Not sure theres much to be learnt in duplicating a test the difference is quite marginal as you say similar to using different resistors etc
The bf862 is perfectly fine
An externally hosted image should be here but it was not working when we last tested it.
Hi,
I just found recently this pin 20 CCS story, thank you for that!
Maybe I have some clues about it, I think it's a noise story.
According to the PCM 1794 internal schematic founded in the main thread, the pin 20 resistor converts the voltage present on the pin 20 into a current. So the pin 20 voltage noise will also be converted in a current noise. I measure this voltage noise to be around 60nV/root Hz which is converted into a current noise of 6pA/root Hz with a 10kΩ resistor.
The CCS have a much higher internal resistance, in the megohms range. 100 times the 10KΩ. So the current noise generated by the the voltage noise will be 100 times lower.
But every resistor produces noise by themselves and there is a resistor in the CCS. Resistor voltage noise is about 0.13*sqr(R) and so current noise is 0.13/sqr(R) in nV and nA at room temperature. A 2KΩ gives 6nV/rHZ and 3pA/rHz. So the current noise of the CSS with a source resistor of 2KΩ is 3pA/rHz, half the initial noise without the CCS. A 6dB improvement and it's already largely noticeable to hears!
But is possible to make things better. The dominant source of noise is the resistor current noise and depends on resistor value. For less noise the resistor value must be increased. The voltage ( 2.4v ) available at the pin 20 makes things impossible, an external negative supply must be used. As I already have it in my dac I build a CCS with a 40KΩ source resistor. Current noise is now 0.65pA/rHZ, 10 times less, 20dB improvement. To my hears the sound quality is better, really better compared to the first simple CSS build with a 2SK30 and a 2KΩ resistor.
Attached is the little schematic of the improved CCS designed for my dac, 0.25mA.
Negative voltage is regulated on board, voltage reference for the V/I conversion is heavily filtered, the source resistor is splitted in a 4 resistors string to reduce excess 1/f noise and as I have plenty of voltage available the CCS fet is cascoded.
So the negative voltage noise is not critical.
Hopes this helps and hopes my poor english makes it understandable nevertheless...
Joël
I just found recently this pin 20 CCS story, thank you for that!
Maybe I have some clues about it, I think it's a noise story.
According to the PCM 1794 internal schematic founded in the main thread, the pin 20 resistor converts the voltage present on the pin 20 into a current. So the pin 20 voltage noise will also be converted in a current noise. I measure this voltage noise to be around 60nV/root Hz which is converted into a current noise of 6pA/root Hz with a 10kΩ resistor.
The CCS have a much higher internal resistance, in the megohms range. 100 times the 10KΩ. So the current noise generated by the the voltage noise will be 100 times lower.
But every resistor produces noise by themselves and there is a resistor in the CCS. Resistor voltage noise is about 0.13*sqr(R) and so current noise is 0.13/sqr(R) in nV and nA at room temperature. A 2KΩ gives 6nV/rHZ and 3pA/rHz. So the current noise of the CSS with a source resistor of 2KΩ is 3pA/rHz, half the initial noise without the CCS. A 6dB improvement and it's already largely noticeable to hears!
But is possible to make things better. The dominant source of noise is the resistor current noise and depends on resistor value. For less noise the resistor value must be increased. The voltage ( 2.4v ) available at the pin 20 makes things impossible, an external negative supply must be used. As I already have it in my dac I build a CCS with a 40KΩ source resistor. Current noise is now 0.65pA/rHZ, 10 times less, 20dB improvement. To my hears the sound quality is better, really better compared to the first simple CSS build with a 2SK30 and a 2KΩ resistor.
Attached is the little schematic of the improved CCS designed for my dac, 0.25mA.
Negative voltage is regulated on board, voltage reference for the V/I conversion is heavily filtered, the source resistor is splitted in a 4 resistors string to reduce excess 1/f noise and as I have plenty of voltage available the CCS fet is cascoded.
So the negative voltage noise is not critical.
Hopes this helps and hopes my poor english makes it understandable nevertheless...
Joël
Ah, I see, you were talking fets used in a buffer, not in a CCS...
I also faced the variance in impedance issue, but found that high input impedance at the amp-end, and low interconnects' capacitive load, produced very good results. I use M Cap Supreme silver/oil 4.7uF / 800VAC
Nick
Doede has included 2sk208s for the pin20 CCS in his new dddac1794 boards yes, you can make them out in the image here:
An externally hosted image should be here but it was not working when we last tested it.
The board looks good, thanks for posting. Are those 10uF caps still needed though?
Nick
The 2 per channel underneath each dac chip in that picture definitely need to stay as they are the vcom caps, rather than just power supply decoupling ones. The others? I guess it's easy enough to try them with and without
CCS Pin 20 Noise
Noise was not the concern with the 6K resistor or CCS circuits to pin 20 nor did I ever hear any significant noise from the DDDAC.
The first consideration was improved sq which is evident to everyone who has tried the "standard" CCS circuit.
The second consideration was a zero tempco so that the CCS current does not change with temperature which was also achieved with the 2SK208O.
It is interesting that you are using 0.25ma when Doede's PCM1794 tests showed that 0.40ma gave the best sound from the PCM1794 DAC.
The circuit with 5 transistors and an external -23 Volts is interesting, but is too complex for me under the circumstance. My DDDAC already sounds analog and "right" so I am going to do other DIY projects.
Noise was not the concern with the 6K resistor or CCS circuits to pin 20 nor did I ever hear any significant noise from the DDDAC.
The first consideration was improved sq which is evident to everyone who has tried the "standard" CCS circuit.
The second consideration was a zero tempco so that the CCS current does not change with temperature which was also achieved with the 2SK208O.
It is interesting that you are using 0.25ma when Doede's PCM1794 tests showed that 0.40ma gave the best sound from the PCM1794 DAC.
The circuit with 5 transistors and an external -23 Volts is interesting, but is too complex for me under the circumstance. My DDDAC already sounds analog and "right" so I am going to do other DIY projects.
Of course not! The dac output noise is out of concern here.
Noise in pin20 current affects sound quality, certainly some kind of intermodulation. As do noise or ripple on dac supply.
In my current dac increasing Iref to 0.4mA implies a lot of changes in the I/V output stage. Too late for big changes. Too bad...
Joël
Appreciate the report
Very interesting
What voltage is on vcom from gnd and what the iv configuration?
The dddac analig section is over voltaged to 8v
It really starts to sing over 7.5v
But you need the right combination of iv analog voltage and ccs current
I've read your post carefully. It appears that the reason why "the first simple CSS build with a 2SK30 and a 2KΩ resistor" sounds worse compared to the 4-resistor one (that you provided the schematics for), is the cascoding and in particular the on-board, fast, wide bandwidth voltage regulation with "better" PCB trace impedance in combination with capacitors' value (10uF only) and type used for decoupling. I don't think your assumption that the 4 resistors have lesser noise compared to one has a merit. The noise calculation also has to include thermal noise calculations - hence, single 40K resistor would induce less thermal noise harm, compared to 4 X 10K. I'd like to hear what you think about this because I find your ccs schematic really interesting.
Regards,
Nick
was there not also a reason that using a jfets that needs say a 2k r versus a 220r to obtain correct current may not be optimal?
i know from what i tested i didnt like the r's in the 500r and above
that may be coincidence though
picking jfets for this wasnt easy
i know from what i tested i didnt like the r's in the 500r and above
that may be coincidence though
picking jfets for this wasnt easy
I don't know the voltage on Vcom, I can measure it but as I'm using the PCM1794 with 5V and 0.25mA it should be the datasheet value.
The I/V is based on a transconductance stage with a very low input impedance.
I haven't yet try to overvolt the analog section. It should be possible with minor modifications.
Maybe I'm wrong but I suppose the right combination of analog voltage and Iref current is sensitive with a plain resistor for the I/V? Vcom has to be close to Vsupply/2?
Joël
Thermal noise depends only on resistor value and temperature. No matter if the resistor is a single one or a string of many as long as the ohmic value is the same.
Some other noises depends on the voltage across the resistor so splitting the resistor decrease the excess noise as the voltage is spread across the resistors.
A quite good introduction to resistor noise:
https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf
In the schematic the strong filtering and the cascode are needed because the -V available is quite noisy. Filtering the reference voltage suppress noise and cascoding increase the ccs internal impedance a lot. I measured many GΩ. High impedance means low current noise conversion from supply noise.
Joël
Interesting but I don't know the answer!
Maybe measuring the current noise will gives a clue.
I go to the noise way because around the pin 20, voltage an current are constants. No swings at all. The only variation is the noise.
Jfet ccs reduce the current noise and the sound quality improves. Better ccs gives lower current noise and sound quality improves again.
So current noise seems to be involved in the story.
But there's maybe other effects I don't know...
Joël
would be interesting for me if you measured it
as far as i remember higher resistor values rose the vcom voltage
yes vsupply/2 is the max for vcom it will clip so need to be a little under
im not sure i remember exactly (its i while since i thought about this chip)
if higher vcom voltage has any benefits other than higher output i think it didnt
some guys would suggest using lower value precision iv resistors and adding gain rather than trying to get max output out of the chips
when i tried it that was the feeling i got but the difference wouldn't be massive
the bulk of the magic is overvolting and rising ccs accordingly (with some fine adjustment)
nige
Last edited:
my guess is that its current stability and noise
the lm334 we tested was great only its a noisy bugger
the 2sk170bl was lovely and quiet but dynamics and speed was slower
It appears that you are mixing thermal and current noise: the thermal noise increases with a larger resistance value, while the current noise decreases.
Thermal noise will be directly proportional to an amount of resistive material used, i.e. to a number of resistors.
Nick
For those of you running 8 or more DAC modules can you guys please share with me the PSU design you used for the 12v?
Cheers
Cheers
If you do a search of my posts, I have uprated the power supply to run 16 Dac modules.
I am in the process of drawing the circuit in Eagle which I will post when done 🙂