[Les Cargill[_4_]]

Gareth Magennis wrote:
How might you obtain an impulse set for a Space Echo?

I'd start with a single sample impulse at various numbers
of repeats and speeds, same way you do a guitar amp.

--
Les Cargill

[Scott Dorsey]

You seem to be angling that tape echo is inherently nonlinear; I'd have
to profess some agnosticism of that.

It's pretty severely nonlinear, as the distortion spectrum should
demonstrate. Listen to a repeated echo... by the fourth time around you
can't even make out what the original signal was.

How might you obtain an impulse set for a Space Echo?

You put an impulse into the input and you record what comes out, and you
get an okay nonlinear time-invariant model out.

Problems with this approach include the fact that it's not time-invariant
at all and there is significant FM due to speed variations going over the
heads.

Another big one is that the Echoplex is subject to external influences and
can't be modelled in isolation. Put it on top of a bass cabinet and the
cabinet vibrates the tape going by and you get another set of (mostly FM)
modulations from the external mechanical source.

Sometimes people will push and pull on the tape with their fingers to
get still different effects.

Sure... you could probably get a pretty good digital model of the thing, but
it would be a lot of work, and people would still prefer the real thing
because of the sheer aesthetics of it and ability to control it mechanically.
People aren't logical, but that's fine. That's how we get music.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Les Cargill[_4_]]

Scott Dorsey wrote:
You seem to be angling that tape echo is inherently nonlinear; I'd have
to profess some agnosticism of that.

It's pretty severely nonlinear, as the distortion spectrum should
demonstrate. Listen to a repeated echo... by the fourth time around you
can't even make out what the original signal was.


That's just a grainy iterated low pass filter. I have a couple of cheezy
bucket brigade pedals that do a decent job of this. It's not
exactly the same thing.

How might you obtain an impulse set for a Space Echo?

You put an impulse into the input and you record what comes out, and you
get an okay nonlinear time-invariant model out.

Problems with this approach include the fact that it's not time-invariant
at all and there is significant FM due to speed variations going over the
heads.


Well, sure.

Another big one is that the Echoplex is subject to external influences and
can't be modelled in isolation. Put it on top of a bass cabinet and the
cabinet vibrates the tape going by and you get another set of (mostly FM)
modulations from the external mechanical source.

Sometimes people will push and pull on the tape with their fingers to
get still different effects.

Sure... you could probably get a pretty good digital model of the thing, but
it would be a lot of work, and people would still prefer the real thing
because of the sheer aesthetics of it and ability to control it mechanically.

Of course.

People aren't logical, but that's fine. That's how we get music.
--scott


--
Les Cargill

[Scott Dorsey]

Les Cargill wrote:
Scott Dorsey wrote:
You seem to be angling that tape echo is inherently nonlinear; I'd have
to profess some agnosticism of that.

It's pretty severely nonlinear, as the distortion spectrum should
demonstrate. Listen to a repeated echo... by the fourth time around you
can't even make out what the original signal was.

That's just a grainy iterated low pass filter. I have a couple of cheezy
bucket brigade pedals that do a decent job of this. It's not
exactly the same thing.

That "graininess" is distortion. There's a lot of it. That distortion is
not a linear function and therefore requires modelling if you want to
effectively model the device.

The bucket brigade filters do something else, they actually have a carrier
frequency and aliasing effects from anything above twice the carrier
frequency, and no reconstruction filters either so plenty of weird aliasing
going on there too. The BBD quantizes time but not amplitude so it has some
attributes of a digital device but not all of them. You can model that too,
but the math is very different.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Les Cargill[_4_]]

Scott Dorsey wrote:
Les Cargill wrote:
Scott Dorsey wrote:
You seem to be angling that tape echo is inherently nonlinear; I'd have
to profess some agnosticism of that.

It's pretty severely nonlinear, as the distortion spectrum should
demonstrate. Listen to a repeated echo... by the fourth time around you
can't even make out what the original signal was.

That's just a grainy iterated low pass filter. I have a couple of cheezy
bucket brigade pedals that do a decent job of this. It's not
exactly the same thing.

That "graininess" is distortion. There's a lot of it. That distortion is
not a linear function and therefore requires modelling if you want to
effectively model the device.


I still think of it as more like something like chiaroscuro for sound.
Lots of specific techniques, but which is "best" is not clear.

I suppose somebody could prefer this version or that version, but
it's just a preference.


The bucket brigade filters do something else, they actually have a carrier
frequency and aliasing effects from anything above twice the carrier
frequency, and no reconstruction filters either so plenty of weird aliasing
going on there too.

Yes. Of course, this is a guitar pedal, so the assumption is
that the amp drops off rapidly past 5k.


The BBD quantizes time but not amplitude so it has some
attributes of a digital device but not all of them. You can model that too,
but the math is very different.
--scott


--
Les Cargill

[Gareth Magennis[_3_]]

How might you obtain an impulse set for a Space Echo?

You put an impulse into the input and you record what comes out, and you
get an okay nonlinear time-invariant model out.

Problems with this approach include the fact that it's not time-invariant
at all and there is significant FM due to speed variations going over the
heads.






I think that pretty much sums up my point, in that each instance of Tape
Echo is unique.


I think you said a while ago that tape echo's are actually more of an
instrument than an effect.




Gareth.

[Scott Dorsey]

Les Cargill wrote:
Scott Dorsey wrote:

That "graininess" is distortion. There's a lot of it. That distortion is
not a linear function and therefore requires modelling if you want to
effectively model the device.

I still think of it as more like something like chiaroscuro for sound.

If you're doing the math, you think about things that are linear and things
that are nonlinear. If frequencies are being generated that weren't in the
original, then we describe that as distortion and there is a specific way
to address that.

Lots of specific techniques, but which is "best" is not clear.

There are a lot needed because there are a lot of different effects in the
system that need to be modelled.

I suppose somebody could prefer this version or that version, but
it's just a preference.

Objectively I can say that this function matches the original better than
this other function. At some point we have to make some subjective evaluation
once we get to the point where it doesn't match the original very well, but
the goal is to stay away from the point where there is any need to worry
about that.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Scott Dorsey]

Gareth Magennis wrote:

How might you obtain an impulse set for a Space Echo?

You put an impulse into the input and you record what comes out, and you
get an okay nonlinear time-invariant model out.

Problems with this approach include the fact that it's not time-invariant
at all and there is significant FM due to speed variations going over the
heads.


I think that pretty much sums up my point, in that each instance of Tape
Echo is unique.

No, it just says that the model required to describe the system is pretty
complex.

The inconsistency of tape echo is different (and that can be modelled too,
but it makes the model even more complex).

I think you said a while ago that tape echo's are actually more of an
instrument than an effect.

People play them.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Les Cargill[_4_]]

Scott Dorsey wrote:
Les Cargill wrote:
Scott Dorsey wrote:

That "graininess" is distortion. There's a lot of it. That distortion is
not a linear function and therefore requires modelling if you want to
effectively model the device.

I still think of it as more like something like chiaroscuro for sound.

If you're doing the math, you think about things that are linear and things
that are nonlinear. If frequencies are being generated that weren't in the
original, then we describe that as distortion and there is a specific way
to address that.


Wait; we're no longer in reproduction-land here; were talking
about intentionally torturing waveforms to some nefarious end.

I bet there are more devices for sale now that add distortion than
try to eliminate it. "Warm" is a code word for distortion, after all.

And yeah - various sort of nonlinearities have varying degrees of
success in being modelled. At some point, the sheer face of the
math keeps the merely interested largely out of it.

Lots of specific techniques, but which is "best" is not clear.

There are a lot needed because there are a lot of different effects in the
system that need to be modelled.

I suppose somebody could prefer this version or that version, but
it's just a preference.

Objectively I can say that this function matches the original better than
this other function. At some point we have to make some subjective evaluation
once we get to the point where it doesn't match the original very well, but
the goal is to stay away from the point where there is any need to worry
about that.

So what really happens is that you end up with an ... electric piano,
which is not really a piano at all. Then there are emulations of
electric pianos, and those become ...resepctable* and on and on.

"[expletive deleted] and ugly buildings get respectable if they last
long enough." - Noah Cross, "Chinatown".

--scott


--
Les Cargill

[Scott Dorsey]

Les Cargill wrote:
Scott Dorsey wrote:
Les Cargill wrote:
Scott Dorsey wrote:

That "graininess" is distortion. There's a lot of it. That distortion is
not a linear function and therefore requires modelling if you want to
effectively model the device.

I still think of it as more like something like chiaroscuro for sound.

If you're doing the math, you think about things that are linear and things
that are nonlinear. If frequencies are being generated that weren't in the
original, then we describe that as distortion and there is a specific way
to address that.

Wait; we're no longer in reproduction-land here; were talking
about intentionally torturing waveforms to some nefarious end.

We never were. This is a thread about tape echo machines.

I bet there are more devices for sale now that add distortion than
try to eliminate it. "Warm" is a code word for distortion, after all.

Probably, although when the Echoplex was new, a lot of effort was put into
reducing coloration. Now the coloration is what people like about it. Go
figure.

And yeah - various sort of nonlinearities have varying degrees of
success in being modelled. At some point, the sheer face of the
math keeps the merely interested largely out of it.

The nonlinearity isn't so bad, a lot of the issue is that on top of the
time-invariant nonlinearity you also get all that FM modulation.

So what really happens is that you end up with an ... electric piano,
which is not really a piano at all. Then there are emulations of
electric pianos, and those become ...resepctable* and on and on.

"[expletive deleted] and ugly buildings get respectable if they last
long enough." - Noah Cross, "Chinatown".

It's that kind of world, yeah.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."