Yes, and I'm sure that a SEN layout stackable with the TPA Buffalo would make a recipe for success. Unfortunately, I'm not capable of doing this......
Nic
more noise but cap free - did you compare how sounds to SEN / CEN
maybe a schematic with more BJT and Rload will be less noiser
BTW what Is measured noise In real compared to simulation at post 95
Last edited:
> more noise but cap free
If you have read and understood the article, you would agree the cap has a different role here than in e.g. Zen IV or D1.
Any distortion due to the cap has a lot less influence, if any at all.
> did you compare how sounds to SEN / CEN
Yes, but I will not make comments of my circuits against ones not from me.
Except only to say that my own default circuit for PCM1704 is still SEN.
> what Is measured noise In real compared to simulation at post 95
The actual noise measurements can be seen at the start of the thread.
-100dBV using 2x NiMH batteries 8.4V each, no filtering caps other than those in the circuit.
No exotic regulators either.
We also listen with headphones to compare one to another.
Patrick
.
If you have read and understood the article, you would agree the cap has a different role here than in e.g. Zen IV or D1.
Any distortion due to the cap has a lot less influence, if any at all.
> did you compare how sounds to SEN / CEN
Yes, but I will not make comments of my circuits against ones not from me.
Except only to say that my own default circuit for PCM1704 is still SEN.
> what Is measured noise In real compared to simulation at post 95
The actual noise measurements can be seen at the start of the thread.
-100dBV using 2x NiMH batteries 8.4V each, no filtering caps other than those in the circuit.
No exotic regulators either.
We also listen with headphones to compare one to another.
Patrick
.
Last edited:
Default circuit for PCM1704 is SEN with R_iv = 2.2k.
Yes, you should read the article , especially if you want to adapt the circuit for your specific needs.
No, we have not measured the noise of CEN BJT.
I don't have access to Jan Didden's Audio Precision every evening.
But we have listened with headphone and compared to SEN, which we have measured.
Patrick
Yes, you should read the article , especially if you want to adapt the circuit for your specific needs.
No, we have not measured the noise of CEN BJT.
I don't have access to Jan Didden's Audio Precision every evening.
But we have listened with headphone and compared to SEN, which we have measured.
Patrick
Have you listened to them? Can you hear the difference between R-2R and delta-sigma?
What's the best value/price R-2R, in your opinion?
I have a sample ES9018 here lying around, I have many versions of ES9022.
I also know 2 persons who have built both 9018 with D1 and 9022, and prefer the 9022 when properly done.
We also had various other DACs before, like AK4396, ...
So yes, I have listened to top range Sigma Delta's.
Our current SEN / CEN protos run on an AD1865. Many of R2Rs are no longer available. Very difficult to manufacture.
PCM1704 is expensive. If you do not want to over-sample (we do OS), then PCM56 is still affordable.
The cost of the DAC in the total cost of a built is less << 10%, and it is SQ deciding.
So I do not save money on them.
Just my personal choice,
Patrick
.
I also know 2 persons who have built both 9018 with D1 and 9022, and prefer the 9022 when properly done.
We also had various other DACs before, like AK4396, ...
So yes, I have listened to top range Sigma Delta's.
Our current SEN / CEN protos run on an AD1865. Many of R2Rs are no longer available. Very difficult to manufacture.
PCM1704 is expensive. If you do not want to over-sample (we do OS), then PCM56 is still affordable.
The cost of the DAC in the total cost of a built is less << 10%, and it is SQ deciding.
So I do not save money on them.
Just my personal choice,
Patrick
.
Last edited:
Someone mentioned that SEN / CEN is still with a cap, even though I explained in the article why it is unimportant here compared to ZEN or D1.
And he mentioned that most discrete IVs for ES9018 also have a cap.
If you really want to have no cap at the expensive of complication and higher distortion, I shall publish my folded cascode IV in due course (earliest October). That has 8 JFETs for 1 single ended channel, including one for a DC servo. here is an earlier proto in SMD
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298006
The one I posted at the beginning of the ZEN IV thread based on JC's BT is a good starting point.
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298531
Patrick
.
And he mentioned that most discrete IVs for ES9018 also have a cap.
If you really want to have no cap at the expensive of complication and higher distortion, I shall publish my folded cascode IV in due course (earliest October). That has 8 JFETs for 1 single ended channel, including one for a DC servo. here is an earlier proto in SMD
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298006
The one I posted at the beginning of the ZEN IV thread based on JC's BT is a good starting point.
http://www.diyaudio.com/forums/pass-labs/173291-zen-i-v-converter.html#post2298531
Patrick
.
Last edited:
I have been asking by PM why I do not add the offset current source, or a reference voltage, ..., on the same PCB.
Let me reiterate.
I have no ambition to make a product from this, and certainly not one that is universal for all thinkable applications (which is always a compromise). The evaluation pack is, as its name implies, a means to make it easier for you to evaluate the working principle of my circuits.
Yes, one can add the additional current source on the same PCB. One can also add a offset voltage reference, charge transfer floating power supply, shunt regulators, perhaps then (even better) the DAC itself and a discrete XO, …. Where should I stop ? I might end up doing e.g. what Spencer does – selling you the complete DAC on one PCB.
Maybe I’ll do that one day, when I have the time (and the need) to make a living from audio.
If I were to use a DAC on a separate PCB that requires a bias DC current, I would, personally, place the biasing current source close to the Dac chip itself, and not at the IV Converter board. A very small daughter board will do easily.
Another specific question relates to whether to use an external (JFET) current source to replace the current source at Pin 6 of a PCM63. While one might be able to make a better current source externally, one would most likely encounter more drift problems with an external CCS than one on chip on the same substrate. Toshiba JFETs are not really super stable thermally in the 10uA range. But that is only a personal opinion.
Patrick
Let me reiterate.
I have no ambition to make a product from this, and certainly not one that is universal for all thinkable applications (which is always a compromise). The evaluation pack is, as its name implies, a means to make it easier for you to evaluate the working principle of my circuits.
Yes, one can add the additional current source on the same PCB. One can also add a offset voltage reference, charge transfer floating power supply, shunt regulators, perhaps then (even better) the DAC itself and a discrete XO, …. Where should I stop ? I might end up doing e.g. what Spencer does – selling you the complete DAC on one PCB.
Maybe I’ll do that one day, when I have the time (and the need) to make a living from audio.
If I were to use a DAC on a separate PCB that requires a bias DC current, I would, personally, place the biasing current source close to the Dac chip itself, and not at the IV Converter board. A very small daughter board will do easily.
Another specific question relates to whether to use an external (JFET) current source to replace the current source at Pin 6 of a PCM63. While one might be able to make a better current source externally, one would most likely encounter more drift problems with an external CCS than one on chip on the same substrate. Toshiba JFETs are not really super stable thermally in the 10uA range. But that is only a personal opinion.
Patrick
I come to this thread late in the game, having already gotten my copy of LA vol. 2. However, reading through the posts, I'm wondering how the elephant in the living room has been ignored: the actual output impedance of the given DAC. In the LA article a good deal of effort is described around constructing a V-I test source, and in the end how a 20k resistor is used for tests with Jan's Ap. But when I went to see what the actual output characteristics were of the one DAC mentioned in the article, the PCM1704, the answer is 1k! This is hardly what one would think of as a current source. In fact, in this context "current output" is quite misleading, almost a misnomer. But it's easier than saying "output which will produce a specified current into a virtual ground".
I did a brief survey of other DACs on the TI/Burr Brown site, and there is NO output impedance specified for the other "current output" ones. Finally found one other specified part, the obsolete AD1865, which states 1.7k, plus/minus 30%. Still not exactly a current source! And consulting Samuel Groner, he mentioned that parts he works with are around 200 ohms!
The problem now is discovering what the effect is of termination in a not-very-low input impedance, for one thing. And what is the code dependence of the effect, if any? Also, noise and distortion results will get appreciably different (worse) when these more realistic source impedances are used.
No one can dispute the sonic results if they are appealing to a given individual, so I certainly don't mean to discourage anyone from playing with these topologies. But rather than writing yet another letter to LA, I felt I ought to air these concerns somewhere.
Brad Wood
I did a brief survey of other DACs on the TI/Burr Brown site, and there is NO output impedance specified for the other "current output" ones. Finally found one other specified part, the obsolete AD1865, which states 1.7k, plus/minus 30%. Still not exactly a current source! And consulting Samuel Groner, he mentioned that parts he works with are around 200 ohms!
The problem now is discovering what the effect is of termination in a not-very-low input impedance, for one thing. And what is the code dependence of the effect, if any? Also, noise and distortion results will get appreciably different (worse) when these more realistic source impedances are used.
No one can dispute the sonic results if they are appealing to a given individual, so I certainly don't mean to discourage anyone from playing with these topologies. But rather than writing yet another letter to LA, I felt I ought to air these concerns somewhere.
Brad Wood
Very interesting comments. In this case this URLs are of interest:
Digi Scoop 1
IV conversion
What means the term "LA article"?
No comments to my post #41 (where can I find an article about the perceived audible differences between this various I/U topologies)?
Last edited:
Vielen Dank tiefbass for those links! The second one is especially good, including the expression of frustration with the unresponsive manufacturer :/ I'm not sure if TI will pick up the phone when I call these days either 
And thanks to heinz1 for the clarification. LA does have a lot of other common meanings, although I'm sure Jan Didden will be pleased if we think of his bookzine as one of the first to occur!
Brad
And thanks to heinz1 for the clarification. LA does have a lot of other common meanings, although I'm sure Jan Didden will be pleased if we think of his bookzine as one of the first to occur!
Brad
Thank you for pointing out the issue of output impedance of R2R DACs. Very good point.
And I hope some other experts in DAC internal constructions would join the discussion.
Firstly, when assessing the performance of an IV converter, one has to be able to quantify the performance of that circuit on its own. That is why we thought it was important to use a high impedance to simulate a current source. The issue of input impedance was explained in some detail in the article. Due to the limited space in the bookzine. A part of that had to be removed to keep the article from being too long winded.
It was also mentioned in the article how this finite input impedance (about 15R using a single pair of complementary JFETs) can be further reduced, at the expense of increase of distortion of the IV circuit itself, and that had to be balanced against performance gain in the DAC distortion due to the lower impedance that the current output sees.
It is correct to say that distortion at the DAC would be lowest when the current output sees a virtual Gnd, as in an Opamp IV. It is also true in saying that a BJT type current conveyor circuit (the famous one being the one from Jocko) would have a lower impedance compared to JFETs (in the order of 2~3R). So the JFET circuit in combination with a limited DAC output impedance will produce some additional distortion. But the non-linearity of the JFET is, as opposed to the BJT, largely quadratic in nature and thus cancels out in balanced mode. The CEN circuit is even better in this sense as it has some complementary cancellation already in single ended circuit. This is one key reason why I prefer JFETs.
So the distortion mechanism described here is not due to the IV circuit itself, but due to the input / output impedance interface between IV and DAC. All the same, it is an issue to be taken into account in the overall system design.
One can half the input impedance of the IV circuit by doubling the no of JFETs, and the distortion will rise by about 6dB (theoretically). One can also increase the bias current which helps to reduce the non linearity of the effective input impedance, without changing its value.
If you look for pure THD figures, nothing will beat a high bandwidth opamp. And I never claim that they sound awful.
I also happen to have a circuit similar to SEN using a pair of LU1014s in triode mode. The input impedance is 1R, and very very linear. The penalty is 48V floating supply and 200mA bias current. I am waiting for the AT1401 to try with.
Best regards,
Patrick
.
And I hope some other experts in DAC internal constructions would join the discussion.
Firstly, when assessing the performance of an IV converter, one has to be able to quantify the performance of that circuit on its own. That is why we thought it was important to use a high impedance to simulate a current source. The issue of input impedance was explained in some detail in the article. Due to the limited space in the bookzine. A part of that had to be removed to keep the article from being too long winded.
It was also mentioned in the article how this finite input impedance (about 15R using a single pair of complementary JFETs) can be further reduced, at the expense of increase of distortion of the IV circuit itself, and that had to be balanced against performance gain in the DAC distortion due to the lower impedance that the current output sees.
It is correct to say that distortion at the DAC would be lowest when the current output sees a virtual Gnd, as in an Opamp IV. It is also true in saying that a BJT type current conveyor circuit (the famous one being the one from Jocko) would have a lower impedance compared to JFETs (in the order of 2~3R). So the JFET circuit in combination with a limited DAC output impedance will produce some additional distortion. But the non-linearity of the JFET is, as opposed to the BJT, largely quadratic in nature and thus cancels out in balanced mode. The CEN circuit is even better in this sense as it has some complementary cancellation already in single ended circuit. This is one key reason why I prefer JFETs.
So the distortion mechanism described here is not due to the IV circuit itself, but due to the input / output impedance interface between IV and DAC. All the same, it is an issue to be taken into account in the overall system design.
One can half the input impedance of the IV circuit by doubling the no of JFETs, and the distortion will rise by about 6dB (theoretically). One can also increase the bias current which helps to reduce the non linearity of the effective input impedance, without changing its value.
If you look for pure THD figures, nothing will beat a high bandwidth opamp. And I never claim that they sound awful.
I also happen to have a circuit similar to SEN using a pair of LU1014s in triode mode. The input impedance is 1R, and very very linear. The penalty is 48V floating supply and 200mA bias current. I am waiting for the AT1401 to try with.
Best regards,
Patrick
.
Last edited:
Here are two links which you might find useful :
Understand the differences between R2R and String DAC architectures
Modifying CD player DAC circuits
Patrick
Understand the differences between R2R and String DAC architectures
Modifying CD player DAC circuits
Patrick
Thanks for those links. The contrast between the R-2R and "String" versions is a good, if extreme, example of the drawbacks of potentially variable impedance. The tubish one does have an interesting variant discussed (in which the object is to make the load on the reference constant, rather than the output impedance of the network).
It would indeed be good if some actual designers could weigh in with details of what is really going on in some of the decent DACs. Unfortunately this may violate their sacred oaths of non-disclosure One could also piece things together from patents, although in that case the art of the patent is to protect without actually giving away how to do something!
I actually like the use of the floating voltage source in your (and similar) design(s). And if one does not regard negative feedback as anathema, closing a loop around the basic structure is an example of what Arbel calls enhancing feedback (see his great book Analog Signal Processing), in which the shunt feedback further reduces the already low input stage impedance.
Brad
It would indeed be good if some actual designers could weigh in with details of what is really going on in some of the decent DACs. Unfortunately this may violate their sacred oaths of non-disclosure
One could also piece things together from patents, although in that case the art of the patent is to protect without actually giving away how to do something!I actually like the use of the floating voltage source in your (and similar) design(s). And if one does not regard negative feedback as anathema, closing a loop around the basic structure is an example of what Arbel calls enhancing feedback (see his great book Analog Signal Processing), in which the shunt feedback further reduces the already low input stage impedance.
Brad
Last edited: