RIAA DIfferent in 96 KHz vs 44.1 KHz

[Marc Hildebrant]

Gang,

If I have a 96 KHz file and use the "Flat to RIAA" Virtual Phone Preamp the sound is different than the same operation done on the file converted to 44.1 KHz.
The 96 KHz seems to have much more bass and volume than the 44.1 KHz version.

If I use the Paragraphic to perform the flat to RIAA, the 96 KHz and the 44.1 KHz file sound the same.

Same result with DC 8.5

Marc

[Craig Maier]

Hi,

Rick and I tested the Virtual Phono Preamp response at 44.1, 48 and 96 kHz back when we developed it. My recollection is that the response of the system was following the RIAA calculated values mathematically perfectly at 48 and 96 kHz, and had a very slight error at 44.1 kHz sample rate (fraction of a dB at 20 kHz as I recall). As far as I know, we have changed nothing in that algorithm since we created it years ago.

Craig

[Marc Hildebrant]

Pictures of the "bug". Song was "Check Your Bucket". Notice the increase in amplitude with the Virtual Phono Preamp.
You have to click on the photo's to see all of them. I have 4 shown but the forum shows three ?

Marc

[Craig Maier]

Hi Marc,

The overall gain error would not be in the VPP, but it is likely an error in the Paragraphic EQ.

Craig

[Craig Maier]

Hi Again

Here is a third party technical evaluation of the various systems for creating the RIAA curve. Note that the Diamond Cut Virtual Phono Preamp beats all other methods in terms of gain and phase accuracy vs frequency by a long shot:

http://www.tdl-tech.com/arscj.pdf

[dbanani]

Hi Marc
I just duplicated the same thing you are seeing.
Using fast edit mode with two identical files(different sample rate/resolution)
Displayed in Windows Tile view, the difference is very noticeable.
D

[dbanani]

Hi Craig
I only used the Virtual Phono Preamp, no Paragraphic EQ
DC8.5
D

[Craig Maier]

Maybe there is an offset error between the two, but the relative gain vs frequency should be exact for each sample rates.

Craig

[dbanani]

Took the two files and identified a segment 936.5 ms in each using easily identifiable reference points.
I then gain normalized each segment to -0.2
Using the Spectrum analyzer, the higher frequencies are noticeably different for each segment
The 44.1/16 file has a significant reduction in high frequency response, especially above 10k
All processing for each file was exactly the same except for changing sample rate/resolution
D

[dbanani]

I just tried the Spectrum Analyzer on a file before and after converting to 44.1/16 - NO VPP.
I get the same problem without the VPP processing of the file
The change sample rate/resolution function seems to be where the problem is occurring.
Additionally, the change sample rate/resolution process changed the length of this file from 472.5 ms to a new length of 481.0 ms a difference in length of about 1.79 % in the newly created 44.1/16 file

[Marc Hildebrant]

Another interesting test :

Using "Make Waves" I made a 1000 Hz sine wave at 96 KHz. At 1000 Hz, the RIAA curve is 0.0 dB, so the Paragraphic EQ gave the expected output , while the Virtual Phono Preamp was 6 dB higher.

With a sine wave at 44.1 KHz, no difference between the two filters.

I do not know if the difference is just a gain change or a frequency difference.

Marc

[Craig Maier]

The only thing that is important is the relative amplitudes of the various frequencies associated with the phono EQ curves. If there is an overall gain error, the VPP has an overall gain control so that you can set it to whatever you need or want.

Craig

[dbanani]

I used the Spectrum analyzer in DC10 in high precision mode
Then compared traces with two identical files after converting one to 44.1/16 and then using the VPP on both files
The 44.1/16 file shows about 14 dB less at the lower frequencies and tapers to about 5 dB difference at the higher Freq.

With the rumble filter off, the differences are not as pronounced - about 10 dB below 1000 Hz tapering to around 5 dB difference above 10000 Hz

[Craig Maier]

Use the make waves generator in conjunction with this chart of calculated values to measure the system performaNce. Note that the reference is 1KHz.



RIAA Curve Table of Values

Frequency in Hz	Level in dB referenced to 0 dB @ 1 kHz*	Frequency in Hz	Level in dB referenced to 0 dB @ 1 kHz*
20	+ 19.3	800	+ 0.7
30	+ 18.6	1,000	0.0 *
40	+ 17.8	1,500	- 1.4
50	+ 17.0	2,000	- 2.6
60	+ 16.1	3,000	- 4.8
80	+ 14.5	4,000	- 6.6
100	+ 13.1	5,000	- 8.2
150	+ 10.3	6,000	- 9.6
200	+ 8.2	8,000	- 11.9
300	+ 5.5	10,000	- 13.7
400	+ 3.8	15,000	- 17.2
500	+ 2.6	20,000	- 19.6

Note: The RIAA EQ system operates in Constant Amplitude mode below the 500 Hz Turnover Frequency. It also operates in Constant Amplitude mode above the 2120 Hz Rolloff Frequency. The system operates in Constant Velocity mode between the Turnover and the Rolloff
Frequencies.

[Craig Maier]

As an aside, the best way to down-sample to 44.1 kHz is to start with a file that is the exact double of that value which is 88.2 kHz. There is no interpolation error when doing it that way. 88.2 KHz can be recorded using the 'custom' feature in the record sample rate selector. After custom is selected, just enter the sample rate in the open field.