Emulate component tolerances and software patents

DSP, Plug-in and Host development discussion.
plusfer
KVRist
61 posts since 1 Apr, 2016

Post Tue Dec 07, 2021 9:36 am

What happens with patents if you want to do something like "TMT" for a saturation plugin, but in another way, for example changing frequencies in the filters, different biases, different parameters to form the waveshaper, etc?

I agree that ideas have to be protected in some way, but patenting something that already exists in nature doesn't make much sense to me, but it's just my opinion.

My question, if achieving the same result but doing it in another way, it would be legal.

DaveClark
KVRist
314 posts since 8 May, 2007

Post Tue Dec 07, 2021 10:04 am

Patents are supposed to cover either structures or processes, but patent attorneys do often try to cover both to save their clients some money. Read the actual claims to determine if you're following all the steps of a process or are building the same structure.

Often at least some claims will be overly broad, attempting to cover anything conceivably like what was actually built, so some judgment about this is necessary because those overly broad claims may be enforceable. Sometimes patents talk about preferred embodiment. This is such an attempt to stretch the claim by saying that it's not limited to the materials or the exact steps mentioned (*). The mere fact that one obtains the same result is not a problem unless that same result is the same structure as was patented, or close enough as to be covered by implied extensions to the claim. If it's strictly a process patent, then one can build the same structure with a different process.

If in doubt, you should consult a patent attorney who has experience in the particular area in question who can advise you as to whether you are infringing or not. If you go to someone without experience in that particular area, you could easily receive bad advice.

(*) Example of preferred embodiment and extensions: "A touching B where A is silicon dioxide and B is silicon. A may also be silicon nitride, insulating oxides of other elements, or other insulators."

User avatar
Urs
u-he
26227 posts since 8 Aug, 2002 from Berlin

Post Wed Dec 08, 2021 1:14 am

It's a fallacy to assume that part randomness automatically means good sound because it's somewhat "alive". So IMHO there's a much greater reason to avoid what the "randomness of part tolerances" claimed in patents, beyond its protection as IP.

To clarify, if part tolerances had really such a big impact, engineers would have known what the individual channels of their consoles were good for. So they'd say for instance "I love the base drum on channel 7, and I always run vocals through group 2". Such that, there would not be any randomness at all in the choices of the engineer or musician. Such that, if you applied randomness for every instance of, say, your channel strip plug-in, you'd completely defy the purpose. It would be randomly either good for the track, or worse, at a 50% chance for either. Such that, the better way of dealing with this would be a set of presets with real world part tolerance levels which are not random at all, but categorised for the kind of program that they enhance: "Here, use this character setting for pads and stuff"

One could argue that minuscule differences in tuning of filters etc. create an overall "less sterile" mix, but I'd argue that even then, randomness is the wrong way to approach this. Instead you'd want carefully chosen differences between the channels whatsoever to guarantee that the effect works as effectively as possible. As an example, in my experience minimal voice detune in a synthesizer works best if thoughtfully chosen for each voice rather than just randomising things a few cent up or down. Making reproducible choices helps to improve the behaviour, which purely random settings can't.

User avatar
Urs
u-he
26227 posts since 8 Aug, 2002 from Berlin

Post Wed Dec 08, 2021 1:30 am

(as for this particular patent, if you know the math of your analogue model, you can simply avoid patent infringement by randomising the parameters directly, instead of the parts. So in an EQ you would randomise the center frequency and Q instead of resistance and capacitance. If doing so can't match the desired result, you simply introduce a trim pot to your circuit, make it a hidden parameter and randomise that. If the patent covers that, then anyone here can provide ample prior art...)

mystran
KVRAF
6758 posts since 12 Feb, 2006 from Helsinki, Finland

Post Wed Dec 08, 2021 1:56 am

Urs wrote:
Wed Dec 08, 2021 1:14 am
One could argue that minuscule differences in tuning of filters etc. create an overall "less sterile" mix, but I'd argue that even then, randomness is the wrong way to approach this.
Randomness is not necessarily the wrong approach, but there are many forms of "random" and not all of them are created equal.

Take an oscillator for example. If it runs at a perfectly stable frequency, it will sound quite "sterile" (which is great if you're going for a modern digital sound, not so great if you're trying to fake a real analog).. but if you throw some randomness into the "component tolerances" here you'll probably end up with what is still an equally sterile oscillator, except now it's slightly out of tune. Similarly, if you randomized the "component tolerances" of the filters of several voices, what you'll probably end up with are things like uneven resonance from one voice to another... or poor tracking, or both.

However, that doesn't necessarily mean "randomness" is bad, but rather that you really need to think about where it's applied. Taking the sterile oscillators as an example, adding a tiny bit of time-varying randomness into the frequency (or preferably phase, to avoid actual tuning drift) of your oscillator will pretty much instantly destroy the "sterile" sound by giving every harmonic some random sidebands, by preventing two oscillators from exactly locking with each other (or an oscillator itself locking with a delayed copy of itself, eg. if you apply a delay), etc..

So my personal take on this is that some carefully thought out randomness (especially randomness over time) can do wonders if you're trying to avoid a modern digitally precise sound... but "random component tolerances" most of the time just make stuff sound objectively worse, without necessarily having much impact on the character of the sound... and just having a per-voice random modulation source for the user to route where they want does pretty much the same thing where desired.
Preferred pronouns would be "it/it" because according to this country, I'm a piece of human trash.

User avatar
Urs
u-he
26227 posts since 8 Aug, 2002 from Berlin

Post Wed Dec 08, 2021 2:10 am

Great point!

plusfer
KVRist

Topic Starter

61 posts since 1 Apr, 2016

Post Wed Dec 08, 2021 5:45 am

Urs wrote:
Wed Dec 08, 2021 1:30 am
(as for this particular patent, if you know the math of your analogue model, you can simply avoid patent infringement by randomising the parameters directly, instead of the parts. So in an EQ you would randomise the center frequency and Q instead of resistance and capacitance. If doing so can't match the desired result, you simply introduce a trim pot to your circuit, make it a hidden parameter and randomise that. If the patent covers that, then anyone here can provide ample prior art...)
I agree, and in the end the same result could be achieved without modifying the internal parts of the filter.

DaveClark
KVRist
314 posts since 8 May, 2007

Post Wed Dec 08, 2021 9:00 am

mystran wrote:
Wed Dec 08, 2021 1:56 am
... some carefully thought out randomness ...
Randomness should never be applied randomly, from the choice of pseudorandom algorithm to the application of it in a particular situation. One of my physics profs and friends made somewhat of a career out of showing how "signals" that had been detected by other folks, who then published papers, were actually signals that were contained in the generator and manifested by interactions. For those with a cavalier attitude about pseudorandom number generators, Knuth's chapter on "Random Numbers" (especially on testing) in his book Seminumerical Algorithms is eye-opening. An important lesson from that it to test the actual application. Pseudorandom means not random. An application may bring out the non-random nature, and things may indeed be made worse.

DaveClark
KVRist
314 posts since 8 May, 2007

Post Wed Dec 08, 2021 9:12 am

plusfer wrote:
Tue Dec 07, 2021 9:36 am
... patenting something that already exists in nature doesn't make much sense to me, but it's just my opinion.
Sorry for not addressing this earlier.

It's possible to patent a novel, non-obvious process that produces granite rocks which we all know exist in nature. One could also patent a process for producing clouds. And so on...

Patenting the structure of granite rocks or clouds would probably be impossible, but I wouldn't bet on it. All you have to do is get past the patent examiner, then defend against office actions (challenges to validity based on prior art, etc.). Being a legal matter that doesn't necessarily have a one-to-one correspondence with reality, there is no telling what might happen.

mystran
KVRAF
6758 posts since 12 Feb, 2006 from Helsinki, Finland

Post Wed Dec 08, 2021 9:55 am

DaveClark wrote:
Wed Dec 08, 2021 9:00 am
Randomness should never be applied randomly, from the choice of pseudorandom algorithm to the application of it in a particular situation.
Oh yeah.. that too.

Nothing is worse than wasting time debugging something that turns out to be a statistical problem with a poor PRNG and it's even worse if the randomness wasn't strictly necessarily to begin with. :)
Preferred pronouns would be "it/it" because according to this country, I'm a piece of human trash.

Return to “DSP and Plug-in Development”