Improved FIltering For 78 's ?

[Marc Hildebrant]

Group,

When I clean-up a 78 record (electric recording), my current method is to :

1. Record at 44.1 KHz 16 Bits

2. Use EzImpulse and various continuous noise filters.

3. Enhance with various effects like the VVA


I have found that the adjustment of the EzImpulse filter is critical to removing the noise without making brass (trumpets) sounds distorted...even with the solo\brass box checked.

Last night, I tried a different approach that seemed to help.

Before the usual multifilter that I use for 78's, I did a pre-filter. The pre-filter is to :

1. Convert to 96 KHz. 16 bits
2. Run the EzImpulse that I use for LP's
3. Convert back to 44.1 KHz

This pre-filter seems to clean up a lot of the noise and thus let the 78 filters that I used in the past work better and not have as much distortion with the brass. While the best method may have been to make my initial recordings at 96 KHz, I didn't and since I have 6000 plus songs....I will not be re-doing the original recordings.

We have talked in the forum before about how the impulse filters work better at 96 KHz, and it seems that the DC8 software and make a pseudo 96 KHz sampling pretty good from the 44.1 KHz material.

Also, the ratio between 44.1 and 96 isn't an integer, maybe that even better for some reason ?

Comments ?

Marc

[Craig Maier]

Hi Marc,

I wonder how well it works if you change it to 48 or 88.2 KHz rather than 96 kHz?


Craig

[Marc Hildebrant]

Craig,

Different sample rates would be different....maybe down the road some I will try.

Others can too !

Marc

[Craig Maier]

Marc,

I understand why higher sample rates would produce better results with impulse noise because the bandwidth of the system will be higher. That would better preserve the harmonic content of the noise impulses, yielding a better signal/noise and thus better discrimination. I can only guess that up-sampling from a lower sample rate produces interpolations which may be useful to the algorithms discrimination ability, but I really am not positive about it. It's certainly interesting.

Craig

[Craig Maier]

Marc,

Can you post or send me your multifliter so I can try your experiment? You can post it on the presets sharing section or send it some other way, whatever is easiest for you.

Craig

[Marc Hildebrant]

Craig,

Sent via E-Mail...

Marc

[Craig Maier]

Marc,

Thanks,

Craig

[Craig Maier]

Marc,

Quick question - are you applying the proper 78 turnover curve (with the Virtual Phono Preamp) before any of the other processes? Ideally, that is how the system is designed, not to say that it will not work other ways too.

Craig

[Marc Hildebrant]

Craig,

The 78 turnover curve is applied in the 78 multifilter.

Recordings are "flat" and the first filter at 96 KHz is on the raw, flat recording.

Marc

[Craig Maier]

Marc,

That's what I am wondering about. That may work, but the system was designed to have a balanced audio spectrum applied to the impulse filters. Not sure what happens when it is weighed more on the top-end. The impulse filter has several control loops and they interact with one another, so it was designed with "balanced tone" so to speak to operate the way that we intended. As I said earlier, it may work other ways too, but it may also do some unexpected things.

Craig

[Marc Hildebrant]

Craig,

My thought was with 78's and no roll-off, the EQ on or off wouldn't matter. Is that wrong ?

Marc

[Craig Maier]

Marc,

That is incorrect - - - one of the loops is a leveling loop that relies on an even energy distribution across the audio spectrum. That loop will be effected by a change in the applied tonal balance. Perhaps there is a way to fool it, but it is looking for a balanced spectrum else the leveling loop will react differently than anticipated. There is another control loop (that I can not describe for ip reasons) that also involves a normal eq across the entire audio spectrum and involves variations in other impulse filter parameters.

So, I am not too sure exactly what happens when non-equalized (ie - without turnover applied first) electrical 78s are processed via the EZ Impulse filter.

Craig