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

[Dan McDonald]

With just a few runs, it looks like:
96kHz can be more aggressive than 44.1. Especially with narrow crackle. Also getting good results with 88.2 sampling rate, which this card also supports. I don't see any difference in performance between the 88.2 and the 96 kHz sampling in terms of any distortion.

If I remember correctly, there was a suggestion in a thread at one point that 88.2 might be a better rate to use if the end product was a CD because of potential errors in downsampling, although I think there were minimal benefits.

48kHz sampling rate is providing good results, but you have to be less aggressive with the de-clicker.

If you use the VPA before de-clicking, the same settings on EZ-Click produce fewer click totals. The good information out of this is that it means you get less distortion because it's operating on fewer potential clicks. I can't really hear a difference in terms of click removal, but it's very evident during processing.

One thing I noticed, and I'm not sure what this means:
When I use the VPA to convert from the flat preamp file to a 'normal' file, at the higher sampling rates, I get a much 'louder' signal (according to the pkf). I'm not sure why that would be. That is, keeping the same setting and running the 4 files (I didn't touch the levels when I made the recordings) one after the other, I gain-normalized at -1.5dB.
gain-normalizing the 96kHz file lowered it (you could see the pkf 'shrink')
88.2 shrunk a little bit
48kHz increased considerably
44.1 increased more.
These were visible in the pkf file before I did the normalizing (and why I did the normalizing). I just did it again to make sure I was accurate, and got the same results. Would this be expected?

Dan

[Dan McDonald]

also.. regarding the gain normalizing issue, they all look about the same before the VVA alters them.

[Marc Hildebrant]

Dan,

With regard to the VPA ; you are using a flat preamp (Let me know if I'm wrong) and with the RIAA recording, the raw signal has much more high frequency content then the same signal after the RIAA conversion in the VPA.

With more high frequency information, the impulse filters should have "better" information to make decisions about noise. While I do not know the details of the algorithums....I'm assuming that the extra high frequency info converts to fast rise times in the time domain and hence provides more info about noise (which is high frequency ).

Marc

[Dan McDonald]

Hi Marc - Yes, I'm using the flat preamp.


Are you saying that the VPA and normalizing makes sense to you?
If the VPA converts the high frequency material to lower frequency, shouldn't the material that has the most high frequency info (noise) be most affected? It seems like it is doing the opposite, so that the material that has the least high frequency content is most affected by converting the high frequency via the VPA.

I probably don't know enough about how the VPA works to have any idea about the explanation, but I thought it was interesting and wanted to report it in case it had implications.
For example, that might be an advantage to 24-bit recording if you are using the lower sampling rates. If gain-normalizing is going to increase the signal, it might be useful to have additional bit depth.

Dan

[Marc Hildebrant]

Dan,

Let me try and describe the events and see if it makes sense to you/me/Craig/anyone else for that Matter.....

The original recording has the high frequencies boosted and the low frequencies attenuated per the RIAA curve.
When the song is first subjected to the impulse filters prior to the VPA stage, the filters are working on a song that has more high frequencies than "normal".

After the VPA stage, the song is converted to the "normal" sound by using the RIAA curve again but this time the curve is opposite to that used first. Thus the RIAA effect from the original recording is cancelled out.

If you use the impulse filters on the normal sound, then they have good info for clicks/pops (clicks/pops are high frequency events). If the impulse filters are used on the extra bright sound (i.e. before the VPA), then they may have better information to find the clicks/pops because all signals from the song are sent out with extra high frequency energy.

If you look at the RIAA curve in the paragraphic EQ filter, you can see the boosting of signals.

Let me know if it makes sense. It's only a theory at this time to explain why you get different amounts of noise events with your experiments.

Marc

[Dan McDonald]

Hi Marc -

Yes, I understood that in the same way you did. The puzzle I was having (which I wasn't explaining well) was that when you use the VPA on files with high sampling rates, the resulting file is close to 0 db in gain (this is for my files, with the levels all set the same). WHen you do that same process with a file that is recorded at 48 or 44.1 kHz, the result is a much lower dB (something like -12 dB, just guessing from the picture), and you have to gain-normalize it to get up to the level of the high frequency files.

That was what I was having trouble understanding. That was why I was suggesting that there may be some benefit to recording at 24-bit for low sampling rate files (because you will need to normalize when you are finished) as compared to higher sampling rate files.

Does that make sense?

Dan

[Marc Hildebrant]

Dan,

I'm as puzzled as you are with regard to the change in gain.
I have seen some "strange" gain changes in the past also.

Marc

[Craig Maier]

Hi,

Here is a theory - - - signals recorded with wider bandwidth (ie 96 kHz sample rates and 48 kHz uppper frequency response limits) contain more High Frequency signal content (ultrasonic frequencies) and, thus create a taller peak file. The signal that we can not hear as humans can still produce taller peak file signals that the software does respond to both with regard to the filters and the display and gain normalize algorithms too.

But, that is a theory. Does anyone think that could explain it?

Craig

[Dan McDonald]

I was thinking along those lines, but then I couldn't figure out how/why applying the VPA would make the smaller frequency files look smaller in the pkf (and are lower in dB) than the higher frequency files.

Well, I think I might see how this could work, but not sure. I guess the question would be what would happen with the files if you changed the sample rate from 96kHz to 44.1 after running the vpa -- would the two files look the same? It seems like they should if it has to do with high frequency information, right?

Dan

[Craig Maier]

Yes, that seems reasonable to me -

Craig

[Craig Maier]

And, I wanted to answer Marc's question - - -

"I have seen some "strange" gain changes in the past also."


----------------------

There are some filters that (on the surface) do not seem that they should exhibit gain, but do have some gain. For example, the IIR Low Pass Filter, High Pass Filter and Bandpass Filter(s).

If those are used with Higher Order (higher than first order), there will be some peaking that will occur at the corner frequency and in the pass-band, especially the Chebyshev response ones. That is just part of their mathematical definition and so it can not be avoided. But, I am not sure if those are the ones that you were referring to, but they jump out as examples of this sort of phenomenonen.

Craig

[Dan McDonald]

Just tried to change the sampling rate of the 96 file after VPA - they still look very different.

However, if I change the sample rate before applying the VPA, they both look the same (lower dB). The second part makes sense to me, since they should ve virtually identical when I apply the VPA, but I don't understand what the VPA does that makes so much difference in the file when you're using high sampling rates.

Dan

[Craig Maier]

Hi,

I do know that the closed form EQ curves portion of that algorithm do involve calculations that include sample rate as part of their solution. It is possible that there is an overall sample rate scaling error within the VPA, but I am not sure. The only thing that I remember checking with it was the relative accuracy of it with respect to the various EQ curves (which showed it to be basically as perfect as the measurement systems would allow). It is possible that the absolute scaling is effected by the sample rate, however.

Craig

[Dan McDonald]

OK - well I'm not going to worry about it too much. I think I'll use 24-bit, 96 kHz with my little recorder for my transfers. That looks like where all the evidence is pointing to as a good thing to do. the 24-bit may not be necessary, but I figure I might as well, since I have it.

It's too bad about the soundcard issues.

Dan

[Craig Maier]

Hi again,

I just created a 1000 Hz sine wave with the make waves generator at 96 kHZ. I ran it through the VPA and observed that the signal passed to the destination file with no change in amplitude. Then, I converted that file to 44.1 kHz and repeated the VPA application with the same result. Dan, can you try that experiment?

Craig