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

[Marc Hildebrant]

Craig,

Yes, I do believe that it's some of both. I.E. more data for the algorithum and more "useful" data.

With regard to the best sound card...I will not be keeping my old system much longer. As such, I have no idea about the specifications for sound cards.

Any suggestions ? Also, the preamp/cartridge part of the system should be checked for frequency response.

Interesting Stuff !

Marc

[Craig Maier]

Hi Marc,

I do not know much about soundcards these days. The last time I looked into it was quite a few years ago. But, I would conclude that one would want at least a true 96 kHz sample rate performance, which is probably not too hard to find these days. I suspect that your preamp has adequate bandwidth (Marantz did things correctly as I recall). As for the cartridge, that probably makes a difference in performance in terms of frequency response beyond 20 kHz. But, I have found that I tend to use a cheap old Stanton 500 just because I have so many different styli types for it. I do have much better cartridges here, but they only have 0.7 mil styli, so they are no good for non-microgroove work. The cartridge that you mentioned that you have is probably pretty good as I recall. I guess the biggest limitation you have in your system is just the soundcard. Maybe just updating that would be a useful investment.

Craig

ps - interesting experiment - - - take your 44.1 kHz file from the jazz test record and just up-sample it using the sample rate converter in Diamond Cut (change sample rate under the edit menu). Then, clean up that file. I wonder if that simple conversion improves the quality of the restoration - - -

[Marc Hildebrant]

New Experiment (Interesting One)


Record "tough" song (12th Street Rag) at 44.1 K. Convert to 96 K with DC8 software.
Use previous filters
Play at 44.1 or 96 on system

No distortion present. Same results as sampling original at 96 K.


Now, I do not change the filters settings during all of these tests, so, another thought is that for each sample rate, and "optimum" value is found.

Marc

[Craig Maier]

Hi,

That is interesting - - - so the "more data" helps theory seems to have some merit (the sample rate converter does perform differing types of interpolations to create the new file - - - (your choice of CD Quality, Pro Quality or Master Quality which involves varying degress of complexity for the interpolator)). It would be interesting to see what happens if the controls are optimized for the two different sampling rates as you suggest. If the best settings that you can get for 44.1 kHz produces an inferior result compared to the best settings that you can get for 96 kHz, then that really says something in favor for the use of the higher sample rate.

Craig

[Dan McDonald]

Hi All... I just checked my soundcard using Craig's method and I get a drop at 21kHz for 44.1, a drop at 24 with 28kHz, and a drop at 21 with 96 and 192. With 192 I notice that the drop is from -92dB to -112 or so, then a flat line out to 48kHz. Is that pretty much what you mean by a brick wall? It looks like there's not "nothing there" - I can see what looks like electrical spikes or something like that all the way out to 48kHz.

When I upsample the 48kHz to 96kHz, I do get almost a completely flat line out to 48.

Very odd stuff.

[Craig Maier]

Hi Dan,

That is the same thing that Marc got. I was able to get up to roughly a 48 kHz brick wall using 96 kHz sampling rates here using two different soundcards. Dan - - - how old is your soundcard? Mine are roughly 4 or 5 years old. I think that Marc's are a little older than mine.

Craig

[Dan McDonald]

Mine's about 5 years old I think. I just tried using my 'editing' computer, which is about a year and a half old, and only has an onboard sound card. I have DC8 on here, but don't use it for anything but editing audio and video. I 'made waves' at 85000 Hz and checked with the spectrum analyzer. There was a big spike where it should be, and the same was true for 96 kHz and 1 45000 Herz spike I created, so the onboard soundcard seems to be much more capable.

Although I didn't try 'make waves' with the soundcard I use for recording. Maybe it can 'read' but can't 'write'?

Dan

[Marc Hildebrant]

Group,

More tests today. The "optimum" setting for the impulse filters is very much different between the different sampling rates. So far, I can use a "more aggressive" value with the 96 than the 44.1 files.

Also, I haven't found, at this time, differences between the varies methods to create the 96 sample file with the DC tools...i.e. PRO/Master/CD

Interesting stuff.

By the way, this "stuff" didn't come up with all the 78 RPM and Edison Diamond Disc work. It seems to be important only for the LP work.

Marc

[Craig Maier]

Hi,

Based on Marc's and Dan's soundcard frequency response (or bandwidth) measurements, I think that there is a compelling reason to try half speed mastering (16 rpm transfer of 33.3 rpm LPs). Both of their soundcards brick wall around 22 kHz without regard to sample rates of 96 kHz. I only say this to folks who have soundcard with upper frequency response limits of 22 kHz even when sample rates are set for 96 kHz.

So, I would like to offer the fact that transferring an LP half speed effectively doubles the bandwidth of a soundcard. So, signals that were at 20 kHz become 40 kHz after the transfer and the speed change conversion. You can effectively increase the bandwidth of your soundcard using the following process.

The process would be different between RIAA preamp setups and flat preamp setups. First, here is the procedure for RIAA preamp setups:

1. Transfer the record with the soundcard set for a 96 kHz sample rate with the record playing at 16 rpm. (it does not matter that your soundcard brick walls at 22 kHz).

2. Convert the transfer to flat via the Virtual Phono Preamp with Hardware set for RIAA and record type set for acoustical

3. Now, convert the speed to double with the change speed filter.

4. Re-convert the resultant file back to RIAA with the Virtual Phone Preamp - this time using the Harware = flat and the record type = RIAA

5. Now, go about restoring the recording as always.

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

For flat preamps:

1. Same as Step 1 above

2. Convert to RIAA using the Virtual Phono Preamp

3. Restore the recording as always.

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

Does this make sense? It should effectively double the bandwidth of your soundcard up to rougly 44 kHz.

Craig

[Craig Maier]

Hi Marc,

The different interpolation methods may be hard to hear in the sample rate converter, but they make a difference when measured with the spectrum analyzer. But, the difference in quality is way down in the mud of the signal, so it could be easily argued that it does not matter much. In the old days, the "Master Quality" used to take more processing time, but with todays machines, it probably makes little difference.

Craig

[rickc]

Wow, good (and long) thread here. Figured I could add some confusion to the mix.
There are reasons why the click filters would act differently when used at 96kHz vs 44.1, but each of the filter have their own
set of differences and so your results will depend on which filters are in the mix

1. Expert Impulse: Depending on the mode, some components of the filter will scale with sample rate and some do not.I would expect that this filter is highly dependent on sample rate, and should be optimized for each setting.
2. Big Click filter: Parameters are based on time, so I would not expect this filter to be depending on sample rate.
3. Narrlow Crackle filter: Some scaling on sample rate, but I would suspect this is also affected by sample rate.

So there are the differences based directly of the scaling of filter frequencies and sample sizes as outlined above, then there are the unintentional differences where the higher sample rate allows for the digital filters to become more ideal digital filters.
As any digital filter is operated at near its Nyquest frequency, the response becomes much less like and ideal filter and can differ significanly from its theoritical shape.
So making the sample rate 96kHz vs 441k, even if all other parameters were truly independent of frequency would still make a audible difference in the click filters

Now none of this says that it will make the filter perform any better or worse at either sample rate, but it does go back to what has been discussed here as to having more high frequency info to work with will enable the filter to better discriminate between noise an music.

[Marc Hildebrant]

Rick,

How does the Ez Impulse filter behave with sample rate ? Same as Expert ?

Marc

[Craig Maier]

Marc,

As I remember, they both (EZ and Expert) scale similarly regarding sample rate.

Craig

[Marc Hildebrant]

Group,

So.....The simple answer seems to be, record at 96K samples a sec. and perform the impulse noise at those settings. Convert to 44.1 K samples a second for follow on work. (I think that this is the recommended method in the DC users guide).

The reason for this approach is :

1. Impulse filters benifit from more data during the decision about noise vs song.
2. Some additional high frequency data is present at the higher sampling rates due to the imperfections in pratical low pass filters ahead of the A/D(This is related to the fact that low pass filters have to be causal ...i.e. do not have negative time).
3. Storage today is "cheap" so the ability to use 96 K rates is not a significant cost.

Well....it's been interesting.

Marc

[Craig Maier]

Good summary -

Craig