Your Vinyl Transfer Workflow (sharing best needledrop practices)*

I’ll put up my entire routine this evening, but I wanted to give a really quick response to this.

I do it with Izotope RX, but I’m sure similar methods should work with any editing software. In RX I do it one of two ways. First, I make sure my levels are set after my declicking and other clean up. That way I have the correct signal-to-noise ratio. I’ve noticed, however, that usually the noise floor of the vinyl is higher than even 12 bits of noise floor. From here I’ll either just dither the file to 12bits, as I’m fairly certain that unless you export it as a 12bit file, it’s still 32bit float with added 12bit dither noise. The other method, and the one I use more often, is to duplicate my file, shift-s to silence it, dither the silence to 12bit so you have the same length of noise as your original file, then copy the noise, and paste it (using mix) into the original file. The nice thing about this method is that you can make the dither noise louder or softer if you feel you need more or less by just adjusting the gain of the dither noise.
I’ll post something more in depth with some samples. I’m really glad others are interested!

 

Yeah, I use RX too. Post your process because if you create a, say, 12-bit file, it's going to automatically be converted to whatever bit-rate your file is. It's not possible to have elements of a different bit-rate within another.

 

My second method works and keeps everything at the 32bit float working bit depth.

Take your entire file and silence it (shift-s). Take that silence, and dither it to 12bits. That gives you a white noise hovering around -100dB. Select all, copy. Then, go to your history state before you silenced the file. Select all, edit - paste special - mix. That will take the 12bit dither white noise from your silenced file, which is the exact same length as the file you're working on, and mix the dither noise in with it, keeping the file at 32bit float. You could also use a copy of the file in another tab rather than going back to a previous history state. Basically you're just copying and pasting noise over your file, you're not changing the bit depth.

This method is good for a few reasons. You can bring the gain up or down on the noise itself if you feel you need less or more. Depending on the source, sometimes you want more. Also, you can noise shape it using the EQ, turning it into blue, brown, pink, or any other noise profile you want. I haven't really done that, except for a roll off at around 16kHz for some older source material.

 

I wouldn't recommend that you brickwall your High Pass & Low Pass Filters. You will introduce phasing and other issues, not immediately perceptible, but noticeable particularly with a good headphone set up. I also think you get more transparent results using the Digital EQ in RX rather than the Analog setting, but that's my opinion. I find that the 12dB/oct curve is the cleanest for Low & High Passes using the RX Eq.

 

I don't use this procedure. I'm with ghostrider - I declick, denoise and press on. I was trying to illustrate better for the person asking about it. I believe BriliantBob uses this but I may be wrong. Speaking of Bob, haven't seen him in awhile.

BrilliantBob - you out there?

 

@BrilliantBob

 

I tried the steps mentioned above from numanoid and I'm gobsmacked by how well it worked, at least in RX8. It really does help remove tons of the noisefloor and artifacts.

 

This means your dither noise was also converted to 32-bit.

 

Whatever it is it works pretty well.

 

But that shouldn't matter, especially since you still want to remain in 32bit float since you're not done processing. You just need the white noise generated by the dither, it shouldn't matter what the bit depth really is, it's just your new noise floor. It's also worth mentioning that the reason why I use 12bits is because Izotope has a limited number to choose from, but 12bits seems to be right where the noise floor of most vinyl and source tapes is anyway. There are a few releases that I could probably use 10bits or 11bits instead. Also, as I mentioned a few posts back, it really seems like if you use the dither function, it says your file is 12bits, but I'm fairly certain that until you actually export your file, it's still 32bit float with dither noise applied. You just need the white noise generated by the dither.

I also forgot to mention that when you use the dither on a silent file, you need to set the noise shaping to none, dither amount to high, and auto blanking to never. Especially that last one, if you have that selected, you'll have a silent file.

Bit depth is just noise floor, or digital hiss. 16bit dither is virtually inaudible which is why I go with 12bits.

 

I'm glad that you had good results with it. It seriously has been a game changer for me, taking all groove noise and crackle out of a recording.

BTW, what's the best way to post some before and after files?

 

Sign up for a Mega.nz account.

 

Okay, here's my method:

The gear I use:

• Pro-Ject Debut Carbon with 2M Blue. I used to have a VPI Scout with an Ortofon Cadenza bronze, and I gotta be honest, I get much better results with this more mid-fi set-up. Is the high end stuff just so detailed that it brings out more flaws? Maybe, all I know is I don't like any of the transfers I did with my VPI. I didn't like the Debut with the 2M Red, but the blue does a great job.)
• Pro-Ject Phono Box. I used to have a Lehmann Black Cube, and it was very good. But, I didn't hear a $900 difference between that and a $100 phono pre. It seems like diminishing returns kick in pretty fast with the mid-fi gear. I'm sure plenty of others will disagree.
• Apogee Duet Firewire. This got great reviews when it came out, it's since been replaced by a newer model, but I see no reason to upgrade. This thing sounds great!
• I record in Izotope for ease of use. Any program should do. Your converter makes far more difference that recording software.
• I record at 24/96 directly from the phone pre into the Duet using RCA to XLR cables, with my setting at +4dB which bypasses the gain adjustment. This seems to be the line level, as 12" singles are hot, long LP sides have lots of headroom. at 24bit you have so much headroom that low recording levels will never be an issue. I actually think 24/96 is overkill especially for vinyl, but all programs work internally at 32bit float (24bit with padding) so you might as well record at that bit depth. I'd probably pick 48kHz, but I like the flexibility that 96kHz affords you if you need to do speed corrections. Also there's a bit of future proofing. As of right now I haven't decided if higher sample rates make the audio editors work better. But maybe in a few years it will matter. Also, 96kHz is the native sample rate of the Duet, so you might as well go for it so it doesn't have to resample internally. I resample to 48kHz for editing. It's much faster processing.

The process:

• Clean the vinyl on my VPI using AIVS 4-step solution.
• Record the vinyl with the above gear and steps
• I record in one long file, flipping sides in the middle. 2xLP I'll do two files.
• Trim the digital silence at the beginning, middle from flipping, and end of the file after the stylus raises. Save this as a FLAC file labelled Raw

Restoration (EQ, Declick, Decrackle):

• Resample the file to 48kHz for faster processing. It doesn't seem to make any difference whether you use the tools in 96kHz vs 48kHz.
• EQ - I cut everything below 20kHz. I've noticed if you change the frequency scale to extended log you can really see where the musical content ends and the noise begins. There's almost always a hard line. You can EQ below that line, with most records the noise starts between 20Hz and 35Hz. Some have nothing below 70Hz.
• Place track markers now so I can...
• Select each song individually, running the Azimuth module, using Suggest then Render to correct for gain mismatch and realign the tracks. You'd be shocked at how far off the azimuth can be, especially on older records where the tape heads may not have been aligned or a copy of a copy that just keeps getting worse.
• Run the Phase module at the same time. You can get a lot of headroom back just by having the phase of each track optimized.
• I run Declick on all records. Nothing is click free. The settings change of course, but there's always clicks that need to be removed. Most of the time 1.5-2.0 works perfectly, removing most everything without touching transients. If there's an area I need to work around, like a count off with drum sticks, I'll just avoid it and do the rest of the file. I run declicker much higher on fades and silence, usually 4.0-5.0. That noise is easily heard and needs to go. Without transients in those areas you can really go in hard.
• Decrackle - mostly just fades and silence between tracks. Since I do these low amplitude areas, I use a strength setting of 3-4, with amplitude skew of -10.
• Any large clicks or noises the automation missed I'll highlight and use the Interpolate module to repair. Obvious big clicks or pops I'll do this before the automatic declicker.
• Convert the file to M/S or Mid/Side mode. Lots of noise is buried in the sides and it makes it easy to see things you might have missed. Also you can hit the side channel hard with the declicker without adversely affecting the audio.

Restoration (Noise Removal):

This is where my method really shines. When I started doing it this way, needle drops went from sounding like needle drops, to sounding like a tape source, especially when you're listening on headphones. As long as the source is free of tracking errors, tracing errors, distortion, etc., nearly all of the noise that vinyl imparts can be removed. To do this you need to add dither noise which smooths out the random broadband noise of the vinyl source. More noise for records that come from a noisy tape source, less for cleaner, newer pressings, perhaps made from digital files. I try to match the noise floor of the original source, you need to have a way of looking at the spectrum for this to work. Izotope, Adobe Audition, I'm sure a few others have this feature. Audacity I don't believe will work.

• Normalize the files to peak around -0.30 dB. This will give you plenty of signal-to-noise ratio. If you keep the file at, let's say, -18dB peaks, this will be 18dB closer to your new noise floor. You don't want that! Get the audio levels as close to zero without clipping.
• These tips are for Izotope, so the commands may change in other programs, but the method is the same. Take your entire file and silence it (shift-s). Take that silence, and dither it to 12bits. That gives you a white noise (think tape hiss) hovering around -100dB. Select all, copy. Then, go to your history state before you silenced the file. Select all, edit - paste special - mix. That will take the 12bit dither white noise from your silenced file, which is the exact same length as the file you're working on, and mix the dither noise in with it, keeping the file at 32bit float. You could also use a copy of the file in another tab rather than going back to a previous history state. Basically you're just copying and pasting noise over your file, you're not changing the bit depth. When you use the dither on a silent file, you need to set the noise shaping to none, dither amount to high, and auto blanking to never. Especially that last one, if you have that selected, you'll have a silent file.This method is good for a few reasons. You can bring the gain up or down on the noise itself if you feel you need less or more. Depending on the source, sometimes you want more. Also, you can noise shape it using the EQ, turning it into blue, brown, pink, or any other noise profile you want. I haven't really done that, except for a roll off at around 16kHz for some older source material.
• Now you have all of that groove noise between tracks, with the added dither noise which will smooth out the random broadband noise of the vinyl. The crackle, the low level "whoosh", all of the noise the vinyl imparts. Open your Denoise module and select all of the vinyl noise between tracks. Learn that noise. If you're familiar with this step, the first thing you'll notice is how much cleaner and more even your broadband noise it, the random spikes, and unevenness of the noise floor is now smooth and equal, at least in the high frequencies. Now from here, you'll need to remove this noise depending on your situation. But I've found that with the dither noise, you get more noise reduction at lower settings than you without. Usually I leave the reduction set at the default of -12dB, and the threshold right in the middle. This method really makes it easier for the program to differentiate between noise and low Level signal. Since I added noise to the entire file, I denoise the entire file. Sometimes I'll use a reduction curve, usually not. Sometimes whitening, but usually not. The defaults give great results using this method.
• I've tried doing this step in Mid/Side mode and I think I get better results, as most of the "whoosh" is actually in the side channel. It seems I can denoise the side more if necessary for the few times I've had a record with lots of bass that's covered by the low frequency "whoosh" sound. But it's only come up when a record is cut at very low volume.
• I redo fades at this point. I silence between tracks, and usually redo fades to digital black using the linear fade mode. Lately I've been redoing the fades, fading to digital black and then doing this...
• Reapply 12bits of dither noise. Why? Because it really cleans up and glues together the original source material. I've never found a record with a noise floor above 12bits. The added dither helps to smooth out any artifacts that may be left from declick and decrackle. If you want to get fancy, leave the new 12bit dither between the tracks and it sounds like you're listing to a high quality tape source. But what I do here is let the music fade to the new 12bit noise floor, which is what it did originally rather than digital black, and leave a bit of silence at the end of the fade, then fade that to digital black with a quick half second or so cosine fade. This keeps the integrity of the fade for all of you purists while actually redoing the fade. Some fades from the 50's and 60's were really clunky, like an engineer was good at slowly fading until the last second when he got bored and cut it quick.
• There's probably a few things I'm forgetting but I'm sure I can add more later.
• I save this file called Clean as 24/48 FLAC. You should save a few times so you can go back and try different things. I rely on history states, but there have been times I'll want to try a different order (like add the dither noise before declicking, something I've been messing with lately). Azimuth and Phase adjustment are tedious and saving after them is probably a good idea.
• I will usually run a low pass filter starting around 20kHz because there is nothing but noise up there. Old recordings I'll start at 15-16kHz. You really can see the hard line where the frequencies fall off. This means anything there is distortion from playback.
• Make some AAC files for my car. Yeah, all of that work and I make a lossy file. Lol.

Well that took longer than I intended. I might make a video or post some screen shots. This method has really changed the way I do transfers. Now I want to redo all of those terrible "professional" needle drops myself and I'm positive I can get better results.

One takeaway from this is that it seems that 16bit audio really is all you need. 16/48 is what I make for lossless files I intend to play. There's no reason for 24bit once you're done editing. I use 48kHz because all playback devices upsample 44.1kHz to 48kHz anyway, or even oversample further. Unless you're burning a CD, I don't see any reason not to use 48kHz.

 

Is there a good way to determine the bit depth of rips?

 

It's all about the noise floor, if you can hear hiss from the original source, the noise floor is well above even the 16bits that we typically use.

I usually try to match the noise during fade outs, with older tape sources you can really see, in a spectrum view anyway, where the noise floor is. I've found it generally hovers around -100dB at 1kHz, I usually use this as my baseline for when I do noise removal also. Generally one pass of dither noise (once the file is normalized) will match the broadband noise of the record. I've had a few really noisy sources where I've added much more, by increasing the gain of the noise. I'd probably select 10bits in these instances if Izotope had that option. Either way, you can't really go wrong right about 12bit. Also, this method is really good for REMOVING the tape hiss from the original source. I usually try to keep it in, but I've had a few recordings that had a very high tape hiss, and once removed it sounded much better so I didn't add any back in.

I'll try to post some screen shots tomorrow.

 

The thing you guys aren't getting is that dither only makes a difference when you go down to 16-bit or lower. If you apply it to a bit-depth higher than, say, 20 bit, it has no effect because the noise floor is so low.

 

Not in this case. I'm not using the noise to fix quantization errors. I'm also not using 16bit dither, exactly because you can't hear it. I recommend taking your silent file, adding 12bits of dither noise (basically increasing the noise floor way above digital black) and listen to it, but remember to use the method I stated above, the default setting won't add dither to silent portions of the audio so it may seem as though nothing is happening or the dither is in audible. It sounds and acts just like tape hiss. The trick is that this noise HAS TO BE AUDIBLE, and 12bit is essentially tape hiss. You mix this into your file and it is very audible. Your "technical" noise floor is still -140dB or whatever, but you've now filled it with noise. This noise is loud enough to be used the same way that low level 16bit inaudible noise will smooth out quantization errors. Don't think of it as dither, think of it as white noise right around -100dB.

I'll post audible samples and screen shots.

Normally you'd be right, 24bit dither is inaudible. 16bit is inaudible. 12bit, audible, 8bit is VERY AUDIBLE. The bit depth is where that noise floor is i.e. how audible it is. Dither is just noise, you can use various noise shaping but the one I'm using is white noise TDPF, hiss.

 

hello all - question about the image below. I have just upgraded my TT and cart - VPI Prime Scout and Soundsmith MIMC Star. now when I record an LP I get thousands of 'micro' clicks as you can see in the image. Never seen this with my old table and MM carts. still trying to dial it in but wondered if anyone else has had this experience with MC cart. If I run a very light de-click on a full album side I get 20-50k clicks detected and cleaned. I really try to only de-click what I can hear and I know most of these are in the 20-30kHz range which I can't hear but still...

thanks

 

Could it be static electricity causing the clicks?

 

I suppose but I do clean my albums and shoot them with my ZeroStat. Never saw anything this bad with my other table and carts having done many, many albums. Playing with the setup, cables, etc. so, we'll see.

 

Does it happen with other LPs? Ensure the connections at the cartridge are solid - fully seated. Also, check the ground point for the TT. Make sure it is good. No paint under the connection.

 

Contact Soundsmith about that too, with that picture. Possibly there is some known issue at some times. Or your MC preamp has a problem you never detected before.

 

RX has a de-click setting that deals nicely with those, and with digital clicks.

 

I have recorded multiple LPs, some new, some old, 80's cheap-o vinyl and Tone Poet 180 grm all with the same problem. Honestly, you can't really hear anything while it is playing but still there.

@JohnO - will be reaching out to SS just to be sure it's not something I am doing or if they have seen anything like this. thanks

@Grant - yep, run the de-click but rather I don't have too. thanks

 

ok, well, mystery solved. It was my ADC (Motu M2 connected and powered via USB-C). switched over to my other ADC (Tascam UH-7000) and the problem went away. hmmm.... go figure. never happened with my other TT/cart. whatevs, feel much better now that I know it's not my new rig.

thanks.