struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm
My first needle drop - lessons learnt
jazzfan
jazzfan's picture
Offline
Last seen: 5 years 1 week ago
Joined: Sep 8 2005 - 8:55am

Struts,

Great write up - a very informative and fun filled read. Hopefully by revealing your somewhat painful learning curve others will be able to avoid suffering many of the same headaches.

Out in cyberspace there seems to be a growing movement of people doing both high resolution (24/88.2 or 24/96) and CD resolution (16/44.1) "needle drops" and uploading the audio files to various file sharing services, such as Rapidfile. The interesting thing is that many of these needle drops are being made from the vinyl versions of recent CD releases, even those of fully digital recordings. The reason being that many people feel that the masters used for vinyl releases DO NOT suffer, or suffer much less, from the dreaded affects of the "Loudness Wars", in other words the vinyl has better dynamic range. But that's a discussion better left for another thread.

Thanks again for very entertaining and informative post.

struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm

My pleasure jazzfan, glad you enjoyed. Happy holidays!

mrlowry
mrlowry's picture
Offline
Last seen: 6 years 11 months ago
Joined: May 30 2006 - 1:37pm

Struts-

That's a very interesting write up. Especially the part about the recording having a higher noise floor and the theory that it was because the turntable wasn't on a proper stand.

Wouldn't you have to have three passes to properly record an album? One to set levels, one for the recording process, and one for any noise reduction that was necessary.

ncdrawl
ncdrawl's picture
Offline
Last seen: 12 years 11 months ago
Joined: Oct 18 2008 - 9:18am


Quote:
Struts-

That's a very interesting write up. Especially the part about the recording having a higher noise floor and the theory that it was because the turntable wasn't on a proper stand.

Wouldn't you have to have three passes to properly record an album? One to set levels, one for the recording process, and one for any noise reduction that was necessary.

with 24 bit recording there is no reason whatsoever to set levels above say -10(at peak).. you could have peaks at -22 and still surpass the range of any redbook cd.. so that being said, in my DAW, i play the loudest track, and set levels according to that. I would never "ride fader" or pencil in automation curves... the dynamic range is already on the record as the conductor intended it, so as long as I set my levels right(and by that I mean set them ONCE so that the loudest track does not go into clipping), the dynamic range will be preserved..... for the noise reduction... .with vinyl the noise is pretty consistent, so what I do is have my denoise algorithm sample a "fingerprint" of a selected passage and apply it across the whole recording..of course clips, ticks, etc, spikes...have to be manually penciled out, but as I said, the noise pattern with a lot of vinyl is spread across the whole record.. denoising and editing can be overdone very easily, so I try to leave it be if I can get away with it.

maybe with tape you would require multiple passes, but not digital.!

mrlowry
mrlowry's picture
Offline
Last seen: 6 years 11 months ago
Joined: May 30 2006 - 1:37pm

ncdrawl-

I think you have misinterpreted what I was saying because you've pointed out the same steps.

1. Setting a proper level
2. Recording
3. Noise reduction

That's three passes through the music.

ncdrawl
ncdrawl's picture
Offline
Last seen: 12 years 11 months ago
Joined: Oct 18 2008 - 9:18am

Yes, my mistake. I got the impression that you were implying "passes" to be a bad thing, IE each pass causing a degradation .. (of course with digital, it doesn't matter how many times you go through...whether 1 or 1,000, youd still get bit perfect results) I am an analogue tape head, so here, "pass" means the act of actually capturing the data, committing it to tape.. at any rate, yes, I guess it would take 3 times going through the music(though I would never go through the whole thing just to set levels..I would use one track for that and roll with it)

struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm


Quote:
ncdrawl-

I think you have misinterpreted what I was saying because you've pointed out the same steps.

1. Setting a proper level
2. Recording
3. Noise reduction

That's three passes through the music.

Sam,

Yes, I guess in theory you would need three full passes although in practice I used ncdrawl's approach of just setting the level based on an arbitrarily chosen subjectively loud passage. Unfortunately the one I chose didn't quite represent the highest peak.

I found a function in Audacity, 'Find Clipping' which shows that my transcription clips in two places (I'd missed one earlier), however examining the waveforms these are very marginal clips where it looks like the peak would have only gone a fraction of a dB over. I have replayed both again and again and try as I might I can't hear them.

However the point remains that for a 'perfect' result I would have to record the whole thing again a few dB down, effectively ending up at your three passes. Mea culpa.

Editor
Editor's picture
Offline
Last seen: 13 years 3 months ago
Joined: Sep 1 2005 - 8:56am


Quote:
I had run the record through the VPI twice before starting, following Stephen's recommended 'two-fluid regime' (see his blog - sorry I couldn't find the link, the 'Search' function doesn't seem to cover the blogs - weird!) ...

Great account, Struts. Just wanted to address this point: for historical reasons, the blogs and forums run on a different server using a different content management system to the news and archives. The website's Search engine only looks for keywords on the latter server, unfortunately. One day, we shall consolidate the different systems, I hope.

BTW, A/D converters tend to produce higher levels of distortion as they approacch 0dBFS, so it is best to under-record a little, not peaking above -3dBFS. Afer you have done all the noise reduction and clip removal, you can then normalize, though I recommend doing so to -0.1dBFS rather than 0dBFS, to leave just a little breathing room for when you downconvert to Red Book (if that's your final goal).

John Atkinson
Editor, Stereophile

struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm

Thanks John.

When I started playing with the spectrum plot I noticed that there is a whole bunch of content between about 42-44 kHz (but nothing at all between 21-42 kHz). I am pretty sure this didn't come from the record so I presume it is some kind of artifact from ADC.

1. Can anybody explain?
2. I assume since it isn't program it would be wise to filter it out, right?
3. Given that if I do this the program only extends to 21 kHz is there any benefit to be had ripping vinyl at 88.2 kHz rather than 44.1 kHz?

Thanks in advance for any clues for the clueless.

Editor
Editor's picture
Offline
Last seen: 13 years 3 months ago
Joined: Sep 1 2005 - 8:56am


Quote:

When I started playing with the spectrum plot I noticed that there is a whole bunch of content between about 42-44 kHz (but nothing at all between 21-42 kHz). I am pretty sure this didn't come from the record so I presume it is some kind of artifact from ADC.

The most likely cause is that it is the tip-mass/groove-wall compliance resonance of your phono cartridge.


Quote:
I assume since it isn't program it would be wise to filter it out, right?

If you can do so without affecting the baseband. But if you are going to convert to Red Book, it will be filtered by definition.


Quote:
. Given that if I do this the program only extends to 21 kHz is there any benefit to be had ripping vinyl at 88.2 kHz rather than 44.1 kHz?

You will get better coding of clicks, which can have significant HF content, meaning they can be eliminated more cleanly. You are also captruing the tip-mass resonance without it aliasing down into the audioband, again making it easier to eliminate later.

John Atkinson
Editor, Stereophile

Editor
Editor's picture
Offline
Last seen: 13 years 3 months ago
Joined: Sep 1 2005 - 8:56am


Quote:

Quote:
Given that if I do this the program only extends to 21 kHz is there any benefit to be had ripping vinyl at 88.2 kHz rather than 44.1 kHz?

You will get better coding of clicks, which can have significant HF content, meaning they can be eliminated more cleanly. You are also [capturing] the tip-mass resonance without it aliasing down into the audioband, again making it easier to eliminate later.

Just to expand on this point: because of their transient energy, LP clicks and scratches will all give rise to ringing from the A/D converter's anti-aliasing filter at the Nyquist frequency. This ringing will persist even if the offending noise is itself removed in post-processing. With a sample rate of 88.2kHz, this ringing will occur at 44.1kHz, so if you capture the audio at 88.2kHz and do all the processing at this sample rate, when you downsample to 44.1kHz, you will eliminate the spuriae.

Similarly, if you use any aggressive filtering, such as a deep, narrow notch at 60Hz or 120Hz to eliminate hum, these filters will also ring on transients. Best to do do these operations at a higher sample rate.

You should also do all the signal processing with 24-bit or 32-bit data, which will ensure that any mathematical artefacts of the processing will be eliminated when you decimate the data to to 16 bits. Downsampling and decimation should therefore always be the last operations you perform on your needle drops.

So, the work flow for a needle drop should be:

1) Capture the audio data at 88.2kHz with 24-bit word length. (I recommend 88.2kHz rather than 96kHz, because not all sample-rate converters perform well with 96-44.1kHz conversions.) Keep the maximum level no higher than -3dBFS.

2) Perform all the click removal and other processing of the data with the maximum bit depth supported by your program, most probably 32 bits.

3) Normalize the data to -0.1dBFS peak (not RMS)

4) Make a copy of the master file for archival purposes.

5) Downsample the data to 44.1kHz, using your program's SRC at its highest quality level.

6) Decimate the word length to 16 bts, making sure you have the dither and noise-shaping options set optimally.

7) Enjoy the music.

John Atkinson
Editor, Stereophile

struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm

Very helpful, thanks John!

fricc
fricc's picture
Offline
Last seen: 4 years 3 weeks ago
Joined: Dec 1 2009 - 10:56am

Note: on request by struts, this text is copied from another thread to this one which is more relevant.

I just read the excellent posting by struts about figuring out needle drops. It's been a few months now since I started doing this myself and I just got to a point where I start to be quite satisfied with the results. I thought I might share a few things with you guys.

The System:

Thorens TD145 MKII + Shure V15 Type V + Pickering XSV3000
ProJect Phono Preamp
Primare I30 Integrated Amplifier
EMU 0202 Audio Interface
Imac 24" + lots of drive space

Audacity freeware software for audio recording and editing
ClickRepair + DeNoise to make things quiet

The Action:

The first difficulty has been figuring out that the EMU 0202 has no input attenuators, you can only increase the input gain from its fixed base input level. This is inconvenient because the ProJect preamp (a really nice piece of gear, despite its low price) has a fixed (40dB) gain too.

Reducing the gain via software on the "Audio Midi Setup" panel won't work either because it will only reduce the gain of the digital subsection of the audio interface, the input amplifiers will saturate anyway. Dumb.

My vintage (perfect conditions) Pickering XSV3000 won't do with this setup: its output is just about loud enough to saturate the audio interface inputs in the loudest transients. I Would have to use some -3dB RCA attenuators out of the preamp to use this cartridge.

Fortunately my other vintage cartridge (the stellar Shure V15 Type V) is some 3dB less loud than the pickering and using it I have yet to find an album that overdrives the audio interface. I also slightly prefer the sound of the Shure over that of the Pickering

zeb
zeb's picture
Offline
Last seen: Never ago
Joined: Jan 2 2009 - 6:01am

Thanks for the thread - very interesting, specially that I am thinking to attempt this for the first time this weekend.

I've got an Macbook Pro and use an Apogee Duet as a sound source. In the Duet disc, they show you how to digitise LPs using the Output of the phono stage or Tape out into the Duet, and using Garageband (which comes for free with Macs) as the recording software.

Would Garageband be good enough or would I be much better off with something like Audacity, which is also available in Mac version? Any info on the pros and cons of each would be great.

Thanks

alrmad
alrmad's picture
Offline
Last seen: 13 years 2 months ago
Joined: Dec 20 2007 - 2:17am

Thanks for some great accounts of archiving vinyl, a theme mainstream mags seem to have slightly overlooked in my humble opinion.
When I set out to digitize my vinyl-collection (around 200 lps), I first looked at sound-cards, lap-tops and the like. Having read Michael Fremer preferred the Alesis Masterlink solution, I began to look at dedicated digital recorders instead. In my settings, not having to set up a computer, even a lap-top near my stereo-equipment is an obvious advantage, and you can always import files into your favourite wave-editor afterwards.

I ended up buying a Korg DSD-recorder to do the job. Just like major record-labels, if you archive your stuff in DSD, any PCM-resolution will be available subsequently. I found the recordings astonishingly indistiguishable to the vinyl rig. My route is:

Rega P9 w/Lyra Argo i
Holfi Battriaa
Korg MR-recorder (w. balanced inputs)

I can perform basic native DSD-editing using Korg's Audiogate-software (divide songs, normalize, dither, fade and transcribe to PCM), and the resulting files can be burned directly from AudioGate as a DSD-disc which can be played back by the Sony XA SCD-5400ES player, Kalman Rubinson has been raving about recently (Stereophile Class A+).

The only real down-side for me is I can' get rid of surface-noise in native DSD, but this has never been much of an issue for me anyway.

I would love to hear other readers' experiences with digital recorders for this purpose.

fricc
fricc's picture
Offline
Last seen: 4 years 3 weeks ago
Joined: Dec 1 2009 - 10:56am

Audacity is free too and is way more flexible than GarageBand for LP recording.

The thing is that it is hard to just do a straight recording from an LP, you always need to tweak stuff (clicks, pops, etc.) Audacity will give you all the tools you'll ever need for that. Try it out, see if you like it...

- Fabio

struts
struts's picture
Offline
Last seen: 2 years 9 months ago
Joined: Feb 1 2007 - 12:02pm

Hey Fabio,

I've been taking a look at ClickRepair and DeNoise. You stated in your post that you decided to keep everything at 24/96 when you realized what wide bandwidth your LPs had, however as far as I can tell (the documentation appears to be non-existent) both tools only work on 16/44 AIF or WAV files.

Please could you elaborate a little on the workflow you used between Audacity and ClickRepair/DeNoise and how you maintained the 24/96 resolution? I would also be interested in any non-default parameters you found to work particularly well.

Log in or register to post comments
-->
  • X