Hi Philip,
Without a circuit (at least one channel) and input impedance of your power amp it's nearly impossible to help you. The AKM D/A converter has a balanced output. I've no idea how they convert it to a unbalanced output. This might be the main problem of this french mod. The balanced circuit shown in post 858 would be OK with load impedances above R10k.
The filter depends very much on the load (input impedance of your power or volume control). Low impedances will cause problems on the low and high frequency side (higher attenuation).
In general all passive solutions have a slow HF (+ 100kHz) roll off problem.
Frank
Without a circuit (at least one channel) and input impedance of your power amp it's nearly impossible to help you. The AKM D/A converter has a balanced output. I've no idea how they convert it to a unbalanced output. This might be the main problem of this french mod. The balanced circuit shown in post 858 would be OK with load impedances above R10k.
The filter depends very much on the load (input impedance of your power or volume control). Low impedances will cause problems on the low and high frequency side (higher attenuation).
In general all passive solutions have a slow HF (+ 100kHz) roll off problem.
Frank
I do not know how the DCX processes it's data, but the assumption that the 24bit capability retains sufficient data when an input bit chopping procedure is applied seems poor.pgruebele said:
I would expect that the MSB in 16bit is set to match MSB in 24bit.
Chop off 30db and you will lose about 5bits leaving just 11bits of data left. If DCX moved all the bits down by 5bits using levels 5 down to 21 then DCX has managed to maintain all the data when gain is reduced by 30db. But I don't think this is how they do it.
But the remainder at the 15 or 16 bit level is not very accurate (we are only consumers paying consumer prices). similarly when claimed 24bit electronics is specified any level below about 20bits is becoming progressively inaccurate and with today's technology probably impossible to achieve 23bit or 24bit in consumer targeted electronics.
Finally, is -30db enough, mindfull that MSB is at +22dbu and -8dbu is a signal of over 300mV (about 9Wpk into 8ohm with a low gain amp)? That is still fairly loud on some program material.
The amp has 32KOhms impedance.
I agree that any digital attenuation is probably chopping bits off the original data. However, I don't hear any degradation with volume. If I listen with input attenuation at 0db (no chopping?), I hear the same problems as with -15 attenuation. The DCX actually does not have -30 attenuation, just +-15 anyway.
At any rate, if this is audible bit chopping is audible it should have been worse with the stock board since I had to keep levels VERY low in order to keep from overdriving the amp...
I also don't think that humans can really hear the difference between 14 bits and 16 bits as long as a filter removes digital noise after the DAC.
Therefore I still don't understand why this board does not sound better. It seems that changing the caps to smaller and better ones is still an option?
Philip
I agree that any digital attenuation is probably chopping bits off the original data. However, I don't hear any degradation with volume. If I listen with input attenuation at 0db (no chopping?), I hear the same problems as with -15 attenuation. The DCX actually does not have -30 attenuation, just +-15 anyway.
At any rate, if this is audible bit chopping is audible it should have been worse with the stock board since I had to keep levels VERY low in order to keep from overdriving the amp...
I also don't think that humans can really hear the difference between 14 bits and 16 bits as long as a filter removes digital noise after the DAC.
Therefore I still don't understand why this board does not sound better. It seems that changing the caps to smaller and better ones is still an option?
Philip
Hi Giulio & Pg,
are you confirming that DCX does a bit chopping or bit down shifting when the digital attenuators are activated?
I have an ulterior motive. I have recently acquired a pair of active Tannoy monitors. They have dual analoque and digital inputs. But the digital side needs digital volume control before reaching the Tannoy. They too claim 24bit ability but I want to preserve as much of the data stream as possible and I need to know/understand what to look for when the digital route becomes my input preference.
are you confirming that DCX does a bit chopping or bit down shifting when the digital attenuators are activated?
I have an ulterior motive. I have recently acquired a pair of active Tannoy monitors. They have dual analoque and digital inputs. But the digital side needs digital volume control before reaching the Tannoy. They too claim 24bit ability but I want to preserve as much of the data stream as possible and I need to know/understand what to look for when the digital route becomes my input preference.
My cables are jut 1.5 feet long, so I don't think that's the issue. The only thing I can think of is to replace the caps for the mid and high channels with better/smaller caps...
Regarding volume control: Since the 16bit data is kept in the most significant of the 24 bits sent to the dac (at full volume), I think that it is not necessary to implement post-DCX volume control. The are 8 bits available for shifting down without loosing information. A post-DCX volume control would decrease things like digital noise and dithering, but I would think that this would be negligible compared to other issues. Of course in reality you probably only have about 6 bits to play with since the dac is probably not accurate to 24 bits itself...
Philip
Regarding volume control: Since the 16bit data is kept in the most significant of the 24 bits sent to the dac (at full volume), I think that it is not necessary to implement post-DCX volume control. The are 8 bits available for shifting down without loosing information. A post-DCX volume control would decrease things like digital noise and dithering, but I would think that this would be negligible compared to other issues. Of course in reality you probably only have about 6 bits to play with since the dac is probably not accurate to 24 bits itself...
Philip
Hi Andrew,
I am no digital whizard but I would say that the digital volume control does exactly what you say. The dcx can control gain at either the input or the output. If attenuation is done at the input I am not clear whether it takes place before or after upsampling (my ignorance). In the first case one would end up with less than 16 bit (if feeding a redbook CD digital signal) to start with.
Philip,
of course you can change mid and tweeter DC coupling caps to lower values as Jan suggests in the AudioXpress article linked above. But the way I understood your problem is that you are not complaining about some kind of tonal balance problem (you say the system sound wonderful on vocals and guitar), but about it sounding muffled (?) on complex musical passages. Now, I would use muffled to describe a tonality problem (lack of high freq extension) but in that case it should be there with all kind of music (keeping constant the frequency content). When you talk about complex passages instead, I immediately think about problems with dynamics, suggesting that one might want to buffer the output of the dcx (the same problem that people with passive resistive preamp often complain about). Talking about caps, those Mundorf M-Caps are supposed to be pretty good. Replacing them with something better (I do not know how much smaller you can go on the mid channel) may be rather expensive...
A third explanation is that the output stage is more transparent and revealing crossover setting problems that the previous output stage was masking. Have you experimented with the crossover points, slope, phase, etc. ?
Do keep us posted.
Best
Giulio
I am no digital whizard but I would say that the digital volume control does exactly what you say. The dcx can control gain at either the input or the output. If attenuation is done at the input I am not clear whether it takes place before or after upsampling (my ignorance). In the first case one would end up with less than 16 bit (if feeding a redbook CD digital signal) to start with.
Philip,
of course you can change mid and tweeter DC coupling caps to lower values as Jan suggests in the AudioXpress article linked above. But the way I understood your problem is that you are not complaining about some kind of tonal balance problem (you say the system sound wonderful on vocals and guitar), but about it sounding muffled (?) on complex musical passages. Now, I would use muffled to describe a tonality problem (lack of high freq extension) but in that case it should be there with all kind of music (keeping constant the frequency content). When you talk about complex passages instead, I immediately think about problems with dynamics, suggesting that one might want to buffer the output of the dcx (the same problem that people with passive resistive preamp often complain about). Talking about caps, those Mundorf M-Caps are supposed to be pretty good. Replacing them with something better (I do not know how much smaller you can go on the mid channel) may be rather expensive...
A third explanation is that the output stage is more transparent and revealing crossover setting problems that the previous output stage was masking. Have you experimented with the crossover points, slope, phase, etc. ?
Do keep us posted.
Best
Giulio
Hi Philip,
has anyone thought about the question why AKM designed a DAC with balanced outputs? What you should amplify is the difference between these two outputs and not the difference between one output and ground. That's the reason why I would prefer the Jan Didden design. Also Jans design supports mute.
The circuit shown in post 858 is OK too.
The Mundorf caps are preety good. You will hardly find better ones. Reducing values won't improve sound quality, but might increase phase shift on the low side. For a tweeter with a 3kHz crossover a 0.47uF cap would be sufficient (load not below 10k). The mid (250Hz) should have at least 2.2uF for not too high phase shift
Use balanced cables (microphon) and use ground (XLR: 1) for shielding only. Connect your power amp between the hot (XLR: 2) and cold (XLR: 3) output. That's not possible with the (poor) 'french' design you have.
Good luck,
Frank
has anyone thought about the question why AKM designed a DAC with balanced outputs? What you should amplify is the difference between these two outputs and not the difference between one output and ground. That's the reason why I would prefer the Jan Didden design. Also Jans design supports mute.
The circuit shown in post 858 is OK too.
The Mundorf caps are preety good. You will hardly find better ones. Reducing values won't improve sound quality, but might increase phase shift on the low side. For a tweeter with a 3kHz crossover a 0.47uF cap would be sufficient (load not below 10k). The mid (250Hz) should have at least 2.2uF for not too high phase shift
Use balanced cables (microphon) and use ground (XLR: 1) for shielding only. Connect your power amp between the hot (XLR: 2) and cold (XLR: 3) output. That's not possible with the (poor) 'french' design you have.
Good luck,
Frank
The french design implements the filter symmetrically on both +ve and -ve output. The filtered -ve output is left floating but could be easily connected to pin 3 of the XLR output plug. I have left it floating since this gives me 2.4 rather than 4.8 V on the output.
Am I right in assuming that given the input impedance of my pream is >20K I can safely assume that the AKM dac sees roughly the same output load (~1K) at its +v and -ve output?
And the DCX implements the mute on the AKM chip. The mute is still there. In fact, there is a menu option for it to be engaged at switch on.
Best
Giulio
Am I right in assuming that given the input impedance of my pream is >20K I can safely assume that the AKM dac sees roughly the same output load (~1K) at its +v and -ve output?
And the DCX implements the mute on the AKM chip. The mute is still there. In fact, there is a menu option for it to be engaged at switch on.
Best
Giulio
Hello,
To help you tuning this board, i decided today to publish schematics of my board.
I hope it will be usefull for DIYers
Enjoy :
http://forum.audiophonics.fr/viewtopic.php?t=12
I hope you will share your modifications with all user to arrive to a good finished board.
Kind regards.
FERHAT Nicolas.
Audiophonics.fr
To help you tuning this board, i decided today to publish schematics of my board.
I hope it will be usefull for DIYers
Enjoy :
http://forum.audiophonics.fr/viewtopic.php?t=12
I hope you will share your modifications with all user to arrive to a good finished board.
Kind regards.
FERHAT Nicolas.
Audiophonics.fr
Hi,
The audiophonics HF filter (1k5, 2.2nF) has an attenuation of about 0.4 dB and 17° phase shift at 20 kHz. The 6.8uF AC coupling cap causes a 0.1dB attenuation and 7° phase shift at 20Hz.
With 1k resistors you would get 0.2dB attenuation and 12° phase shift at 20 kHz.
To get the outputs symmetrically just put in 6 additional AC coupling caps between XLR pin 1 and the filter (left side). See post 858.
Reasonable values for the AC coupling caps are 10uF for the low (20Hz), 2.2uF for the mid (250Hz) and 0.47uF for the high (3000Hz) frequency channel. Higher values are no problem.
For cabling see post 870.
I would use the digital input of the DCX, since the modded analog inputs have the same problem as the outputs. Unbalanced opamps connected with a balanced ADC. For mods of the analog inputs see post 830 (for 6dB higher gain compared with standard DCX use R6 and R8= 1k, R7= 16k2).
Good luck,
Frank
The audiophonics HF filter (1k5, 2.2nF) has an attenuation of about 0.4 dB and 17° phase shift at 20 kHz. The 6.8uF AC coupling cap causes a 0.1dB attenuation and 7° phase shift at 20Hz.
With 1k resistors you would get 0.2dB attenuation and 12° phase shift at 20 kHz.
To get the outputs symmetrically just put in 6 additional AC coupling caps between XLR pin 1 and the filter (left side). See post 858.
Reasonable values for the AC coupling caps are 10uF for the low (20Hz), 2.2uF for the mid (250Hz) and 0.47uF for the high (3000Hz) frequency channel. Higher values are no problem.
For cabling see post 870.
I would use the digital input of the DCX, since the modded analog inputs have the same problem as the outputs. Unbalanced opamps connected with a balanced ADC. For mods of the analog inputs see post 830 (for 6dB higher gain compared with standard DCX use R6 and R8= 1k, R7= 16k2).
Good luck,
Frank
Giulio,
Perhaps muffled is not the right word. The problem is not in the highs. It's that during complex passages, all the sounds coming out of the mid driver somehow seem less clear and mashed together a bit. It looses the clarity and beauty that the sound has on simpler passages.
I have played with crossover settings to no avail. Once again, my feeling is that this problem can be isolated just by listening to the mid driver, so crossover should generally not be the issue anyway. Perhaps the B&K 200.7 amp or the GR60 speakers are the weak link. I'm just confused becuase the DCX mod did not appear to improve things at all, and perhaps made it a little worse...
Oettle,
If those caps are supposed to be good then perhaps it is a futile pursuit for me to try others... :-( I'm not an EE, but wouldn't lower cap values create less phase shift, not more? I guess if I use different caps for the lows, mids, and highs, I would have to play with the phase shift capabilities of the DCX since the drivers would be oit of phase at the crossover points.
Nicko,
Thanks for posting the schematics.
Philip
Perhaps muffled is not the right word. The problem is not in the highs. It's that during complex passages, all the sounds coming out of the mid driver somehow seem less clear and mashed together a bit. It looses the clarity and beauty that the sound has on simpler passages.
I have played with crossover settings to no avail. Once again, my feeling is that this problem can be isolated just by listening to the mid driver, so crossover should generally not be the issue anyway. Perhaps the B&K 200.7 amp or the GR60 speakers are the weak link. I'm just confused becuase the DCX mod did not appear to improve things at all, and perhaps made it a little worse...
Oettle,
If those caps are supposed to be good then perhaps it is a futile pursuit for me to try others... :-( I'm not an EE, but wouldn't lower cap values create less phase shift, not more? I guess if I use different caps for the lows, mids, and highs, I would have to play with the phase shift capabilities of the DCX since the drivers would be oit of phase at the crossover points.
Nicko,
Thanks for posting the schematics.
Philip
Hi Frank,
let me just try to get this straight.
I need to make even the LP filter symmetrical. But can I then tap into just half of the signal (pin 2 + ground). There is no way I can fit 12 MKP caps in those values inside the Behringer. My idea would be to use electrolytics on the -ve and leave it floating.
I understand this must sound rather silly, but I have never been able to figure out how to work out the Thevenin equivalent as soon as there is a series capacitor.
TIA
Giulio
let me just try to get this straight.
I need to make even the LP filter symmetrical. But can I then tap into just half of the signal (pin 2 + ground). There is no way I can fit 12 MKP caps in those values inside the Behringer. My idea would be to use electrolytics on the -ve and leave it floating.
I understand this must sound rather silly, but I have never been able to figure out how to work out the Thevenin equivalent as soon as there is a series capacitor.
TIA
Giulio
Hi Giulio,
Forget ground at all. It's good for shielding only. If you want half output levels just subtract 6dB on the DSP side (ABC input or output 1-6).
Don't use electrolytic caps due to sound quality, but you can use smaller values for mid and high frequency channels. The diameter of your 6.8uF value should be 23 mm. The 0.47 to 1.5 uF values have diameters between 13 and 14 mm.
Have a look at your own post 858. That's exactly where you should end up.
XLR pin 1 is the shielding for the cable, XLR pin 2 is the hot output and XLR pin 3 the cold output which might be connected to ground of your power amp, if it has an unbalanced input only.
Good luck,
Frank
Forget ground at all. It's good for shielding only. If you want half output levels just subtract 6dB on the DSP side (ABC input or output 1-6).
Don't use electrolytic caps due to sound quality, but you can use smaller values for mid and high frequency channels. The diameter of your 6.8uF value should be 23 mm. The 0.47 to 1.5 uF values have diameters between 13 and 14 mm.
Have a look at your own post 858. That's exactly where you should end up.
XLR pin 1 is the shielding for the cable, XLR pin 2 is the hot output and XLR pin 3 the cold output which might be connected to ground of your power amp, if it has an unbalanced input only.
Good luck,
Frank
Hi,
if you want balanced output from DCX you need all three pins connected.
If you want unbalanced output from DCX, then use pins 1 & 2.
Do not leave the output that should have gone to pin 3 floating.
Fit a resistor, 10k will do, from the output to ground.
maintaining balanced source
The pin 3 should be loaded with exactly the same source resistance as seen at pin2, so that if you ever connect a balanced receiver to this unbalanced output the receiver gets loaded properly at it's balanced input. You need to put a passive network that mimics the pin2 connection. A resistor = pin 2 will sort of do, but better is an exact match. Then you are back to using full balanced output and the extra cap you want to avoid.
if you want balanced output from DCX you need all three pins connected.
If you want unbalanced output from DCX, then use pins 1 & 2.
Do not leave the output that should have gone to pin 3 floating.
Fit a resistor, 10k will do, from the output to ground.
maintaining balanced source
The pin 3 should be loaded with exactly the same source resistance as seen at pin2, so that if you ever connect a balanced receiver to this unbalanced output the receiver gets loaded properly at it's balanced input. You need to put a passive network that mimics the pin2 connection. A resistor = pin 2 will sort of do, but better is an exact match. Then you are back to using full balanced output and the extra cap you want to avoid.