noise category definitions, impulse noise, and workflow

[jfritz]

I've been using DCX since DC3 or maybe 5. It's been long enough to exceed my limited memory capacity. I've always assumed that clicks are short bursts that are easily visible in the amplitude vs. time display, crackle is discrete sounds very close together like popcorn popping rapidly or running the edge of a playing card over coarse sandpaper, and hiss is, well, hissing. Lately I've been wondering if I've made some bad assumptions. I have a terrible time with impulse noise. Interpolation is about the only method I've had any luck with to eliminate clicks. I'm not sure I've ever done much with crackle. Neither the expert nor EZ impulse noise tools work well for me. OTOH, I have, what seems to me, wonderful results with the CNF. I'm assuming that either I'm doing something wrong with impulse noise or my definitions aren't correct.

I don't work with 78s. My main interest is pre-WWII blues mostly from CD reissues and some LPs. I've had very little trouble with post 50s, i. e. high fidelity--LPs of my own but the stuff dubbed from 78s is a lot harder. I know that a lot of these recordings don't meet the extremely good copy criterion for source material. A lot of it, particularly the Document recordings, seems to be dubbed directly from 78s with minimal or no noise reduction. My general work flow is: 1. Convert to mono. 2. Raise the gain to about -.5 dB max. 3. De rumble. 4. Try some impulse noise reduction. 5. Do a CNF noise sample from what's there from the lead in or lead out and use two stage--low then high--CNF processing. 6. Try some enhancements, sub-sonic, harmonics, punch & crunch, and/or exciter. I often don't use any. 7. Put the file back into stereo and call it done.

Is there something else I ought to be doing?

Thanks

[Marc Hildebrant]

I have had poor results with using the Impulse filters AFTER the conversion to Mono.

Try using the EzImpulse on the raw stereo recording and CNF on the Mono version.

Marc

[Craig Maier]

I am guessing that your results are very poor when using CD re-issues, and for reasonable technical reasons. The probable reason for that is because of the fact that the CD producer has already "killed" the top end of the spectrum which the impulse filter relies upon to work its magic. Many CD re-releases of old 78 have everything above around 5 kHz chopped off, which is rough on the EZ Impulse filter discriminator system. It wants to "hear" signals all the way out to the top of the audio spectrum (at least 20 kHz). The impulsive noise energy extends to the limit of the spectrum, and the system needs all of that signal.

Also, what Marc says is also true, but you will not have a true stereo signal (both groove walls) availabe to you via a CD; it already would have been summed to mono - - - this you can not derive the benefit of separate left and right groove wall de-clicking.

As for LP's, if they are transferred properly and no bandwidth limitations have been applied, it should work fine - - - unless - - - the LP is a re-release of 78's wherein the LP producer killed the top end of the spectrum. To make matters worse, LPs of old 78s will include all of the impulse noise of the 78 source (they did not have de-clikcer/impulse filters back then) intertwined with the impulsive noise of the LP itself. It becomes sort of a convoluted mess in this case (78 impulsive Noise, 78 Limited bandwidth due to the re-release on LP, one groove wall available only from the 78 original, and the LP noise added on top of everything).

Some of this can easily be determined (top end of spectrum killed) by viewing the signal on the Diamond Cut Spectrograph. Check it out - - - I will bet there is nothing above around 5 kHz on those re-releases.

Craig

ps - as an aside, when I am working directly with old 78s directly, my de-clicking results are almost perfect. Please goto our website and listen to Novelty Tunes of the Roaring 20s Volume 2 which I personally restored in 2013 using the DC8 EZ Impulse filter. There are 20 snippets there to listen to. I doubt that you will hear a single impulse (and the raw material was loaded with clicks, ticks, crackle, etc).

It can be heard via the following link:

http://www.diamondcut.com/store/inde...8c9b996434cbcb

[Marc Hildebrant]

Craig,

Good points and nice analysis.

With regard to an LP version of 78 records...Yes Indeed! I was able to clean up some of those records, but what a mess. The noise was convolved with the effects of the original recording, the first clean up, and then my recording of the "noisily LP".

The restoration was difficult but it can be done.

Marc

[Dan McDonald]

Sometimes I have had good luck with adding harmonics and expanding the high end of things. using the VVA and the ez enhancer can re-emphasize the clicks and make it a little easier for the software to find them. At least I think it works sometimes.

[Craig Maier]

There are also pre-emphasis curves found under the Paragrapahic EQ (presets) along with their compliments (de-emphasis). One is 75 uSec and the other is 25 uSec. I have never tried using these curves in this way, but perhaps adding some pre-emphasis before de-clicking and then reversing that with the same time constant de-emphasis thereafter may help. Again, I have never tried de-clicking material that had already been re-mastered via CD or via LP. So, I do not know how effective using the paragraphic EQ in that way would be. Maybe a combination of things might help (use what Dan suggested and maybe my idea too). Ultimately, I do not think that you will be able to achive the equivelant results as a direct transfer from the shellac disc, however. Just my hunch.

Craig

[Marc Hildebrant]

I have found that the use of the "flat preamp" also helps to add in much high frequency information for the impulse filters.

For 78 RPM records, the equalization has only turnover effects or low frequency effects, so that the EQ can be done before or after the impulse filter (seems to make sense).

Marc

[Dan McDonald]

With the flat preamp, it should be better for the lp clicks also, so that would help, since you'd have two levels of clicks - the originals and the lp versions. I've found declicking before equalization was very good also, although you have to be a little careful not to get too aggressive. Then, when you eq, what is left over basically disappears.

[Craig Maier]

Quoting Marc:

"For 78 RPM records, the equalization has only turnover effects or low frequency effects, so that the EQ can be done before or after the impulse filter (seems to make sense)."

-------------------------------------------------------------------------------------------------------------------------

That makes sense to me too.

Craig

[Craig Maier]

LP re-issues of 78s probably would have followed the following basic process:

1. Transfer the 78 to magnetic tape via a "turnover-only" preamplifier set for Mono (Left plus Right)

2. Mark the big impulses on the tape with a grease pencil and scratch the oxide off of the tape near the marked area with an exacto knife.

3. Reduce the remaining noise of the 78 by limiting its bandwidth to around 5 kHz (systems were quite basic back in those days - no sophisticated filters were available)

4. Create a Master tape for the LP.

5. Cut the master through an RIAA based lathe

Craig

[jfritz]

The frequency limitation is probably it. While I can see information beyond 5K , the rolloff is pretty dramatic and by 10K it's below -90 bB.

[Craig Maier]

That is a pretty steep rolloff; basically, you are seeing no information above 10 kHz. The software "enjoys" those signals all the way past the top of the audio spectrum to discriminate between music and impulsive noise. Sounds like they may have "brick-walled" above 7 kHz? Using pre-emphasis may help a little - - - worth a try, anyway.

Craig

[Marc Hildebrant]

5 KHz ? Wow....

When I hear songs that are limited that much, they tend to sound "flat" or "dull" to me.

I suspect that the 5 KHz limit comes from the standard AM radio limit of 5 KHz due to channel spacing on the radio dial (10 KHz).

There is still important sound info present above 5KHz even for the Acoustic Recorded records.

Marc

[Craig Maier]

5 kHz is quite low. When I perform electrical 78s restorations, I often leave the bandwidth pretty open up to around 7 to 12 kHz, depending on the material. And, when I do roll it off, I use a gentler slope (12 dB/oct).

Craig

[jfritz]

I've tried preemphasis and a meat axe approach with my own eq settings but haven't gotten back enough highs to make de-clicking work any better. My goal with a lot of this stuff is not to make the recordings sound the way they "ought to," but to eliminate enough of the noise to make listening more worthwhile. A lot of the blues recordings have only one known copy. Many of these are worn almost beyond being usable. With the CNF and some other tools I can get rid of enough of the noise so I can understand lyrics I couldn't before. I'm beginning to think my goal is forensic. <g> Naturally I'd love to produce something like the Archaeophone King Olivers, but I'm happy to hear the lyrics.