Just returned from weekend away from home. Great to see progress continues, thank you! Will have a go one day soon. 'totalCRaP_rev2' was my favourite until I heard '1021filtN3_corr' this evening. May be not quite so open & transparent, talking very small fractions, but somehow cleaner with greater depth to the soundstage. More rounded, may be a bit more subtance to the sound, more 'grown up'? Nice....
Just to note... Using 0.05% board as a prototype, have 0.01% ready to drop in when I finish experimenting. Have 2200uF 6V3 blackgates on each of the +/-5V supply lines, ideally want to try them on the +/-4V ladder references, and 100uF on each of the 1.2 & 3.3V lines. Output taken directly from the ladder into dedicated valve amp (modified capacitors & op-amps) from Tube Technology DAC64. Currently only using the optical input. Unlikely to make huge (if any) difference to filter testing, but one never knows. Decide on a favourite filter then remove my mods and listen again.
I'm at 1021filtN3_corr which I enjoy immensely. Vey nice and apparently even better things down the line
Thanks again spzzzzkt!
//
A pleasure.
I concur on the Nyquist filters. These are definitely my favourite although with the FIR1 and FIR2 changed to Nyquist there is something going on, I am not quite sure about what.
All the previous flat filters have been had fairly aggressive sibilance on certain recordings. The sibilance is in the original recording and can be seen in a spectrogram, so it's not a fault as such, it just that it is much more "in yer face" than on other DAC's I have access to - MBP headphone out, D1 Clone, TC Konnect 8.
In contrast the N3 filter makes it sound as I would expect -- present but not prominent.
The Nyquist filter is pretty flat, the -6dB point is determined by the sampling rate and number of taps, and is fixed at 22050Hz. I can vary the bandwidth and window applied to tweak but the -6dB point is immutable as far as I can see. With the bandwidth set to 10kHz and using a kaiser window the filter is -0.035dB @ 19kHz, -76dB @ 26kHz, and -162 at 27.05kHz. Not exactly what you'd call rolled off in the audio band.
So not sure why that is happening without measuring the DAC output.
Will get the scope out again tomorrow and see if I can uncover anything. It will measure/display FFTs too, but I'm puzzled as the noise floor sits about -60/70dB down without the dac connected, too high for my liking. If all else fails, 'read the instruction manual'! It seems to think there are numerous harmonics in the audio band on the rev2 filter when pushing the input up to +3dB as indicated on the tone gen on the PC. I really need to take more time to sort out the test set up I'm using here. I'll also reconnect the Tube Tech DAC in the next couple of days and have a listen, a ladder DAC of Rob Watts design, be good to get the scope on that too once I've read said manual...!!
I'd be cautious about reading anything much into +XdB values. It's very likely that either the signal gen or the DAC is clipping/distorting. I'd stick with 0dBFS output, or lower if necessary.
Are you shorting out the probes when you check the noise floor?
Sorry I'm having too much fun listening to Infected Mushroom's The Gathering. At 16/44.1
I haven't listened to this for a few years and had forgotten just how good tracks like Psycho are ... http://youtu.be/dLWXSsYJoWY
Cool.
I'll look forward listening to your clones of the BAD filters then. You've got the patent and plenty of information on tools and people who can give you pointers on DIYing the filters.
TNT posted up some steep roll off filters earlier in the thread that you could use in the interim.
This is just the last filter with a bunch of quick n dirty Nyquist filters added at the higher sample rates.
I haven't really made any effort to optimise or tweak beyond widening the transition band. With a bit of time and thought I can make up a better batch but for the moment I'll upload the "Just For Jaffar" higher rate filters.
Done a quick test with my limited selection of hi-res:
Kip Hanrahan - Coup de Tete (88.2/24)
Kraftwerk - Minimum Maximum (88.2/24)
Miles Davis - Kind of Blue (192/24)
Sounds fine...
Cool.
I'll look forward listening to your clones of the BAD filters then. You've got the patent and plenty of information on tools and people who can give you pointers on DIYing the filters.
TNT posted up some steep roll off filters earlier in the thread that you could use in the interim.
Not only that; I'll borrow Alpha DAC today to listen where it is soundwise vs DAM.
Many thanks!
Very much appreciated!
I'm currently crafting my own pre/post-ringing optimised version of the IP filter.
Using SoX and... guess what... MS Excel
I'll upload it soon.
Had another go with my scope - but still to read the manual!
All inputs off, superimposed on the analogue signal coming from the ladder output is high frequency noise, something like 8mV pk-pk at around 25-30MHz with around 4mV pk-pk of wideband noise. Interestingly, the 25-30MHz noise looks to be peaking every 1750 uSecs, circa 2.6MHz? Not terribly exciting and well beyond the audio band, but may go some way to explaining why I felt I needed to read the manual when using the FFT function to look at the frequency spectrum. Said FFT with window set to 'Hanning' (alternatives are 'Flat top' & 'Rectangular') clearly shows quite wide peaks in the noise floor at 14 & 28Mhz, especially the latter, at around 70dB down, ie. what I measured above using the standard scope input. Who needs the manual after all??!!
Firing up the FFT function looking at the audio band, with the latest filter installed - no signal input... I'm measuring peaks on the ladder at around 70-75dB down, the noise floor is around 95dB down.
Locked on at 44k, examples are...
3.7k @ -70dB, 4.3 @ -73, 7.3 @ -78 & we've another strong one at 21.6k -72dB.
Locked on at 96k, examples are...
2.6 @ -70, 3.1 @ -73, 4.5 @ -75 & 22.6 @ -67.
They are still there with the default filter...
44k: 1.42 @ -65, 2.84 @ -70 --- 96k: 1.18 @ -60, 3.06 @ -75.
What does this tell us? Surely we have to question the detrimental impact noise above 10-20Mhz will have on sound quality? I'm sure too I can pick holes in my measuring technique without trying too hard, but nevertheless I can clearly identify the peaks noted above in the frequency spectrum within the audio band. For me however, it comes down to 'what I hear'. I like to think my ears are the best instruments for measuring audio quality, music, etc. The latest couple of filters are now my favourites regardless of what I see on my scope. Good fun, all the same.
All inputs off, superimposed on the analogue signal coming from the ladder output is high frequency noise, something like 8mV pk-pk at around 25-30MHz with around 4mV pk-pk of wideband noise. Interestingly, the 25-30MHz noise looks to be peaking every 1750 uSecs, circa 2.6MHz? Not terribly exciting and well beyond the audio band, but may go some way to explaining why I felt I needed to read the manual when using the FFT function to look at the frequency spectrum. Said FFT with window set to 'Hanning' (alternatives are 'Flat top' & 'Rectangular') clearly shows quite wide peaks in the noise floor at 14 & 28Mhz, especially the latter, at around 70dB down, ie. what I measured above using the standard scope input. Who needs the manual after all??!!
Firing up the FFT function looking at the audio band, with the latest filter installed - no signal input... I'm measuring peaks on the ladder at around 70-75dB down, the noise floor is around 95dB down.
Locked on at 44k, examples are...
3.7k @ -70dB, 4.3 @ -73, 7.3 @ -78 & we've another strong one at 21.6k -72dB.
Locked on at 96k, examples are...
2.6 @ -70, 3.1 @ -73, 4.5 @ -75 & 22.6 @ -67.
They are still there with the default filter...
44k: 1.42 @ -65, 2.84 @ -70 --- 96k: 1.18 @ -60, 3.06 @ -75.
What does this tell us? Surely we have to question the detrimental impact noise above 10-20Mhz will have on sound quality? I'm sure too I can pick holes in my measuring technique without trying too hard, but nevertheless I can clearly identify the peaks noted above in the frequency spectrum within the audio band. For me however, it comes down to 'what I hear'. I like to think my ears are the best instruments for measuring audio quality, music, etc. The latest couple of filters are now my favourites regardless of what I see on my scope. Good fun, all the same.
Measuring the Tube Technology DAC64 at the output from the digital board to ground... The output is balanced with the input to the analogue board (disconnected) having an analogue low-pass filter.
With the input locked to 44k, the noise floor measures around 65dB down up to 540kHz when it ramps up to between -50 to -60dB down at 960k. Noise hits -51dB between 1.2-6.3MHz, after which it rises and falls up to 44MHz where is hit -41dB. Noise floor is around -95dB as before.
In the audio band, there are peaks at 5.9k @ -47dB and 7.1k @ -40dB with a 44k input. Switching to 96k I see only 7kHz @ -30dB. I’d argue these measurements are inferior when compared with ‘our dac’. I’ve not listened to the TT for a while, so perhaps that’s the next test.
No ideas what filters it employs. I can say it was (allegedly) designed by Rob Watts, one reasons why I was keen to try our dac given it is fundamentally of the same design. Thus far I’m not disappointed, and wonder how far we can all take it with perhaps, more help from our designer?
I should perhaps say my listening reference is my analogue set up comprising a Jan Allaerts Formula One cartridge mounted in an SME 12 inch arm (arm cable modified and mounted outside the arm) mounted on an SME 30/12 deck, a Tom Evans Mastergroove with specially selected input devices, a Tom Evans Vibe preamp (unmodified!), and my own build EAR859 with the latest mods implelented over the past couple of days. The speakers were made for me by Ted Jordan, photos of mine (with the EAR amp) are on his web site. I also have a pair of MingDa 3008A power amps, somewhat modified again, which the EAR can now compete with. All makes for a tough test for the dac.
With the input locked to 44k, the noise floor measures around 65dB down up to 540kHz when it ramps up to between -50 to -60dB down at 960k. Noise hits -51dB between 1.2-6.3MHz, after which it rises and falls up to 44MHz where is hit -41dB. Noise floor is around -95dB as before.
In the audio band, there are peaks at 5.9k @ -47dB and 7.1k @ -40dB with a 44k input. Switching to 96k I see only 7kHz @ -30dB. I’d argue these measurements are inferior when compared with ‘our dac’. I’ve not listened to the TT for a while, so perhaps that’s the next test.
No ideas what filters it employs. I can say it was (allegedly) designed by Rob Watts, one reasons why I was keen to try our dac given it is fundamentally of the same design. Thus far I’m not disappointed, and wonder how far we can all take it with perhaps, more help from our designer?
I should perhaps say my listening reference is my analogue set up comprising a Jan Allaerts Formula One cartridge mounted in an SME 12 inch arm (arm cable modified and mounted outside the arm) mounted on an SME 30/12 deck, a Tom Evans Mastergroove with specially selected input devices, a Tom Evans Vibe preamp (unmodified!), and my own build EAR859 with the latest mods implelented over the past couple of days. The speakers were made for me by Ted Jordan, photos of mine (with the EAR amp) are on his web site. I also have a pair of MingDa 3008A power amps, somewhat modified again, which the EAR can now compete with. All makes for a tough test for the dac.
Very nice system! And you also liked the 1021filtN3_corr., in my opinion the best filter so far and the closest to analog sound. The only problem I still have is that every album reflects clearly the recording space,technique, preferences in the mix (this is more obvious with classical and Jazz music) while the damR2R with this filter makes all recordings more or less sound the same regarding the space. So it seems the space which is unique with every recording and reflected in analog playback systems is in digital systems a mere function of the filter characteristics and not a sign of perfect reproduction of the original digital recording.Do you have a similar experience or is it just me?
After I got to work on the analogue stage, my Tube Tech was then able to expose poor quality recordings, likewise bring out the best from well recorded and produced source material. While an enjoyable listen, no, it has never got close to Jan’s cartridge, Tom’s electronics, Ted’s speakers and a whole load of my stringing everything together for having the ability to reproduce the dynamics, scale and sheer depth in the soundstage recreating what was happening in the studio or in the auditorium. I think ‘our dac’ shows great promise, although perhaps it will need a firmware update or two in time to reach its very best, and plenty more experimenting with the guys on here to get the best from the filter options available. And, it’s fun!
I think it is a bit too dark and shy in the high end, at least in my maggies, making a lot of small details difficult to hear.
I agree, although I wonder if the soundstaging will be improved over previous ideas? Still to sit and listen.
I've a couple of very large (in size) 4uF & 8uF caps shunting one of the two drivers which are connected in series in Ted's speakers (see above). This has the effect of 'lifting' the top and middle-top ends depending on the capacity I select, currently both. All subjective stuff, great we can choose a filter to best match our listening tastes. Hoping to build DC power supplies tomorrow and hoping the input boards will arrive too. Experience tells me the quality of the power supplies is crutial in achieving quality audio. There may be options which could complement the work the guys are doing here. Apologies, I digress, must stick with talking filters!
I've a couple of very large (in size) 4uF & 8uF caps shunting one of the two drivers which are connected in series in Ted's speakers (see above). This has the effect of 'lifting' the top and middle-top ends depending on the capacity I select, currently both. All subjective stuff, great we can choose a filter to best match our listening tastes. Hoping to build DC power supplies tomorrow and hoping the input boards will arrive too. Experience tells me the quality of the power supplies is crutial in achieving quality audio. There may be options which could complement the work the guys are doing here. Apologies, I digress, must stick with talking filters!