No, you will burn things.... But if you look at the dac1321 picture on the soekris website, there is a J7 connector just after the switching supply's transformer, that could be used to supply +- 7V directly....
@Granite:
If I understand correctly the J7 connector needs a +7V/-7V supply, but the Uptone Audio LPS-1 does deliver +7V only.
Matt
Hi,
thanks for the input. Seems like I would need two LPS then. You could use them centre tapped to get +/- 7 V. That would be a psu double the price of the DAC. That is a wee bit of am exaggeration I suppose. Beside the fact that combining the two psu to get +/- 7 V is beyond my skills. On the other hand the 5 V 1.1 A through the barrel connector might be a tad low, 5.5 W for max. 3 W power consumption.
Regards
thanks for the input. Seems like I would need two LPS then. You could use them centre tapped to get +/- 7 V. That would be a psu double the price of the DAC. That is a wee bit of am exaggeration I suppose. Beside the fact that combining the two psu to get +/- 7 V is beyond my skills. On the other hand the 5 V 1.1 A through the barrel connector might be a tad low, 5.5 W for max. 3 W power consumption.
Regards
The dac1321 is the most interesting one for me as well because I never listen to headphones and do not need balanced outputs.
IMO, the use of two LPS-1 for 800 USD is overkill.
But one LPS-1 with the 5V output should be ok for the power requirements of the 1321.
It is a pity that the DAC section of the 1321 has lower specs than that of the 1421.
Matt
Last edited:
Hi Matt,
thanks for your take on the LPS-1. If you do not use the digital volume control or if you do not listen at very low volume this should not be a problem in my opinion. 25 bits gives you lots of headroom. If the music is recorded with a mic it will not have more then 15 bit. That leaves you 10 bit or 60 dB of attenuation. Two more bits would be nice but are they necessary? The price difference is also considerable.
Best Regards
thanks for your take on the LPS-1. If you do not use the digital volume control or if you do not listen at very low volume this should not be a problem in my opinion. 25 bits gives you lots of headroom. If the music is recorded with a mic it will not have more then 15 bit. That leaves you 10 bit or 60 dB of attenuation. Two more bits would be nice but are they necessary? The price difference is also considerable.
Best Regards
Thanks Granite,
maybe you are right and the bits of 1321 are enough.
Certainly this DAC is in the sweet spot for me.
I think I will buy one when it is available and use it as an USB bus powered DAC.
If the SQ suffers I will buy a LPS-1.
Would be interesting to compare 1321 with LPS-1 vs. 1421.
All the best
Matt
Hi Matt,
ifi power might be an intermediate step.
Hi TNT,
that is what dither is for. As you said this applies to most digital volume controls, whether you start out with 27 (1451 and 1421) or 25 bits (1321), that is the reason why I like my passive transformer based attenuator.
I do not need the additional imputs the 1421 has, actually USB is all I need. So 25 bits should be ok, 0.8 or 0.3 pS clock jitter not sure I would hear any difference.
Best Regards
ifi power might be an intermediate step.
Hi TNT,
that is what dither is for. As you said this applies to most digital volume controls, whether you start out with 27 (1451 and 1421) or 25 bits (1321), that is the reason why I like my passive transformer based attenuator.
I do not need the additional imputs the 1421 has, actually USB is all I need. So 25 bits should be ok, 0.8 or 0.3 pS clock jitter not sure I would hear any difference.
Best Regards
or 16 bits or 163 bits. These "27 bits" etc are of course internal bit depths and should not be confused with the play-backed PCM bit depth. A larger internal bit depth makes the offending calculation be more accurate but don't think for a moment that there is no effect... the original samples are recalculated.
24 bit (true - not fake "resampled ones") lend themselves better to digital volume control. 16 bit ones not so much - these will quite soon only be 12-13 bit PCM streams. For every 6dB, out goes one bit.
Yes - my conclusion is also that analog sounds superior if you can get hold of a proper quality potentiometer.
Unfortunately I believe that the DAM volume is not dithered
//
24 bit (true - not fake "resampled ones") lend themselves better to digital volume control. 16 bit ones not so much - these will quite soon only be 12-13 bit PCM streams. For every 6dB, out goes one bit.
Yes - my conclusion is also that analog sounds superior if you can get hold of a proper quality potentiometer.
Unfortunately I believe that the DAM volume is not dithered
//
Dithering is something you need only when reducing bit deep. The dams and dacs do not reduce bit deep unless you turn the volume way down, and then it's down where you cannot hear it....
Something needing dithering is easy to recognize, you get limited dynamic range and distortion at low volume, t.ex. a dithered 16 bit signal can easily have a 120 db dynamic range, while a non dithered 16 bit signal are limited to -96 db.
Over at SBAF they did a lot of measurements on the dac1541 using a DScope III, still clean signal at -140 dB, his comments were "Double wow!".... The 20 bit schitts dacs gave up at -120 dB.
But when the dacs and dams are running internal at 35 bit and output 27-28 bit, I don't see any need to add dither, and I consider the digital volume control to be perfect.
Something needing dithering is easy to recognize, you get limited dynamic range and distortion at low volume, t.ex. a dithered 16 bit signal can easily have a 120 db dynamic range, while a non dithered 16 bit signal are limited to -96 db.
Over at SBAF they did a lot of measurements on the dac1541 using a DScope III, still clean signal at -140 dB, his comments were "Double wow!".... The 20 bit schitts dacs gave up at -120 dB.
But when the dacs and dams are running internal at 35 bit and output 27-28 bit, I don't see any need to add dither, and I consider the digital volume control to be perfect.
Dithering is needed every time recalculation has been performed. Volume control is re-calcultion. If one do an operation on a 16 bit word, carried in a 16 bit deep system, the bit depth is still 16 bit after the operation but dither is still required to get rid of the errors introduced by te operation.
A 16bit signal can not have 120 dB clinical dynamic range.
From: Dynamic range - Wikipedia :
"In digital audio theory the dynamic range is limited by quantization error. The maximum achievable dynamic range for a digital audio system with Q-bit uniform quantization is calculated as the ratio of the largest sine-wave rms to rms noise is:[19]
snip...
The 16-bit compact disc has a theoretical undithered dynamic range of about 96 dB,[20][d] however, the perceived dynamic range of 16-bit audio can be 120 dB or more with noise-shaped dither, taking advantage of the frequency response of the human ear.[21][22"
NB: "can be..." the DAM don't incorporate dither so for a 16 bit signal it will be 96 dB or less if the digital volume control is engaged and set to attenuate. This is theory and can't be argued. For 24 bit signals, it's a better situation. Just to be really clear - for 16 bit audio the 25/27 bit deep processing "lanes" in the DAMs don't matter.
The SBAF test was made at 24 or maybe even 32 bit PCM wide payload (test signal)?
//
A 16bit signal can not have 120 dB clinical dynamic range.
From: Dynamic range - Wikipedia :
"In digital audio theory the dynamic range is limited by quantization error. The maximum achievable dynamic range for a digital audio system with Q-bit uniform quantization is calculated as the ratio of the largest sine-wave rms to rms noise is:[19]
snip...
The 16-bit compact disc has a theoretical undithered dynamic range of about 96 dB,[20][d] however, the perceived dynamic range of 16-bit audio can be 120 dB or more with noise-shaped dither, taking advantage of the frequency response of the human ear.[21][22"
NB: "can be..." the DAM don't incorporate dither so for a 16 bit signal it will be 96 dB or less if the digital volume control is engaged and set to attenuate. This is theory and can't be argued. For 24 bit signals, it's a better situation. Just to be really clear - for 16 bit audio the 25/27 bit deep processing "lanes" in the DAMs don't matter.
The SBAF test was made at 24 or maybe even 32 bit PCM wide payload (test signal)?
//
Last edited:
Eeeh, like what I said.... Except it don't have anything to do with the response of the human ear, statistics do the trick with basic dither, and noise shaping moves the noise from dither out from the human ears sensitive range.... Don't trust wikipedias simplified articles to much....
Not correct, the dams and dacs have 35 bits internal, not 16 bits, input bit lengths get expanded immediately to 35 bit and first reduced when outputting to the R-2R resistors networks. Again, you only need to dither when reducing bit lengths, you will not get there with a 16 bit input until the volume is down min 11 bits (66 dB) and then you can't hear anything anyway....
The SBAF testing was probably with 24 bit, I don't have any products taking 32 bits in.... In your theory the signal should be ruined and the dac not able to reproduce a -140 dB signal that clean, as at 0 dB the dac1541 actually have the digital volume at -1 dB and therefore active....
I see now that the measurements was done for 24 bit signals. I think you are not correct in thinking that only bit depth reduction requires dither. My 16 bit example did not try to describe your products internal processing depth but rather the property of the signal.
If you want to digitally add 10 dB gain to a 24 bit signal as in your products you will need 2 bits more the accommodate the new result -> 26 bits.
I play 97% good ol CD quality and 3% hirez.
Please explain how 25/27/35 bit internal processing depth helps when I want to digitally attenuate a 44/16 (red book) signal 19 dB? And maintain -140 dB signal integrity? This is how one might read your statements above....
These are DAM measurements for 16 bit signals I would like to see:
1) 0 dBfs signal with 0 dB digital attenuation 300R 1kHz
2) 0 dBfs signal with 19 dB digital attenuation 300R 1kHz
3) 0 dBfs signal with 19 dB digital attenuation 300R 1kHz with dither (not possible today...)
4) -20 dBfs signal with 19 dB digital attenuation 300R 1kHz
//
PS. The handling of 24 bit signal is indeed impressive. DS.
If you want to digitally add 10 dB gain to a 24 bit signal as in your products you will need 2 bits more the accommodate the new result -> 26 bits.
I play 97% good ol CD quality and 3% hirez.
Please explain how 25/27/35 bit internal processing depth helps when I want to digitally attenuate a 44/16 (red book) signal 19 dB? And maintain -140 dB signal integrity? This is how one might read your statements above....
These are DAM measurements for 16 bit signals I would like to see:
1) 0 dBfs signal with 0 dB digital attenuation 300R 1kHz
2) 0 dBfs signal with 19 dB digital attenuation 300R 1kHz
3) 0 dBfs signal with 19 dB digital attenuation 300R 1kHz with dither (not possible today...)
4) -20 dBfs signal with 19 dB digital attenuation 300R 1kHz
//
PS. The handling of 24 bit signal is indeed impressive. DS.
IMHO:
If your input, or computational result, has more bit-resolution than your output/intermediate signal, dithering adds "noise" to the last bit(s) which is chosen such that, for periodic signals, in average, you get the lost bits back.
Thus for measurements that do some averaging (such as fft) with periodic signals (e.g. pure sine) dither looks good.
For music the question is for which time interval music looks approximately like an periodic signal so that the averaging approach can be assumed working. And for which time interval the human ear does something like averaging?
I agree with Soren that, the only place where dithering would make sense (if any) would be when mapping the internal representation to the DAC ladder. The internal computation can be assumed to be exact up to the 30+ bits of the internal representation. Dither would (with the above restrictions) recover some the bits lost by mapping the internal representation to the 28 bits of the ladder. Dithering previously would have only an effect to the beyond the 30+ bits which is lost most likely anyway.
P.S. I meanwhile have some doubt in the usefulness of extensively long filters, for the same reasons: They assume that the signal is periodic to achieve the the figures to which you designed them. So if the music can not be viewed as approximately periodic for at least as many samples as the filter is long it will not perform as designed.
If your input, or computational result, has more bit-resolution than your output/intermediate signal, dithering adds "noise" to the last bit(s) which is chosen such that, for periodic signals, in average, you get the lost bits back.
Thus for measurements that do some averaging (such as fft) with periodic signals (e.g. pure sine) dither looks good.
For music the question is for which time interval music looks approximately like an periodic signal so that the averaging approach can be assumed working. And for which time interval the human ear does something like averaging?
I agree with Soren that, the only place where dithering would make sense (if any) would be when mapping the internal representation to the DAC ladder. The internal computation can be assumed to be exact up to the 30+ bits of the internal representation. Dither would (with the above restrictions) recover some the bits lost by mapping the internal representation to the 28 bits of the ladder. Dithering previously would have only an effect to the beyond the 30+ bits which is lost most likely anyway.
P.S. I meanwhile have some doubt in the usefulness of extensively long filters, for the same reasons: They assume that the signal is periodic to achieve the the figures to which you designed them. So if the music can not be viewed as approximately periodic for at least as many samples as the filter is long it will not perform as designed.
No, no, it is just that case with the volume control.
You have 16 or 24 bit input signal. This is mapped lossless to the 31 bit internal representation. Multiplication (volume control, FIR filters) is done 35-bit, if I recall correctly. So there you could use dither when mapped back to the 31 internal bits. But finally it is all mapped to the 28-bit ladder. So bit 32 is not that important if you have to care about bit 29 first.
16-bit input is not more dither needing than high res in this context.
Switch power on dam 1021
Looking at the way Soeren has powered the 1541, I cannot avoid thinking whether one could do the same with the 1021. Two Mean Well 12V for that, and a 5v MeanWell for the USB? I admit that I am a newbie in this, so please tell me why this will not work as well with the 1021 as with the 1521
Looking at the way Soeren has powered the 1541, I cannot avoid thinking whether one could do the same with the 1021. Two Mean Well 12V for that, and a 5v MeanWell for the USB? I admit that I am a newbie in this, so please tell me why this will not work as well with the 1021 as with the 1521
to dither or not to dither
Hi Guys,
the discussion about the neccessity of dithering is an interesting one, the only caveat for me is that this thread might not be the right one to continue the discussion. Havn't been around here for too long, so I do not know the etiquette. Is it possible that a moderator opens a new thread an moves the corresponding messages there?
For me concerning dithering these links were helpful:
Dithering Explained: What it is, When to Use It, and Why it’s Important
When should you use dither ? - Production Advice
When in doubt, dither - Production Advice
Dither or distort ? Listen and decide for yourself - Production Advice
Concerning the fact that DAC may increase the bit depth from 16 (or 24) bitsnto a higher level and then use digital volume control the question for me is whether this leaves the original 16 (or 24) intact. As far as I know when increasing bit depth is done by adding zeroes to the original code, when I reduce volume do I then only strip away the added zeroes and do not change the signal until I reach the original bit depth?
It is clear to me that with increasing bit depth I have more level in which to divide the signal (therefore higher resolution). If I reduce a 32 bit signal to 16 bits I will run into trouble because certain levels disappear and I need to put the signal in a level which does not correspond to its initial one introducing quantisation errors, which cause distortion (there you ate TNT it is so easy to give more distortion to the people). What I do not know what happens if I take a 16 bit signal, increase bit depth (to 32 for instance) and then reduce bit depth to sixteen again. Initially all samples are in 16 bit levels of course. Then they are put into 32 bit levels and then back into 16 bit levels. If I do not manipulate the signal (samples) they should (in my mind) still be in levels which exist in both bit depths and they should find there initial place meaning this would not introduce quantisation error. If I did manipulate the signal in the meantime levels which exist in the 32 bit world might be populated and when going back to 16 bits I will introduce quantisation error. What happens when I attenuate the signal I do not know.
The question is of interest to me because I upsample all my music to 32/96 kHz on the fly apply DRC via a filter and hand over the 32 bit signal to the DAC. Now in theory USB Receiverchips like Amanero, XMos and CMedia are able to handle 32 bits (but not all of them), still a lot of DAC manufacturers specify 24/384 as the capability of the USB input. That makes me womder whether they just truncate an incoming 32 bit signal, which if I am not mistaken also introduces errors or whether they convert the 32 bit incoming data to 24 bit internally by applying dither (then everything would be ok). Next question would be, if a DAC manufacturer specifies 32/384 does it mean the DAC handels 32 bit signals throughout the digital domain or does it only mean that the USB receiverchip is able to handle 32 bit data but the DAC still works with 24 bit data internally. Food for thought.
Regards
Hi Guys,
the discussion about the neccessity of dithering is an interesting one, the only caveat for me is that this thread might not be the right one to continue the discussion. Havn't been around here for too long, so I do not know the etiquette. Is it possible that a moderator opens a new thread an moves the corresponding messages there?
For me concerning dithering these links were helpful:
Dithering Explained: What it is, When to Use It, and Why it’s Important
When should you use dither ? - Production Advice
When in doubt, dither - Production Advice
Dither or distort ? Listen and decide for yourself - Production Advice
Concerning the fact that DAC may increase the bit depth from 16 (or 24) bitsnto a higher level and then use digital volume control the question for me is whether this leaves the original 16 (or 24) intact. As far as I know when increasing bit depth is done by adding zeroes to the original code, when I reduce volume do I then only strip away the added zeroes and do not change the signal until I reach the original bit depth?
It is clear to me that with increasing bit depth I have more level in which to divide the signal (therefore higher resolution). If I reduce a 32 bit signal to 16 bits I will run into trouble because certain levels disappear and I need to put the signal in a level which does not correspond to its initial one introducing quantisation errors, which cause distortion (there you ate TNT it is so easy to give more distortion to the people). What I do not know what happens if I take a 16 bit signal, increase bit depth (to 32 for instance) and then reduce bit depth to sixteen again. Initially all samples are in 16 bit levels of course. Then they are put into 32 bit levels and then back into 16 bit levels. If I do not manipulate the signal (samples) they should (in my mind) still be in levels which exist in both bit depths and they should find there initial place meaning this would not introduce quantisation error. If I did manipulate the signal in the meantime levels which exist in the 32 bit world might be populated and when going back to 16 bits I will introduce quantisation error. What happens when I attenuate the signal I do not know.
The question is of interest to me because I upsample all my music to 32/96 kHz on the fly apply DRC via a filter and hand over the 32 bit signal to the DAC. Now in theory USB Receiverchips like Amanero, XMos and CMedia are able to handle 32 bits (but not all of them), still a lot of DAC manufacturers specify 24/384 as the capability of the USB input. That makes me womder whether they just truncate an incoming 32 bit signal, which if I am not mistaken also introduces errors or whether they convert the 32 bit incoming data to 24 bit internally by applying dither (then everything would be ok). Next question would be, if a DAC manufacturer specifies 32/384 does it mean the DAC handels 32 bit signals throughout the digital domain or does it only mean that the USB receiverchip is able to handle 32 bit data but the DAC still works with 24 bit data internally. Food for thought.
Regards
That could work. What you don't see is what filtering and such Soren does to the power before it goes to the regulators after the power supply. So while you could just hook up the meanwell power supply and power the Dam-1021, it may not sound as good as Soren's implementation in the 1521.