Your Vinyl Transfer Workflow (sharing best needledrop practices)*

Just a little experiment I thought I'd slap together:



Flat phono amp (no RIAA) from DC to 100kHz+. 62kOhm load + 70pF of just cabling.

then wet-play transfer at 80% speed, adjust sample rate from 48kHz to 60kHz, resample to 96kHz and do software RIAA.

What it sounds like: sassa-sample.mp3 (compare to any youtube recordings at 2:40, all too loud)

 

What SNR are most of you guys getting? When my audio interface (Presonus AudioBox iTwo) inputs are zeroed out I’m getting ~ -63 dBFS of noise when running phono amp -> audio interface, and ~ -59 dbFS of noise when running phono amp > pre-amp > tape out > audio interface. I’d prefer using the second setup because it means I can monitor playback but if it’s making my rips noisier then I don’t think I want to use that signal chain.

 

Well, I built the non-RIAA phono preamp from the breadboard with better tolerances, with some batteries patched in to see how quiet it can be (and because I haven't conjured how to mount a PCB s-video-type receptacle for the power jack yet.)



Average music signal (green) to phono noise (red) after software RIAA and software 30Hz subsonic (this vinyl mastering has below 50hz sharply cut):


Yellow is noise floor of digital interface with phono off, after RIAA de-emphasis tilt. (frequencies of noise are x1.25 due to reduced-speed recording)

At normal speed, after normalizing a recording with a scratch that pops at 6dB above the music, 30 seconds of the needle-up is:

• Peak amplitude: -70.47/-71.15 dB
• Average RMS power: -82.56/81.64 dB
• Total RMS power: -82.55/-80.88 dB

and that's unweighted at full bandwidth to 48000Hz


For now, rebadged RC5532A opamps from 1990 Raytheon that were noise-selected special orders for Peavey consoles (the originals that can take +/-22V).

There's some spiky interference, being on my desk near two monitors, DACs, PC, etc (probably not helped by bandwidth that could receive AM radio). The turntable ground wire disconnected from the preamp adds 50dB of noise. Unequalized audio of a loud 12" single has -9dB peaks below 16dBu full-scale, which should go to -3dB with a balanced line driver installed.

Seems to have obsoleted a phono preamp with pesky RIAA component requirements for accuracy, especially because it makes recording safe at any speed.

 

I’ll measure mine today but wish to use the same method that you did. When you say “zeroed out” what are you referring to (phono inputs tied to ground?). How did you measure those dBFS levels you quoted? I plan on recording a second or two from phono cartridge (wired balanced), through fully balanced phono pre, through ADC to computer WAV file at 96k/24-bit which is my normal needledrop rate. Then I’ll do waveform stats in RX and report results.

 

By "zeroed out" I'm referring to having the input trim/gain on my audio interface at 0 (no boosting or cutting). My measurements were just based on my visual observations of where the noise was averaging on the input meter on my DAW. I did discover that one of the RCA connectors on my turntable is loose and not always making a good connection so I've ordered a replacement and I'll see if that brings the noise down at all.

 

Really great post, thanks for the information.

 

My total RMS levels from 3-5 seconds of recording with no musical signal (tonearm at rest, platter not spinning) was -70.66 dB (left) and -65.86 dB (Right).

 

The spectrum view looks a bit better than the stats imply.

image hosting

 

Thanks for these, I'm going to get some actual RMS measurements from my setup and see how they stack up. The replacement socket for my turntable should be in tomorrow. Once that's in I'll do another round.

 

Here are my numbers and spectrum. Motu M2 connected to a Kenwood C1 control amp with the built in phono amp selected. Sample recorded at 192K.

 

Very impressive! You have nothing to be worried about. Super tight L/R balance.

 

I've thought about bypassing my pre amp and capturing directly from my TT for needle drops, then apply the RIAA Curve with software. At what peak level would you capture the signal. I capture at -6 dB peak normally and increase gain post capture. Would I still capture at a peak of -6 dB? Howw much would the gain increase when the RIAA Curve was applied?

In regard to wet play, doesn't it have an adverse effect on your stylus?

 

A phono input still needs an appropriate preamp. "Capturing direct" isn't really an option, except as an experiment destined to fail.

A turntable needs a specific load that most audio interfaces won't provide. An E-Mu 0404USB, for example, has a 1 megaohm line-level input (a DI guitar input's load) or 1500 ohm mic input. Unstated capacitance. Other interfaces that put guitar DI on a separate jack may still have an unsuitable line-level load such as 10kOhm.

Then, the recorded signal will extremely quiet, with significant noise under it. Reminder that a typical phono preamp with equalization has a 40dB gain at 1kHz, which increases to near 60dB in lower frequencies. That's 60dB closer to your noise floor without that booster.



The flat phono preamp I implemented has 35dB gain across all frequencies, which is a massive boost for the high-frequency-dominated cartridge output that looks like the above, allowed by the +16dBu capabilities of my pro audio interface. It also has input designed for MM phono, with attention to current noise and other particularities.

The software RIAA equalization step may clip the recording (record as loud as practical), but it is a simple thing to reduce the volume of 32-bit audio first or to use gain control in the command-line option.


You'd be hard-pressed to find any internet thread "the cooling, friction-reducing effect of wet play ruined my stylus". Water-soluble adhesives are not used in construction. I'm mindful to dip the cart tip in magic eraser after a play, to pull some water off and clean while still damp.

 

I don't think you'll get anywhere near -6dB without your preamp.

 

Thanks for the reply. I must try the wet play for needle drops. I assume distilled water is used?

 

This is the way. -18 dBFS is what we calibrate to in the archive. Set the gain on your ADC so the 1k is hitting -18 and then you don't have to bother with steps 2-5. WRT needledrops, you could then in theory not apply any normalisation and then all of your needledrops would retain their inherent peak values, making for a more accurate playback experience of the needledrops.

 

Does the 1k tone need to come from the TT? Could I generate a 1k tone from my phone into my ADC for the same purpose?

 

That is a reply to the middle of a misguided conversation long ago. You are not producing streaming audio for television productions to match dialog's 1-minute loudness levels to commercials. You are making digital files to compete and integrate with music ripped from CDs and digital downloads in your library.

Record as loud as you can while leaving a few dB of margin to avoid clipping or overages, even on dust clicks or scratches. This will ensure best signal-to-noise and digital reproduction in recording.

Then in your own "mastering" you can get closer to maximum level. It's best to do both album sides exactly the same, perhaps combining them into one file before splitting to tracks. Normalization to -1dBFS peaks is a good guideline of maximum level, to ensure you have headroom for the level changes that dithering or lossy compression may affect. The only exception might be mellow music that inherently should be quieter than the rest of your library.

Then put your album tracks in shuffle between other music. See if you did a good job. See if you want to get out the L2 ultramaximizer to downsample.

 

I'm sure I'm not understanding something but achieving -18 is easy, just do waveform stats and simple math to figure out exactly how much to adjust the gain. I usually set both channels to hit -18. There are exceptions but they are rarely off more than a little bit.

It has long been my understanding that gain and normalizing are not destructive to the audio, if they used in ways that don't clip.

 

Yes, ideally you'd want it to come from your analogue source. The CBS Laboratories Professional Test Record (STR 100) has 1 kHz tones recorded @ 0 VU. Play them, and set the gain on your ADC so the tones hit -18 dBFS. Then you don't need to worry about clipping the input, or fiddling about trying to guess the balance. You now have your gain set to an analogue standard. In the archive, we like our musical peaks to hit between -6 and -1 dBFS, so sometimes we'll add +3 or +6 dB to the ADC gain in order to achieve this. At home, I'm happy to leave my ADC gain at -18 and let the peaks fall wherever they happen to. I can always normalise in software afterwards.

NOS copies of CBS Labs STR 100 are available on eBay.

 

If you're still reading this, just a note on using the PulseAudio sound server in Linux. This is like the equivalent of the Windows sound mixer (allowing different programs to use audio simultaneously) and which most people here like to avoid for best results (using ASIO drivers or WASAPI). PulseAudio is set by default (for everything that is run through it) to 16-bits and 44.1 kHz or 48 kHz sampling. Searching the Net you'll find some tips to amend the etc/pulse/daemon.conf file, but it's still not clear to me what goes on there behind the scenes, so I try to avoid PulseAudio for audio (I used to run a PC without it, but nowadays it's so intertwined with everything that it's hard to avoid/uninstall). I prefer to use alsamixer. You can just type it in a Terminal to adjust levels instead of using the PulseAudio settings.

Audacity, which you are using, appears to be avoiding PA as well, since it only offers ALSA as "host" and not PulseAudio. ALSA (Advanced Linux Sound Architecture) includes what would be the "driver" for your hardware in Windows (in this case USB Audio). I'm posting this from a Linux PC (just for fun and to keep up a bit) and using the onboard sound chip Audacity will record in 24-bit and 96 kHz (you can zoom in on the low level samples in the wave-form), without changes to the default PulseAudio settings, so I think you should be fine for recording, except there's someting I noticed just now. With Audacity Preferences set to 32bit-float, I don't see any samples below -96 dB (which would confirm 24-bit audio), whereas with the 24-bit setting I do. I believe Audacity does its calculations when editing in 32-bit float anyway, so I would set it to 24-bit for recording. Alternitavely you could use Ardour (and/or JACK Audio Connection Kit) for recording (overkill for needledrops I know) and do any editing in Audacity which is much more intuitive.

Well, I don't know how experienced you are with this stuff, but optimal sound never has been and still isn't the easiest thing in Linux. I like to use Audacious for playback which just uses PulseAudio as standard too, but you can choose ALSA output as well. But then it still goes through PulseAudio with an "ALSA plug-in", unless you direct it to your hardware directly (check where it does and doesn't show up in the playback tab of the PulseAudio mixer).

Good luck on your needledrop journey, I didn't catch up on the samples you posted but it shouldn't be lacking on your front-end that's for sure.

 

Thank you. My following post is not for the anal purists among us:

Of course, people can do what they want, but, I think people around here make a bit too much about avoiding limiting. No one is talking about boosting and limiting to 6db or more. Just two or three db to get you in the ballpark. People talk about just turning up the volume. Sure, you can live that way, but if you have a very dynamic file(s), why nit just take the time to use th3e limiter in the right way. If you have the tools, learn what settings ound the most transparent. With the tools available today, you can get mighty damn close to transparent! There's no reason to sacrifice sound quality for volume anymore. Just don't how ape crazy with them. It isn't any less "audiophile" to bump it up just a couple of notches, providing you are careful. And, there are lots of little tricks one can use to boost the volume without using a limiter if you are dead set against using one.

I record both sides of the album in one shot. Sometimes one side is a bit louder than the other. One can fix that. Sometimes channels are off balance. That can be fixed, too. Never assume the original disk cutter got it right. Never assume the tapes were mixed correctly.

I have been leaving -.6 db for headroom. But, now I may back off to -1.0 db, as the pros recommend these days, anyway. You don't have to shoot for -14 LUFS, as that may be too low for some material. I have been shooting for -11db LUFS nowadays. If you're going for streaming, -14db may be too low anyway. Something like -12 db or -11 db will not usually trigger the limiter in most streaming services.

BTW errbody: limiting is not clipping. Limiting is used to avoid clipping.

In the older days, instead of using a limiter (digital limerters just weren't that good years ago), I would zoom in on a problematic peak and literally reduce the gain of it, ensuring that it would not sound too obvious.

 

Thanks for the follow-up - I had forgotten I asked! Since then I have been fairly happy with my recording using Audacity with the Motu M2. As you say, ALSA is set as host (the only option) in Audacity but the PulseAudio volume control will affect recording level so, as you suggest, there is a ALSA plugin shim. I agree that PulseAudio is "intertwined". However, Audacity is set up and recording at 48kHz and 24 bit. I use Clementine for most playback (and Audacious for quick file playback) and my DAC indicates that playback of files I've recorded are being output at 48kHz. Other files are shown to be 44.1kHz. I checked my /etc/pulse/daemon.conf and no sample rate configuration options are uncommented - so default setup - although I believe I may have fiddled with these in the past.

As it happens, I experimented with Cue files (manually created by taking time stamps from Audacity) with Clementine yesterday for the first time with some limited success - with multiple cues indexing into multiple files where each file was a LP side. It generally worked but I could not get Clementine to honour the overall track numbers so the jury is still out on whether it is worth pursuing that route with Clementine for playback. I also played with the Audacity labelling and splitting facility which is neat.

 

I agree with everything you wrote. iZotope’s Ozone gives me a whole arsenal of limiters to choose from but I only use a couple of them most of the time. Furthermore, their dynamic waveform shows you exactly what if anything the limiter is doing to the music. If a single vocal consonant lasting a millisecond or less gets reduced I know I can’t possibly hear that effect so I let it go. A handful or less of these limiter actions are OK by me as I know that everything else has been normalized linearly to the highest bit resolution I can get out of the source material. If I don’t like the mix balance and it’s a favorite track I’ll do stem splitting with RIPx or iZotope Music Rebalance and remix in Pro Tools to get the qualities I wish to hear. I don’t care a bit about LUFS, and trust my ears because this effort is only for me and some friends I transfer my finished files to.

 

Sounds like all's good for what you require. And yes, the fact that PulseAudio can move the level sliders in Audacity, even though rec/pb sound doesn't appear to pass through it, annoys me. That's why I prefer setting these with alsamixer instead.

I briefly checked Clementine and while it is possible to select ALSA output, I didn't see any option to point it to the soundcard/DAC directly (like there is in Audacious), so it still uses PulseAudio. As long as you stay with 44.1 and 48 kHz files, PulseAudio won't resample these as you've found on your DAC's display. But with anything higher, like a 96 kHz file for instance, I would expect it to still show 48 kHz on the DAC because of the default PA configuration, which only includes 44100 and 48000 Hz sample rates?

I can't help with the cue files, haven't done that, but the individual track labelling and multiple export feature in Audacity is neat as you say.

Enjoy the 'drops (and the process)!