Ulli:
I am totally aware of how the DCX2496 works.
It was the volume control by squarewave I was wondering of.
But as you say it must be the same whatever digital volume you use before the DCX.
I am also temped to use Jans active, just to get the remote volume control, but with my Lundahls both in and out in place of the OP-amps.
But for me the ultimate solution is to use Silk TVCs on the output. I have a custom made remote-controlled Shallco 24-pole switch with 4 decks. Maybe one could add two more decks
.
I am totally aware of how the DCX2496 works.
It was the volume control by squarewave I was wondering of.
But as you say it must be the same whatever digital volume you use before the DCX.
I am also temped to use Jans active, just to get the remote volume control, but with my Lundahls both in and out in place of the OP-amps.
But for me the ultimate solution is to use Silk TVCs on the output. I have a custom made remote-controlled Shallco 24-pole switch with 4 decks. Maybe one could add two more decks
.Hello again, thanks for the single ended output information on the Selectronic piece. I've been leaning towards the pieces from Pilgham's site simply out of familiarity, I've seen the schematics in an audioxpress magazine article. I haven't seen exactly how the Selectronic's I/O board works. Also the information on the active I/O kit from you, Jan is helpful. I like the way your passive I/O board uses the set of relays for mute and that it removes many op-amps from the signal path. Lars, I admire your use of transformers on the outputs, but I'm afraid it would be too big of a leap for me. I need to try an intermediate step before I get to that point. (hope that makes sense) :/
Ulli, As I understand I only need one Vreg that I would use to regulate the voltage to the DACs? And you have two because you use one on the ADCs? That's the best I can figure at the moment. Before I order, my question is since I'll most likely never use analog input, Do I get by with just one Vreg kit? Thanks.
About me: I am most likely not as experienced with the highest of fidelity sound (in terms of owning the equipment to produce such sound and the experience of listening to said equipment for long periods of time) as many of you here are. I am probably more tolerant of sound that is less than high end "audiophile" quality, as most of the people here but I'd say I do fall into the audiophile category as a person simply because I concern myself with musical playback accuracy to a greater extent than does the general population. I do appreciate accuracy and clarity but the little kid in me still likes to play it really loud every now and again. And on a final note, I'm no engineer by any stretch. I'm just a lowly tech. I have a two year degree in electronics from about 20 years ago.
The floating point hardware volume control, kmixer.sys, and kernel streaming isn't as glorified as it sounds. It simply means that I bypassed windows low resolution method for mixing system sounds and media sounds together. When the music is playing on my setup nothing else in the system can access the sound-card. In fact the volume control sliders in Windows Volume Control panel have no effect. Nor does the volume slider control on the media player. I must use the M-Audio software sliders to control the hardware (sound-card) that adjusts the volume of the sound. It adds a little complexity to how I listen because I have to use two applications. A volume control application to control sound level and the media player application to select what song I want to listen to. Kernel streaming is a plug-in for the media player. It sends the audio stream directly to the sound-card for the most direct path possible. The volume is done in the digital domain at 36 bit resolution via hardware DSP on the sound-card. There are no DA to AD conversions along the signal path. There is only one analog conversion and you know where that is. The DCX DACs.
Do I lose resolution at attenuated levels? Most likely, but there's a small chance that I don't. If I listen to 16/44, there probably isn't because the DSP uses a greater bit depth and so does the DCX; thereby carrying the higher resolution through to the end. (this might not be true in reality, I'll explain later) If I play 24/96 source material then I would say I do lose resolution at attenuated levels. But I can't hear it. You guys might but I can't. Most of my recordings are of CD quality. I might have one or two samples of HDCD or a clip from DVD-A but most of my source material is recorded 16/44. I set this up about 4 years ago and been listening to it without any further modification since. Now upon discovery of the mods that are available for the DCX, the curiosity has gotten the best of me. I want to venture into the next level of the sound listening experience.
Lars, and anyone else interested. I have a theory that I do lose resolution by using digital volume control but it could be possible to implement a non degrading method. If the source audio (for example 16 bits) is converted and then attenuated (digitally) at a bit resolution higher (36 bits for example) than that of the original and maintains that resolution to the end (use a 36 bit DAC), I do not believe it would lose resolution to a point to where it got worse than the source because the divisions in the scale have become finer. (36 bits vs 16) However, this is not what happens in the real world device. I would guess even though I have hardware that resolves at 36 bit detail it probably reverts back to 16 bits for transmission to the DCX. In this case I would conclude the system causes a degraded sound. If the sound-card were to revert only to 24/96 and the DCX maintain it then I would guess I am not losing much since the output resolution would remain higher than most of my source material. (16/44) I can conclude that a digital system with an output signal that is the same resolution as its input signal and where attenuation is applied digitally, you end up with degradation based on this simple truth. The "step size" or "divisions of the scale" of any particular digitally quantified system is fixed. The smaller the signal the fewer steps there are to represent that signal. By step size I mean the smallest difference possible in amplitude change at the output of a DAC when the digital input is changed by the smallest possible digital value in decimal notation: one) I'm not at all an expert and there may be sides to it that are unbeknownst to me or that I'm overlooking, thereby not enabling me to reach a full logical conclusion. But based on what I think I know... that's it. And like I said I've listened to it this way for years and have been quite content. I cannot hear any degradation. It sounds clear at low volume as well as high. But then again, maybe, my ears could be made of tin rather than gold. Or that my speakers, amplifiers, and wiring isn't the greatest in the world. There is a possibility that a floating point process can change the amplitude in a non degrading manner but I don't know. To me if the output is converted back to the same resolution as the input, regardless of how the digital attenuation is achieved, there will be some form of degradation, but I simply don't really know. Discussions can be found by searching the net for "floating point vs integer sound quality". I would consider RaneNote Fixed-point vs. Floating-point dsp reliable.
I'll post back on my progress with the kits and try not to write a book next time. Cheers.
Ulli, As I understand I only need one Vreg that I would use to regulate the voltage to the DACs? And you have two because you use one on the ADCs? That's the best I can figure at the moment. Before I order, my question is since I'll most likely never use analog input, Do I get by with just one Vreg kit? Thanks.
About me: I am most likely not as experienced with the highest of fidelity sound (in terms of owning the equipment to produce such sound and the experience of listening to said equipment for long periods of time) as many of you here are. I am probably more tolerant of sound that is less than high end "audiophile" quality, as most of the people here but I'd say I do fall into the audiophile category as a person simply because I concern myself with musical playback accuracy to a greater extent than does the general population. I do appreciate accuracy and clarity but the little kid in me still likes to play it really loud every now and again. And on a final note, I'm no engineer by any stretch. I'm just a lowly tech. I have a two year degree in electronics from about 20 years ago.
The floating point hardware volume control, kmixer.sys, and kernel streaming isn't as glorified as it sounds. It simply means that I bypassed windows low resolution method for mixing system sounds and media sounds together. When the music is playing on my setup nothing else in the system can access the sound-card. In fact the volume control sliders in Windows Volume Control panel have no effect. Nor does the volume slider control on the media player. I must use the M-Audio software sliders to control the hardware (sound-card) that adjusts the volume of the sound. It adds a little complexity to how I listen because I have to use two applications. A volume control application to control sound level and the media player application to select what song I want to listen to. Kernel streaming is a plug-in for the media player. It sends the audio stream directly to the sound-card for the most direct path possible. The volume is done in the digital domain at 36 bit resolution via hardware DSP on the sound-card. There are no DA to AD conversions along the signal path. There is only one analog conversion and you know where that is. The DCX DACs.
Do I lose resolution at attenuated levels? Most likely, but there's a small chance that I don't. If I listen to 16/44, there probably isn't because the DSP uses a greater bit depth and so does the DCX; thereby carrying the higher resolution through to the end. (this might not be true in reality, I'll explain later) If I play 24/96 source material then I would say I do lose resolution at attenuated levels. But I can't hear it. You guys might but I can't. Most of my recordings are of CD quality. I might have one or two samples of HDCD or a clip from DVD-A but most of my source material is recorded 16/44. I set this up about 4 years ago and been listening to it without any further modification since. Now upon discovery of the mods that are available for the DCX, the curiosity has gotten the best of me. I want to venture into the next level of the sound listening experience.
Lars, and anyone else interested. I have a theory that I do lose resolution by using digital volume control but it could be possible to implement a non degrading method. If the source audio (for example 16 bits) is converted and then attenuated (digitally) at a bit resolution higher (36 bits for example) than that of the original and maintains that resolution to the end (use a 36 bit DAC), I do not believe it would lose resolution to a point to where it got worse than the source because the divisions in the scale have become finer. (36 bits vs 16) However, this is not what happens in the real world device. I would guess even though I have hardware that resolves at 36 bit detail it probably reverts back to 16 bits for transmission to the DCX. In this case I would conclude the system causes a degraded sound. If the sound-card were to revert only to 24/96 and the DCX maintain it then I would guess I am not losing much since the output resolution would remain higher than most of my source material. (16/44) I can conclude that a digital system with an output signal that is the same resolution as its input signal and where attenuation is applied digitally, you end up with degradation based on this simple truth. The "step size" or "divisions of the scale" of any particular digitally quantified system is fixed. The smaller the signal the fewer steps there are to represent that signal. By step size I mean the smallest difference possible in amplitude change at the output of a DAC when the digital input is changed by the smallest possible digital value in decimal notation: one) I'm not at all an expert and there may be sides to it that are unbeknownst to me or that I'm overlooking, thereby not enabling me to reach a full logical conclusion. But based on what I think I know... that's it. And like I said I've listened to it this way for years and have been quite content. I cannot hear any degradation. It sounds clear at low volume as well as high. But then again, maybe, my ears could be made of tin rather than gold. Or that my speakers, amplifiers, and wiring isn't the greatest in the world.
There is a possibility that a floating point process can change the amplitude in a non degrading manner but I don't know. To me if the output is converted back to the same resolution as the input, regardless of how the digital attenuation is achieved, there will be some form of degradation, but I simply don't really know. Discussions can be found by searching the net for "floating point vs integer sound quality". I would consider RaneNote Fixed-point vs. Floating-point dsp reliable.I'll post back on my progress with the kits and try not to write a book next time. Cheers.
squarewaves said:
Squarewaves,
This is a very clear explanation, and I can find no fault in your reasoning. I agree.
For the sake of discussion, lets say you go to 20 bits. That means you would end up with, say, the values:
1010101010101010 (16 bit) -> 10101010101010100000 (20 bit)
1010101010101011-> 10101010101010110000
Now you do the volume control, you go to:
10101010101010100000-> 10101010101010100100
10101010101010110000-> 10101010101010110100
Now you can convert this to analog in a 20 bit DAC.
Have you lost resolution? I don't think so. Is this how it would work?
Jan Didden
Intermittent Fault
I have been using the DCX 2496 over the last few months, it has been playing faultlessly.
Only yesterday and also today, after playing for a 1-2 hrs it starts to play up. It stops playing, no display signal/led display on the front panel. After I power off and on again, the units returns to normal operation.
I have not opened up the unit yet, I suspect the cable connections could be the main suspect. Rejiggling the cable connections to the main board may solve the problem. If it does not solve it there could be a "near" dry joint somewhere and it becomes apparent after it the unit heats up.
Anyone experiences similar problem? Any suggestions to fix this problem?
Thanks in advance.
I have been using the DCX 2496 over the last few months, it has been playing faultlessly.
Only yesterday and also today, after playing for a 1-2 hrs it starts to play up. It stops playing, no display signal/led display on the front panel. After I power off and on again, the units returns to normal operation.
I have not opened up the unit yet, I suspect the cable connections could be the main suspect. Rejiggling the cable connections to the main board may solve the problem. If it does not solve it there could be a "near" dry joint somewhere and it becomes apparent after it the unit heats up.
Anyone experiences similar problem? Any suggestions to fix this problem?
Thanks in advance.
I still listen to FM radio. That's why I still use analog input.
Yes, one is for the DACs and the other one is for the ADC.
From the circuit diagram I read:
IC7 feeds +5VA-1 into both ADC chips (IC2 5393 / IC3 5983)
IC8 feeds +5VA-2 into three DAC chips (IC4-6, all 4393)
The picture on page 1 of http://www.linearaudio.nl/Documents/DCX2496 LinSup CG.pdf shows just one Vreg - the one for the DACs.
Ulli
janneman said:
Jan,
I agree in general. But, is that the way conversion is done?
Isn't it, that the analog signal (let's say 1V max) is represented by bits (16 or 20 ones, respectively). The analog range is the same in both cases (here: 1V).
I added values in decimal (which I'm more used to):
1010101010101010 43690
1111111111111111 65535
=> signal is at 2/3 of full range
2/3 of 11111111111111111111 1048575
would be 1048575 * 2/3 = 699050
10101010101010101010 699050
10101010101010100000 699040
This a delta of 0,0014 %.
Of course, it fully depends on how conversion is done.
Ulli
AR2 said:
Hi AR2,
Could you please elaborate further your 6 channel preamp configuration; I'm assuming that you have the input section configured using balanced to unbalanced lundahl. What's after the lundahl? do you have an Alps 6 channel pot, and what's after the attenuator, an active output stage or is it just passively coupled directly to your power amp?
cheers.
Ulli,
If I understand your example correctly, you effectively multiply each sample by (2^20-1)/(2^16-1) (for 20 bits), right.
In my example, adding a bunch of zeroes at the end, is effectively multiplying each sample by 2^4.
I don't see that this is fundamentally different, except for some rounding of the '1' factor?
Jan Didden
If I understand your example correctly, you effectively multiply each sample by (2^20-1)/(2^16-1) (for 20 bits), right.
In my example, adding a bunch of zeroes at the end, is effectively multiplying each sample by 2^4.
I don't see that this is fundamentally different, except for some rounding of the '1' factor?
Jan Didden
Hello Will,
I wrote several time here about my set up, and feel little bit uncomfortable being redundant. I guess when the tread becomes as big as this one than we have to do it again for the newer readers. My apologies to people that are reading this tread from the beginning.
Here is the tread on volume control/preamps:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=76095
Here is the tread about my speakers in order to connect the dots:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=21523
I made some changes in the meantime regarding the amps. On the bottom I have my new DIY Pass Lab X600:
http://home.comcast.net/~burningamp/index.html
and on the top I have Aleph 30 that is closer in output to Aleph 5 (higher rail voltage)
So with all this different amps and speakers I had to have different gains in preamp stage. My whole original output section from DCX is gone. Lundahls are directly connected to AKM chips, balanced. Lundahls are balanced and than signal goes to 6 ch. relay volume control - balanced. After that I have XBosoz balanced preamps for bass channels and for high channels. Mid - tube amp do not need any preamp, so there is a direct connection between my 2A3 amp and volume control. For that I use switch to convert from balanced to single ended.
When I use DCX just as two channel test set up on different speakers I use direct output from Lundahls and volume control. That sounds better than any preamp when added, and that is what I prefer as long as your amp/speaker combination could work with that - if you have enough gain.
As you could see Lundahls are outside of DCX box. They could fit inside obviously without the original in/output board that I am not using anyway. My plan is to make PC board that will hold Lundahls in the DCX box and that will pass digital signal to the processor board. If anyone is interested we should combine the efforts.
I hope this answers your question.
I wrote several time here about my set up, and feel little bit uncomfortable being redundant. I guess when the tread becomes as big as this one than we have to do it again for the newer readers. My apologies to people that are reading this tread from the beginning.
Here is the tread on volume control/preamps:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=76095
Here is the tread about my speakers in order to connect the dots:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=21523
I made some changes in the meantime regarding the amps. On the bottom I have my new DIY Pass Lab X600:
http://home.comcast.net/~burningamp/index.html
and on the top I have Aleph 30 that is closer in output to Aleph 5 (higher rail voltage)
So with all this different amps and speakers I had to have different gains in preamp stage. My whole original output section from DCX is gone. Lundahls are directly connected to AKM chips, balanced. Lundahls are balanced and than signal goes to 6 ch. relay volume control - balanced. After that I have XBosoz balanced preamps for bass channels and for high channels. Mid - tube amp do not need any preamp, so there is a direct connection between my 2A3 amp and volume control. For that I use switch to convert from balanced to single ended.
When I use DCX just as two channel test set up on different speakers I use direct output from Lundahls and volume control. That sounds better than any preamp when added, and that is what I prefer as long as your amp/speaker combination could work with that - if you have enough gain.
As you could see Lundahls are outside of DCX box. They could fit inside obviously without the original in/output board that I am not using anyway. My plan is to make PC board that will hold Lundahls in the DCX box and that will pass digital signal to the processor board. If anyone is interested we should combine the efforts.
I hope this answers your question.
AR2 - I would leave the trasformers out of the actual DCX box or extend the box to suit the transformers. I did experiments with Lundahl LL7902 with DCX placing them in the case .... checking the output with spectrum analysis showed a considerable amount on noise picked up compared to passive / active solutions. So even though the Lundahls seem well shielded there is still too much RF in DCX case and transformers pick it up.
I liked the sound of the transformers, but the inner geek in me was not happy so I kept experimenting.
Todays status is that I have ended up with Jans active mod also. More on that at some later time.
Ergo
I liked the sound of the transformers, but the inner geek in me was not happy so I kept experimenting.
Todays status is that I have ended up with Jans active mod also. More on that at some later time.
Ergo
You are right for using two samples or strings of bits (words) with a difference between them. And I think the system would simply shift the 16 bit word into the most significant bits of the 20 bit register thereby leaving the four least significant bits at zero. I did the math the hard way to make sure it returned the exact same results as simply appending four zero bits. As for adding the "100" to each value that would not change the amplitude of the single cycle waveform that your example used here would represent. It would change only its dc offset. I also might mention that if these two samples are in succession then the frequency of your example is the highest that the system can produce. Volume or amplitude is the result of the difference between two samples more so than their absolute values as a percentage of full scale. However the "100" that you added to both words could represent a small slice of a lower frequency being represented. Take, for example, the complexity of music. High frequencies ride on the waves of low frequencies. Each frequency has its own amplitude noted by the numeric values of the peaks and troughs. So in your example I took the 16bit words (one at a time), converted to decimal, and divided by the maximum number of possibilities for 16bits (65536) to get a fraction. Then I took that fraction and multiplied it by the maximum number of possibilities for a 20bit system (1048576). Converted that value back to binary to get the 20bit equivalent of the 16bit word. Please take note I use maximum possibilities in decimal for full scale not the maximum value of full scale. That's what accounts for the difference of one. It goes like this:
1010101010101010 (43690) / 1111111111111111 (65536) = 0.666656494140625.
Then I take 11111111111111111111 (1048576) * 0.666656494140625 = 699040.
699040 in binary is 10101010101010100000.
Apply the same process to your second 16bit word to get its 20 bit equivalent:
1010101010101011 (43691) / 1111111111111111 (65536) = 0.6666717529296875.
Then I take 11111111111111111111 (1048576) * 0.6666717529296875 = 699056.
699056 in binary is 10101010101010110000.
Your two 16 bit samples become the two following 20 bit samples.
1010101010101010 (43690) -> 10101010101010100000 (699040)
1010101010101011 (43691) -> 10101010101010110000 (699056)
Okay that's the conversion but the volume hasn't changed yet. As described earlier adding a fixed value will only create an offset. You have to multiply both by a fraction to attenuate the volume. Let's use 0.7071. The two 20 bit samples in decimal are 699040 and 699056. Multiply each by 0.7071 and round off the fractional remainder (induce tiny error) to get 494291 and 494302. Convert back to binary. After volume change is applied...
10101010101010100000 (699040) -> 1111000101011010011 (494291)
10101010101010110000 (699056) -> 1111000101011011110 (494302)
As long as we maintain 20bit resolution by using 20bit DACs then it is safe to say that no resolution is lost.
But another thing I'd like to add because it might be relevant. I guess in a 16 bit system (65536 decimal) the half way point between peak and trough of a maximum amplitude signal is 32768 which represents the zero crossing point of said signal. However at the output of the DAC there's a dc offset. 0 would be 0.0Vout. 32767 would be almost 0.5Vout. 32768 would be just over 0.5Vout. And 65535 would be 1.0Vout. (I just realized there's no true halfway point) A capacitor filters out the DC which returns the signal to it's original bias which is for it to swing plus and minus of zero. The reason I mention this is because I'm not sure if the signal is represented as I just said. Because if numbers plus or minus a few digits of 32768 represent quiet, then during a recording session of either a live performance or a transfer of analog tapes into a digital recorder would have to do the same. However it could also be the case that both recording and playback use a numeric system of -32767 to +32767 instead of 0 to 65535. I don't know. Not looking for answers either, just thinking out loud because it might lead somewhere profitable.
Thank you.
I like how you think, it helped me do the calculations earlier that I wouldn't have thought of doing the way I did had you not mentioned this. I'd say to carry out the precision a little further because 43690 isn't 2/3rds of 65536. Using greater accuracy comes up with 699040, exactly the same as compared to Jan's result of 699040 by simply adding four zero bits. Yes I was surprised by the result myself because I was thinking like you, it had to be a percentage of full range. Here I did all that extra math and all I really had to do was add four zeros. But it was nice to verify it.
Anyway... I felt like giving a little of what I know in exchange for everyone's help in selecting the mod kits for my DCX. I ordered the passive I/O kit, the SCR/Clock kit, and Vreg kit from Ward at Pilgham Audio. I can't wait to hear the differences. I might do each kit one at a time and listen in stages for improvements.
Cheers.
ScottGardner said:
Hey... I don't think I like what you're implying. In fact I resemble that remark.
If you let me know what offended you I will be happy to explain. I guess something gets the different feel when is written compared to when is spoken in person. I am assuming you didn't like my sense of humor, so I apologize for that. To make it clear, here is what was funny to me:
It is only here in States that one week off work is considered a vacation. Normally in various places in Europe, July or August or both months are considered dead months since everyone is on vacation. When I say everyone that means majority of business are closed, retail stores are closed with blinds lowered, and streets are visibly vacated. August is certainly not time to do your shoes shopping in Italy. During those summer months there is a big, massive move to various islands and costs in Italy, Spain, France, Greece or ex Yugoslavia's Adriatic. That is the time when people are tanned, relaxed and happy, and are having real fun. Not to mention topless girls en masse on those beaches!
Here in the States there is none of that. I mean none. July here is just like March or November. Business as usual. Like I said people consider " vacation" when they take 3-4 days off work. Do not take me here wrong, I am not saying that people here are not having fun, I am just saying that there is none of that feel when everyone around you is leaving to go somewhere. Because of that people here are not use to the fact that stores are closed, and that things aren't done during that time in majority of Europe.
It is always funny to me to see that clash of cultures and habits. I have numerous friends that came back from trip to Europe and they just couldn't believe that something like that is possible. Some complain, some are jealous, but it is always interesting and funny to me to see the reaction. I hope this gives you idea what was on my mind.