What's under the hood??

Hey,

I just watched a demo of this amp profiler...I guess it's a kind of
system identification algorithm running under the hood? Any guesses as
to what type of algorithm is used to identify the amp transfer
function?

Demo:
http://www.youtube.com/watch?v=NyPgBzqKlUs
Thanks

On 4/3/2012 10:10 AM, Mauritz Jameson wrote:
> Hey,
>
> I just watched a demo of this amp profiler...I guess it's a kind of
> system identification algorithm running under the hood? Any guesses as
> to what type of algorithm is used to identify the amp transfer
> function?

My guess about what he does isn't worth writing down. Just know that, 
since the amps being profiled in general are nonlinear, there is no 
transfer function as textbooks define it,

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

> My guess about what he does isn't worth writing down. Just know that,
> since the amps being profiled in general are nonlinear, there is no
> transfer function as textbooks define it,

Are you saying that you can't come up with a qualified guess?

Yeah, I figured that the amp system is non-linear..but maybe it can be
approximated by a linear system ? And if not, what type of algorithm
do
you use for estimating the non-linear function which maps input to
output?

On 4/3/2012 1:01 PM, Mauritz Jameson wrote:
>
>> My guess about what he does isn't worth writing down. Just know that,
>> since the amps being profiled in general are nonlinear, there is no
>> transfer function as textbooks define it,
>
> Are you saying that you can't come up with a qualified guess?

Sure I can guess, but my guess is not demonstrably better than yours.

Nonlinear circuits don't have a transfer _function_, but they do have a 
transfer _characteristic_. Transfer characteristics don't support 
superposition because of nonlinearity, but in many cases they are 
independent of frequency over a range of interest. Many transfer 
characteristics can be adequately approximated by equations, and all can 
have their essences captured (frequency dependencies and power-supply 
droop aside) in a LUT.

> Yeah, I figured that the amp system is non-linear..but maybe it can be
> approximated by a linear system ? And if not, what type of algorithm do
> you use for estimating the non-linear function which maps input to
> output?

You measure the input/output transfer characteristic using a variety of 
amplitudes and frequencies and go from there. You probably want to 
sample the output at 16 or 32 sample per cycle. From what you learn, you 
construct a model that lets you compute (volts out)/(volts in) for any 
input signal. Auxiliary algorithms can simulate power droop during long 
loud passages. The innovation lies not so much not in what to do as in 
how to do it.

As far as I am personally moved, it has no interest. I learned 
electronics in the 50's building (later designing) home hi-fi sound 
systems. The whole point was minimizing distortion and other 
nonlinearities. I generally prefer the sound of acoustic guitar both 
because I don't need to wear my gunnery-range ear protectors and because 
the barbaric distortion that many guitarists like is absent.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:
> On 4/3/2012 10:10 AM, Mauritz Jameson wrote:
>> Hey,
>>
>> I just watched a demo of this amp profiler...I guess it's a kind of
>> system identification algorithm running under the hood? Any guesses as
>> to what type of algorithm is used to identify the amp transfer
>> function?
>
> My guess about what he does isn't worth writing down. Just know that,
> since the amps being profiled in general are nonlinear, there is no
> transfer function as textbooks define it,
>
> Jerry


People do use the term "transfer function" to describe "graphs
of  input-vs-output". They even talk about "nonlinear transfer
functions."

I don't know if being around guitar amps causes this, or if it's simple
terminology abuse.

--
Les Cargill

> I don't know if being around guitar amps causes this, or if it's simple
> terminology abuse.
>
> --
> Les Cargill

Hahahaha...That's funny Les. Put your glasses on and read my question
again.

Les Cargill <lcargill99@comcast.com> writes:

> Jerry Avins wrote:
>> On 4/3/2012 10:10 AM, Mauritz Jameson wrote:
>>> Hey,
>>>
>>> I just watched a demo of this amp profiler...I guess it's a kind of
>>> system identification algorithm running under the hood? Any guesses as
>>> to what type of algorithm is used to identify the amp transfer
>>> function?
>>
>> My guess about what he does isn't worth writing down. Just know that,
>> since the amps being profiled in general are nonlinear, there is no
>> transfer function as textbooks define it,
>>
>> Jerry
>
>
> People do use the term "transfer function" to describe "graphs
> of  input-vs-output". They even talk about "nonlinear transfer
> functions."
>
> I don't know if being around guitar amps causes this, or if it's simple
> terminology abuse.

Yeah, I read transfer function and thought "output vs input." I guess
context auto-corrected my interpreter to think of that meaning versus
the linear system one. Or maybe it's just abuse, as you say. Al
Sharpton, where are you now? :)
-- 
Randy Yates
DSP/Firmware Engineer
919-577-9882 (H)
919-720-2916 (C)

Mauritz Jameson wrote:
>
>> I don't know if being around guitar amps causes this, or if it's simple
>> terminology abuse.
>>
>> --
>> Les Cargill
>
> Hahahaha...That's funny Les. Put your glasses on and read my question
> again.
>
>
>


I couldn't make it two minutes through the guy's schpeil. I have No
Idea What He's On About. Sorry. It's just a lousy video; the product
could be fantastic, but I'd never guess that from the video.

For guitar amp modelling:

1) There's convolution. It works well enough. I don't know how
he could possibly have done a deconvolution like that.

2) There's waveshaping. it's... okay, in places. Look up
FuncShaper. It about covers it.

Anything else is probably proprietary.

--
Les Cargill

On 4/3/2012 6:52 PM, Les Cargill wrote:
> Jerry Avins wrote:
>> On 4/3/2012 10:10 AM, Mauritz Jameson wrote:
>>> Hey,
>>>
>>> I just watched a demo of this amp profiler...I guess it's a kind of
>>> system identification algorithm running under the hood? Any guesses as
>>> to what type of algorithm is used to identify the amp transfer
>>> function?
>>
>> My guess about what he does isn't worth writing down. Just know that,
>> since the amps being profiled in general are nonlinear, there is no
>> transfer function as textbooks define it,
>>
>> Jerry
>
>
> People do use the term "transfer function" to describe "graphs
> of input-vs-output". They even talk about "nonlinear transfer
> functions."
>
> I don't know if being around guitar amps causes this, or if it's simple
> terminology abuse.

That sort of terminology abuse is unfortunately common. I'm probably 
guilty of it myself from time to time. I don't really appreciate the 
quibble about a Dirac delta not being a function, so I speak of delta 
functions. I'm quite familiar with those graphs of input vs. output from 
the days when I designed audio and RF tube amplifiers. It was natural to 
call them transfer characteristics because they were derived from sets 
of curves called plate characteristics.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On Tue, 3 Apr 2012 07:10:53 -0700 (PDT), Mauritz Jameson
<mjames2393@gmail.com> wrote:

>Hey,
>
>I just watched a demo of this amp profiler...I guess it's a kind of
>system identification algorithm running under the hood? Any guesses as
>to what type of algorithm is used to identify the amp transfer
>function?
>
>Demo:
>http://www.youtube.com/watch?v=NyPgBzqKlUs
>Thanks


There have been a number of papers published on using Volterra series
to model non-linear amplifiers.  They have also been used to add that
Tube amp sound to digital amplifiers.  You might start there for a
better understanding?  I have read a number of papers on measurement
methods used to create a Volterra model of an unknown amplifier.

Here's a good reference.
http://www.amazon.com/Volterra-Wiener-Theories-Nonlinear-Systems/dp/1575242834/ref=sr_1_2?ie=UTF8&qid=1334521232&sr=8-2

Mark DeArman