I would bet as much speaker related as the electronics ..... The problem with studio monitoring is the poor speakers used in monitoring, where most changes are not noticiable until a fair amount of lift is used.
.
And the third- when archiving the sound (the goal of some types of recordings), fidelity is paramount. That's an aspect of recording that many producers/engineers don't get.
I am not a critic of persons, just critical towards ideas that don't fit into my conceptual framework and/or hands on experience with the matter at hand. Since I am here to learn, I am more than prepared to modify my ideas, given proof of my errors in thinking. In this instance, this has yet to occur.
There are a number of reasons for that. The first is my observation, mentioned before, that the supply rails of an amplifier under load fed by a linear ps look much nastier to me than the same fed by a SMPS. The second are the size and composition of the nasties; with a linear ps the deviations are larger and more inside the audio band than with a SMPS, where they typically are way above the audio band and relatively small. This makes filtering them out not that hard. In short, with a well regulated and well designed SMPS, less work will be required from the feedback loop to reject power supply irregularities.
vac
You might be right in the case of a stepped sine sweep, but it should show up in MLS, since the impulse to excite the driver contains all frequencies without discontinuities.
My take on sibilance is that diffraction might be a primary cause. The FR at the top end may flip flop around a great deal when a loudspeakers is measured at different angles off-axis, even taking very small angle steps, diffraction being the cause of this. Different wavefronts alternatively combining constructively and destructively will lead to much higher local peaks than any mechanical form of resonance in a driver can cause.
The ribons as mentioned by Thorsten and DBMandrake may suffer a bit less from this as long as the listener stays within the beam.
vac
Is it a valid concept to look not directly at the rails, but at the amp output in evaluating the effect of the supply, be it SMPS or linear? This looks at it as a system for the case at hand, not as a general rule. One amp may be far more tolerant of issues from one or the other so there may not be a simple answer for which is best. "it depends"
Its only the amp output which connects to the speakers (apart from any induction between cables).
There is more to a PSU than output voltage - output impedance will vary with frequency too. A regulated supply, including SMPS, is likely to have a rising impedance with frequency because of the need to stabilise the feedback loop. A conventional unregulated supply will have a mainly falling impedance.
There is more to a PSU than output voltage - output impedance will vary with frequency too. A regulated supply, including SMPS, is likely to have a rising impedance with frequency because of the need to stabilise the feedback loop. A conventional unregulated supply will have a mainly falling impedance.
sure, and you must have enough bins in the subsequent fft to analyze the response, but my point is that even narrow peaks will show up, given the right methodology.
Although I agree that diffraction at high frequencies is bad (quite bad in fact) and can be a source of "harshness", I'm not sure that I would pin it as the sole culprit for sibilance.
What you're saying implies that it's the perturbations in the frequency response caused by diffraction that are to blame - yet EQ, even accurate narrow band EQ doesn't seem to eliminate it on drivers which have it.
My take is that it's a time domain phenomenon, whether that be one or more high Q resonances that ring for a long time, or perhaps the time delayed signal from a diffractive edge - both stretch out the original signal in time.
Significant diffraction which is centered in the sibilance range could indeed be a contributor, but I think not the only culprit.
With the ribbon you have both a very clean CSD, as well as greatly reduced diffraction effects, assuming its a wave-guide type with directivity control. Whatever the reason, sibilance isn't an issue, even if the treble is turned up.
I take it you have not built many speakers with cheap metal cones like the Daytons. I can assure you, they can cause breakup issues that make any diffraction or complementary interference totally trivial. Run, do not walk to the nearest exit. Of course, a "proper" crossover will keep you out of those troubles and they are not bad mid price drivers. By proper, that means low and steep. I would agree to "significant contributor"
I am in no way discounting interference issues as we are very sensitive to peaks and comb effects that should not be there, or the need for good room management. I agree with you there completely. Just it ain't the whole picture. I suspect it is not discussed as much outside the sub forum as the resolution of most analyzers may not pick it up. I still do pure tone sweeps and walk around in a room to augment all the fancy instrumented tests. The strangest things can show up.
What to have some fun? Put in about a 12K tone in one speaker and walk around the room. Not only does it go in and out, you ability to localize it changes. Try it.
I fully agree with most what you say, there is not only one possible cause for this problem and some systems may suffer from a combination of ills.
I just want to use this occasion to set out that the kind of refraction I am referring to cannot be EQ-ed out, unlike in the case of BSC on the low end. On the high end, refraction can lead to a much more complicated interference pattern, which cannot be compensated electronically. The reason is that it is very local (as verified by measuring speaker response at small off axis intervals). In speakers that suffer from this problem, even minute off axis shifts of the measuring mic may cause the peaks and valleys in the FR to shift around considerably. That's probably also the reason why manufacturers of speakers tend to average the response over a number degrees off axis, this in order to optically flatten out this part of the publishable FR.
Anyways, it is my experience that there is a correlation between how smooth this interference pattern measures, and the absence of sibilance problems.
vac
Yup. That's the kind I was talking about. Hence the suggestion about walking around and room treatments. It is an environmental problem, so it must be dealt with environmentally. What it is that bothers my wife's hearing is not this simple as without changing anything in the room, switching amps has a repeatable effect. This is why I suggest contributory, not causal.
Speaker manufactures average plots as any measurement not in your room is basically irrelevant and the number of folks who know what a CSD is is not that high. What else are they to do?
I may note, she has not liked any ribbon speaker but did like the ML Sequel. The quintessential do-no-wrong Vanderstein also passed muster. I am not sure we have ever heard a good ribbon. I did not care for the Apogee. That was the last one I heard.
Speaker manufactures average plots as any measurement not in your room is basically irrelevant and the number of folks who know what a CSD is is not that high. What else are they to do?
I may note, she has not liked any ribbon speaker but did like the ML Sequel. The quintessential do-no-wrong Vanderstein also passed muster. I am not sure we have ever heard a good ribbon. I did not care for the Apogee. That was the last one I heard.
ML sequel ..... 
Why would speaker measurements be irrelevant , actually they tell the tale and if you set any speaker up and listen in the nearfield you can eliminate enuff of the room acoustics to see if it is the speaker or the room .
Better yet take it outside....
I don't believe Sibilance is just from refraction, it is an combination of multiple error's , refraction and driver resonance (frequency deviation) this could be from the driver it self or brought on by poor matching x-over design.
This abberation may be visable in the imp mag/phase and different amplifiers will react to this differently, diffraction will play an lesser role, it's effects are more that of smearing...
Best to voice drivers in at 1 watt input no xover , there naturnal positives and negatives will show here...
.
Why would speaker measurements be irrelevant , actually they tell the tale and if you set any speaker up and listen in the nearfield you can eliminate enuff of the room acoustics to see if it is the speaker or the room .
Better yet take it outside....
I don't believe Sibilance is just from refraction, it is an combination of multiple error's , refraction and driver resonance (frequency deviation) this could be from the driver it self or brought on by poor matching x-over design.
This abberation may be visable in the imp mag/phase and different amplifiers will react to this differently, diffraction will play an lesser role, it's effects are more that of smearing...
Best to voice drivers in at 1 watt input no xover , there naturnal positives and negatives will show here...
.
Last edited:
I use different (switching) power supplies with T-Amps etc.
(SPSs preferable with big transformers instead 'transformerless' ).
Compared to linear PSUs they produce a tighter bass,
probably because of their lower output impdance at low frequencies.
From my little knowledge, the new(est), ultra low loss
switching regulators strongly mechanically resonates (sounds like 1kHz) and exactly this 'switching tone' can also be heared out of the speakers. I can't easyly remove it with filtering because it's probably ground induced.
You are most probably right that there could be a grounding problem, which sometimes happens when both the SPMS and the audio source are grounded independantly. When the SPMS is double insulated, the ground is not necessary for electric safety, but is there to reduce RF emission. You could try what happens if you ground the SPMS through the audio source (remove ground pin from SPMS). Most likely the problem will go away.
vac
No argument there. Diffraction at treble frequencies is a serious and under appreciated problem, and probably is one reason why many manufacturers average their published responses over a +/- 15 degree window.
At the very least its going to cause small/unstable sweet spot (significant amplitude imbalances in the left-right treble at critical spot frequencies as you move your head location just a few inches) as well as harshness due to comb filtering peaks and valleys, which as you point out will shift all around the place as the listener moves, so there is no way it could be EQ'ed. (Even if you EQ'ed it for one exact location you're then trying to apply a frequency domain fix for a time domain problem anyway)
The problem with this averaging over a small angular window is that it only reveals the average spatial tonal balance and hides the effects of diffraction which we will still be hearing - after all our ears are each only at one location at a time, so a spatially averaged measurement doesn't represent what we would hear.
Likewise, attempts to "fix" diffraction through asymmetrical driver placement on the baffle are misguided - the response might be flatter on axis but there will still be large amounts of diffraction present, just redistributed slightly in time. Asymmetrical placement perhaps makes some sense at low frequencies (1Khz and below in the baffle step region) because the wavelengths involved are long, thus the diffraction "ripples" are spread out over a very wide angular range (thus fairly benign) but it makes no sense at treble frequencies where wavelengths are so short and the response ripples vary so much for such small angular changes.
This is one reason I like wave-guide loaded tweeters so much. As well as other advantages there is a large reduction in treble baffle diffraction for basically no extra work, (even on a plain flat rectangular baffle) and yes I think that reduction in treble diffraction from the cabinet really does make quite a difference to imaging and perceived smoothness. (smoothness being the opposite of harshness)
Last edited:
Best to voice drivers in at 1 watt input no xover , there naturnal positives and negatives will show here...
I am not following you here. 90% of what I would call vlocing a speaker and is tweaking the crossover in a multi-driver system where you have not just environment issues, but direct driver interference. (Unless of course, you are in the Louther camp) I do start at one watt. Or are you talking about room tuning and don't want the extra variables of multi-driver interference confusing the results? I have not had any luck going outside as I voice my speakers for the exact room where I am using them. A luxury OEM's do not have. Advantage DIY.
I am not following you here. 90% of what I would call vlocing a speaker and is tweaking the crossover in a multi-driver system where you have not just environment issues, but direct driver interference. (Unless of course, you are in the Louther camp) I do start at one watt. Or are you talking about room tuning and don't want the extra variables of multi-driver interference confusing the results? I have not had any luck going outside as I voice my speakers for the exact room where I am using them. A luxury OEM's do not have. Advantage DIY.
No S..it Sherlock - the return of the hissing sibilants...
Folks,
Today a little parcel arrived for me. It contained a DAC that I helped to develop (final production version).
This allowed me some cute experiments. This DAC actually features two seperate DAC Chips (one "hybrid - read multibit/delta-sigma combo" modern Chip and an antique 16 Bit direct conversion DAC).
The "direct conversion" DAC Chip has options for being fed oversampled and upsampled signals plus modes with no digital filtering and only some analog filtering (including one that corrects the 3dB rolloff at 20Khz from not using digital filters).
The Modern DAC does not allow filter-less operation (rather, it does actually, it just sounds awful that way) but has been given some interesting digital filters in addition to the traditional "Brickwall" filter, including minimum phase filters, sub-halfband (apodizing) filters and soft rolloff filters. Actually the filters that where retained for the production version are all "combinations" of at least two of the above characteristics.
So I did some listening using music that has a tendency to show this issue off, the recording was the eponymous first album from Italian duo Musica Nuda. It has minimal if any processing (certainly no de-essing) but is very closely miked for the vocals.
Using my personal preference (no oversampling or upsampling, 16 Bit DAC, analogue filter for flat high frequency response) this recording does not "hiss" in any unnatural way, but the small details such minimal lipsmacking by the singer and "teeth sounds" was very clear and distinct (I would have likely used more acoustic filtering and distance for the singer to avoid capturing these noises).
Switching off the high frequency correction made the sibilants sound subdues and a little unclear.
Switching to the modern DAC immediately made the sibilants hiss.
Among the filters available the brickwall filter caused most hissiness and spittiness. Second worst was paradoxically the filter that combined minimum phase, soft rolloff and sub-halfband filtering.
Least offensive was the filter that is minimum phase and sub-halfband which I prefer slightly to straight Non-Os (without HF rolloff correction), with minimum phase & soft rolloff coming in third.
I did not really listen to any other aspects of the sound quality, only to hissy sibilants, for this my ranking is:
1) Non-OS + HF-rolloff corrected (< -1dB @ 20KHz)
2) Oversampled DS with minimum phase sub-halfband filter
3) Non-Os without HF rolloff correction (> -3dB @ 20KHz)
4) Oversampled DS with minimum phase soft rolloff filter
5) Oversampled DS with minimum phase sub-halfband soft rolloff filter
6) Oversampled DS with brickwall filter
This may give some datapoints.
The best visual pulse/step response came in third, the second choice shows a LOT of HF ringing, but only post ringing, no pre-ringing, the first choice has a single cycle of post-overshot in the response. The worst choice has massive pre- and post- ringing...
Ciao T
Folks,
Today a little parcel arrived for me. It contained a DAC that I helped to develop (final production version).
This allowed me some cute experiments. This DAC actually features two seperate DAC Chips (one "hybrid - read multibit/delta-sigma combo" modern Chip and an antique 16 Bit direct conversion DAC).
The "direct conversion" DAC Chip has options for being fed oversampled and upsampled signals plus modes with no digital filtering and only some analog filtering (including one that corrects the 3dB rolloff at 20Khz from not using digital filters).
The Modern DAC does not allow filter-less operation (rather, it does actually, it just sounds awful that way) but has been given some interesting digital filters in addition to the traditional "Brickwall" filter, including minimum phase filters, sub-halfband (apodizing) filters and soft rolloff filters. Actually the filters that where retained for the production version are all "combinations" of at least two of the above characteristics.
So I did some listening using music that has a tendency to show this issue off, the recording was the eponymous first album from Italian duo Musica Nuda. It has minimal if any processing (certainly no de-essing) but is very closely miked for the vocals.
Using my personal preference (no oversampling or upsampling, 16 Bit DAC, analogue filter for flat high frequency response) this recording does not "hiss" in any unnatural way, but the small details such minimal lipsmacking by the singer and "teeth sounds" was very clear and distinct (I would have likely used more acoustic filtering and distance for the singer to avoid capturing these noises).
Switching off the high frequency correction made the sibilants sound subdues and a little unclear.
Switching to the modern DAC immediately made the sibilants hiss.
Among the filters available the brickwall filter caused most hissiness and spittiness. Second worst was paradoxically the filter that combined minimum phase, soft rolloff and sub-halfband filtering.
Least offensive was the filter that is minimum phase and sub-halfband which I prefer slightly to straight Non-Os (without HF rolloff correction), with minimum phase & soft rolloff coming in third.
I did not really listen to any other aspects of the sound quality, only to hissy sibilants, for this my ranking is:
1) Non-OS + HF-rolloff corrected (< -1dB @ 20KHz)
2) Oversampled DS with minimum phase sub-halfband filter
3) Non-Os without HF rolloff correction (> -3dB @ 20KHz)
4) Oversampled DS with minimum phase soft rolloff filter
5) Oversampled DS with minimum phase sub-halfband soft rolloff filter
6) Oversampled DS with brickwall filter
This may give some datapoints.
The best visual pulse/step response came in third, the second choice shows a LOT of HF ringing, but only post ringing, no pre-ringing, the first choice has a single cycle of post-overshot in the response. The worst choice has massive pre- and post- ringing...
Ciao T
Simon, yea, wave guides are one of the next things I want to play with, but for another advantage as the efficiency allows even greater padding . It does separate the driver centers more and the transition can be very problematic and construction gets a lot harder. Ever try and carve a full flair out of MDF?
I was surprised how big a difference a prototype made when I went from 3/8 inch radius on the box corners to 3/4 radius with a bit of additional sanding. Makes for a bigger baffle though; a problem I am dealing with doing a narrow tower for my Nephew.
I have to disagree the asymmetrical placement is not useful. Over-rated I will grant you. I made two otherwise identical speakers. Sharp corner boxes. In one, the tweeter was equal distant from three edges. The other followed the 3-5-7 fraction. It measured smoother. I only measure at 0, 22, and 45 degrees. I then took my big router and radiused the corners. This improved both, more on the symmetrical one. So, it does help, but it is just one of many factors. One method distributes the problem, the other reduces it in the first place. Neither cure it. So I try to do both if I have the space. Direct driver interference is still dominant.
Just so others may not get confused, we are combining a lot of different but related things. Edge diffraction, room reflections and far more important driver to driver interference all play a part in the constructive/destructive interference that varies with location, axis, and everything in the environment. We need to address all of them. Of course, the environment does these things to live sound as well. It is when our brain gets confused that the patterns do not make sense to what ambient sounds behave like we find it objectionable. My theory.
I was surprised how big a difference a prototype made when I went from 3/8 inch radius on the box corners to 3/4 radius with a bit of additional sanding. Makes for a bigger baffle though; a problem I am dealing with doing a narrow tower for my Nephew.
I have to disagree the asymmetrical placement is not useful. Over-rated I will grant you. I made two otherwise identical speakers. Sharp corner boxes. In one, the tweeter was equal distant from three edges. The other followed the 3-5-7 fraction. It measured smoother. I only measure at 0, 22, and 45 degrees. I then took my big router and radiused the corners. This improved both, more on the symmetrical one. So, it does help, but it is just one of many factors. One method distributes the problem, the other reduces it in the first place. Neither cure it. So I try to do both if I have the space. Direct driver interference is still dominant.
Just so others may not get confused, we are combining a lot of different but related things. Edge diffraction, room reflections and far more important driver to driver interference all play a part in the constructive/destructive interference that varies with location, axis, and everything in the environment. We need to address all of them. Of course, the environment does these things to live sound as well. It is when our brain gets confused that the patterns do not make sense to what ambient sounds behave like we find it objectionable. My theory.
- Status
- Not open for further replies.