Thanks. It wasn't clear what your reference for comparison was, and that answers my question.
Obviously listening to DAM -> PCM4222 -> FireFace -> Gustard x10 (ESS9018 based) -> amp -> Speakers is going to be a very different experience to Fire Face -> spdif -> DAM -> amp -> Speakers.
The main issue (apart from playing back a recording of the DAM via a Sigma Delta DAC) is that the filtering is additive, so you are listening to the effects of a filter which is the convolution of DAM + PCM4222 + Gustard x10 filters. This means that when you say "closest to the source" you are effectively selecting the filters that have the least influence on sound when combined with the filters in the PCM4222 and x10.
If that is the exact setup up you are planning on using it's entirely valid way of comparing, but it will not be an accurate reflection of how the filters behave with the DAM as a standalone.
cheers
Paul
Interesting point that hop.sing makes about referencing to a source, but I'm not sure that comparing to another DAC as all DACs add some colouration to the sound.
About the best I can think of is the approach used in one of the dCS papers which is to use an analog source as reference. About the only way I think I can feasibly do this is:
TT -> phono stage -> preamp -> amp -> speakers
TT -> phono stage -> ADC -> spdif -> DAM -> preamp -> amp -> speakers
The unavoidable downside is that it introduces an additional conversion step in the form of the ADC, but gives you a non-digitised source which can be used as a reference. It shouldn't matter that the TT/phonostage has it's own coloration as what is at issue is the ability of the ADC/DAC to accurately reproduce the signal it is fed.
cheers
Paul
About the best I can think of is the approach used in one of the dCS papers which is to use an analog source as reference. About the only way I think I can feasibly do this is:
TT -> phono stage -> preamp -> amp -> speakers
TT -> phono stage -> ADC -> spdif -> DAM -> preamp -> amp -> speakers
The unavoidable downside is that it introduces an additional conversion step in the form of the ADC, but gives you a non-digitised source which can be used as a reference. It shouldn't matter that the TT/phonostage has it's own coloration as what is at issue is the ability of the ADC/DAC to accurately reproduce the signal it is fed.
cheers
Paul
I will concentrate on NOS "filter" at first upsampling via PC to get the basic impression of the DAC. What do you mean with pretty soft? A slow filter? No filter could be very interesting too. Deemphasis - Are you detecting the use in the bitstream and applying the filter in this case only?
I would apreciate very much an intermediade firmware release (comming soon) only fixing the sample-rate-change clicks and power-off plops.
Hello leehan
I don't think I recommended any specific centering for the impulse per se.
The amplitude/phase relation of the target will dictate the way ringing is distributed around the impulse peak (pre/post).
Windowing (centering+number of taps+windowing algorithm) should be chosen to accommodate that target as much as possible.
Energy centering "should" give the best result, but you can also play with percentage values (especially if you know the pre/post ringing behavior).
Of course you can remove some pre ringing by forcing a small centering value (in effect truncating it), but this is probably not the best method -not the intended method at least- and it will alter your result curve and amplitude/phase relation.
If you use minimum-phase filters you will get no pre ringing and only post ringing, and if you use linear-phase filters you will get both pre and post ringing.
So by mixing minimum and linear phase filters, possibly at different frequencies and with different slopes and shapes, you can somewhat "distribute" the ringing around the impulse peak and choose how much pre and post ringing you want, and at what frequency it will appear.
Yes, these are the Problems I am facing with this method, but still, The Gustard is on both files, so the variable is the pcm4222 ADC.... Of course, the filters build up, so results get blurry.
The closest to the source question with converters is so hard to answer... Because all files are at least converted once, probably even more often.
So I think, if this is going to be your Listening DAC, it is completely valid too use a filter that sounds best and not closest to the digital source.
Thinking about it... since all digital files are already butchered with at least one conversion, NOS might even be closer to the analog source, because it might restore some dynamics, that got lost in ADC. Just thinking loud....
But you really would need the very best ADC, but than that could work well, but only with records that never saw a conversion filter before
But you really would need the very best ADC, but than that could work well, but only with records that never saw a conversion filter before
That is true - the TC isn't bad for what it is, but . I also have a Sound Devices 722 which has excellent preamps and fairly good convertors. John Marks did a review of sorts in Stereophile which put it just short of a Grace 201/Tascam DV-RA1000 recording in DSD mode. This was back in 2006 however.
It shouldn't matter if there are conversion filters if they are in the source - just as you can't control the filtering done when producing CD's or digital downloads. Anyway I'm sure I can find a pre digital pressing or two - I have very little vinyl pressed after 1990, I ain't a new skool vinyl hipster
The issue I was pointing in your configuration is that DAM, ADC and x10 all ADD filtering to the original source. Anyway I think we'll have to agree to disagree on the validity of your testing.
I found that if you use a linear filter with adjustable slope it's reasonably straight forward to control the pre/post ringing by tweaking the slope of both Linear and Minimum phase filters.
There is an AES paper from 1995 written by Heylen and Hawksford which presents an algorithm for continuously adjustable phase in FIR filters. The paper includes a supposedly complete but barebones C++ implementation, but after entering the code I've discovered there is at least one macro missing for a critical step, and a second macro that is not used in any of the code. The two macros have different numbers of parameters so it's not a case of one being misnamed...
There are volume differences between many of the filters, so you'll need to recalibrate on a per filter basis.
The first conversion (and a record will have two conversions at least, if recorded digitally) is the most destructive one, in my experience, so you probably will hear more differences with true analogue soundfiles.It shouldn't matter if there are conversion filters if they are in the source - just as you can't control the filtering done when producing CD's or digital downloads. Anyway I'm sure I can find a pre digital pressing or two - I have very little vinyl pressed after 1990, I ain't a new skool vinyl hipster
The issue I was pointing in your configuration is that DAM, ADC and x10 all ADD filtering to the original source. Anyway I think we'll have to agree to disagree on the validity of your testing.
Yes, my testing is seriously flawed, but has definitely advantages (together with disadvantages) over subjective listening in an uncontrolled environment, but through a shorter signal path, you can't deny that.
That's what I meant
Tobias
Yes, quite odd indeed. Doesn’t work on my dac either. We ARE talkng about the same dac board, aren’t we? Mine is 0.01% FW rev. 0.9.
Is there a way that uManager can confirm the actual filter?
Hi Pos, thanks for your explanation.
Not a specific one, but I interpreted your suggestion in this post (http://www.diyaudio.com/forums/vend...magnitude-24-bit-384-khz-158.html#post4219492) as using 1% for MP, 50% for LP and any value intermediate values for IP.
May be I avoided "energy" centering as it sounded a bit mysterious to me Does it correspond to a particular percentage value or is it a different algorithm (sounds like a dynamic/variable thing)? And yes, my intention at that moment was to achieve a 30%/70% distribution of pre/post ringing respectively.
What would be the difference between these 3?
A. Minimum phase filter with centering = middle
B. Minimum phase filter with centering = 1%
C. Minimum phase filter with centering = energy
I expect at least 2 of them to be the same or very close.
Ok, LP and MP parameters are the main tools for adjusting pre/post ringing ratio, not centering.
Cheers Pos!
Not a specific one, but I interpreted your suggestion in this post (http://www.diyaudio.com/forums/vend...magnitude-24-bit-384-khz-158.html#post4219492) as using 1% for MP, 50% for LP and any value intermediate values for IP.
May be I avoided "energy" centering as it sounded a bit mysterious to me
Does it correspond to a particular percentage value or is it a different algorithm (sounds like a dynamic/variable thing)? And yes, my intention at that moment was to achieve a 30%/70% distribution of pre/post ringing respectively.What would be the difference between these 3?
A. Minimum phase filter with centering = middle
B. Minimum phase filter with centering = 1%
C. Minimum phase filter with centering = energy
I expect at least 2 of them to be the same or very close.
Ok, LP and MP parameters are the main tools for adjusting pre/post ringing ratio, not centering.
Cheers Pos!
Yes, 1% if your target curve is MP, not to turn a filter into MP (even if it can work, in a way).
Energy centering tries to find the best window position (ie centering) around the impulse peak to maximize the ringing energy inside it.May be I avoided "energy" centering as it sounded a bit mysterious to me
For example it *should* return something close to 0 for MP targets, and taps/2 (middle) for LP targets.
For MP targets It does not work as intended most of the time, for some reason that I'll have to investigate...
Yes, 1% if your target curve is MP, not to turn a filter into MP (even if it can work, in a way).
Energy centering tries to find the best window position (ie centering) around the impulse peak to maximize the ringing energy inside it.
For example it *should* return something close to 0 for MP targets, and taps/2 (middle) for LP targets.
For MP targets It does not work as intended most of the time, for some reason that I'll have to investigate...
So the ideal method of developing filters would be:
1. Middle centering for LP
2. Try both energy and manual centering for IP (intended ratio) and MP (1%), and test which one performs better.
Many thanks Pos, both for your explanations and for developing rePhase of course
After listening to a few hundred albums of various genres, here is an interim impression report:
There is a BIG difference from recording to recording. So I'd suggest reviewers to listen to many, not just your most favourite one.
Of course it's important to make our favourite recordings sound good. Then, we are in the excitement territory rather than being true to the source. So I'm glad that we'll have multiple filter options. As long as each filter is developed properly, all will have a use.
For the record, my set up is the most basic: Amanero USB > dam1021 > single ended buffered output > Beyerdynamic DT880 600ohms. This way I can eliminate contribution of amplifiers and crossovers in speakers.
Dam is powered by a transformer. It turned out that buffered output impedance and voltage level I use matches well with my headphoes for low distortion.
[Off topic blurb]
This setup should be scoring really high on price/performance in our hobby of diminishing returns.
To whet some more appetite for dam1021, with those good sounding recordings, I can easily figure out the distance between the mic and the acoustic instrument, whether a guitarist is using his nails or fingertips to play the guitar, and the fretbuzz of electric basses that I haven't heard or paid attention before (up to the point of being slightly disturbing ). Not to mention the realism of the instruments. This is new to me, so I'm excited. Others who are already used to this quality may not be so. Bad studio setup or mastering is equally obvious on some others.
There is a BIG difference from recording to recording. So I'd suggest reviewers to listen to many, not just your most favourite one.
Of course it's important to make our favourite recordings sound good. Then, we are in the excitement territory rather than being true to the source. So I'm glad that we'll have multiple filter options. As long as each filter is developed properly, all will have a use.
For the record, my set up is the most basic: Amanero USB > dam1021 > single ended buffered output > Beyerdynamic DT880 600ohms. This way I can eliminate contribution of amplifiers and crossovers in speakers.
Dam is powered by a transformer. It turned out that buffered output impedance and voltage level I use matches well with my headphoes for low distortion.
[Off topic blurb]
This setup should be scoring really high on price/performance in our hobby of diminishing returns.
To whet some more appetite for dam1021, with those good sounding recordings, I can easily figure out the distance between the mic and the acoustic instrument, whether a guitarist is using his nails or fingertips to play the guitar, and the fretbuzz of electric basses that I haven't heard or paid attention before (up to the point of being slightly disturbing
). Not to mention the realism of the instruments. This is new to me, so I'm excited. Others who are already used to this quality may not be so. Bad studio setup or mastering is equally obvious on some others.How about using HQ Player to upsaple to 384 Khz ?
I have been deep into DSD 256 and HQ Player the last 8 months and so have completely missed out on this exciting R2R arrival.
During my explorations with HQ Player (available for Win, Mac and Linux), I have found that the upsampling of 16/44 to 24/352 or 384 is done very well. There are lots of choices for filter types tochoose from. It will all run on a i5 no problem for PCM upsampling. HQ Player also has a NAA = Network Audio Adapter which was only available for Linux, but now is available for Win and Mac as well, so no problems with USB drivers. The NAA can run on a separate computer, and needs only to be seeing a common DHCP on the LAN as the HQ Player Desktop machine. We have also found that the NAA PC can be optical isolated on the LAN using a pair of inexpensive Fibre Media Converters which has a very big effect on shutting out noise coming over the CAT 5 (although you have to be careful about the power supply going to the FMC attached to the NAA PC)
So by using HQ Player to do the umpsapling to 352 or 384, you would only need to have a nice 352 or 384 filter, a much easier task than a filter that does a good job on 16/44
Apologies in advance if I am repeating a previous post, I have not read the whole thread
I have been deep into DSD 256 and HQ Player the last 8 months and so have completely missed out on this exciting R2R arrival.
During my explorations with HQ Player (available for Win, Mac and Linux), I have found that the upsampling of 16/44 to 24/352 or 384 is done very well. There are lots of choices for filter types tochoose from. It will all run on a i5 no problem for PCM upsampling. HQ Player also has a NAA = Network Audio Adapter which was only available for Linux, but now is available for Win and Mac as well, so no problems with USB drivers. The NAA can run on a separate computer, and needs only to be seeing a common DHCP on the LAN as the HQ Player Desktop machine. We have also found that the NAA PC can be optical isolated on the LAN using a pair of inexpensive Fibre Media Converters which has a very big effect on shutting out noise coming over the CAT 5 (although you have to be careful about the power supply going to the FMC attached to the NAA PC)
So by using HQ Player to do the umpsapling to 352 or 384, you would only need to have a nice 352 or 384 filter, a much easier task than a filter that does a good job on 16/44
Apologies in advance if I am repeating a previous post, I have not read the whole thread
Regardless of where it is done you have to deal with the imaged generated by resampling at some point.
Any kind of upsampling or oversampling of 44.1kHz still has a 2.05kHz transition band between the audio band and fs/2.
The DAM basically does in the FPGA what HQ Player does in software.
Last edited: