Optimum settings for recording LPs with further editing (sample rate and bit depth)

[eddie the eagle]

Hi All,

For those that haven't looked at them, I'd recommend the Tracer DVDs on how to use Diamond Cut - very informative.

One thing that I picked up was that you need to record an LP at settings higher than 44.1kHz to avoid artifacts (which are very obvious. I'm running an ESI Juli@ card that's capable of 192kHz 24bit sampling, which crashes DC8 and is too high a sampling rate for the big click filter (rejects it.) After about 8 seconds of sound, DC8 hangs and then crashes.

Rather than experiment, I believe that someone out there would have done the hard work. Is anyone able to recommend a starting point for recording LPs to save a day's mucking around?

THanks in advance,

Andrew

[Dan McDonald]

I use 24-bit, 96 khz and that seems to work just fine. I've done transfers and cleanup that people can't tell it was from an LP. I think the high sampling rates are mainly used to eliminate the frequencies of clicks that are there but we can't hear, to help the de-clickers do their jobs. I think the 24-bit is mainly there for other filters that need a little room to operate.

I'm guessing the 192 problem you are running into has something to do with your system resources.

[Marc Hildebrant]

Dan,

Have you ever found an application for LP's where more than 96/24 is needed ?

Marc

[Dan McDonald]

No - I haven't. Have you? 22kHz is higher than most of us can hear, so 44.1 is fine for that. The only issue is whether the filters are improved or not, and I haven't seen any improvement above 96kHz. I don't know if I even noticed it above 48kHz, but I still do it because I figure there's probably some benefit.

Dan

[Marc Hildebrant]

Dan,


While my hearing has held up good (15KHz) over the years, the rest of me has worn out !

My question was in regard to removing noise with the various filters. I had tried removing noise from a "Wooly Bully" recording at 16 bits/44.1KHz sample and wasn't having a good result. When I tried at 16 bits/96 KHz, the results were much better...hence my question.

Marc

[Dan McDonald]

hmmm. That's interesting. I never found much benefit. Was that record in pretty bad shape? Sam the Sham & the Pharoahs, I assume

[Craig Maier]

Hi,

There are ultrasonic signals contained in the impulsive noise that are "heard" by the software and used to help discriminate between the noise and the good audio. So, using 48 kHz or 96 kHz will improve the relative levels of the impulsive noise signals (so long as the higher sample rates are not emulated {some low cost soundcards do emulate these sample rates rather than actually sample at those rates}). That means that you should get better discrimination between imulsive musical events vs ticks, clicks and pops with the EZ Impulse Filter, the EZ Clean Filter or the Expert Impulse Noise Filter with 48 kHz or even 96 kHz sampling rates (but I think that the law of diminishing returns and not benefit is realized at higher rates than 96 kHz).

As for signal depth, I just use 16 bits myself becaue that represents 96 dB and the records only have around 50 dB of Signal to Noise ratio.

But, I will say that it does not hurt anything to use a higher resolution.

Craig

[Craig Maier]

Hi Again,

BTW - just as a generality, I usually transfer my 45s and LPs at 48 kHz and 16 bits so that my archive database does not become too large. On certain occasions, I will transfer at 96 kHz when I think that the quality of the LP could benefit from the doubling of file size.

Craig

[Marc Hildebrant]

Group,

Well, good old Wooly Bully seemed to benifit from more samples (96 vs 44), but, I have just started down the LP road at this point and I'm not an expert at all.

I have spent a lot of time getting the 78's in good shape, and I'm finding that the "modern" lp is much different to clean up. It seems to be in the nature of the impulses between the two technologies.

Marc

[Craig Maier]

Hi,

Yes, the dv/dt, the length and crest factors of the pulses are much different between 78s and LPs (and 45s). LPs (well, microgroove recordings) will benefit more from the higher sample rates (such as 48 kHz or 96 kHz) than 78s. I recommend de-clicking with the EZ Impulse filter with the settings starting at around 50/50/50. If there are up-front vocals or brass instruments, turn on the Solo/Brass feature which helps to discriminate between certain waveshapes produced by those types of audio signals and impulsive noise. I use the Big Clicks feature only on severly gouged recordings.

Let us know how you make out.

Craig

ps - it is also important to use the correct EQ curve. Generally, that is not an issue as RIAA was pretty standard after around 1956 or something like that (details are in the users guide). If it is non-RIAA, it is a good idea to convert from the RIAA (or flat) transfer to the correct curve before de-clicking the recording. This EQ Correction can usually be accomplished using the Virtual Phono Preamp found under the filter menu or via the Paragraphic Equalizer. There are lots of presets therein to help out with that task.

[Craig Maier]

Here is a little more on bit depth pertaining to record transfer and clean-up. Since dB is a log scale, and assuming that LPs or 45s have a 56 dB signal to noise ratio (being generous), recording with 16 bits of resolution (96 dB) produces much more resolution than required. For example, given the above assumption, the headroom in dB between the LP recording and the 16 bit digitization bit depth will be 96 dB - 56 dB = 40 dB. 40 dB of headroom does not mean that it is roughly 2:1 in advantage. From the digitization process perspective, it represents 100 : 1 (on a linear scale) more than required to accurately capture the signal.

Craig

[Dan McDonald]

Craig (or Marc or anyone) - so if LPs benefit the most, is there a relationship between speed and sampling rate benefit such that 78s would benefit the least, 45s next and LPs last or is it as much an interaction of speed and the material the record is made of?

I'm trying to figure out why Marc found it for Sam the Sham. Maybe an experiment is in order, with a useless lp being played at 33, 45 and 78 speeds, or a useless 78 played at 78, 45, 33 and 16?

Or is all of this just physics and no reason to do the experiment?

Dan

[Craig Maier]

Hi Dan,

Well, as they say, one experiment is worth a thousand expert opinions. I would never use theory to discount a well thought out experiment.

But, in theory, it comes down to the tangential velocity of the record surface coupled with the stylus width. Somewhere on this form (some time back) I did a little analysis of the max theoretical bandwidth of various record formats coupled with groove width/stylus width. And since the signal to noise is represented as a ratio, I would not expect that an LP played at 78 would improve the de-noising process. Actually, you would need a flat preamp and an exceptional phono cartridge to do that properly and they would have to have a bandwidth of roughly 50 kHz or more to work properly.

[Craig Maier]

Hi Again,

I just found that analysis of record format frequency response. If you are interested, click here:

http://www.diamondcut.com/vforum/sho...ical+bandwidth

Craig

[Dan McDonald]

OK - I was thinking more of the 'signal to noise' ratio of the ultrasonic signals generated by the needle hitting the crack or pit in the record. That was why I was thinking that the experiment would use the same medium at different speeds and different sampling frequencies. I was thinking that that would hold everthing constant except speed and sampling values. Then you'd have pretty definitive information about the best way to approach it.
Something like:
Speeds: 33 45 78
Format LP
45
78
at 3-4 different sampling frequencies.
Like I said, I've never found benefit above 48 kHz but tend to go with 96 just because I can and I figure it doesn't hurt (except for storage).
Dan