New guitar effects processor technology: who is interested?

[Jot]

Guitar > no or a few stompboxes > amp is how many people like to play, but there's also a sizeable population of people who buy digital processors. Case in point:

Brian Eno has said that he's a big fan of the Eventide Harmonizer effects processor, mostly because of the user-friendly and intuitive interface: there's a top control layer that's effortless to use, a second layer that provides a few simple options, a third layer for subtler control, and a fourth, deep layer for nuts-and-bolts-level manipulation. He said that that's the only processor he's used that seems to be designed for musicians rather than for software engineers. He wrote a fan letter to Eventide about the unit and later learned that the engineer who designed the Harmonizer framed the letter and posted it on the wall above his desk.

He went on to say that one option he'd love to see on an effects processor is a randomizer button that would scramble all of the settings, allowing the user to audition lots of different sounds without having to go through the tedium of carefully tweaking parameter values. Along with the randomizer button, he said that he'd want a control that would offer the option of constraining the value ranges for further randomization of some or all of the parameters. That way, once he'd stumbled upon an interesting sound, he could start homing in on a final set of parameter values worth saving to a preset. Finally, he'd want an undo option, for backing out of blind alleys.

[horruh]

Guitar > no or a few stompboxes > amp is how many people like to play, but there's also a sizeable population of people who buy digital processors. Case in point:

Brian Eno has said that he's a big fan of the Eventide Harmonizer effects processor, mostly because of the user-friendly and intuitive interface: there's a top control layer that's effortless to use, a second layer that provides a few simple options, a third layer for subtler control, and a fourth, deep layer for nuts-and-bolts-level manipulation. He said that that's the only processor he's used that seems to be designed for musicians rather than for software engineers. He wrote a fan letter to Eventide about the unit and later learned that the engineer who designed the Harmonizer framed the letter and posted it on the wall above his desk.

He went on to say that one option he'd love to see on an effects processor is a randomizer button that would scramble all of the settings, allowing the user to audition lots of different sounds without having to go through the tedium of carefully tweaking parameter values. Along with the randomizer button, he said that he'd want a control that would offer the option of constraining the value ranges for further randomization of some or all of the parameters. That way, once he'd stumbled upon an interesting sound, he could start homing in on a final set of parameter values worth saving to a preset. Finally, he'd want an undo option, for backing out of blind alleys.

I recognise this very much.
I like the layered control approach. I am very well aware of the fact that there are in principle two types of potential users of my product-to-be:

1) those people who are just interested in playing. For these people it is indeed very important to have a very simple user interface with only little control. For this group of people it will be very interesting to hook new developed (by the second group of people, I'll describe those in in minute) high quality effect modules into their box. These modules will be shared on the internet: that's my vision.

2) those people who want to develop novel guitar effects and share (or sell) these among the community. This category of people do not want to be bothered by complicated DSP assembly language: they want to enter mathematical audio acoustic models (which represent guitar effects) in a graphical way, which is then translated by the compiler into code which can be loaded into the effects processor.

If I would rank myself: somtimes I belong to category (1), and sometimes to category (2)

[Hurricane]

RE : So,....... I'll bite

Thanks for your feedback sofar

I am thinking into the direction of a multi-effects guitar processor which is to be delivered as a complete product (containing all nice high-quality and usefull effects, like the ones already mentioned).

On top of that it would be very nice to have kind of an open source community in which guitarists/programmers can develop and share their own exciting guitar effects. These effects can then be loaded into the box, on top of or replacing the standard effects.
I am aware of a few existing products which look like this, but these seem to be very difficult to program.

My technology is very versatile. It is possible to model all kinds of effects. But it can also go way beyond that.

And, Hurricane: yes I am definitely interested in your experience/improvements!

============================================

Judging from the responses I see so far , it looks like your headed in the right direction .

I have a busy plate today and tomorrow ( gigging in a recording session for Mr. Hippy Hippy Shake himself " Chan Romero ) so I will think hard and thoroughly on this issue believe me because it's is the future of guitar and it's sound and it's recording !

Easy and thank you for your interests in my thoughts pertaining to this topic .

Hurricane Ramon

[Chuck H]

As far as 'what's marketable to the average guitarist', (You do need a decent demographic to make this worth while, right?). All I can say is:

Q: What does it mean when the guitar player is drooling from both sides of his mouth?

A: The stage is level.



Caviar sales in the mid western United States are low too. You must consider that the buying public is far more interested in how to play the game and make $$$ than how to program an effects unit. Otherwise they couldn't afford to buy your product in the first place... Hmmm?!?

Marketing is a bitch.

Chuck

[bob p]

I've done more than my fair share of signal deconvolution. (I started back in the Dark Ages when processors sucked and nothing complicated could be calculated in real time). Having a good deal of experience, I'm one of those guys who doesn't want to let any kind of DSP anywhere near my guitars.

As much as a digital guy as I may be, I want analog guitar signal processing. Why? Not because DSP is complicated to me, and not because analog processing frees you from having to think about DSP -- I could do that DSP my sleep. To me the problem is that I know too much about convolution/deconvolution, to the point that DSP convolved musical signals just sound awful to me. I want to put a gun to my head every time that I hear a recording that was made with a POD.

Then there's the learning curve problem -- I don't think that the average guitar player wants to know that a convolution product is expressed as (f * g), that they want to know how to calculate an impulse-response transfer function, or that they'll know enough to understand the difference between a FIR filter and an IIR filter. To use a DSP effects processor, an end user is either going to have to climb a serious mountain of a learning curve, or some serious technology "layering" will have to be put in place so that musicians don't have to have to learn / understand what "convolution" means, so that they don't have to learn about the frequency domain vs. the time domain, Fourier series, Bessel functions, yada, yada, yada. Most guitar players aren't mathematicians and most of them won't want to think in terms of complex (real/imaginary) numbers. For a product to be successful, the math has to be buried in a layer that's far below the user interface.

A large part of the reason that I like analog is because it sounds better to my ears. Part of the reason that I like analog is because the technology doesn't get in the way of making music the way that digital technology seems to. To me, much of the value of DSP is negated by its objectionable tonal qualities and the effort that it takes to use it. Even though I am a complex number array manipulator and a mathematician, I find beauty in an analog stompbox and a valve amplifier. The tone just can't be matched.

Just my $0.02.

[horruh]

I've done more than my fair share of signal deconvolution. (I started back in the Dark Ages when processors sucked and nothing complicated could be calculated in real time). Having a good deal of experience, I'm one of those guys who doesn't want to let any kind of DSP anywhere near my guitars.

As much as a digital guy as I may be, I want analog guitar signal processing. Why? Not because DSP is complicated to me, and not because analog processing frees you from having to think about DSP -- I could do that DSP my sleep. To me the problem is that I know too much about convolution/deconvolution, to the point that DSP convolved musical signals just sound awful to me. I want to put a gun to my head every time that I hear a recording that was made with a POD.

Then there's the learning curve problem -- I don't think that the average guitar player wants to know that a convolution product is expressed as (f * g), that they want to know how to calculate an impulse-response transfer function, or that they'll know enough to understand the difference between a FIR filter and an IIR filter. To use a DSP effects processor, an end user is either going to have to climb a serious mountain of a learning curve, or some serious technology "layering" will have to be put in place so that musicians don't have to have to learn / understand what "convolution" means, so that they don't have to learn about the frequency domain vs. the time domain, Fourier series, Bessel functions, yada, yada, yada. Most guitar players aren't mathematicians and most of them won't want to think in terms of complex (real/imaginary) numbers. For a product to be successful, the math has to be buried in a layer that's far below the user interface.

A large part of the reason that I like analog is because it sounds better to my ears. Part of the reason that I like analog is because the technology doesn't get in the way of making music the way that digital technology seems to. To me, much of the value of DSP is negated by its objectionable tonal qualities and the effort that it takes to use it. Even though I am a complex number array manipulator and a mathematician, I find beauty in an analog stompbox and a valve amplifier. The tone just can't be matched.

Just my $0.02.

You have a respectful background.

I am also educated in signal processing, so I fully understand your explanation on learning curve which is indeed huge if a standard guitar player would want to develop digitized guitar effects.

However, in my professional life I never practiced my signal processing knowledge and it has been 20 years ago since I finished my study on electronics.

Despite this, it did not take much sweat to develop the digitized effects I mentioned earlier in this thread. I did not have to dig in complex mathematics in order to code the algorithms. The essence is that digitization of almost every popular guitar effect is very well explained and described on many corners of the internet. For almost every effect I designed I just had to "re-code" these descriptions in my language.

And if I can do it, than many others can do it as well.....

Still I realize that only a small group of people would be jumping into this game. The majority of the population are the users.

[Steve Conner]

4. I have yet to hear one of these that does a good TS-9.

I think this is because the diodes in the TS-9 give hard clipping that generates a lot of high-order harmonics. The exponential law of a diode generates an infinite series of harmonics, and in a sampled system that'll run up against the Nyquist limit pretty soon. You have four options:

1: You try to model it faithfully and get a lot of aliasing, giving a nasty, inharmonic high end.

2: You soften the clipping to avoid aliasing, maybe using cubics, Chebyshev polynomials or the like. These solve the aliasing problem because they have finite harmonic series, but then you don't have a faithful model.

3. You model it faithfully, oversampling massively beforehand and decimating back down. This is very expensive in processing power: I'm not sure the $5 DSP inside the average Line6 thingy could do it.

4: Some compromise combination of 2 and 3.

Digital distortion is one of the hardest things to do properly, and you really have to understand the non-linear math.

This category of people do not want to be bothered by complicated DSP assembly language: they want to enter mathematical audio acoustic models

I can't really see the difference. A DSP program IS a mathematical audio acoustic model. How would you express an issue like the one I described above in your language?

[horruh]

I think this is because the diodes in the TS-9 give hard clipping that generates a lot of high-order harmonics. The exponential law of a diode generates an infinite series of harmonics, and in a sampled system that'll run up against the Nyquist limit pretty soon. You have four options:

1: You try to model it faithfully and get a lot of aliasing, giving a nasty, inharmonic high end.

2: You soften the clipping to avoid aliasing, maybe using cubics, Chebyshev polynomials or the like. These solve the aliasing problem because they have finite harmonic series, but then you don't have a faithful model.

3. You model it faithfully, oversampling massively beforehand and decimating back down. This is very expensive in processing power: I'm not sure the $5 DSP inside the average Line6 thingy could do it.

4: Some compromise combination of 2 and 3.

Digital distortion is one of the hardest things to do properly, and you really have to understand the non-linear math.


I can't really see the difference. A DSP program IS a mathematical audio acoustic model. How would you express an issue like the one I described above in your language?

For non-linear modeling I use your option (4) with 10x oversampling. My reference is a non-linear differential equation. An example is the following paper, which describes digitized models of the BOSS DS1 and Ibanez tube screamer:
https://ccrma.stanford.edu/~dtyeh/pa...distortion.pdf

Many DSP assembly languages I know of require quite a lot of the DSP architecture and memory architecture details in order to write efficient programs.

My DSP language does hide all these details. Hence, the programmer only has to deal with the DSP algorithm itself.

[Enzo]

I certainly wouldn;t discourage you from trying anything that interests you. But if the question is about marketing it - as opposed to could you pull it off technically - I'd be doubtful.

A lot of guys find three knobs of active EQ too complicated to adjust. And plenty guys don;t understand the difference between guitar-distortion-echo-amp, and guitar-echo-distortion-amp. We techno-geeks are a very small minority of guitar amp-dom.

[bob p]

Digital distortion is one of the hardest things to do properly, and you really have to understand the non-linear math.

That's the take home point. Distortion is really hard to model unless you have a lot of processing power. Even with oversampling it's not a trivial task.

We talked about this a few years ago on the old site when the Line 6 first came out. Back then I mentioned that I thought some of the significant hurdles to getting a good distortion model involved the Gibbs Effect and aliasing. But that was a long time ago when it comes to technology. Many things have changed, but many others haven't.

But maybe I shouldn't be so pessimistic. Realistically speaking, I think that my biases should fade away as processing power continues to improve. When I tinkered with the idea, the processors just weren't capable of doing things in real time. Back then we were using general purpose CPUs, not purpose built DSP processors. But then DSP chips became available and the situation got a lot better. When POD came out, the technology was a lot better than it was when I gave up on the idea. Today the best processors are a lot more capable than what's in the POD. Tomorrow processors will be even better. Someday we'll have a new reason for heat rising out of our amps -- and it will be coming from the CPU instead of a set of tubes. <shudder>

I'm sure that someday there will be enough processing power to accomplish much more musically appealing convolution products. IMO even when processors make the computational tasks easy, the biggest hurdle will remain: designing an easily tweakable package that the average guy who isn't an EE will be able to understand and embrace. As processing power gets better, we'll probably end up complaining less about the underlying SDP and focus more on the user interface. If you can build a good user interface on top of good sounding DSP, I think you'll have a winner on your hands.

[bob p]

Digital distortion is one of the hardest things to do properly, and you really have to understand the non-linear math.

That's the take home point. Distortion is really hard to model unless you have a lot of processing power. Even with oversampling it's not a trivial task.

We talked about this a few years ago on the old site when the Line 6 first came out. Back then I mentioned that I thought some of the significant hurdles to getting a good distortion model involved the Gibbs Effect and aliasing. But that was a long time ago when it comes to technology. Many things have changed, but many others haven't.

But maybe I shouldn't be so pessimistic. Realistically speaking, I think that my biases should fade away as processing power continues to improve. When I tinkered with the idea, the processors just weren't capable of doing things in real time. Back then we were using general purpose CPUs, not purpose built DSP processors. But then DSP chips became available and the situation got a lot better. When POD came out, the technology was a lot better than it was when I gave up on the idea. Today the best processors are a lot more capable than what's in the POD. Tomorrow processors will be even better. Someday we'll have a new reason to explain the heat coming out of our amps -- it won't be from a set of valves ... it will be from the CPU.

I'm sure that someday there will be enough processing power to accomplish much more musically appealing convolution products. IMO even when processors make the computational tasks easy, the biggest hurdle will remain: designing an easily tweakable package that the average guy who isn't an EE will be able to understand and embrace. As processing power gets better, we'll probably end up complaining less about the underlying SDP and focus more on the user interface. If you can build a good user interface on top of good sounding DSP, I think you'll have a winner on your hands.

[bob p]

1: You try to model it faithfully and get a lot of aliasing, giving a nasty, inharmonic high end.

I think this may be a problem that's disliked more by people with vintage tone ears. I can imagine a future generation of guitar-toting whippersnappers who will be totally unfamiliar with tube amps, and who will embrace the tone of DSP harshness as being musical. It's happened before.

Oh no! I must be getting old -- I used the term "whippersnapper."

[Mark Hammer]

Distortion has remained the hardest thing to model digitally, because it depends on a moment-to-moment reclaibration of the software. DSP does a fabulous job of things that can continue on their own withut any recognition of what the dynamic properties of the signal are right now. So, any filtering or time-based stuff has always been terrific. Compression certainly relies on moment-to-moment detection, but it is applied over a much longer timespan of the signal. Distortion requires application of a VERY complex algorithm on a millisecond by milli9second basis.

Now, certainly processing power is a big part of that, and the use of dedicated DSP chips whose sole function is to produce distortion, has helped immensely. But the absolute bottleneck to any of it at the moment is really the algorithmic description of what takes place to the signal when it is at this amplitude, that amplitude, or had transitioned from this one to that one over such and such a period of time. What happens in distortion circuits is very dynamic. The models are certainly getting better, but they don't seem to have the feel of analog circuits yet. It's like the difference between a very good painted portrait, and a hi-res photograph: certainly recognizable, certainly realistic, but also certainly not the real thing.

[kingjj]

sounds like its going to all be about sculpting an attractive image for a given target. i wouldnt mind having one to play with but like most of the guys here, it would probably collect dust after awhile. the younger players would probably be more receptive as would be sound guys and producers. but for the average player we could care less about the math. you just need to find the right packaging.