Filter brewing for the Soekris R2R

Hello all.

It stated about the dam1941, that it "..upsamples everything to the final 2.8/3.1 Mhz DAC sample", If that is the case (and also is valid for the dam1021), why try to get anything more out of it? DSD128 and DSD256 will be downsampled to DSD64 anyway. Or?

Regards
Phy1970
 
I recently acquired a DAM1941, finished the hardware construction, added an excellent power supply and an AudioNote buffer with Miflex copper capacitors. Now I would like to try some filters. The thread is very long, so I dare to ask.
Which filters are the most audible compared to the original ones? What I care most about is a wide and deep stage. Maybe someone can help me?
 
  • Like
Reactions: psilonaught
AN clone of preamp A2 with tube 6sn7 and tube power supply PCB form Ali, power supply from muzgaudio.com, output transformers 2,2 uF copper Miflex, dispaly Audiozen lite

but this off topic only once, wana know more write to me
1718209220577.jpeg
 
I recently acquired a DAM1941, finished the hardware construction, added an excellent power supply and an AudioNote buffer with Miflex copper capacitors. Now I would like to try some filters. The thread is very long, so I dare to ask.
Which filters are the most audible compared to the original ones? What I care most about is a wide and deep stage. Maybe someone can help me?
Personally I still use TNT’s latest filters.

Concerning your Audio Note buffer, the general opinion seems to be not to use a buffer but simply use the raw output. Sounds absolutely amazing directly into my tube pre amp.
 
  • Like
Reactions: living sounds
SebasianL
Could you please write me the number of the entry containing the package of filters you use?
AN is a preamplifier and low gain and works exactly like yours, the RAW output is connected directly to the tube preamplifier, I also want to try this Aikido Anode follower.
 
I haven't calculated the exact lowest possible coefficient, but in several steps I have seen:

  • 0.000000000029109 appears as 01 00 00 00
  • 0.000000000029100 appears as 00 00 00 00
I will use coefficients with a maximum of 10 decimal zeros to calculate filters in Matlab. Those that already have 11 zeros will appear as zeros in MKROM.

This can be verified using the program https://hexed.it/ by viewing the “1021filt_120.skr” coefficient file.

For example, the filter 07 Soft Minimum FIR1 has approximately 651 effective coefficients out of the 3999 coefficients shown in the 1021filt_120.txt file. The rest are zeros in the “.skr” file.
 
I already have made two filters that theoretically is better those presented by the manufacturer. I need to make adjustments to equalize all the levels of all the filters, and compare the ringing in the passband and the attenuation in the stopband, as well as the measured distortion of each one.

I still need to compare the filters I have at 2k with those from the manufacturer at 4k, and with the two I will propose: one added to a NOS and a very short QuasiNOS with minimal ringing but higher high-frequency response.
 
Yes, I considered using it, but it wasn't suitable for several reasons. I had to createMatlab program custom-made to generate graphs showing detailed oscillations in the passband. While making the filters, it allows me to read the Soekris filters for quick comparisons and see if there are any improvements. I can adjust the same gain to compare listening and distortion measurements. I will publish the Matlab program because I had to do many iterations to achieve good results.
 
One of the two filters i have, has a zero every 8 coefficients. These generate the least distortion because, 1 out of every 8 times of oversampling, they set the exact value of the original sample, and the other 7 are interpolations. Let's say it is "Bitperfect" at the sampling point. These are filter of the Nyquist type. The NOS and QuasiNOS are also of this type.

In contrast, the Soekris filters are not. I am interested in measuring them to see if there is an improvement. I will provide the measurements. The QuasiNOS 2k showed this improvement in distortion for that reason.

I have achieved a significant reduction in oscillations in the passband compared to the original Soekris filter, which is very good. The Soekris filter has oscillations similar to Paul’s best filter in 2K.

You can ask for more performance from the Matlab program, with attenuations above 200 dB in the stopband, but they will not be reflected in the "skr" if there are more than 10 zeros in the coefficients.

I need to keep fine-tuning to obtain the best results.