Your Vinyl Transfer Workflow (sharing best needledrop practices)*

Perhaps you can post a sample of a record that we likely all have. I'm reworking my Pink Floyd dsotm a 8 year old reissue and it has some minor sibilance in a few spots. I think it would be hard for my TT to slow down that much to try it plus I'm talking minor tracking issues it might not even be worth the effort. About 60 seconds into this sample there is some vocal sibilance I don't know if your method would solve it.

Dropbox - Allen Toussaint the bright mississippi.flac - Simplify your life

 

Actually I was just going to comment on harby's posts about half-speed needledropping. I recall reading about it about a decade ago from someone who has success using it for taming the sibilance in the 30th anniversary reissue of DSOTM. I've tried it a couple of times over the years, but didn't notice much of a difference. There are so many factors that affect vinyl playback, not the least of which is stylus type, setup, alignment, tonearm damping, etc., that maybe I just didn't have the right combination of factors to get noticeable results. The theory certainly makes sense.

 

Your filter experiments are interesting. I used Brian Davies' free Equalizer program when I tried half-speed needledropping.

Here's something you might look into since you seem to know your filters! Many albums we needledrop have been cut using something called an elliptical filter to sum lower bass the mono. Since this was an analog filter, it also affects the lower midrange, in the Side Channel, and narrowing the perceived width of the soundstage. I think this is one reason a lot of people found CDs so jarringly different when they first came out (aside from bad early A/D and D/A converters). Someone created a filter for Audacity that does an amazing job of replicating the elliptical filter. What I'd like to do is find a way to reverse the effect of the elliptical filter to restore the original width of the master before it was run through the elliptical filter when cutting the LP. I've had some success with iZotope Ozone's EQ in analog mode applying a low shelf boost to the Side channel, but at times it's just not quite right as it overemphasizes the ambience and width, depending on the recording. However, when it works, reversing an elliptical filter can really open up the sound of a needledrop not only in the low end but also well up into the mids.

Here are some technical details: The Secret Society of Lathe Trolls

 

Short of installing Audacity and that filter, what are the exact settings you use in Ozone?

 

Actually the Audacity filter won't reverse the filter used during vinyl cutting; it only emulates it. Here's what I use in Ozone:



Note that this is in Mid-Side mode with the curve only in the Side channel. If this causes too much bass boost in the resulting file, I add a second band to cut down on the lower bass:


These were mostly arrived at through trial and error, just comparing the width and placement of instruments between needledrops and digital versions until the needledrop with EQ seemed to have about the same placement of hard-panned sounds.

 

That is the best decrackle plug in I've used. I don't like the Waves X-Click, though.

 

An elliptic filter is simply a filter with a higher Q, one with more peaking. Also generally known as Chebyshev; contrast to Bessel with lower Q. While the poles of a Butterworth filter (Q:0.707) lie on the S-plane unit circle, higher Q lies on a squished ellipse.

Saying "elliptic" doesn't tell us much about the filter. A 2nd-order Chebyshev 0.01dB ripple filter has a Q of 0.724 (similar to Butterworth); one with 0.25dB passband ripple is Q: 0.8093. My filter graph some posts back shows the result of higher Q.

A second order elliptic filter (as described being used in this mastering processor) has one ripple peak to go along with it's faster cutoff slope (higher filter orders, that don't seem to be employed in mastering, actually have multiple ripples.) If there was such a peaking filter applied to the side channel, it would actually increase the width above the cutoff frequency.

In fact, according to the table in the linked post, the Neumann bass filters have no "peaking" - translation: non-elliptic.

-

The link is about mastering M-S processors, specifically the design of one, and as such, talks in terms of results: crosstalk, vertical modulation, etc. One has to decode the language. Simply, it is extolling the benefit of a 2nd order filter vs a 1st on the side channel.

Attempting to reverse such bass-summing filtering and potential effects on higher frequencies is problem-fraught:

- the bass-summing used in mastering is for limiting the amount of vertical modulation; it is not an encoding-decoding scheme like RIAA EQ,
- while RIAA de-emphasis reduces the amount of surface noise, reversing a bass-summing filter increases the amount of out-of-phase noise caused by the medium, noise we've already discussed is beneficial to remove,
- its use is unknown. We can't reverse an unknown ripple without knowing the frequency and Q. The link's table even indicates dynamic processing on a Neumann 84, where the frequency changes as demanded, like a limiter - impossible to reverse.
- the filter has a phase response, which causes M-S cancellation and non-cancellation when summed back


We talk about the stereo image on vinyl. We primary have two factors in play on playback:

- crosstalk from the cartridge, usually about -20dB, that narrows the stereo image
- distortion that differs between channels, noise that increases the apparent width (and messes up imaging)

You can't increase the former (as simple as turning up the side channel a bit) without increasing the latter. Reversing a bass-sum filter has more potential ill-effects than a little stereo widening to correct crosstalk.

Then - do we need to widen this bass anyway? We have a mono subwoofer in our room that crosses over at 85Hz. Our ears don't have the distance between them to distinguish the phase of bass, and the magnitude of the pressure is little-reduced by needing to travel to the other side of our head. Put on headphones, and that low piano note only in your right ear couldn't be produced by any real piano.

I say: narrow the bass even more yourself in the side channel. Your own highpass; not reversing theirs. We want to reduce pops in the bottom of the groove, not emphasize them. We don't want to warble our woofers more from up-and-down warps.

By increasing the filter Q you use when applying highpass to the side channel, you can, however, create a hump above the cutoff that does do some stereo widening; which according to "crosstalk" graphs of the mastering EQ, should still of very low magnitude to not go into artificial widening.

If it is stereo presentation you find lessened, start first at a touch of "stereo widener" to compensate for the crosstalk of vinyl.

 

Thanks! I replicated the settings and made a preset. I tested it and it sounds good. But, after I read @harby's post, my head is spinning. I gotta start getting more sleep!

 

Don't sweat it. This was an experiment that produced some results I find to be quite nice. Let's not get bogged down too much in filter theory.

 

Actually, rather than address all your points, I'll simply provide some links that explain this better than I can.

I read about the Audacity add-on that emulates an elliptical filter and tried it. I found that it was quite amazing how applying it to several CDs of material I also have on vinyl resulted in perceived width that lined up really well with the vinyl. Note that the author of this filter based it on a the Neumann EE module, which is a first-order 6dB/octave filter.

More here: https://forum.audacityteam.org/viewtopic.php?t=54088

Here's the script of the filter:

;nyquist plug-in
;version 1
;type process
;categories "LV2 "
;name "Bass to Center..."
;action "Mixing bass to center..."
;info " By Jvo Studer <[email protected]>. Released under GPL v2.\n\n Frequency-selective filter to crossfeed (mix) bass frequencies to center (mono).\n Works only with stereo tracks."
;control eeqfc "Crossover Frequency" int "Hz" 150 10 500
;control level "Bass Feed Proportion" int "%" 95 20 100
;control gcomp "Bass Boost" real "dB" 0.5 0 6
;; basstocenter.ny by Jvo Studer, V1.1 March 2011
;; Released under terms of the GNU General Public License version 2
;; GNU General Public License v2.0 - GNU Project - Free Software Foundation
;; Simulates the "Elliptic EQ" filter found on vinyl mastering
;; consoles by means of a first order highpass filter in the
;; side-channel (L-R difference).
(setq eeqfc (max (min eeqfc (/ *sound-srate* 2.0)) 1))
(setf factl (* level -0.005))
(setf factr (* level 0.005))
(setf fcomp (* eeqfc 0.74))
;; elliptic EQ
(defun eefilt (sig eefc)
(lp (diff (aref sig 0) (aref sig 1)) eefc))
;; bass shelf EQ
(defun eq-low (sig fl gain &optional (slope 1.0))
(multichan-expand #'eq-lowshelf sig fl gain slope))
(if (arrayp s)
(eq-low (vector
(sum (aref s 0) (prod factl (eefilt s eeqfc)))
(sum (aref s 1) (prod factr (eefilt s eeqfc)))) fcomp gcomp 0.5)
(format nil "Plug-In Error:\nSelected tracks must be stereo."))
​

I'm not really a programmer so most of this is beyond me, but I would really love to find a way to reverse this. Yes, I know vinyl has crosstalk, but it's not linear. Most cartridges specs how much wider separation at 1kHz then they do at 10khZ for example so applying a stereo width enhancer doesn't really counteract the effect of narrowing the lower midrange from the use of an elliptical filter when cutting vinyl (which I've read in a couple of places online amounts to a reduction of 6dB in width). You can clearly see the effects of this if you convert a file to Mid/Side then compare the frequency curves of each channel. The Side will be lower in amplitude but on more recent audiophile releases such as stuff done by our host, the distance between them remains more or less the same. On older LPs however, there's a clear roll-off starting in the midrange and dropping gradually down and then really dropping off in the bass region.

Here are a couple of other useful links.

Elliptic filter - Wikipedia
Elliptic Equalizer Boards and Kits
Maselec MTC elliptical filter

 

All I know is that when I applied the settings to a couple of needledrops, they sounded fuller. Sure, it increased the bass content, and I personally did not detect and separation of the bass frequencies, but it added a certain presence.

It seems to me that once the lower frequencies are summed at the cutting stage, you can't really return that to what it was before.

 

If you can find something with material that's hard panned, you'll notice it's panned even wider.

The bass isn't completely summed to mono, except at the very lowest frequencies. Where reversing the elliptical filter helps is in the lower mids as it sort of restores something closer to the original width.

 

What we have in mastering is NOT an elliptic filter from the standpoint of filter design. It has nothing to do with the poles and slopes and ripples of a filter.

Instead, it is perhaps a misuse, early use, or mistranslation from German. Here, I think "elliptic" originates from the stereo matrix: phase display or a goinometer display becoming less circular as the side channel magnitude decreases:



We see it used in German, on the name of an EE77 Mastering equalizer



Here, as on others, it is a 6dB slope single-order Butterworth highpass filter applied to the side channel (actually bass subtracted from both channels out of phase). The Audacity "mastering plugin" is also simply a 6dB lowpass Butterworth filter done to the side channel, and an unsubstantiated optional bass boost.

The problem is the massive amount of gain required to reverse it - becoming infinite gain at 0Hz.

While below I show the 150Hz highpassed green signal, and the missing audio that would reconstruct the full bandwidth (like a speaker crossover), the filter we would need to apply for a "correction" is like the green filter turned upside down in dB. It's not your "shelving" filter.



You'd be inserting the highpass filter again, but as a negative feedback stage, or multiplying by the correction needed.

You can imagine turning up the out-of-phase information at 50Hz by 20dB - multiplying it by a factor of 10, is going to result in massive rumble, especially since bass is already boosted by the RIAA curve.

And even if you know someone used exactly an EE77 clone and not Kirkwood's 12dB/oct slope processor, good luck guessing the setting:



One might draw just a little bit of the inverse curve into an FFT equalizer and apply it to the side channel, but that doesn't correct the phase, which might be the only place with actual benefit.

 

Could be. According to the KA-Electronics website I linked to in my last message:

Yes, I realized that early on when experimenting with this. However, I don't really want to "de-sum" the low bass frequencies, I just want to try and counteract the narrowing of the lower midrange as this has the most impact on the soundstage due to the elliptical filter being analog. As I wrote before, the shelving filter I did in Ozone was the closest I came to achieving soundstage positioning in my needledrops that seemed close to hires and CD digital versions of the same material. It might not be completely accurate, but for most material, I like the results. As Grant mentioned, it has a certain presence. I also find it often makes cymbals sound less congested and "digital." Go figure!

Agreed. This one reason why I'm a big proponent of denoising in M/S and focusing in the noise in the Side channel because that's normally where most rumble lives. When I convert to M/S and check the RMS values, there's typically a 7-10dB higher level in the Side channel. Just reducing that so that the RMS noise level in the side channel matches that of the Mid channel really helps a lot. As for the boost from my Ozone filter, I've found that with proper denoising of the lows in the Side channel, there's rarely a problem (I rarely try to denoise the highs as that's where denoising really reveals its weaknesses).

 

Inspired by true elliptical equalizers, I've come up with a filter that will deliver maximum passband response, correcting the mastering vertical modulation bass filter, while still giving us the additional out-of phase noise reduction.

First we run the same mastering filter in reverse to cancel it's phase shift. Makes its response a 2nd-order filter with Q:0.7071

We then treat that as one of the 2nd order stages in a higher order Chebychev filter sequence. A 0.01dB ripple 7th order Chebyshev design table, for example, lists Q: 0.7028, 1.3798, 4.5208, and then a 0.48x Fo single pole.

Not ripply-enough for correction. Some tweaking and I make a pretty flat filter. But dang, I also need to run the identical in reverse. Have to go beyond the rules..

Here's where I'm at simply with a 4th order the same frequency and a 4th order half the frequency of the "encoder" with hand-adjusted Q. The solid is the roll-off before (here a 150Hz setting), and the green line is the roll off-after - flat response much closer to the cutoff. The dark area under green is where we've restored the stereo.


Wrote a batch file to again do it as one command to a stereo wav. Simply type in the guessed turnover frequency.

That's still conventional thinking. I can probably use peaking filters at lower frequency ratios to bring up response below the turnover, while still maintaining the goal of a strong side-channel high-pass that cuts vinyl artifacts...

Of course if the mastering didn't use such a filter, only kicked in to limit peaks, or has a lower F3 setting, you get too much widening. And we're still losing stereo above a subwoofer's crossover.

 

Very cool. Keep at it!

 

Wow! Sh!t got real in this thread recently. Where's that book on string theory?

 

Yep. We're trying to keep the Beatles freaks from turning this into yet another fab four thread.

 

This thread makes me feel like a simpleton. I don't use such elaborate methods on my needledrops. I just hit the Magic Wizard button and let it fix my recordings.

 

Maybe another 10 pages on setting the level, then?

Another topic, as long as we are peeling back the layers of mastering manipulations to get the music back...I've noticed two different LPs recently where in the middle of a song, the level just drops, and then back up again, and then down... Like they are letting a monkey play with the lathe's volume..

 

Here's an example of strange volume changes - and of the power of just 44.1/8 bit depth to hold vinyl...

wackyvolume.flac (1.8mb)

 

Could be an electrical issue at the cutting lathe. Could have been the recording engineer riding the gain, something producer/engineer Gus Dudgeon did quite often.

On a lot of 80s albums that I have, the first track starts out pretty loud, then suddenly drops in volume anywhere from a minute to through about half of the song. That was a deliberate practice done to make the song jump out at radio programming managers in the belief that the louder volume would impress them. It's the same logic used for using compression on 45s and CDs.

 

Yep, with the right music, 8-bit audio sounds just fine.

 

You and I are in the same ballpark. I like how the vinyl sounds I'm sure some could be engineered better. I don't think I will ever get that good.

I like how it sounds. I'm still stuck on 24/96

Here is my latest LZ 2 the latest reissue. This is from my recently replaced TT and I loved how the drop sounded "today". I know I can que up the old needle drop and it will sound the same. Perhaps my mood is everything today I think this is very good. I'm posting this sample to make a point about declicking the whole file. First I my be wrong but listen to the sample at 7.1, 8.1 21.6 and 23.3 are all drum ticks that sound like clicks so running declick even at level 1.0 removes them completely. I'm also happy with my denoise process. I mainly remove resonance and little else. I use a 3 pass process similar to prior posts customized for every record.

Dropbox - Led Zeppelin 2.flac - Simplify your life

 

The ticks seem like they are part of the music. See here, the first peak 10% in is a drum tick, followed by reverb, while at 90% you have an unheard vinyl tick.



However, I suspect they sound tick-like only in a few instances and have the sharp impulse response into the ultrasonic because of mistracking. See here the uncharacteristic burst of high-frequency noise:



Here's what we've got: perhaps an acceleration that leaves the cartridge behind. It lands again. The plastic of vinyl is elastic and has an ultrasonic springy-ness. You have little high frequency warbles only seen in one channel while the other is smooth. Then the high-frequency peaks after, seen in both channels but about 90 degrees out-of-phase, 8, 9, 10, 11 samples in period (which could also be a 10Khz resonance in a drumstick and not a cartridge.)


Another ticky sounding one. Here I posted both channels. You can see abruptly where one channel doesn't look like the other, followed by little high-frequency ripples.


One must be mindful in analysis that this is not what the groove looks like - a maximum amplitude peak comes about when the stylus is moving sideways the fastest.