CHOW Tape Model by Jatin Chowdhury

[chowdsp]

Just checking - current nightly is v2.8.2. There was no v2.8.1 release... correct.?

Yes, right now nightlies are v2.8.2 (unreleased), but will be switching to v2.9.0 verrry soon. And correct, there was no 2.8.1 release.

[chowdsp]

This filter has the same sound most 24 dB filters do: It's not so about how much low-end it cuts as it does the job properly, but the track feels narrow and boomy somehow.
Try this: Get TDR NOVA and try using a HPF on a track. Listen for the differences between 12 dB and 24 dB. IMO 12 dB sounds more open and transparent.

That's fair, I tried A/B-ing the 12 and 24 dB/Oct filters (I tried both NOVA and the stock Ableton EQ), and I do hear what you're saying, especially once I get myself into more of a "mastering" headspace. I'm curious what creates that phenomenon, my guess is it has to do with the fact that we can hear down to 20 Hz, so when the DC blocker cuts too much of it, we do notice ever so slightly.

I tried switching the DC blocker back to a 12 dB/Oct filter, and felt that was okay. I think some of the stability improvements since I first created the DC-blocking filter have improved the low-frequency rumble situation.

How about this: I'm planning to release v2.9.0 pretty soon. If you try out the latest nightlies, and are okay with them, then I'm happy to do the 2.9.0 release with the 12 dB/Oct filter.

Thanks for sharing your attentive ears!
- Jatin

[heavymetalmixer]

This filter has the same sound most 24 dB filters do: It's not so about how much low-end it cuts as it does the job properly, but the track feels narrow and boomy somehow.
Try this: Get TDR NOVA and try using a HPF on a track. Listen for the differences between 12 dB and 24 dB. IMO 12 dB sounds more open and transparent.

That's fair, I tried A/B-ing the 12 and 24 dB/Oct filters (I tried both NOVA and the stock Ableton EQ), and I do hear what you're saying, especially once I get myself into more of a "mastering" headspace. I'm curious what creates that phenomenon, my guess is it has to do with the fact that we can hear down to 20 Hz, so when the DC blocker cuts too much of it, we do notice ever so slightly.

I tried switching the DC blocker back to a 12 dB/Oct filter, and felt that was okay. I think some of the stability improvements since I first created the DC-blocking filter have improved the low-frequency rumble situation.

How about this: I'm planning to release v2.9.0 pretty soon. If you try out the latest nightlies, and are okay with them, then I'm happy to do the 2.9.0 release with the 12 dB/Oct filter.

Thanks for sharing your attentive ears!
- Jatin

For sure! I'll try to current nightly and compare it to the stable version and other plugins. Let me know when a newer nightly or stable version is up.
Btw, the same "issue" with the filter can also happen with LPFs but I'm not sure if it does happen with oversampling so I can't say anything for sure. I've heard that this change in tone is due to "phase distortion", the steeper the EQ move, the bigger the distortion, and filters are the most agressive EQ moves.

PD: I don't like 6 dB per octave filters either, given that they're even less steep than 12 dB I have no idea why.

[heavymetalmixer]

Diferences between 2.8.0 and 2.9.0 Nightly
1) Oh, so you added a Compression section.

2) I checked with Plugindoctor and the HPF is indeed 12 dB per octave. Also, the Linear Phase Oversampling mode still changes the phase with the LPF, though the phase change is different than with Minimum Phase.

3) Something weird on both: On Oversampling at 1x the EQ curve shows the plugin boosting the high frequencies a lot.

4) The 2.9.0 Nightly seems to use less CPU under exactly the same settings as 2.8.0. For example: While set on NR8, at 16x Oversampling (Minimum Phase) 2.8.0 was at 4.2% of CPU while 2.9.0 Nightly at 3% (I'm using a Ryzen 3600).

5) According to Reaper the 2.9.0 Nightly is reporting more delay in samples.

6) After checking only the high frequencies on some mixes with 12 dB and 24 dB HPFs on EQs, it seems the 24 dB ones are more transparent (a little bit darker) while the 12 dB ones make the sound brighter. This darker color might be the cause if why the 24 dB filters might feel "narrower", as high frequencies are the best way to feel direction and openess.

As to why a HPF is changing the high frequency behavior, I have no idea.

I tried rising the Treble on the 2.8.0 a little bit and it really felt more open than the 2.9.0 Nightly. (Treble from 0.07).


After reading the last point you already know my opinion: Right now the 24 dB HPF actually sounds better (I don't like harsh sounds).

7) Linear Phase Oversampling sounds way brighter than Minimum Phase. I didn't like it.

8 ) I like the Compresion feature, it helps compress peaks a little bit more. Although, the fact that the "Ammount" is basically another Input Gain knob makes the plugin harder to use as the Output Knob would be turned down a lot because of it, plus the moves on Saturation and Drive.

A Threshold-kind of control could be a better idea IMO.

9) Both versions by default have a lower volume than the original track. It would be better if the volume before and after the plugin is around the same, so any users don't become biased by loudness (right now I guess some people don't like the plugin because it makes the sound instantly quieter).

10) I didn't find any bugs on any of the two versions on my tests.

[chowdsp]

Thanks much for the thorough testing and feedback!

3) Something weird on both: On Oversampling at 1x the EQ curve shows the plugin boosting the high frequencies a lot.

Yeah, at 1x oversampling the equation solvers that are used for solving the magnetic hysteresis equations don't converge so well, which causes the high frequency jump. I had actually planned on removing the 1x option a while back because of this behaviour, but ended up keeping it around since it turned out that some people were using that setting specifically for that effect, and I didn't want to break their saved settings.

4) The 2.9.0 Nightly seems to use less CPU under exactly the same settings as 2.8.0. For example: While set on NR8, at 16x Oversampling (Minimum Phase) 2.8.0 was at 4.2% of CPU while 2.9.0 Nightly at 3% (I'm using a Ryzen 3600).

This was actually a big part of the effort for version 2.9.0, since probably the complaint I've heard most often from users is that the CPU performance isn't great compared to many other plugins out there. I've felt this pain myself, since I mostly use a sightly older Dell laptop for my music, and typically can't use more then 4-5 instances of ChowTape in a mix. So I spent a bunch of time working to optimize the plugin processing code. I wrote a blog about it here if anyone's curious.

5) According to Reaper the 2.9.0 Nightly is reporting more delay in samples.

This makes sense since the compression processing adds a little bit of extra delay.

6) After checking only the high frequencies on some mixes with 12 dB and 24 dB HPFs on EQs, it seems the 24 dB ones are more transparent (a little bit darker) while the 12 dB ones make the sound brighter. This darker color might be the cause if why the 24 dB filters might feel "narrower", as high frequencies are the best way to feel direction and openess.

As to why a HPF is changing the high frequency behavior, I have no idea.

I tried rising the Treble on the 2.8.0 a little bit and it really felt more open than the 2.9.0 Nightly. (Treble from 0.07).


After reading the last point you already know my opinion: Right now the 24 dB HPF actually sounds better (I don't like harsh sounds).

Interesting... I have no problem going back to the 24 dB/Oct filter .

8 ) I like the Compresion feature, it helps compress peaks a little bit more. Although, the fact that the "Ammount" is basically another Input Gain knob makes the plugin harder to use as the Output Knob would be turned down a lot because of it, plus the moves on Saturation and Drive.

A Threshold-kind of control could be a better idea IMO.

Glad you like it! The "Amount" knob is actually a mix of input gain, threshold, and ratio in a "traditional" compressor.

The way it works is that if you set the "Amount" knob to 5 dB, the input signal gets boosted 5 dB, but then any peaks above -10 dB (determined by -2 * amount) will get compressed so that a peak which was originally at 0 dB will remain at 0 dB. The idea was to implement an algorithm for "ideal tape compression" so that its function is orthogonal to what's happening in the "tape hysteresis" section. To some extent the "amount" control is baked into the algorithm, so I'm not sure how easy it would be to extract out a pure "threshold" parameter. Besides, I've actually found the control quite useful, since if I set my input signal to peak at 0 dB, I can set the "drive" and "saturation" parameters, and then play with the compression amount, knowing that the peak level going into the hysteresis won't change.

In other words, I sometimes find myself having to turn down the output volume because of the "drive" and "saturation" (see the next item), but not because of the compression amount. If the compression used a "threshold" parameter I feel like I would want another gain knob to control how hot the signal is going into the hysteresis section (whereas I don't feel like I need that knob with the current configuration).

9) Both versions by default have a lower volume than the original track. It would be better if the volume before and after the plugin is around the same, so any users don't become biased by loudness (right now I guess some people don't like the plugin because it makes the sound instantly quieter).

Yeah, volume normalisation has been kind of tricky for this plugin, particularly because of the complexity of the hysteresis algorithm. I do have some ideas around this, but they're going to have to wait for the next version .

Anyway, I'll change the DC blocking filter back to 24 dB/Oct, so the nightly builds should have that once the build pipeline runs overnight tonight.

[bungle]

Do you know the chowdsp website is unreachable, git is still fine.

Yeah, I've been having trouble with my web server lately. It should be back up now.

I'm in the process of migrating to a more stable platform, hopefully that should be wrapped up in the next week or so.

If you ever get stuck again, let me know, we provide free super fast server space for indie developers, no ads, no attribution, just paying it forward

[Hink]

thats really cool of you bungle

[crickey13]

Hello Jatin, how long until the next official update?

[Obsolete462444]

PD: I don't like 6 dB per octave filters either, given that they're even less steep than 12 dB I have no idea why.

I guess there is a sweetspot for the highpass filter slope, where it removes a significant amount of subbass rumble, without affecting too much of the lower frequencies you don't want to remove.

Say for example: 6 dB might not be effective enough (not steep enough), 24 dB too steep, so 12 dB makes sense as an ideal middle ground in that scenario. TDR Nova GE for example has 18 dB / octave filters, which work even better in some scenarios.

[Echoes in the Attic]

Just discovering this and good timing because I was looking at the tape effects in the IK group buy, however those have a lot of latency and this one doesn't seem to. HAs anyone done much testing against the IK tapes? I have tested those much at all.

[heavymetalmixer]

PD: I don't like 6 dB per octave filters either, given that they're even less steep than 12 dB I have no idea why.

I guess there is a sweetspot for the highpass filter slope, where it removes a significant amount of subbass rumble, without affecting too much of the lower frequencies you don't want to remove.

Say for example: 6 dB might not be effective enough (not steep enough), 24 dB too steep, so 12 dB makes sense as an ideal middle ground in that scenario. TDR Nova GE for example has 18 dB / octave filters, which work even better in some scenarios.

After the tests I did yesterday I noticed that 12 dB filters make the track brighter just by setting it ON, and 6 dB filters make it even brighter. Finally I get the reason why I don't like them.
Btw, I'm not sure if the same applies for side-chain HPFs in compressors, like with TDR Kotelnikov that lets you choose the slope.

[chowdsp]

If you ever get stuck again, let me know, we provide free super fast server space for indie developers, no ads, no attribution, just paying it forward

Wow, thanks! That's very kind . I'll be sure to reach out if that happens.

[heavymetalmixer]

Thanks much for the thorough testing and feedback!

Anyway, I'll change the DC blocking filter back to 24 dB/Oct, so the nightly builds should have that once the build pipeline runs overnight tonight.

You're welcome, your plugins are awesome and more people should know about them.
Also, what was the idea behind the Linear Phase oversampling?

[chowdsp]

Hello Jatin, how long until the next official update?

Not long at all! I'm hoping to publish the v2.9.0 release before the end of the week. Right now the main outstanding tasks are to go through my usual testing cycle and sanity checks, then submit a build for the iOS App Store review, generate the official release builds and installers, and then make the release on GitHub.

[chowdsp]

Thanks much for the thorough testing and feedback!

Anyway, I'll change the DC blocking filter back to 24 dB/Oct, so the nightly builds should have that once the build pipeline runs overnight tonight.

You're welcome, your plugins are awesome and more people should know about them.
Also, what was the idea behind the Linear Phase oversampling?

Thanks! I had some users requesting a linear phase option, and since I was overhauling the oversampling menu anyway, it seemed like a good time to add it. I use the oversampling implementation from the JUCE framework which has easy-to-use options for choosing between a lower-latency polyphase IIR anti-aliasing filter or an linear phase FIR filter.