Abraxalito's DAC, which features a 7th order analog reconstruction filter that's down -17dB @ 25KHz in my modeling of it. I don't readily recall what he exactly specs. for his filter. That should prevent IMD from any half-decent amplification component. Most amplifier schematics which I've seen feature input filtering to prevent that. Even should some amplifiers still suffer IMD, that certainly is not the case for the majority of audio amplification components. For those amplifiers which do suffer IMD, why should those listeners find that IMD pleasant and relaxed sounding? No other IMD which I'm aware of is pleasant to the ear. Rather, it's typically a very non-musical sounding distortion. Perhaps, I'm still missing your point, however.
What is wrong IMO is that this person seems to have no mathematical background and is looking at a Fir filter as something that is smearing the sound.
The basic intention for using a Fir filter is to low pass filtering while keeping the content in the passband intact and not to smear it.
Depending on the quality of the realisation in length and calculation accuracy, some filters approach this ideal better then others.
Hans
Status of our Suspect List
I think that this would be a good time to review the status of our suspect list. Here it is, with some comments by me in italics:
B) RECONSTRUCTION/IMAGE-BAND HANDLING
===============================================
2) Lack of an FIR interpolation-filter, freeing the DAC from certain processing 'artifacts' , such as:
a) time-domain signal echoes produced within Equiripple on-chip FIR filters.
b) impulse response ringing (pre or post)
c) half-band filters plainly violating Nyquist
d) are prone to clip on peak sample normalized recordings - the intersample overshoot issue.
e) purely analog image-band suppression inherently sounds different than digital suppression?
I believe that all items under B2 are still active suspects. We are just now beginning to conceive or suggest experiments to identify which items may be effectively eliminated from suspicion. More to come soon on those experiments.
3) Phase-modulation of the baseband signal due to insufficiently suppressed image-bands. In other words, because the signal waveform is not fully reconstructed according to the sampling theorem requirements.
I suggest that we consider removing item B3 above from the active suspect list. While not absolutely conclusive, acoustic filtering pretty effectively removes all ultrasonic (>20KHz) information for male listeners above age 40. Which therefore includes the image-bands, inherently reconstructing any digital audio signal.
4) The unsuppressed image-bands are, somehow, producing audible IM products directly within the ear.
B4, is an item which we've yet to give much thought to. The notion here is that although the ear filters ultrasonic signals, is may be possible for the the structure of the inner ear to inter-modulate them down to being audible. I think this unlikely, but is conceivable as the ear is a non-linear acoustic structure, so one never knows until a search of the existing research. An additional reason that seems to make this unlikely to produce pleasant NOS sound, however, is that there is no link, of which I'm aware, between IMD and it producing more pleasant sounding music. Just the opposite.
C) ALTERED JITTER IMPACT
=============================================
5) Different jitter impact due to fewer D/A conversion cycles per second.
6) Reduced supply and ground noise due to slower clock rates.
As most of you already realize, Catagory C is not trivial to assess the impact of. Since we are, however, only concerned with the affect of jitter as a function of sample rate, perhaps that will allow us perform a simple experiment or analysis to remove it from suspicion. Therefore, I'm interested in your thoughts and reasoning on whether this category is primarily responsible for OS and NOS sounding different and worthy of some sort of effort to demonstrate that.
D) SAMPLE-PERIOD RELATED QUANTIZATION ERRORS
==============================================
7) Converter settling-time becomes a smaller percentage of each conversion period as the conversion rate is made slower.
8) Harmonic-distortion may be sample-rate dependent.
Category D7, has been receiving a good amount of engineering attention. Hans Polak has confirmed the issue of settling-time as producing slew-rate distortion in simulations. There are two main practical ways to mitigate this distortion, RC filter it (the easiest way), or utilize a Sample-and-Hold circuit to completely block it (the more complex way). Both settling-time solutions interface with the I/V section, and Hans, as well as Marcel have been identifying interesting new solution approaches there, as well.
As for D8, it should not be too difficult to identify, via their respective data sheets, which converter chips produce a significantly different THD figure as a function of sample rate. If those converters are not the same which are commonly utilized in NOS DACs, then D8 can likely be removed rom further suspicion as the reason why OS and NOS sound different.
Please, feel free submit any thoughts or questions about the list or my suggested changes.
I think that this would be a good time to review the status of our suspect list. Here it is, with some comments by me in italics:
B) RECONSTRUCTION/IMAGE-BAND HANDLING
===============================================
2) Lack of an FIR interpolation-filter, freeing the DAC from certain processing 'artifacts' , such as:
a) time-domain signal echoes produced within Equiripple on-chip FIR filters.
b) impulse response ringing (pre or post)
c) half-band filters plainly violating Nyquist
d) are prone to clip on peak sample normalized recordings - the intersample overshoot issue.
e) purely analog image-band suppression inherently sounds different than digital suppression?
I believe that all items under B2 are still active suspects. We are just now beginning to conceive or suggest experiments to identify which items may be effectively eliminated from suspicion. More to come soon on those experiments.
3) Phase-modulation of the baseband signal due to insufficiently suppressed image-bands. In other words, because the signal waveform is not fully reconstructed according to the sampling theorem requirements.
I suggest that we consider removing item B3 above from the active suspect list. While not absolutely conclusive, acoustic filtering pretty effectively removes all ultrasonic (>20KHz) information for male listeners above age 40. Which therefore includes the image-bands, inherently reconstructing any digital audio signal.
4) The unsuppressed image-bands are, somehow, producing audible IM products directly within the ear.
B4, is an item which we've yet to give much thought to. The notion here is that although the ear filters ultrasonic signals, is may be possible for the the structure of the inner ear to inter-modulate them down to being audible. I think this unlikely, but is conceivable as the ear is a non-linear acoustic structure, so one never knows until a search of the existing research. An additional reason that seems to make this unlikely to produce pleasant NOS sound, however, is that there is no link, of which I'm aware, between IMD and it producing more pleasant sounding music. Just the opposite.
C) ALTERED JITTER IMPACT
=============================================
5) Different jitter impact due to fewer D/A conversion cycles per second.
6) Reduced supply and ground noise due to slower clock rates.
As most of you already realize, Catagory C is not trivial to assess the impact of. Since we are, however, only concerned with the affect of jitter as a function of sample rate, perhaps that will allow us perform a simple experiment or analysis to remove it from suspicion. Therefore, I'm interested in your thoughts and reasoning on whether this category is primarily responsible for OS and NOS sounding different and worthy of some sort of effort to demonstrate that.
D) SAMPLE-PERIOD RELATED QUANTIZATION ERRORS
==============================================
7) Converter settling-time becomes a smaller percentage of each conversion period as the conversion rate is made slower.
8) Harmonic-distortion may be sample-rate dependent.
Category D7, has been receiving a good amount of engineering attention. Hans Polak has confirmed the issue of settling-time as producing slew-rate distortion in simulations. There are two main practical ways to mitigate this distortion, RC filter it (the easiest way), or utilize a Sample-and-Hold circuit to completely block it (the more complex way). Both settling-time solutions interface with the I/V section, and Hans, as well as Marcel have been identifying interesting new solution approaches there, as well.
As for D8, it should not be too difficult to identify, via their respective data sheets, which converter chips produce a significantly different THD figure as a function of sample rate. If those converters are not the same which are commonly utilized in NOS DACs, then D8 can likely be removed rom further suspicion as the reason why OS and NOS sound different.
Please, feel free submit any thoughts or questions about the list or my suggested changes.
Last edited:
The Chord in my system "proves" this as i can take it out any time i want, then i play NOS. So it can be heard instantly. No "smearing" of any kind when i put the Chord into the signal chain, it actually improves the spaceous experience without adding a sonic signature. I can even select the amount of TAP's on the Chord and hear the difference as well.
I do not want to advertise for the Chord but just want to state that a good FIR is able to sound better then NOS. I was hoping someone here would try to build a better FIR as this is beyond my capabilities...
I do not want to advertise for the Chord but just want to state that a good FIR is able to sound better then NOS. I was hoping someone here would try to build a better FIR as this is beyond my capabilities...
I'd be one of those. But, these DAC's measure really poorly.
I have seen here that people are talking about minute differences in noise specifications of OP amps... used as an analog filter, but - I do not want an analog filter because every time I tried a filter (passive RCL, or active), the sound lost natural body and harmonics that, to me, are very important. The reason is very simple - every time the sound undergoes any kind of processing (in digital or analog domain), it loses something.... maybe we should talk about what that something is and how & why it is lost? I do have a theory, backed up by few facts and few assumptions; also 3-decade long experience s well...
That's why I tried to set the boundary.... let's focus on the digital section first, and then some 1000 posts down the track, we can include the analog section. This would be something, wouldn't it...
I agree very much with you - but mmmhhh, we are both dddac lovers with passive I/V and no analog stages behind it to get on 1-2V rms right?
The most complete discussion of Chord filters and other thoughts of Rob Watts on data conversion can be found in a series of youtube videos, the first of which is at: Interview with Chord Electronics' Rob Watts - Part 1: R2R vs Delta-Sigma vs Chord FPGA - YouTube
That said, there is a fair amount of other public commentary by Watts, some of it in videos and some in text.
One can make a very long fir filter in software. That is exactly what some of the filters in HQ Player are. It takes a very powerful SOA computer and SOA video card combination to run some of them. Do you find they sound like Chord M-Scaler? Do they sound better?
Some people find they don't sound like M-Scaler. Why? Different algorithm. Watts uses a windowed sinc filter, with a proprietary window taper shape.
Some people find they don't sound like M-Scaler. Why? Different algorithm. Watts uses a windowed sinc filter, with a proprietary window taper shape.
Last edited:
Defenitely the Chord Scaler is doing a incedible good job here. It adds more depth and staging experience... Instruments get more "space" from each other. This is notecible every time i switch it back and forth. (I am not talking about the Chord DAC btw as this i have sold....i use my current source NOS dac)
To get the most out of the Chord i even build a dual mono NOS DAC as now it can handle 352,8kHz (or even 384kHz but that is no use for me)
So, yes it could be created and uploaded on a chip but this is not something i can do myself...
To get the most out of the Chord i even build a dual mono NOS DAC as now it can handle 352,8kHz (or even 384kHz but that is no use for me)
So, yes it could be created and uploaded on a chip but this is not something i can do myself...
Last edited:
Here is a link below (thanks, Don Hill) to, by far, the longest audio FIR filter I have ever heard tell of, selectable up to 8 billion taps! It's a PC/Mac based software implementation which works offline, so not in real-time. You feed in a music file and it outputs an up-sampled ('remastered') version for library storage on a computer file server. The reason it can process a filter with so many taps is that it reads the entire input file before computing the up-sampled output file. Which is why it can not (as I read it) process samples in real-time. Meaning, no disc or internet streaming track play.
A bigger issue may be that it costs $500 to license. Although, it is available for a free 30 day trial. Enough time to permanently convert a copy of an existing music library, should the user be disposed to. In the price scale of home audio equipment, perhaps $500 isn't a large sum, but I really wish it were a fraction of that amount. The free trial combined with the high price indicates to me that the product may demonstrate it's effectiveness quite well.
PGGB - Offline remastering
Last edited:
So then you end up with huge file instead of creating it "on the fly" ?
I like to hardware version more as i do not like big files...
So i would like to "create" a chip that can do this for me... Now i do not need one as i am happy with what i have but i would like to be able to do this and learn from it.
I like to hardware version more as i do not like big files...
So i would like to "create" a chip that can do this for me... Now i do not need one as i am happy with what i have but i would like to be able to do this and learn from it.
Since there's a need for some standard audio files, here's an attempt with a link to three .wav files of rather different nature and good quality to test between various options.
Dropbox - NOS - Simplify your life
Hans
Dropbox - NOS - Simplify your life
Hans
The size increase of the output file mostly depends on the up-sampling ratio. A 2xOS, for example, would essentially double the track's file size, so you would have a 88.2KHz file where before it was 44.1KHz. I also would prefer a real-time solution, but then you could not have an 8-billion tap filter.
The issue for we audiophiles is whether the sound is improved enough to worth the bother.Hi Ken,
That's a very interesting option, specially for this thread, to produce an OS .wav version of your original 44.1Khz source and then compare the 44.1Khz in NOS against the external-upsampled version in a much higher Fs also played in NOS.
With these lengthy Fir filters it could give a lot of answers whether or not a very good OS filter is hurting the sound.
Hans
I think it is worth trying....
At least it betters my standard NOS solution. The Chord i use "fell into my lap" so to speak so i got the chance to play with this device. Now i want more haha...
To upgrade to an M Scaler is just too expensive for me. It is still 4K Euro and as a DIY'er it is difficult for me to spend that kind of money on a box that could be made with some tools and elbow grease... plus think of the fun and joy if this is a succes.
At least it betters my standard NOS solution. The Chord i use "fell into my lap" so to speak so i got the chance to play with this device. Now i want more haha...
To upgrade to an M Scaler is just too expensive for me. It is still 4K Euro and as a DIY'er it is difficult for me to spend that kind of money on a box that could be made with some tools and elbow grease... plus think of the fun and joy if this is a succes.
Did you try a SD card reader, direct I2S VS all that long chain and bigger noisy ground loops ? Maybe you use Toslink or assymetric digital input ?
I like the minimalist way ECDESIGN company ruled the data input.
Would like to know if that heavy signal treatements are not wasted by the substract (poor clock quality ofen, worse as the speed increases, computer with smps chip in board with poluted ground. That's a true problem with playback softwares with personal computer and likes (RaspBeryy, NAS,...)
Yes i have tried i2s directly etc. I found no advantage over standard S/PDIF. Toslink can not handle the high sampling rate needed (stops at 88.2kHz) not because of the fiber but because of the hardware used in standard transports.
I do not use any computers etc. only a transport (SD-card is an option but a hassle, i do have it somewhere to try it one day).
I do not use any computers etc. only a transport (SD-card is an option but a hassle, i do have it somewhere to try it one day).