I would recommend a "best fit" riaa approximation single pole analog eq in the feedback of the preamp ~ 100 Hz low pass corner frequency
this requires ~+/- 6 dB of digital eq, the digital corrector filter coefficients could be adjusted for the measured preamp corner frequency for high final riaa accuracy (actually any single pole preamp roll off frequency between 50 and 212 Hz works nearly as well)
then "wasted" dynamic range is only ~ 12 dB
adjusting gain to fit the record's level to the adc full scale is likely still desirable as real adc electronics aren't going to deliver 24 bits - good current monolithic audio adc's in PC soundcards struggle to deliver 110+ dB dynamic range: <20 effective bits of dynamic range, and then the digital eq 12 dB is going to cut that some more
of course the digital domain math should be done with extended word size to prevent overflows and the accumulation of truncation/rounding errors but your final output can't be better than the S/N of the electronics you originally capture with, minus the headroom wasted by the digital eq and any headroom you allow over the record's peak
this requires ~+/- 6 dB of digital eq, the digital corrector filter coefficients could be adjusted for the measured preamp corner frequency for high final riaa accuracy (actually any single pole preamp roll off frequency between 50 and 212 Hz works nearly as well)
then "wasted" dynamic range is only ~ 12 dB
adjusting gain to fit the record's level to the adc full scale is likely still desirable as real adc electronics aren't going to deliver 24 bits - good current monolithic audio adc's in PC soundcards struggle to deliver 110+ dB dynamic range: <20 effective bits of dynamic range, and then the digital eq 12 dB is going to cut that some more
of course the digital domain math should be done with extended word size to prevent overflows and the accumulation of truncation/rounding errors but your final output can't be better than the S/N of the electronics you originally capture with, minus the headroom wasted by the digital eq and any headroom you allow over the record's peak
Netlist said:
The crest factor is just the peak to rms value. Each LP will be different, if you don't want to clip obviously you have to keep the peak below full scale. Highly compressed music will have a lower crest factor.
Each filtering strategy will probably have a different one. I found over a wide range of LP's I couldn't get safely more than 3dB or so extra gain in front of the A/D even with full RIAA. That is to say after looking at the file, I realized I could have recorded it hotter.
jcx,
I would prefer Scott’s approach as that gives a theoretical gain of 20dB and, in case of the Ono, easier to implement. That, of course if I understood you correctly.
Scott
Thanks, but that requires more labor in that the source material has to be recorded twice.
Would you think that the crest factor can be calculated from the stats (post #14)?
I’ve set the input sensitivity of the soundcard to a good average, knowing that I might loose a few dB's now and then.
/Hugo
I would prefer Scott’s approach as that gives a theoretical gain of 20dB and, in case of the Ono, easier to implement. That, of course if I understood you correctly.
Scott
Thanks, but that requires more labor in that the source material has to be recorded twice.
Would you think that the crest factor can be calculated from the stats (post #14)?
I’ve set the input sensitivity of the soundcard to a good average, knowing that I might loose a few dB's now and then.
/Hugo
I didn’t buy the Voxengo plug-in as the free version was good enough to get a grip on what it does. Instead I experimented with a lot of combinations of fft filter and graphic equalizer settings.
I abandoned Scott’s and jcx's idea for now as I couldn’t get a perfect match between the analogue and digital filters, mostly the gain settings were problematic.
Here’s the (for now) final setup:
I connected the cartridge directly to the balanced inputs of the instrument preamp.
+ to hot, - to cold and the shield only connected to the preamp’s pin 1.
I carefully lined out a very accurate 40 points fft filter and did all the equalization from there.
The preamp has a gain of 66dB and I use about 50 of them (guess, have to measure) which is just enough to keep the S/N level of the card to an acceptable level of –72dB.
I recorded a track and subtracted the pre-recorded soundcard noise from it. I let the fft filter do the job and amplified to about –0.5dBFS.
I have done four sample tracks so far and the result is a gain in dynamics of roughly 2dB on each track, compared to the same recording and same amount of noise reduction as with the phono preamp.
Despite the (very) unconventional setup, I’m pleased with the finer details I hear.
analog_sa, you were correct about the two samples.
/Hugo
I abandoned Scott’s and jcx's idea for now as I couldn’t get a perfect match between the analogue and digital filters, mostly the gain settings were problematic.
Here’s the (for now) final setup:
I connected the cartridge directly to the balanced inputs of the instrument preamp.
+ to hot, - to cold and the shield only connected to the preamp’s pin 1.
I carefully lined out a very accurate 40 points fft filter and did all the equalization from there.
The preamp has a gain of 66dB and I use about 50 of them (guess, have to measure) which is just enough to keep the S/N level of the card to an acceptable level of –72dB.
I recorded a track and subtracted the pre-recorded soundcard noise from it. I let the fft filter do the job and amplified to about –0.5dBFS.
I have done four sample tracks so far and the result is a gain in dynamics of roughly 2dB on each track, compared to the same recording and same amount of noise reduction as with the phono preamp.
Despite the (very) unconventional setup, I’m pleased with the finer details I hear.
analog_sa, you were correct about the two samples.
/Hugo
For those interested, here’s the fft filter.
It works with Audition 2.0 and you have to paste it into the \Documents and Settings\<User>\Application Data\Adobe\Audition\2.0\effect_settings.xml file.
It is only accurate with a sample rate of 48000 and 32bit float .wav files.
Replace “ItemXX” with the next number you’ll find in the row of filters inside the .xml file.
In my case, “Item26”.
<KeyVal Key="ItemXX" Type="string">RIAA inverse 48000,3,39,0,99,126,97,232,96,340,95,447,93,554,91,662,89,769,86,876,83,984,80,1091,77,1198,74,1305,71,1413,68,1520,65,1628,63,1735,60,1842,58,1950,56,2057,55,2164,53,2272,52,2379,50,2486,49,2593,47,2701,45,2808,43,2926,40,3023,37,3130,34,3237,31,3345,28,3452,24,3559,21,3667,17,3774,13,3881,10,3989,6,4096,2,39,0,99,126,97,232,96,340,95,447,93,554,91,662,89,769,86,876,83,984,80,1091,77,1198,74,1305,71,1413,68,1520,65,1628,63,1735,60,1842,58,1950,56,2057,55,2164,53,2272,52,2379,50,2486,49,2593,47,2701,45,2808,43,2926,40,3023,37,3130,34,3237,31,3345,28,3452,24,3559,21,3667,17,3774,13,3881,10,3989,6,4096,2,2,0,2048,1,4,0,0,648,31,831,57,1000,100,3,0,200,-22,20,8192,2,1,1,1,48000,0,4096</KeyVal>
/Hugo
It works with Audition 2.0 and you have to paste it into the \Documents and Settings\<User>\Application Data\Adobe\Audition\2.0\effect_settings.xml file.
It is only accurate with a sample rate of 48000 and 32bit float .wav files.
Replace “ItemXX” with the next number you’ll find in the row of filters inside the .xml file.
In my case, “Item26”.
<KeyVal Key="ItemXX" Type="string">RIAA inverse 48000,3,39,0,99,126,97,232,96,340,95,447,93,554,91,662,89,769,86,876,83,984,80,1091,77,1198,74,1305,71,1413,68,1520,65,1628,63,1735,60,1842,58,1950,56,2057,55,2164,53,2272,52,2379,50,2486,49,2593,47,2701,45,2808,43,2926,40,3023,37,3130,34,3237,31,3345,28,3452,24,3559,21,3667,17,3774,13,3881,10,3989,6,4096,2,39,0,99,126,97,232,96,340,95,447,93,554,91,662,89,769,86,876,83,984,80,1091,77,1198,74,1305,71,1413,68,1520,65,1628,63,1735,60,1842,58,1950,56,2057,55,2164,53,2272,52,2379,50,2486,49,2593,47,2701,45,2808,43,2926,40,3023,37,3130,34,3237,31,3345,28,3452,24,3559,21,3667,17,3774,13,3881,10,3989,6,4096,2,2,0,2048,1,4,0,0,648,31,831,57,1000,100,3,0,200,-22,20,8192,2,1,1,1,48000,0,4096</KeyVal>
/Hugo
KBK said:
Never say never. Such a thing may exist, but then it will be a very rare animal indeed, built from discrete components, and if it has to attain 20-24 bit accuracy excruciatingly expensive.
Present-day ADC chips are manufactured on CMOS IC processes, and these processes run off supplies not exceeding 5V, often (much) less.
Full-scale input for such chips often means 2Vpp. The distance between this and the chip and its surrounding circuitry's thermal noise floor dictates the true dynamic range. Cryogenic cooling may help
Yes,
And fortunately the RIAA defines an almost steady 6 dB/oct droop in gain over frequency. That makes noise life a lot easier in the analogue domain. That advance vanishes if you want to do the RIAA correction in he digital domain IMO and do you need an incredible dynamic range for the front end.
And fortunately the RIAA defines an almost steady 6 dB/oct droop in gain over frequency. That makes noise life a lot easier in the analogue domain. That advance vanishes if you want to do the RIAA correction in he digital domain IMO and do you need an incredible dynamic range for the front end.
Werner said:
Never say never. Such a thing may exist, but then it will be a very rare animal indeed, built from discrete components, and if it has to attain 20-24 bit accuracy excruciatingly expensive.
Present-day ADC chips are manufactured on CMOS IC processes, and these processes run off supplies not exceeding 5V, often (much) less.
Full-scale input for such chips often means 2Vpp. The distance between this and the chip and its surrounding circuitry's thermal noise floor dictates the true dynamic range. Cryogenic cooling may help
No, think current in through a big resistor into a virtual ground. All you need is a current in A/D. Then if want something else to ponder compute the classical shot noise of a 1mA current and then compute the noise of 1mA made by a noiseless 1000V battery across a 1M resistor.
Werner said:
Thanks a lot, Werner. I did some extensive test with Voxengo with different types of music and the results are very good. I now have the cartridge plugged in directly into the soundcard and do everything in the digital domain. I'm truly satisfied with the outcome.
One thing I might do in the future is insert a transformer between the MC and the soundcard.
/Hugo
Turntable -> PC (RIAA EQ done by software filter)
This is a thread from last year on this general subject. The emphasis is on IIR filters for digital RIAA equalization, including coefficients for IIR biquads.
Note that an IIR filter implementation will be minimum phase and stable as long as the z-plane poles and zeroes are between -1 and +1.
This is a thread from last year on this general subject. The emphasis is on IIR filters for digital RIAA equalization, including coefficients for IIR biquads.
Note that an IIR filter implementation will be minimum phase and stable as long as the z-plane poles and zeroes are between -1 and +1.
Pjotr said:And fortunately the RIAA defines an almost steady 6 dB/oct droop in gain over frequency. That makes noise life a lot easier in the analogue domain. That advance vanishes if you want to do the RIAA correction in he digital domain IMO and do you need an incredible dynamic range for the front end.
So, has anybody tried to build a RIAA by using a simple integrator (6dB/oct slope and with sufficient gain) and then a shelving filter afterwards (either in analog or digital domain) to complete the job?
I found some time to play again.
What about making it at the input, since the output of the soundcard is firewire, and I can't see myself poking in there? I found this filter which gives the curves as seen in the picture but how to do it with a cartridge that is wired balanced?
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.