, since then, there has been more research on the audibility of GD distortion. Blauert's 0.4 ms criteria still stands but there have been a lot more Listening Tests, especially on LF phase distortion. I would very much to be able to conduct my own tests on audibility of LF phase response and I suspect so does Tom Danley.
Because we now have the technology to correct it, it seems sensible to re-visit LF phase distortion. But these involve GD distortions of MUCH greater than Blauert's 0.4 ms.
I'll also point out that it is simple to devise test signals that certain people (the true golden pinnae) can pick out as distorted by simple all-pass networks but these signals NEVER appear in speech & music.
Last edited:
I can imagine many useful EQ circuits that have GD issues. They could include those used in room EQ of various types, Bass roll off due to cap coupled and transformer coupled amps (are they still in use?), in multi-stage tube circuitry... RIAA 'rumble' filters. To name just a few flys in the ointment. Now with direct coupled ss circuits the low end GD is not there to such as extent as it was. But we still use EQ heavily in recording side (and in homes).... there is a hidden issue when the THD tests show everything is just super duper. Another variable in one persons evaluation and anothers'. Maybe why it is so rare to find well recorded bass sound IMO.
Thx - RNMarsh
Last edited:
Sorry - I wasnt talking about volume controls specifically but ADC/DAC mostly ---> To test them at all input levels and then compare. -RNM
Last edited:
2SY
I was very "happy" , when WM and PGA become available, datasheets looks very "nice" and I can substitute in more elegant way MDAC.. But real experience pushed me back to MDAC, and than to other, better , noncompromise solutions. It has nothing to do with "instinctively rejection.. I was a lot dissapointed. Results I showed here are not results of bad (incorrect) aplication, this is very "primitive" task . No one presented (or I can not find anything..)better , measured results for realistic conditions - gain different (lower) than 0dB, about -20-30dB, what is the most used region in real use.
According Levinson preamp, how do You know, they used PGA? Are You sure ? In ML383 it was MDAC, and meaurements results suportted this. And where see You information about measurement conditions (actual gain setting)?
What was the previous stage and the following one ? No cap as a charge, like a cable ? No Caps in the signal path ? Separate grounds for command and signal ? Symmetrical or not ?
I bought a kit to evaluate, and i had to modify-it on several points to get it nice.
From Stereophile (ClasséCP-800 D:A preamplifier):
The volume control is implemented with two two-channel Burr-Brown PGA2310 programmable-gain chips, one per channel used as a differential volume control.
Tonally, the CP-800 fell into the camp of the clean and clear rather than the mellow and euphonically colored. In that respect it was somewhat similar to the Parasound Halo JC 2 ($4000), which I reviewed in March 2008: A wealth of recorded detail was laid bare without being spotlit. In level-matched comparisons with the Ayre Acoustics K-5xeMP ($3500; I reviewed it in June 2011), the Ayre sounded slightly veiled, though there was a robustness to its soundstaging that resulted in more fully fleshed-out images within that soundstage.
Last edited:
He knows what he is doing, we have known each other for 10 years. No stupid mistakes. And remember that he measures distortion in front of and behind PGA and values differ in 2 orders (behind PGA it is 100x higher).
And please do not quote audiophile magazines impressions, they are useless.
And please do not quote audiophile magazines impressions, they are useless.
Last edited:
, Classé CP-800 D/A preamplifier Measurements | Stereophile.com , pictures 6 to 9..and 14As i wrote, previous stage was buffer (OA with gain1) and series resistor 33R, distortion at this pont was 2 orders lower than at output. Output was also buffered with OA (LM4562) distortion figures at input/output the same..Grounds separated, conected according recomendations in one point (a "must" condition). Unbalanced in/out conection.
Last edited:
I was amused by the reference to JC2. This said, their measurements (note the levels)
Mark Levinson No.38S preamplifier Measurements | Stereophile.com
Mark Levinson No.38S preamplifier Measurements | Stereophile.com measurements are not at 0dB of the volume level.
About measured distortions of the Classé:
"Although the level of distortion with a full-scale digital input signal was low in absolute terms, it was higher than I expected, the third harmonic lying at –77dB, or 0.014% (fig.6). Reducing the output level by 10dB with the volume control preserved the level of the third and other harmonics (fig.7), whereas reducing the level of the signal data by the same 10dB dropped the third harmonic to –87dB, or 0.005% (fig.8). This suggests that the slight nonlinearity occurs in the D/A conversion circuitry ahead of the volume control.,"
Note that they made measurements at various volume control positions, and the distortion DECREASE with attenuation and levels (as expected and opposite to your remarks).
[edit] Not so sure about their analysis, may-be the switches are the responsible, but at least HD ratio does not increase.
Really, there is something strange in this story: two people, knowing each other, and feeling the same (BAD) about those volume controls, while so many people are satisfied with them, and prestigious manufacturers use them in some of their high end products.
Kind of -120dB distortion syndrome ?
Of course, we all agree a passive stepped attenuator is better, and, as i'm not paid to promote any Ci, i will stop here.
About measured distortions of the Classé:
"Although the level of distortion with a full-scale digital input signal was low in absolute terms, it was higher than I expected, the third harmonic lying at –77dB, or 0.014% (fig.6). Reducing the output level by 10dB with the volume control preserved the level of the third and other harmonics (fig.7), whereas reducing the level of the signal data by the same 10dB dropped the third harmonic to –87dB, or 0.005% (fig.8). This suggests that the slight nonlinearity occurs in the D/A conversion circuitry ahead of the volume control.,"
Note that they made measurements at various volume control positions, and the distortion DECREASE with attenuation and levels (as expected and opposite to your remarks).
[edit] Not so sure about their analysis, may-be the switches are the responsible, but at least HD ratio does not increase.
Really, there is something strange in this story: two people, knowing each other, and feeling the same (BAD) about those volume controls, while so many people are satisfied with them, and prestigious manufacturers use them in some of their high end products.
Kind of -120dB distortion syndrome ?
Of course, we all agree a passive stepped attenuator is better, and, as i'm not paid to promote any Ci, i will stop here.
Last edited:
- Status
- Not open for further replies.