Latency error in EQ+

[Axyom]

There is a latency compensation issue with EQ+.

When duplicating an audio track and adding an EQ+ on just one track, there is a significant change in the sound, but when adding a time shift of -4 samples on the same track, the phasing disappear. This makes the EQ+ unusable for any parallel processing and in particular mid/side processing.

I wrote to the Bitwig's support and their answer is the following :

" no, this is expected. As the EQ+ is not a phase linear EQ, it is not recommended to be used in parallel processing.

And as it is not recommended for parallel processing, the few samples latency it causes, depending on its settings, are not compensated so that it does not create clicks caused by the delay compensation when adding it during playback."

Which I find is quite a bad answer, considering they do mid/side with EQ-5 on some native presets. Furthermore, I think introducing unwanted latency is a bigger problem than clicks when adding an EQ, which won't annoy anybody imo.

What do you think ?

[] Peter:H []

There is a latency compensation issue with EQ+.

When duplicating an audio track and adding an EQ+ on just one track, there is a significant change in the sound, but when adding a time shift of -4 samples on the same track, the phasing disappear. This makes the EQ+ unusable for any parallel processing and in particular mid/side processing.

I wrote to the Bitwig's support and their answer is the following :

" no, this is expected. As the EQ+ is not a phase linear EQ, it is not recommended to be used in parallel processing.

And as it is not recommended for parallel processing, the few samples latency it causes, depending on its settings, are not compensated so that it does not create clicks caused by the delay compensation when adding it during playback."

Which I find is quite a bad answer, considering they do mid/side with EQ-5 on some native presets. Furthermore, I think introducing unwanted latency is a bigger problem than clicks when adding an EQ, which won't annoy anybody imo.

What do you think ?

EQs if not explicitely stated are not linear phase. They even alter phase relationship in your audio.
Linear Phase EQ on the other hand suffer from preringing. PLease feel free to read here
https://audiohertz.com/2018/03/08/what- ... -phase-eq/
https://www.musicianonamission.com/linear-phase-eq/

[Hanz Meyzer]

I think 4 samples compensation already sound much better, at least better than nothing. I can use most EQs in parallel (the effect is more subtle than EQ+), the effect of EQ+ is very drastical, I guess due to oversampling or so. The oversampling buffer samples at least should be compensated. So it is REALLY only about phase alteration, and not about any kind of buffer anymore! I have the impression that the Bitwig team understood this part in a wrong way, but I did not look into the source

I think it should be improved.

[Hanz Meyzer]

...But maybe each filtertype + oversampling will introduce a slightly different delay, so it simply is impossible kind of, to compensate. Only if all filters would have an extra delay filled up to the maximum delay/slowest filter...

[billcarroll]

...But maybe each filtertype + oversampling will introduce a slightly different delay, so it simply is impossible kind of, to compensate. Only if all filters would have an extra delay filled up to the maximum delay/slowest filter...

Some good info:

As you know, EQ plugins let you boost and cut frequencies in a sound. You can fix and change sounds by turning certain parts of it up or down.

But there’s one major thing you may not know about EQs. EQ slightly delays the sound it’s added to.

While changing the frequency content, EQs naturally add latency to parts of the sound.

This latency creates phase smearing. Certain harmonics get delayed, “smearing” across the sound.

Linear phase EQs make sure the entire sound gets delayed simultaneously.

The phase relationship of the sound remains intact. You don’t have to worry about smearing.

https://www.musicianonamission.com/linear-phase-eq/

[Dionysos]

It's only "just latency" if you don't change any of the EQ's bands, which would defeat the point of using an EQ in the first place. As soon as you boost or cut any of the bands, you'll introduce phase shifts that can't easily be compensated for.

So IMO the current behaviour makes sense.

[stamp]

I think we should all stress support. I prefer having clicks whens inserting eq+ over the phase related problems I'll encounter later.

[Hanz Meyzer]

I still think its due oversampling.

Try yourself with any other EQ, e.g. Pro-Q3, in a parallel container with a dry layer, and see/hear yourself.

[billcarroll]

Pro-Q 3 is a more refined EQ with far more options than EQ+, and Pro-Q 3 has Zero Latency, Natural Phase, and Linear phase processing modes.

Out of curiosity, what are you trying to accomplish by making an exact copy of an audio track? Phase issues with EQ are the tip of the issues Iceberg. Manipulation of either audio track affecting the time domain for any frequency range will likely introduce phase issues.

[Hanz Meyzer]

Pro-Q3 also does compensation in non "zero latency" modes. Since EQ+ also uses a "non zero latency" mode, it should be compensated.

There are phasing effects by filters - OK, this is normal, and can't be prevented.

But there is also a delay added by EQ+ (obviously), and I suspect here the oversampling. This should be compensated. As a lot people already found out, it is around 4 samples. These 4 samples should be compensated. The result then is already eons better than without any compensation.

[stamp]

Pro-Q3 also does compensation in non "zero latency" modes. Since EQ+ also uses a "non zero latency" mode, it should be compensated.

There are phasing effects by filters - OK, this is normal, and can't be prevented.

But there is also a delay added by EQ+ (obviously), and I suspect here the oversampling. This should be compensated. As a lot people already found out, it is around 4 samples. These 4 samples should be compensated. The result then is already eons better than without any compensation.

This

[pdxindy]

Pro-Q3 also does compensation in non "zero latency" modes. Since EQ+ also uses a "non zero latency" mode, it should be compensated.

There are phasing effects by filters - OK, this is normal, and can't be prevented.

But there is also a delay added by EQ+ (obviously), and I suspect here the oversampling. This should be compensated. As a lot people already found out, it is around 4 samples. These 4 samples should be compensated. The result then is already eons better than without any compensation.

+1

[billcarroll]

Pro-Q3 also does compensation in non "zero latency" modes. Since EQ+ also uses a "non zero latency" mode, it should be compensated.

There are phasing effects by filters - OK, this is normal, and can't be prevented.

But there is also a delay added by EQ+ (obviously), and I suspect here the oversampling. This should be compensated. As a lot people already found out, it is around 4 samples. These 4 samples should be compensated. The result then is already eons better than without any compensation.

Are you saying there is a bug and the entire signal is delayed due to latency introduced by oversampling, and this is not being compensated? If so, have you tested this and reported it as a bug?

From the response above from Bitwig it appears they are saying certain harmonics get delayed with EQ+ because it is not linear phase, and as expected this causes “smearing” across the sound. Therefore not good for parallel processing.

When one sound is heard across multiple tracks, a regular EQ can change the phase relationship between those tracks. This is one example when you want linear phase.

[Hanz Meyzer]

No, there is no bug, but the EQ+ is delaying the signal by some samples, I guess due the nature of the algorithm.

I think I already explained this above.

[calltheOperator]

Yeah kind of silly bitwig hasn't solved this. What's even more annoying is that the eq is obviously oversampling, but then depending on your activated bands, you can't get sample accurate latencies even if you try to compensate. You can test with simple phase nulling with the time-shift plugin. go sample by sample and it should do a perfect null at some point, but it doesnt. So theres sub sample latency in there... thanks for oversampling bitwig, it's a fine eq. But also, shit sucks sometimes.

Idk why people are getting this confused with linear phase EQs. REgular EQs cause phase distortion at the specific frequencies effected. That phase distortion is created by the delay network that makes up any equalizer, analog or digital. But any frequencies that go unfiltered arrive with no delay.

But the error here is equal to putting a delay plugin before your EQ between 4-12 samples or something like that. It has nothing to do with linear phase or not. The same latency compensation that keeps your tracks aligned when using a distortion plugin or whatever isnt on this plugin. The delay also changes by the amount of bands you have active. So simple compensation is impossible in a normal workflow. Use two in a chain and double the issue. smh.