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

[Craig Maier]

Quoting Dan:

"I think I'll use 24-bit, 96 kHz with my little recorder for my transfers."

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

Neither 24 bit or 96 kHz can hurt the signal - - - they can only help from what I can tell. I have only held the position that 96 kHz is probably more important than 24 bits.

Craig

Craig

[Dan McDonald]

I'm getting very odd things. I tried to generate 96000 Hz at 192000 Hz sampling rate and it said I can't do it because it has to be half the sampling rate (which it was). So I generated it at 95000 HZ and it was fine. Whin I applied the VPA (as if it were a flat preamp), the image disappeared completely in the pkf file. (like it was completely empty). The same thing happened if I converted it to 44.1, which you'd expect.

When I run the make waves at 1000 Hz, I get exactly what you get.

When I run the make waves at 25000 Hz, you see a major reduction when you run the VPA, and you get noise down at the floor (up to 560 or so Hz). I guess that makes sense, since it's emphasizing the bass end of things, so all it has to work with is some very low level noise (around -75 dB after the VPA). I'm a little surprised that the make waves would produce low level noise - maybe you know where that comes from.

So I guess this is saying that the lower tones are emphasized and the upper frequency tones are de-emphasized. I guess the reason you get a different effect with high sampling rate files is that there must be some sort of percentage reduction or some curvilinear process where the further up you go in the file, the more reduction you get.

I'm not sure if this makes sense but I think it's saying that it's based on some sort of proportion of the available range, rather than based on absolute frequency. Does that make sense or is that even possible?

That way, you'd get a lot of reduction in a 96kHz file at the top end, but by the time it got down to 20kHz, there wouldn't be much reduction in the area that actually included the sound you were interested in. However, if you start with around 20kHz at the top end, you'd get quite a bit of reduction in the music portion. I think that would explain all of these results. Now if anyone understands what I am trying to say, then please let me know

[Craig Maier]

Hi Dan,

It is pretty late, but I do understand one thing about your post. If you try to generate 96,000 Hz signals with make waves generator using 192,000 Hz sample rate values, it will not take because that that is the exact Nyquist value. The theory says that the sampled signal must have no components at half the sample frequency and that was exactly half. So, the flag thrown out is effectively correct, but it probably should say that the signal needs to be less than half the sample rate, so that is a syntatactical error in the Diamond Cut software.

I have to think about the rest of your post and will try to respond later after a little shut-eye.

Craig

[Dan McDonald]

Hey Craig - get some sleep!
Anyway, I'm apparently wrong in part of my ideas. I tried 400 Hz and 14000Hz, and they both behave the same, whether it's 44.1 or 192kHz sampling. The 400 is increased by the VPA and the 14kHz is decreased by the VPA.

So the odd things seem to be happening when frequencies are outside of the normal range.

I just tried it again with 18kHz and still get the same results for both sampling rates.

Dan

[Craig Maier]

Hi Dan,

I take it that you are saying that frequencies out of the normal range do it such as ultrasonic signals?

Craig

[Dan McDonald]

Yes, that is what is happening. Anything I've tested above 20kHz has shown the effect; anything below that doesn't seem to make a difference what sampling rate was used. The odd thing is that I still don't understand why the vpa then seems to 'flip' the effect in that a 44.1 sampling rate with music in the file will show a loss of dB but a 96kHz file with music will not show the loss in dB associated with use of the VPA. It seems like that is the opposite of what you'd expect from the analysis of the generated waves.

Maybe I'm just stuck on the problem too much. There must be some sort of logical explanation.

Dan

[Dan McDonald]

For a little additional information, which I think pretty much explains all this. I switched to 88.2 kHz instead of 96kHz for these tests.

If you run the RIAA (higher precision) preset in the paragraphic eq, at 0dB change, the 88.2kHz files act like 44.1kHz files, with essentially no change in gain.

If you run the VPA with the flat preamp preset, it increases the gain of the 88.2kHz files. I have to reduce the gain about 11dB to make the two different ways equivalent.

If you downsample to 44.1 before running the VPA, it's the same as the original 44.1 file.

Both the Paragraphic EQ and the VPA appear to lower the high frequency info about the same amount when run on the 88.2 files. You can see the before-after difference in the spectrogram. There are lots of spikes maxing out at 44.1hHz, but after you run the VPA or the ParaEQ, both max out around 24kHz, with occasional spikes going all the way out to 44.1. I assume this means that both are pretty much doing what they are supposed to be doing, and everything I can see is telling me that the end result should be the same for the 88.2 and the 44.1 files (the spectrogram is almost identical), but something in the action of the VPA (but not the ParaEQ) is increasing the gain when the sampling rate is high.

I think I'm done testing this. Now I just know that the higher the sampling rate, the more I need to reduce gain in the VPA.

Dan

[Marc Hildebrant]

Dan,

Interesting stuff ! Maybe a "tweek" to the code is needed ?

Marc

[Dan McDonald]

That's what I was guessing - that there's some sort of connection between sample rate and gain in the code, but I know so little about programming and the math behind the audio that I really have no clue. I just know that I have to reduce the gain if I have high sampling rates, and that's ok for me.

[Craig Maier]

It could be a overall scaling error in the code, but the relative levels as a function of frequency should be spot on (the relative levels have been carefully checked against the theoretical before we introduced the VPA some time back). There is a big difference between how the paragraphic EQ realizes these curves and the way that the VPA does this. The paragraphic EQ is a set of 10 resonant filters with variable Q, amplitude and frequency. The VPA uses closed form mathemtatical realizations of the various EQ curves shown as checkboxes therein.

Craig

[Dan McDonald]

Thanks Craig - Please don't take that as a complaint or anything. I can't begin to imagine the complexity of DC8 and the interactions that must go on with something like this. I'm just fine scaling the gain down with high frequency files.

Dan

[Craig Maier]

We do not take it as a complaint - - - just something to look at sometime before the next release. We just never noticed it during development and testing. Our focus was always on the relative dB levels vs Frequency vs EQ curve and their relative accuracy.

In the interim, note that the VPA has a Volume Control (Gain). Maybe you can save a special preset or presets with the Volume Control set to compensate for the gain difference along with some other settings of interest to yourself when using 96 kHz sample rates and the VPA.

Craig

[Dan McDonald]

One other question about the implications. Since the high frequency material is scaled down with the VPA, it seems like it would make sense to de-click before the VPA. I know the presets are set up to deal with the post-vpa situation, but I'm wondering if it would make sense to develop some good settings as presets for the pre-vpa.

In terms of presets, I have already started on those. I seem to like a little more treble than is typical anyway. I think maybe my high frequency sound is not as good as it used to be. I like hearing drums and cymbals very prominently. There's a high-frequency component of drums that doesn't come to the forefront in commercial recordings or lps. I do hear it a lot more prominently in jazz recordings than in rock or other forms, but I like that sounds - it makes it more 'live' to me. So I tweak the RIAA a little and add some extra treble end.

Of course, if someone with really good hearing should listen to my music, it might hurt their ears!

Dan

Dan

[Marc Hildebrant]

Dan,

I have found that I like the "highs" if the recording is clean of distortion.

Marc

[Craig Maier]

Dan,

I do not know how to optimize any of the impulse filters for non-standard (flat) transfers. When we developed the routines, we always applied the appropriate EQ first, whether it was involving electrical 78s, 45s or LPs. So, it is uncharted territory for us (using flat transfers directly into the impulse filters). I can say that we intentionally designed into all of those filters a very wide range of adjustability so that almost any situation would be able to be accomodated by a user. In other words, on just about any material that you can imagine as the source, we provided ranges for all the controls that went from almost doing nothing to the signal to creating total distortion. You may have observed that to be the case with the impulse filters (EZ, Expert, Narrow Crackle, Big Click, EZ Clean).

This has turned out to be an advantage especially to Forensics users in extremely unusual restoration situations (radio transmission crackle, etc).

Craig