Noise from EzImpulse Filter

[Marc Hildebrant]

Group,

I was working on a Edison Diamond Disc Song and found a curious effect.

When I'm applying a filter, I use the classic edit mode, so that if a click or scratch is left behind after the filter, I can look back at the source to find the noise and remove it with the "I" key ( or J/K).

I had a case, where I could not find the noise in the source, but, found the noise AFTER the use of the EzImpulse filter ?
If I brought the "crackle" setting from 30's to "teens" then the extra noise event did not occur after using the filter.

Strange.

At the time that the extra noise was produced by the filter, the song had music with lots of rapid changes.

Anyone else find this to occur ? You have to be careful with the filter settings.

Marc

[Dan McDonald]

Yes, sometimes the interpolation can be too aggressive and you need to back it off a little bit. I was using the impulse filters to reduce some static-type noises the other day and found that some parts worked just great and other parts you had to be too aggressive to reduce the static, and that produced a different kind of noise.

[Marc Hildebrant]

Dan,

Thanks for the comment. At first, I was surprised that a "filter" could cause a noise event; but after thinking about it, I could see the problem occuring with an aggressive setting.

By the way, what is "aggressive" for one type of music can be "mild" for another.

Marc

[Craig Maier]

Hi,

Yes, any of the impulse filters (Expert, EZ Impulse, Narrow Crackle, Big Click) can create an interpolation where one is not needed. As noted, using any of them too agressively causes them to find "signals" rather than "noise". When that happens, it will interpolate the "signal" leaving behind a noise event.

I have found that (with the EZ Impulse Filter) settings of any of the controls above around 80 on almost anything will cause incorrect file analysis. If you want to experiment around, I think that one can reliably see this effect with the Narrow Crackle slider by setting it to around 90 - 95 on most any sort of material.

Craig

[Marc Hildebrant]

As a follow on......

For the Hill and Dale records when played on a stereo cartridge, the common method to produce the vertical signal is to use the L-R method. The question comes up, when do you use the impulse filters ? Before the L-R or after ?

I have found that the EzImpulse filter works very differently on the noise before the L-R conversion vs after the L-R conversion. In my latest filter, the best approach seems to use an EzImpulse filter before and after the L-R conversion and tune each filter. After the L-R event, I have to set the crackle value to 0 ; but before the L-R event, it can be at a much higher setting !

My conclusion is that the L-R computation can change the shape of noise signals since in real life recordings, the noise is not the same on the L (left) channel as on the R (Right) channel...thus the sharp spike on the R side with music on the L can get changed into a "sloppy" spike that the filters find hard to remove.

Note : This line of thinking has a different result for a lateral recording played back on a stereo cartridge and the L+R operation.

Comments ?

Marc

[Craig Maier]

Hi -

Yes, that makes sense to me. Most people are working with lateral or stereo sources; only a few work with verticals. So, with lateral material, it is often beneficial to de-click the two channel signal before summing to mono (L+R). Of cours, with stereo records, it is mandetory to keep the channels divided throughout the restoration process.

But, on verticals, I believe that it is best to convert to mono with L-R first before de-clicking because the inpulse noise signals in the vertical plane are not necessarily mirror images on one another as they are with laterals.

Craig.