[sdiy] Guy researches and talks about how aging components change the sound of a synth

[Chris McDowell]

I hope me continuing on this side of the discussion doesn't ruffle feathers, forgive me if it does! you can stop reading now if you get the sense your feathers are in for a ruffling. 

I work with AI -all day-. It helps me find components, convert footprints, write code, build web tools, simulate motor coils, push sketches to oled displays (these are just recent examples). I am super familiar with claude's tics because of this, and that's why I'd bet my house claude wrote that guy's post. it truly doesn't matter though! part of my knee jerk reaction is from personal shame and projection: before I realized why I shouldn't, I also let AI write for me. it is part of learning to use the tool, learning what it is good for and what it isn't. speaking for us is not something it is good at, and I think it's not a model limitation but a misinterpretation of its value. we speak to each other to tell each other what we're thinking. telling each other what the model is thinking as a stand-in for our own communication is simply not solving the same problem, and it is a bigger problem still that it almost seems like it is. using AI to format your thoughts IS a way you can increase the effectiveness of communication, and it's a drag that my own knee jerk reactions (and obviously those of other folks on the internet, here and on that reddit thread) render that usage of it highly suspect, too. 

an ongoing challenge I have with claude is that he is certain he knows what is important and will go on long adventures to prove it, even when he is incorrect. my reaction to this particular project was along those lines. this person, who obviously cares about drum machines and synthesizers and DSP and thus makes them family, does not design electronics. claude confidently told them what claude thought was important about the analysis, and this person followed the threads. it is awesome that they got excited and did a bunch of digging, though my guess here from the style and results of the output is that claude did the digging. if this person could interpret schematics and time constants, they would likely be familiar with component tolerance realities, too. 

if indeed this was original analysis and they used claude to translate it for them, then they sort of misled themselves and claude added insult to injury by presenting their findings in an even less credible way. if what happened is what I think happened (they're making a plugin, claude seeded and carried out an adventure, they asked claude "how do I make sure people know about my product?" and claude said "you need to post something interesting about it in a public place like reddit. here, I'll write it") it would have been more useful to everyone and more interesting in general for them to say "claude had a bunch of ideas about this, here is his adventure".  

what mark said is what I felt when I read the post, and how I should have led, lol. 


SO, back on topic. DO components reliably age in the way they're proposing? I could go ask claude but after all this, I'm going outside instead haha. 

Chris 


> On Sep 1, 2026, at 9:21 AM, Gordonjcp <gordonjcp at gjcp.net> wrote:
> 
> On Tue, Sep 01, 2026 at 05:16:59AM +0200, cheater cheater via Synth-diy wrote:
>>> I kind of feel like his thesis feels like AI, too. Like a drum plugin that traces all behavior back to a component drawn on the schematic, that's something claude or chad would think is important
>> 
>> this has been the defining theme in physical component modelling in
>> virtual analog synthesizers since the late 00's. i know people who
>> build some of the best VSTs of that type and this is exactly the line
>> of thinking that was explained to me, sometime around 2010. long
>> before LLMs existed.
> 
> It's easy to become blind to the function of a circuit when you study the details too long. I've talked to someone who is trying to come up with a really "exact" model of a TB303 filter and they're getting deep into how why it has "low pass filters" everywhere - they're talking about the supply decoupling.
> 
> Yeah, that's not why the 303 sounds the way it does.
> 
>>> Yeah, old drum machines have lots of variance
>> 
>> he's saying the opposite. they don't have a lot of variance. they
>> differ from the schematic, but they all differ in the same way.
> 
> In the article he mentions that a capacitor is 83% lower in value or something. That is well within spec for the components used in a TR909, all Roland's stuff uses cheap crappy components and you have to remember that back then an extremely high-grade electrolytic would be *at best* about 20% tolerance, regardless of what it said on the packet.
> 
> It's kind of like transistors now and then - any two random cheap Chinese allegedly-BC548s you pull out of the bag will be pretty damn close, today, but we used to have to get up to all sorts of tricks to match the really really expensive low-tolerance ones in the olden days.
> 
>>> all of that still reads like I'm criticizing you for sharing, heh, but I'm not! Thank you for sharing. I am criticizing that guy for posting weird AI stuff on reddit
>> 
>> it's easy to fall into the habit of criticizing output that seems like
>> AI to you, but in this case the criticism just doesn't apply.
> 
> It does read kind of AI-y to me as well but it may have been written by someone for whom English isn't a first language. Das erwarte Ich, mein Deutsch ist etwas seltsam klingend für Muttersprachler ;-)
> 
> -- 
> Gordonjcp
> 
> ________________________________________________________
> This is the Synth-diy mailing list
> Submit email to: Synth-diy at synth-diy.org
> View archive at: https://synth-diy.org/pipermail/synth-diy/
> Check your settings at: https://synth-diy.org/mailman/listinfo/synth-diy
> Selling or trading? Use marketplace at synth-diy.org