CHOW Tape Model by Jatin Chowdhury

[rbn777]

...

[chowdsp]

Haha, glad you're enjoying it !

Basically the temperature knob and the age knob kind of work together. So if your age knob is at 500 years, and the temperature knob is at 90C, I measure wear on the capacitor as though the circuit has been running for 500 years at 90C. If you lower the temperature to, say, 30C, the capacitor will age more slowly and last longer.

If I've done my job right, then for every instance of the plugin, each capacitor in the circuit should have a slightly different lifetime. So maybe for the instance you're running above, the lifetime is exactly 500 years at 90.91C, but at 90.9C, the lifetime might be 501 years. Once the capacitor "dies" of "fails" then you get the random voltage-dependent behavior you described!

[Obsolete236871]

Wow, really intesting the modelling of the impact that age and temperature has on the circuit (diode clipper). I just tried it and I really like the tones it can create. Great for not too sterile sounding clipping / distortion / fuzz.

Could you please add an output gain, so that it's possible to compensate for the increase in loudness within the plugin itself? That would be great!

[chowdsp]

You got it! New build should be available now, at the same link.

[rbn777]

...

[Obsolete236871]

You got it! New build should be available now, at the same link.

[BLUE-S-MAN]

Thank you very much for your work.Would CopyEQ get compiled to Win Vst to?

[rbn777]

...

[BLUE-S-MAN]

Aah,thank you for the hint.

[noiseresearch]

/* whitenoise */

[BLUE-S-MAN]

Maybe a download error, try clicking the dll file and then press the download button.

[mcbpete]

Yeah I had the same thing when I first found out about these wonderful plugins - Seemed to only download about 100Kb of each plugin rather than the correct file size. For safety maybe download/clone the whole directory as a zip (clicking the green 'Clone or download' button here - https://github.com/jatinchowdhury18/Com ... inearities then clicking 'Download zip') then extact out the required dll and/or vst3 files as needed

[groundswell]

I have a stupid question: is it the tape or something else that suffers from hysteresis? If it's the tape, then wouldn't hysteresis apply to each segment of the tape separately, while it's under the recording head? If this is the case, then it would seem that hysteresis would only be relevant assuming a previous signal already recorded onto the tape. Yet from the paper that the plugin is based on, I get the impression that we are computing hysteresis as if we were constantly recording to the same piece of stationary tape. What am I missing here?

IIRC yep, it's indeed the tape material that's hysteretic. The tape is a continuous, one dimensional media that's being magnetized from one point to the next, thus it's the previous magnetization at time (t-1) and before that will influence the magnetization at time t, resulting in a kind of "drag".

[mystran]

I have a stupid question: is it the tape or something else that suffers from hysteresis? If it's the tape, then wouldn't hysteresis apply to each segment of the tape separately, while it's under the recording head? If this is the case, then it would seem that hysteresis would only be relevant assuming a previous signal already recorded onto the tape. Yet from the paper that the plugin is based on, I get the impression that we are computing hysteresis as if we were constantly recording to the same piece of stationary tape. What am I missing here?

IIRC yep, it's indeed the tape material that's hysteretic. The tape is a continuous, one dimensional media that's being magnetized from one point to the next, thus it's the previous magnetization at time (t-1) and before that will influence the magnetization at time t, resulting in a kind of "drag".

The magnetic hysteresis is how a tape (or a magnetic disk drive, etc) stores data (analog or digital) and this is why any decent tape material needs to have a fairly wide hysteresis loop (unlike say the recording or playback heads, where we want a materials with as little hysteresis as possible).

Now, the magnetisation of the tape obviously isn't limited to a perfect "point" below the recoding head, but rather there is always a slightly wider area where the magnetic field is applied, which is why a higher tape speed will give you a better high-frequency response.

However, the hysteresis equations given in the paper have no dependence on tape speed. The paper given as a reference for the time-derivative is about inductors/transformers where the magnetic core material doesn't move over time, hence I don't see why it would be relevant either. If we assume that the time-scales relevant for hysteresis are fast enough that the speed of the tape becomes negligible, then the time-dependency becomes essentially irrelevant for audio frequencies and we might just as well model it as a static non-linearity. If we assume that the tape-speed is non-negligible, we should be working with spatial as well as temporal derivatives.

I'm not questioning the existence or the modelling of the hysteresis as such, but rather how it is applied to the specific case of moving tape. The math used just doesn't make any sense to me and I was trying to ask whether I'm missing some intrinsic assumption that wasn't properly explained in the paper.

ps. I'd also like to complain about the fact that the paper uses the same symbols for different things in different equations is also rather confusing, for example "k" refers to the width of the hysteresis loop, but also the wave number, depending on which equation you are looking at. While it isn't particularly difficult to figure out which definition applies, I'm pretty sure there would have been plenty of spare letters to pick a unique symbols for every equation.

[chowdsp]

I think the fundamental answer to you question is in eqn (7) from the paper (this is adapted from Bertram's "Theory of Magnetic Recording"). Basically at any instant in time, the magnetic domains on the tape are magnetized with a certain amount of magnetic flux (M), by the magnetic field present at that time (H). Bertram's book allows you to use any general hysteresis model to compute this equation. I chose to use the Jiles-Atherton model, which is a good model, but was originally presented in the continuous time domain. The paper you mentioned about transformers was simply used as reference for how to discretize the JA model for use in a digital system.

You're correct that the tape is not just magnetized in discrete points, that there is some "smearing" of the signal inherent in the process (I mention this in section 6 of the paper). I went through the math of this smearing a few months after the paper was published: it turns out the frequency dependence of this smearing is pretty much perfectly equivalent to the spacing loss noted in eqn (13), which makes sense since that equation was determined from empirical measurements by Kadis ("Science of Sound Recording"). There is some phase smearing that happens as well, which can be modelled using high-order allpass filters (I mentioned somewhere earlier in this thread), but I'm still working out an efficient way to implement this, plus I haven't found it to make a significantly audible difference.

Finally, I'd like to point out that my hysteresis model presented in the paper was validated by comparison with measurements taken from the machine the model was based on (see section 5.3). In the months since the paper was published, I've extended this model by measuring and comparing with other tape machines (Nagra, Ampex ATR, various cassette recorders), and found that the measurements align very well with the model (I've done my best to remove the effects of the circuits in the tape machines from the measurements to get a fair comparison with the tape model itself).