|
I've been wanting to hear for myself if the "Railsback" adjustments make a difference. I finally took some time to make two sound files, one which uses pure equal temperament and the other which incorporates "railsback" adjustments, using the chart linked earlier that came from a Yamaha digital piano manual.
Folloing are the two links. The timbre used is a pretty common variant of a DX7 string synth patch. It's being played by a Java-based FM synth that I wrote.
6 seconds of a low drone: A1 (midi 33, 55 Hz)
6 second drone of a high drone added, A6 (midi 93, 1760 Hz)
then an arpeggio that spans A2 to A5 comes in.
With the "railsback" version, the pitch adjustments range from -8 cents for the low A1 to +9 cents for the high A6.
The difference between the two tracks is pretty subtle! I can hear it most readily when running both files at the same time, but flipping (solo/mute) back and forth between them quickly. I find it's easiest to distinguish the pitch difference between the high drones. It does seem to me that the "railsback" version sits a little easier, somehow. But I wonder if this high string synth would sound still better if it were yet a few cents sharper.
There's some detuning built into the patch, but the Op stack which predominates is a pure 1 : 1 : 4 (where 1 is the carrier and 1 and 4 are modulators stacked in series). The other stack is a "feedback" loop with the pitch set via a 0.998 factor (i.e., for 440 Hz, the detuned stack is 439.12 Hz, and in that region, 0.247 Hz is 1 cent). I think its overtones phase with the main stack overtones more than they directly affect the perceived pitch. Maybe this stack does have an impact on the sense of the octaves. But I'm not at all clear how, especially since the exact same patch is being used for all the notes being played.
I think the internal phasing of the timbre that results would tend to make clashes between the overtones and fundamentals of concurrent notes less prominent, as there is a baseline of considerable timbral activity regardless of the exact pitches being played.
So, is it worth the trouble to correct for the octave stretch? IDK. I think it depends on the context. The producer (and the musicians involved) need to use their best judgment. Sometimes if a note is slightly "off" or "pitchy" it actually works better. A subliminally sharp note might give it a bit of edge or assertiveness, a subliminally flat note might add a bit of despondency or passivity that fits the moment.
|
Hello Phil,
Thanks for the report. The two sound files are slightly different in terms of amplitude and headroom. The?railsback_stringsynthdrone is slightly louder than the equal temperament one. This is evident when looking at the waveforms. Could you re-create these files by rendering them using the same MIDI file for a better comparison.?
Thanks in advance, E? ?
toggle quoted message
Show quoted text
On Nov 1, 2021, at 10:58 AM, philfrei via < philfrei@...> wrote:
I've been wanting to hear for myself if the "Railsback" adjustments make a difference. I finally took some time to make two sound files, one which uses pure equal temperament and the other which incorporates "railsback" adjustments, using the chart linked earlier that came from a Yamaha digital piano manual.
Folloing are the two links. The timbre used is a pretty common variant of a DX7 string synth patch. It's being played by a Java-based FM synth that I wrote.
6 seconds of a low drone: A1 (midi 33, 55 Hz)
6 second drone of a high drone added, A6 (midi 93, 1760 Hz)
then an arpeggio that spans A2 to A5 comes in.
With the "railsback" version, the pitch adjustments range from -8 cents for the low A1 to +9 cents for the high A6.
The difference between the two tracks is pretty subtle! I can hear it most readily when running both files at the same time, but flipping (solo/mute) back and forth between them quickly. I find it's easiest to distinguish the pitch difference between the high drones. It does seem to me that the "railsback" version sits a little easier, somehow. But I wonder if this high string synth would sound still better if it were yet a few cents sharper.
There's some detuning built into the patch, but the Op stack which predominates is a pure 1 : 1 : 4 (where 1 is the carrier and 1 and 4 are modulators stacked in series). The other stack is a "feedback" loop with the pitch set via a 0.998 factor (i.e., for 440 Hz, the detuned stack is 439.12 Hz, and in that region, 0.247 Hz is 1 cent). I think its overtones phase with the main stack overtones more than they directly affect the perceived pitch. Maybe this stack does have an impact on the sense of the octaves. But I'm not at all clear how, especially since the exact same patch is being used for all the notes being played.
I think the internal phasing of the timbre that results would tend to make clashes between the overtones and fundamentals of concurrent notes less prominent, as there is a baseline of considerable timbral activity regardless of the exact pitches being played.
So, is it worth the trouble to correct for the octave stretch? IDK. I think it depends on the context. The producer (and the musicians involved) need to use their best judgment. Sometimes if a note is slightly "off" or "pitchy" it actually works better. A subliminally sharp note might give it a bit of edge or assertiveness, a subliminally flat note might add a bit of despondency or passivity that fits the moment.
|
I'm not going to be able to do this right away, as the program isn't sufficiently automated for external triggers yet, and my MIDI DAW is on an older XP machine that is in storage.
What sort of analysis are you hoping to perform? Maybe there is a way I can support that more easily?
This weekend I could redo the recording and double check that each step of the process is identical. I may have overlooked something. The timing of the start of each section will be me clicking "Go" while watching a timer. I'll try to use my "musician's" accuracy instead of treating the entrance timings casually.
As I've been listening more, I am noticing the following. I wonder if others hear the same thing?
1) When the high string drone first comes in, it sounds like it could be a little flat in both clips, but more so in the Equal Temperament clip.
2) When the arpeggio plays A3 or higher, the high drone sounds in tune to me, on the Railsback clip, though possibly it sounds a bit sharp when the arpeggio hits A5.
3) When the arpeggio plays A4 or higher on the Equal Temperament clip, the high drone sounds in tune to me.
I'm reminded of optical illusions where the color of a region varies based on the surrounding colors. It seems entirely likely that there exist aural parallels, where concurrent tones can affect perception. But the pitch perception effects here are a great deal more subtle than the optical illusions I've seen.
The implication is kind of stunning to me. It suggests that the pitch of a synth may need to adjust a few cents this way or that, depending on the context, to remain sounding "perfectly in tune". This would be rather difficult to encode! But at the same time, musicians probably are already doing a bit of this (some more successfully than others) automatically in response to what others around them are playing.
-----Original Message-----
From: Eb Myatt <eb.myatt@...>
To: [email protected]; philfrei@...
Sent: Tue, Nov 2, 2021 12:57 pm
Subject: Re: [YamahaDX] Intonation on DX7
Hello Phil,
Thanks for the report. The two sound files are slightly different in terms of amplitude and headroom. The?railsback_stringsynthdrone is slightly louder than the equal temperament one. This is evident when looking at the waveforms. Could you re-create these files by rendering them using the same MIDI file for a better comparison.?
Thanks in advance,
E? ?
On Nov 1, 2021, at 10:58 AM, philfrei via < philfrei@...> wrote:
I've been wanting to hear for myself if the "Railsback" adjustments make a difference. I finally took some time to make two sound files, one which uses pure equal temperament and the other which incorporates "railsback" adjustments, using the chart linked earlier that came from a Yamaha digital piano manual.
Folloing are the two links. The timbre used is a pretty common variant of a DX7 string synth patch. It's being played by a Java-based FM synth that I wrote.
6 seconds of a low drone: A1 (midi 33, 55 Hz)
6 second drone of a high drone added, A6 (midi 93, 1760 Hz)
then an arpeggio that spans A2 to A5 comes in.
With the "railsback" version, the pitch adjustments range from -8 cents for the low A1 to +9 cents for the high A6.
The difference between the two tracks is pretty subtle! I can hear it most readily when running both files at the same time, but flipping (solo/mute) back and forth between them quickly. I find it's easiest to distinguish the pitch difference between the high drones. It does seem to me that the "railsback" version sits a little easier, somehow. But I wonder if this high string synth would sound still better if it were yet a few cents sharper.
There's some detuning built into the patch, but the Op stack which predominates is a pure 1 : 1 : 4 (where 1 is the carrier and 1 and 4 are modulators stacked in series). The other stack is a "feedback" loop with the pitch set via a 0.998 factor (i.e., for 440 Hz, the detuned stack is 439.12 Hz, and in that region, 0.247 Hz is 1 cent). I think its overtones phase with the main stack overtones more than they directly affect the perceived pitch. Maybe this stack does have an impact on the sense of the octaves. But I'm not at all clear how, especially since the exact same patch is being used for all the notes being played.
I think the internal phasing of the timbre that results would tend to make clashes between the overtones and fundamentals of concurrent notes less prominent, as there is a baseline of considerable timbral activity regardless of the exact pitches being played.
So, is it worth the trouble to correct for the octave stretch? IDK. I think it depends on the context. The producer (and the musicians involved) need to use their best judgment. Sometimes if a note is slightly "off" or "pitchy" it actually works better. A subliminally sharp note might give it a bit of edge or assertiveness, a subliminally flat note might add a bit of despondency or passivity that fits the moment.
|
I'm not going to be able to do this right away, as the program isn't sufficiently automated for external triggers yet, and my MIDI DAW is on an older XP machine that is in storage.
What sort of analysis are you hoping to perform? Maybe there is a way I can support that more easily?
Phil,
Since the differences can be very subtle, it is best to eliminate any?unnecessary sources of variation to make an accurate aural?comparison.
This weekend I could redo the recording and double check that each step of the process is identical. I may have overlooked something. The timing of the start of each section will be me clicking "Go" while watching a timer. I'll try to use my "musician's" accuracy instead of treating the entrance timings casually.
A metronome perhaps?
As I've been listening more, I am noticing the following. I wonder if others hear the same thing?
1) When the high string drone first comes in, it sounds like it could be a little flat in both clips, but more so in the Equal Temperament clip.
2) When the arpeggio plays A3 or higher, the high drone sounds in tune to me, on the Railsback clip, though possibly it sounds a bit sharp when the arpeggio hits A5.
3) When the arpeggio plays A4 or higher on the Equal Temperament clip, the high drone sounds in tune to me.
Yes, the high string drone does sound slightly?“off” in the beginning and "falls into place" from A4 onwards. This is one reason I like to use real-time patch edits during mixing and live playing though other sources of variability also need to be considered (below).
I'm reminded of optical illusions where the color of a region varies based on the surrounding colors. It seems entirely likely that there exist aural parallels, where concurrent tones can affect perception. But the pitch perception effects here are a great deal more subtle than the optical illusions I've seen.
The implication is kind of stunning to me. It suggests that the pitch of a synth may need to adjust a few cents this way or that, depending on the context, to remain sounding "perfectly in tune". This would be rather difficult to encode! But at the same time, musicians probably are already doing a bit of this (some more successfully than others) automatically in response to what others around them are playing.
Aurally, other sources of variability like (multi-band) compression, EQ and FX chains also need meticulous adjustment in an actual mixing or live situation.
|
Thanks for confirming the "sounds off" and "falls into place" perceptions that I had. If that can be heard by others, then my main goal has been accomplished.
Interesting to hear about "meticulous adjustment" being expected and practiced in the studio. My main wonder has been if this is well known (pitch adjustments for synths) and commonly done or not. I'm thinking that perhaps the situation is as follows: nothing in particular is done beforehand, in terms of adding "Railsback" adjustments to synths (or comparable set of adjustments), since even these are at best only good first guesses and the changing context may invalidate those choices. So instead, in the mix, the precaution is made that the synths be automated (MIDI or OSC), and audible pitch problems are handled via adjustments as needed.
I'm just someone who works with a home studio, so I don't know what is happening in the pro studios. When I was asking around, for example, at various retailers that sell audio gear, no one mentioned any specific tools used specifically for the purpose of dealing with "stretch" tuning issues.
Therefore, this kind of adjustment relies on skilled operators with excellent ears and their attention to detail, rather than software?
-----Original Message-----
From: Eb Myatt <eb.myatt@...>
To: philfrei@...
Cc: [email protected] < [email protected]>
Sent: Wed, Nov 3, 2021 6:21 am
Subject: Re: [YamahaDX] Intonation on DX7
I'm not going to be able to do this right away, as the program isn't sufficiently automated for external triggers yet, and my MIDI DAW is on an older XP machine that is in storage.
What sort of analysis are you hoping to perform? Maybe there is a way I can support that more easily?
Phil,
Since the differences can be very subtle, it is best to eliminate any?unnecessary sources of variation to make an accurate aural?comparison.
This weekend I could redo the recording and double check that each step of the process is identical. I may have overlooked something. The timing of the start of each section will be me clicking "Go" while watching a timer. I'll try to use my "musician's" accuracy instead of treating the entrance timings casually.
A metronome perhaps?
As I've been listening more, I am noticing the following. I wonder if others hear the same thing?
1) When the high string drone first comes in, it sounds like it could be a little flat in both clips, but more so in the Equal Temperament clip.
2) When the arpeggio plays A3 or higher, the high drone sounds in tune to me, on the Railsback clip, though possibly it sounds a bit sharp when the arpeggio hits A5.
3) When the arpeggio plays A4 or higher on the Equal Temperament clip, the high drone sounds in tune to me.
Yes, the high string drone does sound slightly?“off” in the beginning and "falls into place" from A4 onwards. This is one reason I like to use real-time patch edits during mixing and live playing though other sources of variability also need to be considered (below).
I'm reminded of optical illusions where the color of a region varies based on the surrounding colors. It seems entirely likely that there exist aural parallels, where concurrent tones can affect perception. But the pitch perception effects here are a great deal more subtle than the optical illusions I've seen.
The implication is kind of stunning to me. It suggests that the pitch of a synth may need to adjust a few cents this way or that, depending on the context, to remain sounding "perfectly in tune". This would be rather difficult to encode! But at the same time, musicians probably are already doing a bit of this (some more successfully than others) automatically in response to what others around them are playing.
Aurally, other sources of variability like (multi-band) compression, EQ and FX chains also need meticulous adjustment in an actual mixing or live situation.
|
On Nov 3, 2021, at 3:19 PM, philfrei via < philfrei@...> wrote:
Thanks for confirming the "sounds off" and "falls into place" perceptions that I had. If that can be heard by others, then my main goal has been accomplished.
Interesting to hear about "meticulous adjustment" being expected and practiced in the studio. My main wonder has been if this is well known (pitch adjustments for synths) and commonly done or not. I'm thinking that perhaps the situation is as follows: nothing in particular is done beforehand, in terms of adding "Railsback" adjustments to synths (or comparable set of adjustments), since even these are at best only good first guesses and the changing context may invalidate those choices. So instead, in the mix, the precaution is made that the synths be automated (MIDI or OSC), and audible pitch problems are handled via adjustments as needed.
Quite. Good audio production is also a result of hours of practice. Unbeknown to many, good producers are also highly-trained and/or talented musicians... Brian Eno, Steven Wilson to name a couple. ?
I'm just someone who works with a home studio, so I don't know what is happening in the pro studios. When I was asking around, for example, at various retailers that sell audio gear, no one mentioned any specific tools used specifically for the purpose of dealing with "stretch" tuning issues.
Therefore, this kind of adjustment relies on skilled operators with excellent ears and their attention to detail, rather than software?
Don’t get me wrong, Visionaries like you could make software to facilitate matters. ?
Best wishes, Eb
|
