High Gain, Low Volume, 'classic' guitar sound

Started by chuckles August 29, 2005
This is a bit tricky to explain, I've been playing around with clipping
for a little while using a clean guitar, passed through a clipping
algorithm which produces a square wave at its highest level (when
passed a sinusoid) and a rounded version of the sinusoid for values
0-(max-1). The sound it produces becomes gradually 'fuzzier' as the
level is set higher and higher. Its not great and produces a distortion
which is next door to useless (as it begins to destroy some parts of
the signal at high levels which produces crackling as opposed to a
distorted signal). So my question is just what is it that makes the
guitar playing typically heard in the bluesy music from the sixties and
seventies sound the way it does? As an example George Harrison's
playing on Sgt Peppers Lonely Hearts, Pink Floyd's Money solo etc. I
know there is a great debate about valve amps and the majesty of their
clipping regions but even with my own very cheap setup (fender
stratocaster and 230v epiphone solid state amp) I can get a sound much
closer to that classic sound then with my current clipping algorithm.
Typically I set my gain high while leaving my volume low to produce the
required sound. So is it proper clipping I'm hearing from my amp or
something else all together?

Anyone who could give me an idea about whats going on, that'd be great.


Cathal

Check out this site

http://www.geofex.com/effxfaq/distn101.htm

If you are getting "Squarewaves" your probably clipping too much.

Also Google "Guitar Waveforms"

Eric

chuckles  wrote:

>The sound it produces becomes gradually 'fuzzier' as the >level is set higher and higher. Its not great and produces a distortion >which is next door to useless (as it begins to destroy some parts of >the signal at high levels which produces crackling as opposed to a >distorted signal). So my question is just what is it that makes the >guitar playing typically heard in the bluesy music from the sixties and >seventies sound the way it does?
You want distortion that generates even harmonics rather than odd harmonics. A square wave has only odd harmonics and has a buzz-like sound. Steve
Thanks very much lads!

In  Steve Pope wrote:
> You want distortion that generates even harmonics rather > than odd harmonics. A square wave has only odd harmonics > and has a buzz-like sound. > > Steve
A lot of people say this, but it has little truth to it. Sure, we don't really want a "square" wave--even without the mad aliasing, it would be a little harsh (more like transistors than tubes). But when people talk about even versus odd, they are usually talking about symmetrical (odd harmonics) versus non-symmetrical (both) transfer functions for the distortion. In practice, when you play a guitar (which doesn't have a steaqdily symmetrical output to begin with, even in the best of conditions), you will have an incredibly difficult time convincing yourself that there's any difference in tone between the two, once you've gotten the other necessary elements in place (mainly tone filtering). This is true even if one half of the non-symmetrical transfer function is significantly different than the other half. (Also, square waves are pretty hollow sounding, not buzzy, and even more so after you've softened the corners. Narrow pulse waves are buzzy.)
Good question--I'll put it another way, which may help shed a little 
light:

So many guitar amps sharing the same tubes and amplifier design--why do 
they sound so different from one manufacturer to another (Marshall sound, 
Fender sound...)?

Well, one clue is that you won't get that nice guitar sound overdriving 
a tube amplifier designed for stereo hifi applications. Hint: Guitar 
amps don't have a flat tone response (ever see a guitar amp boasting 20-
20Khz response curves?).

So, try filtering the output of the distortion. The tone control knobs 
of a Fender Twin give you the basic idea.

You can filter it before the distortion processing too. Think about this, 
for instance: You want to design an amp with mega gain--insane 
distortion for screaming leads. But you figure out pretty quickly that 
mega-clipping low frequencies makes the tone flatulant, and mega-
clipping high frequencies screeches like nails on a chalkboard. Nasty. 
Not so bad ont he mids, where you want that punch, so how 'bout dipping 
the lows and highs going into the clipper, then reshaping the tone again 
at the output of the clipper to get back so of the beef at the low end 
and sparkle at the high end. By shaping the tone before the clipper, 
you've chosen to distort the mid more than the lows and highs, giving 
you the scream without the flatulance and screech.



In <1125328070.265241.93310@f14g2000cwb.googlegroups.com> chuckles wrote:
> This is a bit tricky to explain, I've been playing around with > clipping for a little while using a clean guitar, passed through a > clipping algorithm which produces a square wave at its highest level ( > when passed a sinusoid) and a rounded version of the sinusoid for > values 0-(max-1). The sound it produces becomes gradually 'fuzzier' as > the level is set higher and higher. Its not great and produces a > distortion which is next door to useless (as it begins to destroy some > parts of the signal at high levels which produces crackling as opposed > to a distorted signal). So my question is just what is it that makes > the guitar playing typically heard in the bluesy music from the > sixties and seventies sound the way it does? As an example George > Harrison's playing on Sgt Peppers Lonely Hearts, Pink Floyd's Money > solo etc. I know there is a great debate about valve amps and the > majesty of their clipping regions but even with my own very cheap > setup (fender stratocaster and 230v epiphone solid state amp) I can > get a sound much closer to that classic sound then with my current > clipping algorithm. Typically I set my gain high while leaving my > volume low to produce the required sound. So is it proper clipping I'm > hearing from my amp or something else all together? > > Anyone who could give me an idea about whats going on, that'd be great. > > > Cathal
in article 20050831170145607-0700@news.linkline.com, Nigel Redmon at
earlevel@earlevel.com wrote on 08/31/2005 20:01:

> In Steve Pope wrote: >> You want distortion that generates even harmonics rather >> than odd harmonics. A square wave has only odd harmonics >> and has a buzz-like sound. > > A lot of people say this, but it has little truth to it.
i think it has more than a little, Nigel. in article 20050831171709501-0700@news.linkline.com, Nigel Redmon at earlevel@earlevel.com wrote on 08/31/2005 20:17:
> So many guitar amps sharing the same tubes and amplifier design--why do > they sound so different from one manufacturer to another (Marshall sound, > Fender sound...)? > > Well, one clue is that you won't get that nice guitar sound overdriving > a tube amplifier designed for stereo hifi applications. Hint: Guitar > amps don't have a flat tone response (ever see a guitar amp boasting 20- > 20Khz response curves?). > > So, try filtering the output of the distortion. The tone control knobs > of a Fender Twin give you the basic idea.
filtering before the distortion has some effect also. in fact, once you mix non-linear operations with filtering, you get into some alchemy that is difficult to quantify. you could also have non-linear distortion, filtering, *and* feedback. really, if you're gonna model the tubes (with some interelectrode capacitance and inductance), driving circuits with blocking capacitors, output transformers (with hysteresis and other non-linear characteristics) and power supply voltages with 60 Hz ripple and sorta poor regulation (that affects the tube curves when Jimmy Page lays on the power chord and the B+ voltage drops and gets more rippley), all that makes for a difficult to discover witches brew for a model. one thing i would suggest is to take the guitar input signal and upsample it immediately to 192 kHz and do all this nasty non-linear stuff at that high sampling rate, then LPF it before downsampling to 48 kHz or whatever the output is. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
In  robert bristow-johnson  
wrote:
> > in article 20050831170145607-0700@news.linkline.com, Nigel Redmon at > earlevel@earlevel.com wrote on 08/31/2005 20:01: > >> In Steve Pope wrote: >>> You want distortion that generates even harmonics rather >>> than odd harmonics. A square wave has only odd harmonics >>> and has a buzz-like sound. >> >> A lot of people say this, but it has little truth to it. > > i think it has more than a little, Nigel.
Hi Robert, Well, I think you know I wrote the DSP code for a couple of commercial amp simulator products and know enough about the guts of other associated products that I'd better not go into a lot of detail. But I will tell you that I spent a lot of time trying to be as computationally cheap as possible with my first one that had to run on a single slow 56001. The results were so difficult to distinguish from more elaborate methods, was so much cheaper (computationally), and the kicker was that it solved a strange and annoying quirk of asymmetrical distortion plaguing the more complex design, that it became adopted in all products that followed. On one hand, I think I've said too much, but on the other it's really about the tone processing (99%) that makes it sound like a real vintage amp. So, I have to stand by my assertion of "little". In an A-B test, changing only symmetrical versus asymmetrical (including where the shape of one half is drastically different--see "you know who's" patents for an example) distortion, most listers would have a heck of a time tellingthe difference; those who could latch onto the subtle difference would find that it's simply that--a difference--choosing which is "better" is more difficult. And in a mix, forget it. Again, the assertion that it's an important part of getting the guitar amp sound is just not true. Sure, mess around with it when you're at the point of playing with power supply sag and other subtleties and see if you like it, but yeah, it is "little". The OP was having problems sounding anything like a guitar amp--even versus odd harmonics should be the least of his concerns at this point. (I'll reiterate that using a symmetrical "clipper" does not imply that the output is symmetrical--due to DC offset, attack and decay characteristics of guitars, etc.)
>> ... So, try >> filtering the output of the distortion. The tone control knobs of a >> Fender Twin give you the basic idea. > > filtering before the distortion has some effect also...
You probably noticed that the remainder of my post was about pre- filtering, and you're probably directing the comment to the OP, but since you quoted me ahead of your response, I'm not sure :-/
> one thing i would suggest is to take the guitar input signal and > upsample it immediately to 192 kHz and do all this nasty non-linear > stuff at that high sampling rate, then LPF it before downsampling to > 48 kHz or whatever the output is.
Yes, though 192k is probably not quite enough in the general case. Tip to the OP: You just see what the minimum you can live with, after you've gotten everything the way you want it. On the one hand, the more the distortion, the more the aliasing, but then again the more distortion to drown out the aliasing. But the clincher is note bends--nothing like the aliasing moving down when you bend a note up to give it away. Playing a harmonic and bending it with the whammy bar tends to make it pretty obvious. You have to oversample up to the minimum point that the aliasing is a problem. And thank goodness that in most speaker cabinets, the high drop off like a rock, though for best generality you can't count on that (you might want to record direct from the "amp", especially for things like processing vocals). Oh yeah (to the OP): The speaker cabinet is huge--you don't play a Marshal through hifi speakers. You really need to spend some time there, but for a really gross approximation, to get started while you're working ont he other stuff, try a steep (none of this single-biquad stuff) 5kHz-ish lowpass. In the end, you'll prbably want some convolution for the cabs, but you can get in the ballpark shaping the tone with IIRs.
Try reducing the Bass by about 8-20 db from 350 Hz downwards before hitting
the distortion

followed by boosting it up again after the distortion.
Try reducing the Bass by about 8-20 db from 350 Hz downwards before hitting
the distortion

followed by boosting it up again after the distortion.
In  robert bristow-johnson  
wrote:
> > in article 20050831170145607-0700@news.linkline.com, Nigel Redmon at > earlevel@earlevel.com wrote on 08/31/2005 20:01: > >> In Steve Pope wrote: >>> You want distortion that generates even harmonics rather >>> than odd harmonics. A square wave has only odd harmonics >>> and has a buzz-like sound. >> >> A lot of people say this, but it has little truth to it. > > i think it has more than a little, Nigel.
Hi Robert, Well, I think you know I wrote the DSP code for a couple of commercial amp simulator products and know enough about the guts of other associated products that I'd better not go into a lot of detail. But I will tell you that I spent a lot of time trying to be as computationally cheap as possible with my first one that had to run on a single slow 56001. The results were so difficult to distinguish from more elaborate methods, was so much cheaper (computationally), and the kicker was that it solved a strange and annoying quirk of asymmetrical distortion plaguing the more complex design, that it became adopted in all products that followed. On one hand, I think I've said too much, but on the other it's really about the tone processing (99%) that makes it sound like a real vintage amp. So, I have to stand by my assertion of "little". In an A-B test, changing only symmetrical versus asymmetrical (including where the shape of one half is drastically different--see "you know who's" patents for an example) distortion, most listers would have a heck of a time tellingthe difference; those who could latch onto the subtle difference would find that it's simply that--a difference--choosing which is "better" is more difficult. And in a mix, forget it. Again, the assertion that it's an important part of getting the guitar amp sound is just not true. Sure, mess around with it when you're at the point of playing with power supply sag and other subtleties and see if you like it, but yeah, it is "little". The OP was having problems sounding anything like a guitar amp--even versus odd harmonics should be the least of his concerns at this point. (I'll reiterate that using a symmetrical "clipper" does not imply that the output is symmetrical--due to DC offset, attack and decay characteristics of guitars, etc.)
>> ... So, try >> filtering the output of the distortion. The tone control knobs of a >> Fender Twin give you the basic idea. > > filtering before the distortion has some effect also...
You probably noticed that the remainder of my post was about pre- filtering, and you're probably directing the comment to the OP, but since you quoted me ahead of your response, I'm not sure :-/
> one thing i would suggest is to take the guitar input signal and > upsample it immediately to 192 kHz and do all this nasty non-linear > stuff at that high sampling rate, then LPF it before downsampling to > 48 kHz or whatever the output is.
Yes, though 192k is probably not quite enough in the general case. Tip to the OP: You just see what the minimum you can live with, after you've gotten everything the way you want it. On the one hand, the more the distortion, the more the aliasing, but then again the more distortion to drown out the aliasing. But the clincher is note bends--nothing like the aliasing moving down when you bend a note up to give it away. Playing a harmonic and bending it with the whammy bar tends to make it pretty obvious. You have to oversample up to the minimum point that the aliasing is a problem. And thank goodness that in most speaker cabinets, the high drop off like a rock, though for best generality you can't count on that (you might want to record direct from the "amp", especially for things like processing vocals). Oh yeah (to the OP): The speaker cabinet is huge--you don't play a Marshal through hifi speakers. You really need to spend some time there, but for a really gross approximation, to get started while you're working ont he other stuff, try a steep (none of this single-biquad stuff) 5kHz-ish lowpass. In the end, you'll prbably want some convolution for the cabs, but you can get in the ballpark shaping the tone with IIRs.
in article 20050831170145607-0700@news.linkline.com, Nigel Redmon at
earlevel@earlevel.com wrote on 08/31/2005 20:01:

> In Steve Pope wrote: >> You want distortion that generates even harmonics rather >> than odd harmonics. A square wave has only odd harmonics >> and has a buzz-like sound. > > A lot of people say this, but it has little truth to it.
i think it has more than a little, Nigel. in article 20050831171709501-0700@news.linkline.com, Nigel Redmon at earlevel@earlevel.com wrote on 08/31/2005 20:17:
> So many guitar amps sharing the same tubes and amplifier design--why do > they sound so different from one manufacturer to another (Marshall sound, > Fender sound...)? > > Well, one clue is that you won't get that nice guitar sound overdriving > a tube amplifier designed for stereo hifi applications. Hint: Guitar > amps don't have a flat tone response (ever see a guitar amp boasting 20- > 20Khz response curves?). > > So, try filtering the output of the distortion. The tone control knobs > of a Fender Twin give you the basic idea.
filtering before the distortion has some effect also. in fact, once you mix non-linear operations with filtering, you get into some alchemy that is difficult to quantify. you could also have non-linear distortion, filtering, *and* feedback. really, if you're gonna model the tubes (with some interelectrode capacitance and inductance), driving circuits with blocking capacitors, output transformers (with hysteresis and other non-linear characteristics) and power supply voltages with 60 Hz ripple and sorta poor regulation (that affects the tube curves when Jimmy Page lays on the power chord and the B+ voltage drops and gets more rippley), all that makes for a difficult to discover witches brew for a model. one thing i would suggest is to take the guitar input signal and upsample it immediately to 192 kHz and do all this nasty non-linear stuff at that high sampling rate, then LPF it before downsampling to 48 kHz or whatever the output is. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Good question--I'll put it another way, which may help shed a little 
light:

So many guitar amps sharing the same tubes and amplifier design--why do 
they sound so different from one manufacturer to another (Marshall sound, 
Fender sound...)?

Well, one clue is that you won't get that nice guitar sound overdriving 
a tube amplifier designed for stereo hifi applications. Hint: Guitar 
amps don't have a flat tone response (ever see a guitar amp boasting 20-
20Khz response curves?).

So, try filtering the output of the distortion. The tone control knobs 
of a Fender Twin give you the basic idea.

You can filter it before the distortion processing too. Think about this, 
for instance: You want to design an amp with mega gain--insane 
distortion for screaming leads. But you figure out pretty quickly that 
mega-clipping low frequencies makes the tone flatulant, and mega-
clipping high frequencies screeches like nails on a chalkboard. Nasty. 
Not so bad ont he mids, where you want that punch, so how 'bout dipping 
the lows and highs going into the clipper, then reshaping the tone again 
at the output of the clipper to get back so of the beef at the low end 
and sparkle at the high end. By shaping the tone before the clipper, 
you've chosen to distort the mid more than the lows and highs, giving 
you the scream without the flatulance and screech.



In <1125328070.265241.93310@f14g2000cwb.googlegroups.com> chuckles wrote:
> This is a bit tricky to explain, I've been playing around with > clipping for a little while using a clean guitar, passed through a > clipping algorithm which produces a square wave at its highest level ( > when passed a sinusoid) and a rounded version of the sinusoid for > values 0-(max-1). The sound it produces becomes gradually 'fuzzier' as > the level is set higher and higher. Its not great and produces a > distortion which is next door to useless (as it begins to destroy some > parts of the signal at high levels which produces crackling as opposed > to a distorted signal). So my question is just what is it that makes > the guitar playing typically heard in the bluesy music from the > sixties and seventies sound the way it does? As an example George > Harrison's playing on Sgt Peppers Lonely Hearts, Pink Floyd's Money > solo etc. I know there is a great debate about valve amps and the > majesty of their clipping regions but even with my own very cheap > setup (fender stratocaster and 230v epiphone solid state amp) I can > get a sound much closer to that classic sound then with my current > clipping algorithm. Typically I set my gain high while leaving my > volume low to produce the required sound. So is it proper clipping I'm > hearing from my amp or something else all together? > > Anyone who could give me an idea about whats going on, that'd be great. > > > Cathal
In  Steve Pope wrote:
> You want distortion that generates even harmonics rather > than odd harmonics. A square wave has only odd harmonics > and has a buzz-like sound. > > Steve
A lot of people say this, but it has little truth to it. Sure, we don't really want a "square" wave--even without the mad aliasing, it would be a little harsh (more like transistors than tubes). But when people talk about even versus odd, they are usually talking about symmetrical (odd harmonics) versus non-symmetrical (both) transfer functions for the distortion. In practice, when you play a guitar (which doesn't have a steaqdily symmetrical output to begin with, even in the best of conditions), you will have an incredibly difficult time convincing yourself that there's any difference in tone between the two, once you've gotten the other necessary elements in place (mainly tone filtering). This is true even if one half of the non-symmetrical transfer function is significantly different than the other half. (Also, square waves are pretty hollow sounding, not buzzy, and even more so after you've softened the corners. Narrow pulse waves are buzzy.)
Thanks very much lads!

chuckles  wrote:

>The sound it produces becomes gradually 'fuzzier' as the >level is set higher and higher. Its not great and produces a distortion >which is next door to useless (as it begins to destroy some parts of >the signal at high levels which produces crackling as opposed to a >distorted signal). So my question is just what is it that makes the >guitar playing typically heard in the bluesy music from the sixties and >seventies sound the way it does?
You want distortion that generates even harmonics rather than odd harmonics. A square wave has only odd harmonics and has a buzz-like sound. Steve
Check out this site

http://www.geofex.com/effxfaq/distn101.htm

If you are getting "Squarewaves" your probably clipping too much.

Also Google "Guitar Waveforms"

Eric

This is a bit tricky to explain, I've been playing around with clipping
for a little while using a clean guitar, passed through a clipping
algorithm which produces a square wave at its highest level (when
passed a sinusoid) and a rounded version of the sinusoid for values
0-(max-1). The sound it produces becomes gradually 'fuzzier' as the
level is set higher and higher. Its not great and produces a distortion
which is next door to useless (as it begins to destroy some parts of
the signal at high levels which produces crackling as opposed to a
distorted signal). So my question is just what is it that makes the
guitar playing typically heard in the bluesy music from the sixties and
seventies sound the way it does? As an example George Harrison's
playing on Sgt Peppers Lonely Hearts, Pink Floyd's Money solo etc. I
know there is a great debate about valve amps and the majesty of their
clipping regions but even with my own very cheap setup (fender
stratocaster and 230v epiphone solid state amp) I can get a sound much
closer to that classic sound then with my current clipping algorithm.
Typically I set my gain high while leaving my volume low to produce the
required sound. So is it proper clipping I'm hearing from my amp or
something else all together?

Anyone who could give me an idea about whats going on, that'd be great.


Cathal