If you use a balanced input, the 2.5V doesn't matter. The differential stage will subtract the two 2.5v parts from each side and will come out to zero. No need to waste money on caps here.
The DCX as-is can drive 600 ohms, much lower than even the worst amplifier input, so a buffer is also a waste of money and quality. Even with xformers at the output, this still is a non issue, the AKM's can drive 2k so through a 1:1 xformer this is still 2k. ONLY if you would use a passive output (filter), then depending on the impedance values, you *might* have some effect from amp input load. But smart value selection can also avoid that if it would be an issue. There's more about this on my web site if you would be interested.
jd
so simple?
Thank you Michael and Jan!
So, is it really so simple? All that I need is six transformers and the six channels potentiometer? Well, guys, this seems too good to be true!
I like the transformers way, but now please be patient with me: in this thread I have seen so many trafos suggested that i can't get out. Sowther, Jensen, Edcor, Lundhal.... help! Actually the test that Michael reported was very useful, but... which model is best suited for this application
?
By the way, digging in the past posts of this thread, I have seen that Scott Gradner already suggested which is the better way to connect the trafos and potentiometer....
Thank you all for your help!
Ciao
Paolo
Thank you Michael and Jan!
So, is it really so simple? All that I need is six transformers and the six channels potentiometer? Well, guys, this seems too good to be true!
I like the transformers way, but now please be patient with me: in this thread I have seen so many trafos suggested that i can't get out. Sowther, Jensen, Edcor, Lundhal.... help! Actually the test that Michael reported was very useful, but... which model is best suited for this application
?By the way, digging in the past posts of this thread, I have seen that Scott Gradner already suggested which is the better way to connect the trafos and potentiometer....
Thank you all for your help!
Ciao
Paolo
Other options
Six channel tube buffer? Now you are really putting some time and money into it. If you want an active mod the Jan Didden board is very convenient. One of the best sounding mods is also one of the cheapest. Just send the dac outputs direct out through a high quality cap such as the Dayton foil and use passive volume control right at the amps, inside the amps even, with minimal wire following the passive stage to the first active component in the amp for the best sound.
Yes, ditto what Jan and Scott say. You don't really need the buffer stage, as long as your cables lengths out of the 6-way pot are reasonable. I suggest a low value pot, like 20K or 10K. If you could get a 5K pot, that might be even better.
As Scott mentions, the cap mod is the easiest and sounds good. I prefer the transformer sound but it just a matter of taste, really.
It is that simple.
As Scott mentions, the cap mod is the easiest and sounds good. I prefer the transformer sound but it just a matter of taste, really.
It is that simple.
DCX distortion reduces at lower levels
I've spent some time measuring the DCX output in single ended mode (using a simple XLR to RCA adapter). The output isn't too bad at 2 volts and less, but it gets significantly worse as you go toward the "0 dB" level of 9.75 volts. This suggests to me that using an attenuator on the stock output isn't such a good idea, unless you need output much less than 2 volts. I also wonder what gives the DCX so much distortion. I think the opamp used is capable of much better performance. One way to think about the stock DCX is that it has fairly nice 2V output with 13db extra headroom which you hope not to need because it gets more distorted (though still much less than most speakers). Interestingly, THD+N is not specified by Behringer for levels greater than 2.4 volts output (and those specs are much better than what I measured, possibly at much higher impedance).
For test purposes, I have been driving DCX with coax spdif digital from a Juli@ soundcard driven at 44.1kHz with 24 bit signal. So this tests only the DCX output, not the analog inputs (which I don't use). I used the program RightMark Pro. To attenuate the signal (which would normally overload the Julia's input above 2V) and still maintain 10K impedance, I used an Aragon 28K preamp. Residual of Juli@ is about 0.0004% distortion, and the Aragon is also very low in harmonic distortion (it didn't seem to contribute any) but it can add some noise, depending on attenuation. I tried to measure at 0dB, 6dB, 12dB and 15dB, using DCX input attenuation adjustment. However, Rightmark automatically measures THD+N at -3dB, plus you are supposed to set it to start at -1dB in the first place, so I believe the actual attenuations are -4dB, -10dB, -16dB, and -19dB, relative to max 22dbu (9.75V) output.
The THD+N measurements for those attenuations are 0.013%, 0.011%, 0.0069%, and 0.0048%. I believe these are fairly accurate, and they are essentially all harmonic distortion. My S/N measurements are limited by my approximately 100dB residual, but were 104.7, 103.4, 100.8, and 98.6, all still pretty good IMO. BTW, I have measured S/N as high as 107.1dB by removing the Aragon but using fixed 6dB attenuator and that was with 6dB digital attenuation, so I believe the Behringer 112dB S/N specification can be met. Basically I think distortion rather than noise is the worst offender here, so throwing a little dynamic range away for lower distortion is the better approach to using this. Hence my recommendation not to use outboard attenuator. Also, the problem here is almost certainly distortion caused by analog output circuitry, not by the DAC. And there is dynamic range to burn in the DSP with at least 24 bits (or 144dB) available.
Now this is all at 1Khz, but the DCX also has a serious rise in THD at higher frequencies. Rightmark does a sweeps at "-6dB" and "-15dB". In the worse case, "-6dB" (which would actually be -7db because of initial -1dB) I recorded a maximum of 0.058% distortion at 6.3kHz. That's the top frequency you can measure the 3rd harmonic for with 44.1khz sampling, and I can't yet measure with a different sampling rate than I am sending to DCX. Extrapolating, however, it is easy to imagine that distortion of 0.1% distortion is exceeded either at higher levels or higher frequencies.
At low enough levels, the THD+N doesn't rise at higher frequencies, thus being dominated by noise rather than harmonic distortion. But that only seems to occur at levels below -30dB (about 0.3 volts).
Now what does the harmonic distortion look like? Third harmonic is substantially larger than 2nd, and additional odd harmonics fall off "exponentially" at 45 degree angle on dB scale, with even harmonics being much smaller. Not the worst possible (such as flat or increasing odd orders) but not sonically great either.
I've spent some time measuring the DCX output in single ended mode (using a simple XLR to RCA adapter). The output isn't too bad at 2 volts and less, but it gets significantly worse as you go toward the "0 dB" level of 9.75 volts. This suggests to me that using an attenuator on the stock output isn't such a good idea, unless you need output much less than 2 volts. I also wonder what gives the DCX so much distortion. I think the opamp used is capable of much better performance. One way to think about the stock DCX is that it has fairly nice 2V output with 13db extra headroom which you hope not to need because it gets more distorted (though still much less than most speakers). Interestingly, THD+N is not specified by Behringer for levels greater than 2.4 volts output (and those specs are much better than what I measured, possibly at much higher impedance).
For test purposes, I have been driving DCX with coax spdif digital from a Juli@ soundcard driven at 44.1kHz with 24 bit signal. So this tests only the DCX output, not the analog inputs (which I don't use). I used the program RightMark Pro. To attenuate the signal (which would normally overload the Julia's input above 2V) and still maintain 10K impedance, I used an Aragon 28K preamp. Residual of Juli@ is about 0.0004% distortion, and the Aragon is also very low in harmonic distortion (it didn't seem to contribute any) but it can add some noise, depending on attenuation. I tried to measure at 0dB, 6dB, 12dB and 15dB, using DCX input attenuation adjustment. However, Rightmark automatically measures THD+N at -3dB, plus you are supposed to set it to start at -1dB in the first place, so I believe the actual attenuations are -4dB, -10dB, -16dB, and -19dB, relative to max 22dbu (9.75V) output.
The THD+N measurements for those attenuations are 0.013%, 0.011%, 0.0069%, and 0.0048%. I believe these are fairly accurate, and they are essentially all harmonic distortion. My S/N measurements are limited by my approximately 100dB residual, but were 104.7, 103.4, 100.8, and 98.6, all still pretty good IMO. BTW, I have measured S/N as high as 107.1dB by removing the Aragon but using fixed 6dB attenuator and that was with 6dB digital attenuation, so I believe the Behringer 112dB S/N specification can be met. Basically I think distortion rather than noise is the worst offender here, so throwing a little dynamic range away for lower distortion is the better approach to using this. Hence my recommendation not to use outboard attenuator. Also, the problem here is almost certainly distortion caused by analog output circuitry, not by the DAC. And there is dynamic range to burn in the DSP with at least 24 bits (or 144dB) available.
Now this is all at 1Khz, but the DCX also has a serious rise in THD at higher frequencies. Rightmark does a sweeps at "-6dB" and "-15dB". In the worse case, "-6dB" (which would actually be -7db because of initial -1dB) I recorded a maximum of 0.058% distortion at 6.3kHz. That's the top frequency you can measure the 3rd harmonic for with 44.1khz sampling, and I can't yet measure with a different sampling rate than I am sending to DCX. Extrapolating, however, it is easy to imagine that distortion of 0.1% distortion is exceeded either at higher levels or higher frequencies.
At low enough levels, the THD+N doesn't rise at higher frequencies, thus being dominated by noise rather than harmonic distortion. But that only seems to occur at levels below -30dB (about 0.3 volts).
Now what does the harmonic distortion look like? Third harmonic is substantially larger than 2nd, and additional odd harmonics fall off "exponentially" at 45 degree angle on dB scale, with even harmonics being much smaller. Not the worst possible (such as flat or increasing odd orders) but not sonically great either.
Adapter
Is your XLR to RCA adapter built with pin 3 shorted to ground, terminated to ground with a resistor, or pin 3 open? If not, can you snip pin 3 open and try the measurements again?
That sounds quite good... distortion is low at moderate signal levels that wouldn't require any attenuation between the DCX and amps. Living with a bit less dynamic range is something I could easily do. What would you recommend for signal level, input gain and output gain to achieve that, keeping the maximum output levels below 2.5 V or so?
- Doug
- Doug
BTW, I'm pretty sure pin 3 (signal -) is open, but will check. I have in past always used balanced connections between amp and DCX. Before returning it to service, I'm going to switch my Julia to balanced mode (requires reversing sub-card) and re-measure. Of course balanced should be better numerically, but may not help odd orders of distortion much.
Actually even 2 volts output produces more distortion at high frequencies than I am happy with. But it's probably not generally audible and not worth adding additional post-crossover amplification to avoid. It might be worth fixing by modification. I see a number of interesting possible mods, but they generally look too complex, extensive, or expensive.
I don't know, probably for most systems where amplifier requires about 1-2V and preamplifier provides that, just using 1x gain passthrough is fine, and probably the safest guess for low distortion.
Don't worry about not lighting up all the lights. I used to worry about that, and in fact on one of my old systems with a single-ended connection, I use 12dB line attenuator at the output to make sure I'm getting the "full dynamic range" from DCX. I sometimes wondered if it didn't sound worse that way, and now I'm determined to remove the attenuator.
Or, if you want, increase level on input and decrease on output with controls, as needed. But with 144db or more dynamic range in the DSP, lighting up all the lights isn't really necessary. If you simply decrease level digitally on output, be sure that input never gets too high either. I didn't measure it this time, but I think I'd want to never drive DCX input with more than 4v, until at least the point where amplifier would into hard clipping (any many like more do far more than rated power on peaks). So that argues against trying to get too tricky with digital attenuation control on output to maximize S/N.
pot value
argh... I have already a six channel alps, but it is 100k.... is it usable? I'm planning to put it near the dac and the overall cable lenghts will be quite short.
Wow, it's getting simpler and simpler!
About the buffer stage, I've found it often beneficial even if the driving component should have been able to drive easily its load. C'mon, it's only three double triodes and one trafo
But in any case, could you please help me about a suitable transformer?
Thank you very much!
Paolo
argh... I have already a six channel alps, but it is 100k.... is it usable? I'm planning to put it near the dac and the overall cable lenghts will be quite short.
Wow, it's getting simpler and simpler!
About the buffer stage, I've found it often beneficial even if the driving component should have been able to drive easily its load. C'mon, it's only three double triodes and one trafo
But in any case, could you please help me about a suitable transformer?
Thank you very much!
Paolo
I guess this is because a speaker does not always have a constant impedance, an 8 ohm speaker can sometimes present a 5ohm load to the amplifier at certain frequencies so this is why an amplifier designer has to make sure his 4 ohm rated amp will not blow when presented with a 2ohm load...
Thanks for the explanation, I think I'm starting to understand whats going on here, am I right in assuming that with this configuration:
DCX => 20K pot => Amp
That Rload is 20K but Rsource is variable depending on the position of the pot (assuming interconnects are 0 ohm)?
I guess this isn't ideal from the Amp's point of view, so I'm guessing that is a buffer was used between the 20k pot and amp then the amp (Rsource) would then see a fixed impedance?
yes, the DCX sees the 20k pot as Rload.
The amp sees the 20kpot as Rsource that varies from 0r0 to 5k0.
What will the amp do when Rsource is near 0r0 and will it behave differently when Rsource=5k0?
Then you bring in the cable effect.
Will the DCX drive the 20k //cable capacitance?
Will the 20k pot drive Rin//[cable capacitance+amp capacitance]?
Thanks Charles for the great info. The odd order distortion makes sense to me. Thanks for you findings.
Pidigi: As for the 100K pot, it is not the best choice. Unless you use it followed by your tube buffer, then it would be about right.
For a transformer, all you need is a 1:1 transformer capable of 2V RMS. That's 8dBu or 6dBV - you will see some specified that way. A transformer that can do higher levels is fine. I have had good luck with both 600:600 and 10K:10K transformers. Put a 3K resistor across the primary of each transformer to flatten the impedance.
I would suggest something from Jensen, Lundhal or Onetics (my fav). All very nice. Dave Slagle is supposed to make a nice transformer, but I have not heard them.
Pidigi: As for the 100K pot, it is not the best choice. Unless you use it followed by your tube buffer, then it would be about right.
For a transformer, all you need is a 1:1 transformer capable of 2V RMS. That's 8dBu or 6dBV - you will see some specified that way. A transformer that can do higher levels is fine. I have had good luck with both 600:600 and 10K:10K transformers. Put a 3K resistor across the primary of each transformer to flatten the impedance.
I would suggest something from Jensen, Lundhal or Onetics (my fav). All very nice. Dave Slagle is supposed to make a nice transformer, but I have not heard them.
Giving that away, aren't they?
It's a good sign that it costs more the the DCX itself - I definitely want the "gold" version.
Pin 3 clipped...more distortion
OK, it turned out that all my 4 brands of XLR to RCA adapters connect pin 3 to RCA ground. So I did as you said, clipped pin 3, leaving pin 1 to shield and pin 2 to RCA hot. I did this only for one adapter, so had to measure in mono. As mentioned in earlier post, I am measuring analog output with DCX being driven by digital signal at 44.1kHz with 24 bits.
Result at -4dB was increased THD+N, 0.018%. I retested with unmodified adpater and got 0.011%, lower than previous 0.013% probably because of testing in mono vs. stereo. This is all THD, noise is so low it doesn't affect the THD+N number.
It seems like, to it's credit, the DCX handles balanced negative being attached to ground just fine. And when doing that, like most of my adapters, signal level is increased, doubled as you can see. (oops, i removed my incorrect hypothesis about how distortion was decreased this way. Actually, some sort of cancellation is probably involved.) So, you get twice the output (6dB more) with less distortion.
The decreased distortion in mono mode suggests to me that some of the THD is related to power supply limitation. The more the outputs are drawing, the more the voltage supplied to output board sags, which increases distortion. But this is not the primary source of distortion.
OK, it turned out that all my 4 brands of XLR to RCA adapters connect pin 3 to RCA ground. So I did as you said, clipped pin 3, leaving pin 1 to shield and pin 2 to RCA hot. I did this only for one adapter, so had to measure in mono. As mentioned in earlier post, I am measuring analog output with DCX being driven by digital signal at 44.1kHz with 24 bits.
Result at -4dB was increased THD+N, 0.018%. I retested with unmodified adpater and got 0.011%, lower than previous 0.013% probably because of testing in mono vs. stereo. This is all THD, noise is so low it doesn't affect the THD+N number.
It seems like, to it's credit, the DCX handles balanced negative being attached to ground just fine. And when doing that, like most of my adapters, signal level is increased, doubled as you can see. (oops, i removed my incorrect hypothesis about how distortion was decreased this way. Actually, some sort of cancellation is probably involved.) So, you get twice the output (6dB more) with less distortion.
The decreased distortion in mono mode suggests to me that some of the THD is related to power supply limitation. The more the outputs are drawing, the more the voltage supplied to output board sags, which increases distortion. But this is not the primary source of distortion.
Last edited:
Hi,
When you ground that pin3 you essentially go from balanced to single ended output. The total difference output signal doesn't change (1/2 signal on pin 2 and 1/2 signal on pin3) but you now measure the total to gnd instead per phase so now it looks like you have the double value. That also explains the increased THD: some of the THD+N is normally cancelled between the phases (that's the reason for balanced in the first place) so you also see a higher THD. This is even when the output circuits itself don't change in THD etc, which normally they won't anyway.
jd