yes, I did stae before somewhere that phase shift induced by HP filtering at the input could cause audible distortion.
Hmm. At 100Hz, I calculate phase shift as arctan(6/100) = 3.4 degrees. Doesn't seem like "loads." Am I missing something?
Suppose you have a 10% variation between channels, with one channel having a 6Hz zero and the other channel at 6.6Hz. Phase shift at 100Hz for the first channel is 3.4 degrees, for the second channel, 3.7 degrees. That's a 0.3 degree difference. At 100Hz, this represents about 2.5cm displacement. Doesn't seem like very much?
Not much, but the phase difference (although perhaps not the equivalent displacement) will be larger at lower frequencies. My point is that we are not very sensitive to phase, but quite sensitive to phase differences. Hence if a subsonic HP filter is audible then interchannel phase difference might be the cause as it could result in image position being frequency-dependent. However, it would seem to fade into insignificance when compared with typical loudspeaker and room responses.
The next question then might be: is it possible that the HP filter is removing subsonics and near-sonics which might otherwise generate IM in the amplifier or speaker? If so, it could be another example of where people prefer their sound with extra distortion.
The next question then might be: is it possible that the HP filter is removing subsonics and near-sonics which might otherwise generate IM in the amplifier or speaker? If so, it could be another example of where people prefer their sound with extra distortion.
When thinking about power supplies everyone concentrates on 50/60/100/120Hz, but some forget about the subsonic variations in supply voltage. People who suffer from faradaphobia and want DC coupling everywhere, or who are frightened of subsonic filters, are likely to find that they need stabilised supplies. Otherwise their amp and speaker may be forced to process strong sub-Hz signals along with the music. Normal engineering just includes coupling caps and no longer needs stabilised rails.
Picture shows massive sag in waveform that is direct result of changing phase, aka, smearing.
My system does a pretty good approximation of fo below 10Hz.
First trick is getting speaker to play realistic square wave at 40Hz. Speaker driver does a lot more damage to signal smear wise than a humble Butterworth filter. With a little DSP:
Top track is recording of system compensated to get recognizable 40Hz square wave. Turned up it makes the windows flex.
Bottom track is unassisted system attempting 40Hz square wave, in comparison it is more of a buzzing noise that doesn't move much air.
From here the humble phase smearing Butterworth is applied to compensated system. With same settings as before, bass is not as intense, and timbrel balance has got more buzzing to it.
The air just doesn't move the same way.
my subscription expired and it looks like the paper is not freely available anywhere else. what issue does it address exactly?
Crowhurst, N. H., 1957 Some Defects in Amplifier Performance Not Covered by Standard
My favourite bitp195
Some, of course, have discovered for themselves that the audible performance does not appear to be related to the figures on the specification and have concluded that "specifications are valueless---the only reliable test is to listen to it."
From my notes:
Crowhurst, N. H., 1957 Some Defects in Amplifier Performance Not Covered by Standard Specifications JAES 5(4) p195-202
a) provides a model for clipping distorion "creates new sounds", especially driver stage overload with feedback
b) demonstated models with vastly different IMD/THD numbers
c) noted the problem of nested feedback: 2nd & 3rd order distortion plus two nested feedbackloops-> 81st order products
d) additional HF problems of mismatch/non-linearity -> "program modulated, high frequency "noise""
e) how to fake square wave response
f) lovely example of clipping induced bias offset
Linkwitz under his Frontiers in Speaker Design point G provides additional guidance (and further reading, which I've still to do) on useful testing methods
My favourite bitp195
Some, of course, have discovered for themselves that the audible performance does not appear to be related to the figures on the specification and have concluded that "specifications are valueless---the only reliable test is to listen to it."
From my notes:
Crowhurst, N. H., 1957 Some Defects in Amplifier Performance Not Covered by Standard Specifications JAES 5(4) p195-202
a) provides a model for clipping distorion "creates new sounds", especially driver stage overload with feedback
b) demonstated models with vastly different IMD/THD numbers
c) noted the problem of nested feedback: 2nd & 3rd order distortion plus two nested feedbackloops-> 81st order products
d) additional HF problems of mismatch/non-linearity -> "program modulated, high frequency "noise""
e) how to fake square wave response
f) lovely example of clipping induced bias offset
Linkwitz under his Frontiers in Speaker Design point G provides additional guidance (and further reading, which I've still to do) on useful testing methods
Last edited:
If you want to go looking for evidence Linkwitz in Frontiers F discusses his concerns on the relationship between phase shift & group delay. And at what point they will become audible.
found it after all on some Italian forum that had a dead link and then dug it up some more... well, nevermind. gave it a quick read and I wouldn't say that it's exactly revolutionary info.
for instance, the part about a complex nonlinearity that coincidentally nulls some distortion products is too far fetched, that can't happen in real life.
response into reactive loads, sure, it can say a thing or two.
concerning clipping behaviour (sudden or gradual onset and recovery) I guess it does make a difference and I suspect that, just as the author says, explains the "driving ability" of some amps. obviously, it applies when comparing underpowered amps.
for instance, the part about a complex nonlinearity that coincidentally nulls some distortion products is too far fetched, that can't happen in real life.
response into reactive loads, sure, it can say a thing or two.
concerning clipping behaviour (sudden or gradual onset and recovery) I guess it does make a difference and I suspect that, just as the author says, explains the "driving ability" of some amps. obviously, it applies when comparing underpowered amps.
I'm quite familiar with this. But Dr. Linkwitz uses more conventional definitions for things like "transients," and understands the difference between "group delay" and "phase delay," so I have no trouble understanding what he's writing.
Let's put the blinders back on. Phase wrecking capacity of amplifier/speaker system is very well understood, and my pic is great example. That aside, thread describes better sound attribute of amp1 (mains noise amp) as better bass. Initial posts suggest this is happening below 250Hz.
Video capture of sweeps with resistive load, reveals general linearity of both amplifiers.
Posting of wave files (or even decent MP3), including stimulus would be helpful. Log sweep starting from 5Hz or less would be nice. Cool Edit parameters used to generate sweep would be nice too.
Comparison sweep of amplifiers loaded with speaker resulting in perceivable bass differences would then be most helpful.
Video capture of sweeps with resistive load, reveals general linearity of both amplifiers.
Posting of wave files (or even decent MP3), including stimulus would be helpful. Log sweep starting from 5Hz or less would be nice. Cool Edit parameters used to generate sweep would be nice too.
Comparison sweep of amplifiers loaded with speaker resulting in perceivable bass differences would then be most helpful.
Skeptical? Good.
Above are two wave forms with identical spectra. Basis is 40Hz square wave. Both tracks have Butterworth 6Hz 1st order high pass; simulating preamp/amplifier behavior. Both waveforms have everything above 5th harmonic removed by FFT. The woofer used in recordings is Peerless XLS 830500 in about 80L enclosure. For simulation purpose I've applied Butterworth 30Hz 2nd order high pass filter to get basic roll off behavior of woofer setup. Result is reduction of 40Hz fundamental. Since amplitudes of harmonics is virtually undisturbed, compensation in top track using FFT was used to reduce 40Hz fundamental to match level in lower track.
The synthesized waveforms compare most favorably with recorded waveforms. In the recordings, spectra do differ, thus the synthesis.
The equally normalized waveforms put peaks in lower track at 0dB; upper track peaks at -4dB.
Looks can be deceiving.
Playing back the waveforms using cheap, but nicely sealing ear buds reproduces perceptions extremely similar to what I get listening to the woofer under the two conditions.
No, damage to signal by good electronics is trivial relative to effects of transducer systems.
This is an excellent demonstration of when phase is audible. Impact exists for all source with requisite bandwidth, but is highly dependent on content, and how much phase whacking has occurred. Acoustic bass, piano, cello, deep voices, all sound more realistic to me with fully corrected system.
I've rendered above as .mp3, zipped, and attached.
The song remains the same, the spectra remain the same, but the air moves differently.
So are they the same spectrum or not? Your first sentence say they are. The rest appears to say that they are similar but not identical. 'Virtually undisturbed' means disturbed.Barleywater said:
To get two signals with the same spectrum but different phases you need to prepare them using two inverse FFT with identical amplitudes but different phases. Any other method could result in amplitude differences.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.