Line out vs digital in
First of all I'm learning everyday as I've landed a DJ gig. Just bought a DCX2496 and running 2 way Peavey SP-1's (modded) and a sub, in 3 way stereo. I use a PC as source with Foobar 2000 + Wasapi (event) plugin, 0.0 db gain, (rca 75 ohm to xlr to input A ) DCX setup as AES/BEU input.After fiddle f******g with xover's and output gains, My modest amps are barely idling. Question is, you are saying that using an analog output control for master volume ie: Jan Didden's mod would be the best bet to max sound quality for a very limited budget? Regards 2S2P
Exactly. For optimal sound quality, you want to run the D/A near or at full output level (without clipping, of course). This maximizes the use of the digital bits, which is information about the sound to the D/A. If you attempt to use digital volume control or do not run input/outputs at maximum levels, you are effectively throwing bits away, which results in less information about the original sound. Also can increase s/n ratio.
Ideally, you want the hottest signal possible out of the DCX (this doesn't mean crank up the digital gain in the DSP). Your ganged analog volume control then adjusts the signal to each of your amps based on desired audible volume level.
Ideally, you want the hottest signal possible out of the DCX (this doesn't mean crank up the digital gain in the DSP). Your ganged analog volume control then adjusts the signal to each of your amps based on desired audible volume level.
sorry if this has already been asked (searches don't find this info).
what logic level (voltage) is i2s run at, in the dcx2496? is it 3.3v or 5v?
I'm about to try to interface some wolfson wm8805 spdif tx chips by tapping into the i2s lines on the behringer. but it would be nice to know that its already 3.3v (?)
what logic level (voltage) is i2s run at, in the dcx2496? is it 3.3v or 5v?
I'm about to try to interface some wolfson wm8805 spdif tx chips by tapping into the i2s lines on the behringer. but it would be nice to know that its already 3.3v (?)
The DCX2496 is a mixed 3.3V and 5V design but the link between DSP and ADC/DACs is 3.3V only.
Is there a certain reason for converting onboard I2S to S/PDIF? This would need a sample rate conversion (SRC) and a reclocking on the destination side of the S/PDIF line similar to the digital input (CS8420) of the DCX. This will cause a decrease of sonical performance!
I'm not aware of the actual discussions. I set up a list of the available mods for the DCX two years ago. Might be that some links don’t work any more but perhaps it’s helpful for you.
Hello Frank,
I'd like to mention that for my active mod there is now a small optional mic input board.
The digital input XLR connector is replaced with a combined XLR and 6.3mm stereo mic connector on a small board. Ward Maas from Pilgham Audio can supply the kit.
Maybe you could update your overview accordingly:
Retains Mic input ? yes (option)
jan
I'd like to mention that for my active mod there is now a small optional mic input board.
The digital input XLR connector is replaced with a combined XLR and 6.3mm stereo mic connector on a small board. Ward Maas from Pilgham Audio can supply the kit.
Maybe you could update your overview accordingly:
Retains Mic input ? yes (option)
jan
Last edited:
Hi Jan,
I updated my file according to your proposal. I asume there aren't any other new mods for the DCX available meanwhile but prices might have changed? So please look on the sites of the sellers for the latest prices.
Hi Frank.
I know you have a LOT experience with DCX. I want make USB digital input on DCX with Amanero USB to i2s http://www.diyaudio.com/forums/vendors-bazaar/216474-usb-i2s-384khz-dsd-converter.html.
I want ask Amanero to modify drivers so there will be just 96khz option /safety reason/ and resample on pc on source side /digital in "clockless" domain/. For example you can force win7 to resample all sound to 96 khz with decent quality algorithm. All dcx devices will be slaved to amanero 24.576 mhz clock. This way there will be just one physical clock domain and you omit ASRC. What do you think about such solutions? How you solutions react to jitter on source side? Is there some measurable differences between sources or are modern ASRC algorithm clever enough to suppress any problem before them?
Thank you very much for answer and thanks for all your very rational DIY DCX contributions so far.
I know you have a LOT experience with DCX. I want make USB digital input on DCX with Amanero USB to i2s http://www.diyaudio.com/forums/vendors-bazaar/216474-usb-i2s-384khz-dsd-converter.html.
I want ask Amanero to modify drivers so there will be just 96khz option /safety reason/ and resample on pc on source side /digital in "clockless" domain/. For example you can force win7 to resample all sound to 96 khz with decent quality algorithm. All dcx devices will be slaved to amanero 24.576 mhz clock. This way there will be just one physical clock domain and you omit ASRC. What do you think about such solutions? How you solutions react to jitter on source side? Is there some measurable differences between sources or are modern ASRC algorithm clever enough to suppress any problem before them?
Thank you very much for answer and thanks for all your very rational DIY DCX contributions so far.
yes! I want to run my own choice of dacs as the output stage. many people prefer to mod the analog chain on the dcx's output but I'd like to go all digital and have the box just run spdif-in and spdif-outs.
I'm planning on using a chip I've already worked with, the wm8805. I've been able to hack a chip into the audio widget and that's been reliably converting i2s to spdif. my audio widget is now just a usb/spdif box at up to 24/192.
I'm not sure I follow this. no src is needed. the data is 24/96 as it hits the local set of 3 dacs onboard. all I'm doing is taking the 24/96 data and skipping the AK dacs and sending it to 2 or 3 wolfson transmitter chips.
people have done this before:
Behringer DCX2496
I think all the kits of boards that were done years ago are all sold out, so I'm going to try to revisit this idea but with 8805 chips instead of the cirrus ones.
I'm hoping I can use the 3 clocks as-is. I know lrclock and bitclock are 'fine' but I'm not 100% sure if I can just feed the onboard mclock to the wolfson. I'll find out really soon, though, before I build too much.
also, this is where someone else did the same mod as I'm planning:
http://www.diyaudio.com/forums/digital-line-level/36252-dcx2496-mod-digital-out-x-3-a.html
http://www.diyaudio.com/forums/digital-line-level/36252-dcx2496-mod-digital-out-x-3-a.html
Hi tomtom,
No sample rate conversion (SRC) is lossless. There is no difference whether it is done by software or hardware. Depending on the quality of the software it even may be worse than the AD1896 used on my mod. That’s why it makes sense that the output sample rate of your PC audio board is the same as your audio data (most probably 44.1 kHz) so that there is no conversion in the PC.
So if your audio data isn’t already 96 kHz there is no advantage there is always at least one conversion.
If using a USB to I2S converter the common clock source must be close to the DACs. The jitter of the clock for the DACs is very important! That’s the reason why my mod (2.6 ps jitter) is replacing the poor clock generator on the DCX DSP board.
Hi linuxworks,
To my knowledge S/PDIF combine data and clock on one line. I assume at the end of your S/PDIF line is somewher a DAC which needs both clock and data seperated. So you first would need a PLL for this separation and than you need to reclock both it with a sample rate converter (SRC) to reduce jitter.
A solution without reclocking but PLL only would decrease audio quality dramatically because of high jitter.
this is just regular shipping of data over spdif as a transport, same as usual. whatever 'issues' there are in spdif being clock+data, its no different here.
you can ship digital audio over i2s or spdif; that's your 2 choices these days. almost no dacs natively take an i2s box-to-box connection. when you go thru boxes, you go via spdif, almost always.
unless you do interbox i2s *very well*, its just better to use spdif. don't you agree?
I've already decided I want to bypass/ignore the dac side of the behringer. I'd have to replace op-amps, clean up the analog supply and then still deal with the super high voltage level out that swamps my current vol control (cs3318, which does not take anything close to +22db levels that the behringer puts out).
on the receive side of the dacs (my outboard dacs), there may or may not be reclockers. sometimes they can be asrc chip front-ends or lesser reclockers like the wm8804/5 series. some dacs I bought recently do actually have 8805's as their spdif receivers, so if you believe wolfson, there is reclocking going on at the receive side (dac box).
but regardless, even if there was not reclocking going on, so what? if you agree that you want to use outboard dac technology (ie, have the choice to use outboard dacs) then you have only 2 choices for data interconnect and really just 1 if you want everything to be able to connect to it.
I guess I'll find out how 'bad' it is to take the internal i2s, convert to coax level spdif and then into some decent outboard dacs. I suspect that I'll get better performance with this strategy than directly modding the analog inside this noisybox
you can ship digital audio over i2s or spdif; that's your 2 choices these days. almost no dacs natively take an i2s box-to-box connection. when you go thru boxes, you go via spdif, almost always.
unless you do interbox i2s *very well*, its just better to use spdif. don't you agree?
I've already decided I want to bypass/ignore the dac side of the behringer. I'd have to replace op-amps, clean up the analog supply and then still deal with the super high voltage level out that swamps my current vol control (cs3318, which does not take anything close to +22db levels that the behringer puts out).
on the receive side of the dacs (my outboard dacs), there may or may not be reclockers. sometimes they can be asrc chip front-ends or lesser reclockers like the wm8804/5 series. some dacs I bought recently do actually have 8805's as their spdif receivers, so if you believe wolfson, there is reclocking going on at the receive side (dac box).
but regardless, even if there was not reclocking going on, so what? if you agree that you want to use outboard dac technology (ie, have the choice to use outboard dacs) then you have only 2 choices for data interconnect and really just 1 if you want everything to be able to connect to it.
I guess I'll find out how 'bad' it is to take the internal i2s, convert to coax level spdif and then into some decent outboard dacs. I suspect that I'll get better performance with this strategy than directly modding the analog inside this noisybox
also, whatever jitter is there already at the pins or pads I intend to tap into, that wont ever be cured by any reclocking. the reclocking won't help 'embedded jitter' which I think is already 'in there' by the time I'm tapping i2s.
the best I could hope is to not induce *more*, during the transit. and hope that the receiver can faithfully unpack what I just sent.
the 2 boxes (dcx and outboard dac) will be right on top of each other and cable will be short and clean (less than half a foot, probably, and perhaps even using bnc).
the best I could hope is to not induce *more*, during the transit. and hope that the receiver can faithfully unpack what I just sent.
the 2 boxes (dcx and outboard dac) will be right on top of each other and cable will be short and clean (less than half a foot, probably, and perhaps even using bnc).
Some basics:
Copying audio files or transferring them via USB is lossless. That’s because these files are data only plus the information, that the data was sampled e.g. with 44.1 kHz.
Transferring audio data via P2S or S/PDIF is NOT lossless! That’s because clock isn’t an exact and always constant value any more. It has become part of the signal and it is based on a real clock generator which never is exact nor always the same. It isn’t such important that the clock generator is 100% exact. That’s because our brain isn’t a frequency counter. But we can hear clock jitter very well. So this is a real issue.
Transferring audio data via a long I2S or S/PDIF cable adds a lot of jitter. Without reclocking and so removing most of the jitter audio quality would be unacceptable.
So connecting a DAC via S/PDIF without reclocking (SRC) isn’t an option at all.
Hi Frank,
Thank you very much for quick answer. I don't want to argue with you. But there is imo difference with Asynchronous SRC where there are two independent clock domain /the source and the dac/ and data are reclocked to dac clock with source jitter included in process AND resampling in "pure data" domain without any clock at all /resampling data in PC/. So in my view only jitter then is dac clock jitter. I have no knowledge nor experience about audibility of these things.
So my question was if you see any REAL advantage of "offline data" resampling vs ASCR /AD 1896/ resampling. Theoretical advantage is imo obvious.
Many thanks again.
Tomas
The subject you started, is very interesting one and it opens conversation that many are trying to tackle. DCX sans DACs, output, and clock... What is left? Why are we using it? For DSP, because so far there is very little out that does it as well as DCX does. The question is if this is best approach since DCX is limited to 96KHz, and doing all that work with format that is already far behind does not makes much sense in my mind.
But if that is not an obstacle, and I am sure many would say they do not care for anything over 96KHz, than thinking forward how to move signal is a good question. In that light I believe that if DACs are removed, and in their place we substitute some kind of pin holder, so that we could attach another board on the top, than that might be the way to go. On that top board than we could have any new DACs with new integrated Oettle's reclocker. This effort in my mind would worth only if all new electronic is designed and integrated on the board. The good choice might be ESS DACs or DAC. This will allow us to move unchanged signals that were delivered to original DACs, and the distance would be minimally increased, what is imperative if I2S is used.
Other than that - internal solution - no other effort really makes sense. I have not seen anyone made as good DSP based solution that satisfy all requests audiophiles expect.
That way we could use DCX's solid DSP and dramatically improve DACs and output.
But if that is not an obstacle, and I am sure many would say they do not care for anything over 96KHz, than thinking forward how to move signal is a good question. In that light I believe that if DACs are removed, and in their place we substitute some kind of pin holder, so that we could attach another board on the top, than that might be the way to go. On that top board than we could have any new DACs with new integrated Oettle's reclocker. This effort in my mind would worth only if all new electronic is designed and integrated on the board. The good choice might be ESS DACs or DAC. This will allow us to move unchanged signals that were delivered to original DACs, and the distance would be minimally increased, what is imperative if I2S is used.
Other than that - internal solution - no other effort really makes sense. I have not seen anyone made as good DSP based solution that satisfy all requests audiophiles expect.
That way we could use DCX's solid DSP and dramatically improve DACs and output.
well, I'm going to try and see for myself, I guess. I've located the 3 holes for the 3 clocks that I'll need. the 3 smd resistors are the 3 data lines.
I'm still not understanding what you are trying to warn me about. the data is already digital and all I'm doing is placing an spdif transmitter and receiver pair in between the 2 ends of i2s. same as any spdif transport would be!
I understand that i2s clock (at least one clock, depending on the technology used in the spdif receiver end of things) can be critical and if not handled well, it can add jitter. I'll do my best to try to treat those lines carefully in my mod. not sure what else I can do.
I've always believed that the transmitted jitter (introduced by the stage where you have i2s as an input and spdif as an output) can be attenuated at the receiving end by buffering and reclocking. we both agree in this and any dac I'll use will have a good enough input receiver section, including buffering+reclocking.
the inherent jitter that is already in the signal at the place where we pick up i2s traces can't be attenuated; that will be what we have to live with (even the analog guys who use the onboard dacs can't attenuate this built-in jitter). so that's a non-issue in this mod.
anyway, I'll try and see if this gets me acceptable analog results at the output of the dacs that I plan to use. I'll do some rmaa testing (maybe others if I can get access to better test gear) and see if the end to end is as bad as you think it will be. I just don't agree that this is bound for failure, but I'll find out once its built
I'm still not understanding what you are trying to warn me about. the data is already digital and all I'm doing is placing an spdif transmitter and receiver pair in between the 2 ends of i2s. same as any spdif transport would be!
I understand that i2s clock (at least one clock, depending on the technology used in the spdif receiver end of things) can be critical and if not handled well, it can add jitter. I'll do my best to try to treat those lines carefully in my mod. not sure what else I can do.
I've always believed that the transmitted jitter (introduced by the stage where you have i2s as an input and spdif as an output) can be attenuated at the receiving end by buffering and reclocking. we both agree in this and any dac I'll use will have a good enough input receiver section, including buffering+reclocking.
the inherent jitter that is already in the signal at the place where we pick up i2s traces can't be attenuated; that will be what we have to live with (even the analog guys who use the onboard dacs can't attenuate this built-in jitter). so that's a non-issue in this mod.
anyway, I'll try and see if this gets me acceptable analog results at the output of the dacs that I plan to use. I'll do some rmaa testing (maybe others if I can get access to better test gear) and see if the end to end is as bad as you think it will be. I just don't agree that this is bound for failure, but I'll find out once its built
Attachments
Hi tomtom,
I hope i got your question.
When using software for sample rate conversion there are also to clock domains (otherwise it wouldn't be a conversion). So there is no difference compared to the hardware SRC if both have the same conversion quality. The quality of the AD1896 you can easily see in the datasheet but you don't know the quality of your conversion software. So I would prefer the hardware solution. It's also much more flexible using different audio sources.
In any way you need a low jitter clock close to the DACs.
The only device you could avoid is the PLL but only if you use the DAC clock as a source for your USB to I2S converter. I fear the benefit avoiding a PLL is not pretty high.