Hi,
I will surely move from my 1941 to the new dam1021 rev7. There is a friend electronics that is designing a small motherboard for it, nothing fancy, mostly interfacing with different things. We are thinking of offering this board in a gb if someone is interested, but it's too early yet.
Anyway, with my dam i am using the spikestabber's pack, mostly switching between EQHQ and c128dp. I did try to load the gumisb pack once but it's incompatible with the 1941, also the limit of 44.1/48k sample rate is problematic for me as i have high res record in my collection. Apparently the new dam use a different last filter with half the speed so my beloved paul's filters are not compatible anymore ?
If there is a kind soul that can adapt paul's filter to the rev7, i would be always thankfull
I will surely move from my 1941 to the new dam1021 rev7. There is a friend electronics that is designing a small motherboard for it, nothing fancy, mostly interfacing with different things. We are thinking of offering this board in a gb if someone is interested, but it's too early yet.
Anyway, with my dam i am using the spikestabber's pack, mostly switching between EQHQ and c128dp. I did try to load the gumisb pack once but it's incompatible with the 1941, also the limit of 44.1/48k sample rate is problematic for me as i have high res record in my collection. Apparently the new dam use a different last filter with half the speed so my beloved paul's filters are not compatible anymore ?
If there is a kind soul that can adapt paul's filter to the rev7, i would be always thankfull
Hi gumisb,
I like the Idea of the cascaded Biquad you implemented.
The coef. you may have used / the only ones I found from your posts were these:
b0=0.84133546241110235
b1=1.68267092482220470
b2=0.84133546241110235
a1=-1.99985813298937410
a2=-0.99985815140002265
not sure if they are still up to date?
However - cascading these 8 times in MatLab gives the following (magnitude normalized) plot:
(Shown is only the freq. range from 20KHz to fs/2)
Is this what you anticipated?
Fs/2 for 352.8KHz is 172.xxKHz -> impressive attenuation of <-350dB
At 20Khz the attenuation is at -6.2dB
I like the Idea of the cascaded Biquad you implemented.
The coef. you may have used / the only ones I found from your posts were these:
b0=0.84133546241110235
b1=1.68267092482220470
b2=0.84133546241110235
a1=-1.99985813298937410
a2=-0.99985815140002265
not sure if they are still up to date?
However - cascading these 8 times in MatLab gives the following (magnitude normalized) plot:
(Shown is only the freq. range from 20KHz to fs/2)
Is this what you anticipated?
Fs/2 for 352.8KHz is 172.xxKHz -> impressive attenuation of <-350dB
At 20Khz the attenuation is at -6.2dB
IIR filters are nice. If I remember correctly, the last IIR filters I made were based on 15th order elliptic filter.
It really depends what is your goal. IIR biquads are in the middle between FIR1 and FIR2 upsampling(reconstruction) filters.
So everything outputs IIR will be taken by FIR2.
Ideally for me would be situation when I can multiply just one coefficient of 1 by 8 on each FIR stage. That would give opportunity to create true nos filter with IIR as substitute of analog output filter.
It really depends what is your goal. IIR biquads are in the middle between FIR1 and FIR2 upsampling(reconstruction) filters.
So everything outputs IIR will be taken by FIR2.
Ideally for me would be situation when I can multiply just one coefficient of 1 by 8 on each FIR stage. That would give opportunity to create true nos filter with IIR as substitute of analog output filter.
If someone wants to try some new ones:
I tried a different approach here:
I designed the filters in Matlab with Filter designer app
then I took the coeffitients and multiplied them by 8 and 0.94 for a little headroom
in the filter file, the multiplication factor is set to 1.
-- 04: FIR1: Nyquist, 8.5 Band, Window Chebychev, 3998 Order, 3999 coeffs, flat untill 19.4KHz, -130dB @20KHz
-- 05: NOS
-- 06: Nyqist Band 8, Window Bohman, 3999 coeffs
-- 07: NOS
-- 08: Nyquist 8.5 Band, Window Chebychev, 239 coeffs
-- 09: NOS
- - 10: Nyquist 8 Band, Window Blackman Harris, 239 coeffs
-- 11: same as 10
-- 04: NQCH_FIR1 ->08: NQCH_FIR2
-- 05: NOS ->09: NOS
-- 06: NQBOH ->10: NQBMH
-- 07: NOS ->11: NQBMH
I tried a different approach here:
I designed the filters in Matlab with Filter designer app
then I took the coeffitients and multiplied them by 8 and 0.94 for a little headroom
in the filter file, the multiplication factor is set to 1.
-- 04: FIR1: Nyquist, 8.5 Band, Window Chebychev, 3998 Order, 3999 coeffs, flat untill 19.4KHz, -130dB @20KHz
-- 05: NOS
-- 06: Nyqist Band 8, Window Bohman, 3999 coeffs
-- 07: NOS
-- 08: Nyquist 8.5 Band, Window Chebychev, 239 coeffs
-- 09: NOS
- - 10: Nyquist 8 Band, Window Blackman Harris, 239 coeffs
-- 11: same as 10
-- 04: NQCH_FIR1 ->08: NQCH_FIR2
-- 05: NOS ->09: NOS
-- 06: NQBOH ->10: NQBMH
-- 07: NOS ->11: NQBMH
Another year, another Filter
Here's my new Filter set - I finally managed cascading the upsampled FIR1 filter coeffs with the FIR2 coeffs in Matlab, for a far better design study.
Main goal was :
The MatLab Code for upsampling and cascading i wrote is:
fir1_coeff = coeffs(fir1).Numerator;
ups_fir1_coeff = upsample(fir1_coeff,8);
ups_fir1 = dsp.FIRFilter(ups_fir1_coeff );
hcas = dsp.FilterCascade(ups_fir1 ,fir2);
fvtool(hcas, ttt, fir2)
The Filters (44.1 and 48KHz only) are now as follow:
Cascaded 04 & 08:
IR cascaded 04 & 08
Cascaded 05 & 09
IR cascaded 05 & 09
Cascaded 06 & 10
IR Cascaded 06 & 10
Cascaded 07 & 11
IR cascaded 07 & 11
Here's my new Filter set - I finally managed cascading the upsampled FIR1 filter coeffs with the FIR2 coeffs in Matlab, for a far better design study.
Main goal was :
- high attenuation at Fs/2
- as short as possible
- no passband attenuation
The MatLab Code for upsampling and cascading i wrote is:
fir1_coeff = coeffs(fir1).Numerator;
ups_fir1_coeff = upsample(fir1_coeff,8);
ups_fir1 = dsp.FIRFilter(ups_fir1_coeff );
hcas = dsp.FilterCascade(ups_fir1 ,fir2);
fvtool(hcas, ttt, fir2)
The Filters (44.1 and 48KHz only) are now as follow:
| FIR1 | = | = | = | FIR2 | = | Casc FIR1&2 | = | = | = |
| Name | n= | -3dB@ | @22.05KHz | Name | n= | +/-0.1dB | @20KHz | @352KHz | pre-Ringing |
| lin KBD | 196 | 20.05KHz | -22.78dB | lin BMH | 238 | 0...17.9KHz | -3.2dB | -133dB | 5 |
| lin BMH | 481 | 20.05KHz | -32.45dB | lin Kaiser | 149 | 0...18.9KHz | -2.75dB | -140dB | 8 |
| lin Flattop | 1047 | 20.27KHz | -74.3dB | lin BMH | 239 | 0...18.3KHz | -1.59dB | -133dB | 11 |
| lin NQ Cheby | 3999 | 19.51KHz | -129.1dB | inv Sinc Equi | 239 | 0...19.3KHz | -128dB | -50dB | >>32 |
Cascaded 04 & 08:
IR cascaded 04 & 08
Cascaded 05 & 09
IR cascaded 05 & 09
Cascaded 06 & 10
IR Cascaded 06 & 10
Cascaded 07 & 11
IR cascaded 07 & 11
Last edited:
guys this is 124 page thread technically over 1240 replies. Quite an extensive thread. Can anyone upload the best sounding filters with NOS like signature yet very transparent and revealing without fatigue. I know Im asking alot I guess but I like the initial stock filter which came with version 1 one of the first version. Is there anything better that one can suggest requesting to post the filter pack.
@rhythmsandy
The "NewNos" Filter I suggested (post #2475) was also measured by spzzzzkt in Post #356 (page 18) with a very much alike magnitude response as the MatLab plot I posted.
There also is a "TrueNOS" Filter by spzzzzkt: #1209 (page 61)
With the Interp rate set to 1, I think this is a moving average filter - so it should not be used for music?
spzzzzkt wrote this:
dam1021,44100,1,1,1,1
Input FIR, 44.1 Khz Samplerate, bypass
1.000
Then again, a NoFilter should look like this (I guess):
dam1021,44100,8,4,8,1
04 Input FIR, 44.1 Khz Samplerate, bypass
8.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
The "NewNos" Filter I suggested (post #2475) was also measured by spzzzzkt in Post #356 (page 18) with a very much alike magnitude response as the MatLab plot I posted.
There also is a "TrueNOS" Filter by spzzzzkt: #1209 (page 61)
With the Interp rate set to 1, I think this is a moving average filter - so it should not be used for music?
spzzzzkt wrote this:
dam1021,44100,1,1,1,1
Input FIR, 44.1 Khz Samplerate, bypass
1.000
Then again, a NoFilter should look like this (I guess):
dam1021,44100,8,4,8,1
04 Input FIR, 44.1 Khz Samplerate, bypass
8.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
As this is the blanket forum for custom filters, this seems a sensible place to ask if anyone has any insights (or, better yet, a solution) into the apparent inability of my dam1121 to use custom filters (specifically, starting with the, "Paul ensemble.") This issue persists regardless which f/w I install. The filters included with the f/w function as intended across a range of bitrates/sampling frequencies, but uploading anything else produces a failure to lock. For testing/playing purposes I'm running coax into a basic (AK4113?) s/pdif-to-i2s board, so nothing too wild at this stage.
Am I missing something obvious?
Regards,
Eric
Am I missing something obvious?
Regards,
Eric