Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: "ripping" LP's (Read 37093 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

"ripping" LP's

Reply #25
Quote
Well, 70dB is the best you get from an LP.

Even if we ignore rumble, I suspect, eh?

I think you're right here, but there isn't a whole lot of data out there that shows it is. It seems logical that rumble would be of a fairly large magnitude, because it needs to be pretty massive to hear it, and it's a well known noise characteristic of LPs. But there's plenty of other noise in the preamp and the groove itself to mitigate it a lot, and I personally have never heard it or seen it in my rips.

Quote
Now consider, in noise one critical bandwidth wide, we can detect a tone down about 5.5 or 6 dB in energy from that noise, as long as both are above absolute threshold.

Wait. Are you saying that the audible SNR in a critical band can be no less in amplitude than -3dB (being -6db/2)? That seems awfully conservative to me, since level differences of 1dB are ABXable.

Quote
On the other hand, this is just as true of digital recordings as analog recordings, so I don't really understand why you mention this. The noise from the LP will dominate the noise from 16 bit uniform, TPD quantization at all frequencies, I suspect. I even suspect most pre-amps won't do any better than that.


Yeah, you're right; I admit my post is a little semantic. In fact, right now my preamp (phonopreamps.com TC-750 with an upgraded power supply) is punching 80-90dB SNR, so you're pretty spot on in saying that the preamps can make this point somewhat moot. However, there's no reason why a preamp couldn't have 110dB of SNR.

I just get the impression that a lot of people think that because vinyl's noise floor is 70dB that absolutely no signal is useful below that level, when the math doesn't support it.

Quote
Now, your statement about noise floor is true, but it applies to both analog and digital noise floors equally, unless somebody forgot to dither (oops), at least.  Say for 24kHz bandwidth (48khz sampling rate), the noise in a critical band at 1kHz will be at about -22dB re: the total noise, for anything with a white noise floor.

This puts the total "hear down into the noise" at about -28dB, give or take, relative to the wideband noise level.  This is all well and good and all that, but it really doesn't matter because it works the same for analog and digital (except that the analog noise floor won't be quite flat, even if we ignore rumble), so you might get another 10dB out of that, making the 70dB into 80dB.  that's still 16dB of headroom/tailroom, isn't it, now?

Yeah. It is.

"ripping" LP's

Reply #26
Quote
I just get the impression that a lot of people think that because vinyl's noise floor is 70dB that absolutely no signal is useful below that level, when the math doesn't support it.[a href="index.php?act=findpost&pid=327595"][{POST_SNAPBACK}][/a]

Well, I suspect your maths is rather more advanced than mine. I can only go by what I've experienced in over a decade of LP-to-digital transfers.

Vinyl with a noise floor at -70dB is *extremely* rare. If you find such an LP, it's worthy of a little jig in celebration. The noise floor of typical vinyl (ie. the LPs you'd buy in an ordinary record shop) is more like -45dB. But spectrally it's mainly at low frequencies - this is why it's nowhere near as noticable (and hence objectionable) as the more broadband noise you get from cassette tape. So although my knowledge is empirical rather than rigorous, I frankly find it hard to believe that any vinyl LP will have discernable signal much below about -80dB.

The most obvious test I can think of to check for the audibility of quantisation noise when recording vinyl would be to try recording the end of Neptune from The Planets Suite at both 16 and 24 bit resolution, then try to ABX the results. Unfortunately I don't have this work on vinyl, so I can't try it. Perhaps someone else here would be prepared to give it a go.

 

"ripping" LP's

Reply #27
Quote
would be to try recording the end of Neptune from The Planets Suite at both 16 and 24 bit resolution, then try to ABX the results
This might be interesting but you forgot to add that the LP must be of the highest quality vinyl, in pristine condition, properly cleaned, on the best playback equipment. And both recordings done simultaneously, since two separate playings will give slightly different outputs. Then one should be able to more confidently state 'yep, there isn't any difference to be heard'

There is also the complication of how to do manage the 16 bit recording. You don't want to do it on a different, 16 bit only, soundcard or you have already compromised the comparison conditions. Will you get the 16 bits from the 24 bit recording by truncating? or by converting with dithering and noise shaping?

I don't know all the ends and outs of this but the recording software partly controls the soundcard. An obvious example is selecting the sample rate frequency. Recording at 16 bits with a 24 bit card can be with or without dither on some soundcards, via the card's mixer panel, but I'm not sure what happens if the mixer application doesn't provide that option. One should probably expect a difference between a dithered and non-dithered recording -- except that the vinyl noise is likely to make that moot to begin with (but one could simply ABX that as a starting point and find out if it is audible or if the background noise totally obscures any difference).

"ripping" LP's

Reply #28
I'd like to interject here to say that so far, over the last few days, I haven't seen a vinyl rip that had an inter-track noise level lower than 40db from the track peaks, on mostly rock LPs, and the noise itself is surprisingly wideband. Granted a lot of the vinyl is 30 years old, and I do not have a proper cleaning machine, but one LP I bought 3 months ago and never played up until now. So I stand corrected.

"ripping" LP's

Reply #29
Quote
It's true that rounding errors always occur. That's part of the beauty of 32 bit float (or 64 bit in some programs) -- the rounding errors are really far down where they can never effect the music. As a demo I amplified a 55 minute LP recording by 6dB, then I amplified it by -6dB. The peak differences between the result and the original came out at -150dB and the average differences at -190dB. I could have had the same results by amplifying it 100dB, then -100dB. Anyway, that's so far down that it doesn't exist in human terms or even hardware terms.


No, as the numbers become larger the rounding errors of float numbers also become larger.

"ripping" LP's

Reply #30
Unless I missed it, there has not been any discussion here about the subsonic contribution. I've been able to examine about a half dozen recordings made on other systems and they were all similar to my own results. The subsonic level, measured between tracks, is always higher than any of the other background noise. It is often higher than moderate level audio.

The LP recording I'm currently working on is a bit worse than average in most respect, being a well abused 50's recording that I rescued from a thrift shop, but I don't know why the subsonics would be much different. The raw recording (not normalized)  measures Average RMS -50dB between tracks. Running nothing but a subsonic filter over it drops that 9.4dB. Another recording from a much better condition LP, awaiting its turn for processing, measures initially -57dB and then -68dB after the subsonic filter This is an even larger difference.

In general the highest measured lowest frequency levels vary from about 2Hz to 12Hz. One can't exactly hear the difference directly, but that is a lot of wasted amplifier power and cone flapping for nothing good. So, to add another maximum to my collection: filtering out the subsonics is always desirable. Noise reduction by itself does not accomplish this to an adequate extent.

Quote
No, as the numbers become larger the rounding errors of float numbers also become larger.
I concede that, but I think the difference is insignificant when we are talking about music. Making a test on my currently in-process recording, I get an RMS average difference of -195dB using 6dB as the factor and -178dB with 100dB. That certainly is a larger error but no one will ever hear the difference. I guess my pronouncement was more or less equivalent to my other statement about normalization being "absolutely distortion and noise free." Neither is absolutely true, but the difference is absolutely undetectable by human senses.

"ripping" LP's

Reply #31
Quote
Quote
Now consider, in noise one critical bandwidth wide, we can detect a tone down about 5.5 or 6 dB in energy from that noise, as long as both are above absolute threshold.

Wait. Are you saying that the audible SNR in a critical band can be no less in amplitude than -3dB (being -6db/2)? That seems awfully conservative to me, since level differences of 1dB are ABXable.[a href="index.php?act=findpost&pid=327595"][{POST_SNAPBACK}][/a]


I'm not sure what you're saying here, or what 6/2 is supposed to mean.

I'm referring to the well-known noise-masking-tone results from Hellman, etc, that have been tested quite a few times.

I'm saying that you havea tone energy -6dB relative to noise energy (for noise in one critical bandwidth), and then it's inaudible.  That's for a tone and narrowband noise centered on the tone, of course.  That's when the TONE disappears.

If you use a tone to mask noise, you need more like +30dB on the tone to make the same noise disappear.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #32
Quote
The subsonic level, measured between tracks, is always higher than any of the other background noise. It is often higher than moderate level audio.[a href="index.php?act=findpost&pid=327959"][{POST_SNAPBACK}][/a]


Indeed, that would be the "rumble" of a few posts ago, and yes, it is quite large in many cases, due both to pressing as well as playback issues. (and cutter lathe, etc, etc, etc)

As an aside, I wonder how the subsonic filter you use is implimented.  Highpass IIR filters at low frequencies can be quite a pain to impliment, especially in fixed point.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #33
Since I don't do anything in fixed point, none of those issues are relevant to me. CoolEdit (now Audition) has a variety of flter types. Its recommended rumble filter is IIR, a Bessel filter, but there are other choices available that can do the job.

I most often use the filter in the Sonic Foundry NR pluggin. Its construction is not revealed. It seems to go a good job, except in those exceedingly rare cases where there is anything significant at 30Hz. Its cutoff slope is too gentle to fully preserve that frequency.

"ripping" LP's

Reply #34
Quote
Quote
would be to try recording the end of Neptune from The Planets Suite at both 16 and 24 bit resolution, then try to ABX the results
This might be interesting but you forgot to add that the LP must be of the highest quality vinyl, in pristine condition, properly cleaned, on the best playback equipment. And both recordings done simultaneously, since two separate playings will give slightly different outputs. Then one should be able to more confidently state 'yep, there isn't any difference to be heard'[a href="index.php?act=findpost&pid=327896"][{POST_SNAPBACK}][/a]
The point about getting the best quality vinyl/playback equipment is fair enough, in order to demonstrate whether greater than 16 bit recording is *theoretically* necessary. But to a certain extent those who claim that it isn't are in a no-win situation: no matter what vinyl is used to show that 16 bits is enough, the opponents can always claim that there might be better vinyl out there that wouldn't pass the test. The bottom line is that this is a practical issue: what's required for transparent transfer of real-world LPs in someone's record collection?

Quote
There is also the complication of how to do manage the 16 bit recording. You don't want to do it on a different, 16 bit only, soundcard or you have already compromised the comparison conditions. Will you get the 16 bits from the 24 bit recording by truncating? or by converting with dithering and noise shaping?[a href="index.php?act=findpost&pid=327896"][{POST_SNAPBACK}][/a]
It seems to me that the thing which is most likely to demonstrate a real, audible difference would be to simply make a single 24 bit recording, then truncate it to 16 bits without dithering. Are we agreed on this? In that case, it is my personal belief based on experience that the noise level of vinyl is sufficiently high that even this rather brutal approach will still yield inaudible differences. In effect, vinyl surface noise is high enough that it auto-dithers any digital recording way above the 16 bit threshold.

"ripping" LP's

Reply #35
I think yes, if there is some difference to be heard, it should be demonstrated under that condition.

I don't suppose it will ever happen, but I would be quite interested to play with a 'laser' turntable. While it has problems of its own, its output is supposed to be completly free of surface noise and rumble. That should make quite a difference on 78s and could also be a more critical test of 16 bit vs 24 bit capture.

"ripping" LP's

Reply #36
One thing I didn't see mentioned about normalization is that if you are going to do it, you should do it before track splitting so you maintain the "as produced" difference in volume between tracks.

My personal results on s/n are up to 72 dB as reported by Cooledit.  I use an Adcom 710 preamp and the on-motherboard sound (some analog devices ADC.)

"ripping" LP's

Reply #37
Quote
One thing I didn't see mentioned about normalization is that if you are going to do it, you should do it before track splitting so you maintain the "as produced" difference in volume between tracks.

My personal results on s/n are up to 72 dB as reported by Cooledit.  I use an Adcom 710 preamp and the on-motherboard sound (some analog devices ADC.)
[a href="index.php?act=findpost&pid=328063"][{POST_SNAPBACK}][/a]


How do you separate signal from noise with this measurement?
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #38
What filter settings are recommended for a rumble filter? I'm trying to get one working using Nyquist in Audacity. A cutoff frequency of 20hz seems quite reasonable, but given that bass is attenuated in the difference signal, I was wondering if a higher cutoff (say 60hz; some say 140hz is ok) would be useful for it, and 20hz for the mono signal.

"ripping" LP's

Reply #39
EDIT: Got confused by my waveform plots. There is something in one of my vinyl rips which correlates somewhat to sub-20hz signal in the CD version, but I'm not sure if it means anything.

"ripping" LP's

Reply #40
There are a number of filters in CoolEdit that can function for rumble removal. The preset for the Bessel filter I mentioned earlier is actually on the Butterworth tab. It has a 28Hz cutoff and is 18th order, very steep. A frequency analysis plot shows its effect somewhat visible up to about 40Hz. There are several other filters in that set that also seem to work adequately given similar cutoffs.

There is an FFT filter that produces a slope downward from 30Hz when I set its cutoff at 20Hz.

The 30 band Graphic Equalizer does a reasonable job if one just pulls down the first band (31Hz) the full 15dB. Its results looks to slope downward from about 60Hz.

The unspecified filter in the Sonic Foundry NR package produces a somewhat less steep slope starting about 80Hz.

Any of these works well in general. The Sonic Foundry filter especially must be considered in terms of the recording. Most LPs have no music to lose at those lower frequencies, but I recall one recordings in particular with some 30Hz tones that were largely eliminated when I tried the Sonic Foundry filter.

"ripping" LP's

Reply #41
Quote
g?

I also agree that normalizing is best done as the last step. A normalized file can easily go above 0dBFS through other following operations. That doesn't present any serious problem when working in 32 bit float but it means permanent clipping in 16 bit files; in floating point files you will have created the necessity of re-normalizing to get back down to the proper level.

One thing that might give normalizing a bad name is insisting on 0dB. It is best to normalized to -0.2 or -0.3dB to avoid potential clipping problems in the DAC, even though the kind of signals that can do that are not supposed to be very common in music. I've read claims that some CD players have such a problem. Also, my slow memory retrieval finally reminds me that many Creative cards have some kind of built in limiter that really squashes things unpleasantly at the upper signal level limits. I think that starts somewhere around -3dB, so a normalized file could sound significantly off if one is blessed with one of those cards.



More info on this phenomenon here (distortion of intersample peaks -- it's got to do with digital metering as well as D/A reconstruction)

http://www.cadenzarecording.com/papers/Digitaldistortion.pdf

http://recforums.prosoundweb.com/index.php...aa06e677aa7a9ee

http://www.audioholics.com/techtips/specsf...ngtrendsP12.php


If you record 'hot' you might end up with some nasties even if the digital 'VU meter' shows you've stayed below 0 db.


.

"ripping" LP's

Reply #42
Quote
What filter settings are recommended for a rumble filter? I'm trying to get one working using Nyquist in Audacity. A cutoff frequency of 20hz seems quite reasonable, but given that bass is attenuated in the difference signal, I was wondering if a higher cutoff (say 60hz; some say 140hz is ok) would be useful for it, and 20hz for the mono signal.
[a href="index.php?act=findpost&pid=328167"][{POST_SNAPBACK}][/a]



I would set Hp filter for L+R at 16Hz.

I would set Hp filter for L-R at perhaps 80Hz, perhaps even higher.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #43
Quote
EDIT: Got confused by my waveform plots. There is something in one of my vinyl rips which correlates somewhat to sub-20hz signal in the CD version, but I'm not sure if it means anything.
[a href="index.php?act=findpost&pid=328255"][{POST_SNAPBACK}][/a]



Train? Plane? A/C?

I've been able to observe planes flying overhead in the .1 to 10Hz range, inside an interior office in a large office building.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #44
Quote
The preset for the Bessel filter I mentioned earlier is actually on the Butterworth tab. It has a 28Hz cutoff and is 18th order, very steep.

[a href="index.php?act=findpost&pid=328299"][{POST_SNAPBACK}][/a]


I would use a 3rd order Butterworth HP myself.  18th order is way overboard.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #45
Quote
Quote
EDIT: Got confused by my waveform plots. There is something in one of my vinyl rips which correlates somewhat to sub-20hz signal in the CD version, but I'm not sure if it means anything.
[a href="index.php?act=findpost&pid=328255"][{POST_SNAPBACK}][/a]



Train? Plane? A/C?

I've been able to observe planes flying overhead in the .1 to 10Hz range, inside an interior office in a large office building.
[a href="index.php?act=findpost&pid=328440"][{POST_SNAPBACK}][/a]

An honest-to-god album, actually. I took a 30-second clip of a track from Autechre's Untilted, in FLAC format. There was a clearly identifiable noise floor from DC up to about 4 Hz. Then it increased by roughly over 20db/decade before jumping up to peak at 20-30hz. I have reason to believe that the LP version is from the same digital master as the FLAC version (clear lowpass at around 22 khz) and so I would guess that the same subsonics might exist in the LP version. However I don't have evidence that the subsonics in the LP version actually correlate well to the FLAC version. Some intertrack silence on the LP peaks around 10hz, and the same section of music that was tested in the FLAC version loses coherence in its frequency response around that point.

"ripping" LP's

Reply #46
Just curious. Ripping is a term often applied to digital audio extraction. I don't know its origin, whether it has any logic to it or is just a word someone came up with as easier to say. Here people are applying it to recording from analogue, which is a very different process. I have seen the word in the context of audio CDs often enough, but recording seems to be called recording most other places. Is this an ‘in' joke, or some such, around here?

"ripping" LP's

Reply #47
Quote
An honest-to-god album, actually. [a href="index.php?act=findpost&pid=328452"][{POST_SNAPBACK}][/a]


Well, yeah, but the question is from what did the signal arise?

Especially with good condenser mikes, a solidstate preamp, and digital recording, one can be very surprised by the low-frequency content in a room.
-----
J. D. (jj) Johnston

"ripping" LP's

Reply #48
something tells me autechre werent using mics and/or preamps ;]. too passe for those guys.

as for ripping, i always say im ripping records, ripping cds, ripping tapes. putting media into digital form i consider 'ripping'.

i think this discussion rules...coincidentaly i have over the past few days been ripping my vinyl to my harddrive and burning to cd. id like some feedback on the pros/cons of my process:

im using a stanton str8-100 with the spdif digital coax output to my soundcard.
using a shure whitelabel cartridge, brand new (but broken in).
recording to wav file, then burning to audio cd.

does anyone have any experience with the digital tables? I know that alot of quality now depends on the a/d converter within the table, but i am loving the results. a friend of mine said a digital turntable was like a digital wooden spoon :].

does anyone know do i lose any quality by burning to audio cd? i have heard that cda files are just wav files with indexes or something. i would rip back to my comptuer if needed using eac.

love yall

edit: realise this aint the most audiophile friendly way to do things, but just fyi, as far as i can tell, i get pretty damn near amazing results, based on shear listening enjoyment.

"ripping" LP's

Reply #49
Writing your recordings to CD-R does not involve any change in quality from what you have on hard disk if the .wav file is 16 bit 44.1kHz. You are writing them to a different media, which in this case involves a change of format, but the data is not being changed.

The only potential downside is data loss. If the CD-R blanks are not decent quality, it may be harder to read back from them than from hard drives. Errors in the data can result. Some errors are readily corrected perfectly, some more serious ones result in 'guesses' about what is correct. For the most part there isn't any problem.

The only definition I can find for 'ripping' is DAE. This does not involved 'putting' anything into digital form since the data is already digital. Audio data on optical disk is encoded in a different way that when on other digital media. It must therefore not only be read from the optical media but also converted to the 'normal' format for computers -- DAE. This is just a little bit different mechanically from copying from one hard drive to another.

I suspect 'to rip' is pretty much a slang term anyway, so the dictionary police probably won't be knocking on your door anytime soon. You might look at it as similar to the people who insist on calling their automobile engine a 'motor' when in fact there are clear definitions of engines and motors and they are not the same things.

Having the DAC in the turntable or in a box on the equipment rack or inside the computer is more or less irrelevant. It is the quality of the parts and the implementation that is important. If your results satisfy you they do. Could you tell any difference with a more expensive setup? It costs a pretty penny to find out.