I don't see why you wouldn't want to simply use the
I/V converter from the D1. Schematics at www.passlabs.com
I´m afraid, basically it´s the same circuit
exept the using of MAT as matched cascode device.
the reason is that I got a few dozens of them by chance and
MAT03 too, so I play around with them.
Matching Mosfets can be frustrating sometimes, MAT´s are
matched by design.
What I personally don´t like are caps in the signal path
therefore I decided to make the output nullable.
Since the MAT works as a cascode it´s influence on sound is a minor one.
U.
I/V converter from the D1. Schematics at www.passlabs.com
I´m afraid, basically it´s the same circuit
exept the using of MAT as matched cascode device.
the reason is that I got a few dozens of them by chance and
MAT03 too, so I play around with them.
Matching Mosfets can be frustrating sometimes, MAT´s are
matched by design.
What I personally don´t like are caps in the signal path
therefore I decided to make the output nullable.
Since the MAT works as a cascode it´s influence on sound is a minor one.
U.
Combine best designs
The idea of designing a firstclass DAC for DIY is very attractive!
I recently upgraded my Musical Fidelity 24bit 96 kHz DAC with Panasonic FC capacitors. This small step improved depth and spaciousness of the sound enormously, so I guess that when building one from scratch with topnotch components could really bring about a firstclass sound.
Maybe we should compare some presentday DIY designs floating around in the Internet, and pick out the best ideas for our balanced Pass-like design....?
Here are some links:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=9685
http://home.worldonline.dk/hifiside/ecdafsp.htm
http://www.dddac.de/
http://www.quadesl.com/dac.shtml
http://members.chello.nl/~m.heijligers/DAChtml/dactop.htm
http://www.specmail.nl/zelfbouw/dac24/dac24.htm
Then there is this commercially available module form Hoerwege in Germany that, according to tests should sound very good indeed. It is built with highgrade components. Maybe we could learn something there (info only in German, but the pictures epak to themselves):
http://www.hoer-wege.de/dacup24192.html
There must be hundreds of other links as well...
Anyway, using 24 bit 192 kHz chips seems a must since these are no longer very expensive.
Regards,
Lucas.
The idea of designing a firstclass DAC for DIY is very attractive!
I recently upgraded my Musical Fidelity 24bit 96 kHz DAC with Panasonic FC capacitors. This small step improved depth and spaciousness of the sound enormously, so I guess that when building one from scratch with topnotch components could really bring about a firstclass sound.
Maybe we should compare some presentday DIY designs floating around in the Internet, and pick out the best ideas for our balanced Pass-like design....?
Here are some links:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=9685
http://home.worldonline.dk/hifiside/ecdafsp.htm
http://www.dddac.de/
http://www.quadesl.com/dac.shtml
http://members.chello.nl/~m.heijligers/DAChtml/dactop.htm
http://www.specmail.nl/zelfbouw/dac24/dac24.htm
Then there is this commercially available module form Hoerwege in Germany that, according to tests should sound very good indeed. It is built with highgrade components. Maybe we could learn something there (info only in German, but the pictures epak to themselves):
http://www.hoer-wege.de/dacup24192.html
There must be hundreds of other links as well...
Anyway, using 24 bit 192 kHz chips seems a must since these are no longer very expensive.
Regards,
Lucas.
motivation
Dear Nelson,
of course your circuit is in its simplicity AND performance
sensational (especially for one like me who heard a D1 working)
but I believe that motivation for DIYers is not only to copy good and proven designs but to play around a little bit with them and put in some (more or less good) own ideas.
Sometimes its like reinventing the 8 cornered wheel or
something like that, sometimes its an improvement.
Uli
PS: I don´t like caps in the signal path
Nelson Pass said:
Dear Nelson,
of course your circuit is in its simplicity AND performance
sensational (especially for one like me who heard a D1 working)
but I believe that motivation for DIYers is not only to copy good and proven designs but to play around a little bit with them and put in some (more or less good) own ideas.
Sometimes its like reinventing the 8 cornered wheel or
something like that, sometimes its an improvement.
Uli
PS: I don´t like caps in the signal path
D1 I/V converter
Here is my experience with the D-1 I/V at the output of Sony Current Pulse dacs.
I tried it the first time as you see it on the original schematic and found that at the sources of
Q2 and Q5 the voltage was too low to work (don’t know why).
Then the use of constant current sources replacing R28 and R33 comes in help.
The CCS are with Jfets (2sk170) as seen on the Pass article DIYOPAMPS.
Once set the current at 10mA/halve , the voltage at the sources of Q2 and Q5 was around 4 volts.
This is not much because the current that drive the buffer is 4/1500=2mA : a bit low for a irf 610.
(especially remembering that the D-1 manual claims 14V at the sources of Q2 and Q5 for a current at near 10mA) .
Looking forward for this problem I set R27 and R34 (drain resistor for the cascode pair) down to 1000 ohm each. Now the buffers runs at about 6mA .
This electrically speaking.
From a performance standpoint I tried it on 2 different players: a CD and a CD/SACD .
In the first case every thing was quiet even without C15 and C16 (capacitors across drain resistors) .
In the second case (CD/SACD player) can say that the SACD is quieter than CD and a capacitor across the Drain seems necessary.
Regarding matching constant current sources (I used 2sk170) why not a pair of 2sk389dual Jfet …
Ah … the listening … a pleasure.
Here is my experience with the D-1 I/V at the output of Sony Current Pulse dacs.
I tried it the first time as you see it on the original schematic and found that at the sources of
Q2 and Q5 the voltage was too low to work (don’t know why).
Then the use of constant current sources replacing R28 and R33 comes in help.
The CCS are with Jfets (2sk170) as seen on the Pass article DIYOPAMPS.
Once set the current at 10mA/halve , the voltage at the sources of Q2 and Q5 was around 4 volts.
This is not much because the current that drive the buffer is 4/1500=2mA : a bit low for a irf 610.
(especially remembering that the D-1 manual claims 14V at the sources of Q2 and Q5 for a current at near 10mA) .
Looking forward for this problem I set R27 and R34 (drain resistor for the cascode pair) down to 1000 ohm each. Now the buffers runs at about 6mA .
This electrically speaking.
From a performance standpoint I tried it on 2 different players: a CD and a CD/SACD .
In the first case every thing was quiet even without C15 and C16 (capacitors across drain resistors) .
In the second case (CD/SACD player) can say that the SACD is quieter than CD and a capacitor across the Drain seems necessary.
Regarding matching constant current sources (I used 2sk170) why not a pair of 2sk389dual Jfet …
Ah … the listening … a pleasure.
Tschrama, this is what I tried up to now during the trial&error design of my digital pre-amp:
- classic opamp I/V-converter
- Pass D1 output stage and
countless variations thereof
in simulation software and
real components
- Jocko Homo's "easy-to-build-
I/V-stage" and countless
variations thereof in real
components (Jocko's original
circuit wouldn't run in simulation,
only the real-world circuit
worked for me)
- LCAudio ZAPFilter mk II clone
with lots of modifications
(simulation only)
- passive (resistor only)
"I/V-stage"
I tried them with cheap DAC chips, I wouldn't want to use my pricey PCM1704's on test setups.
I haven't yet found something that I think would be a good solution for my DAC. The Pass D1 output stage sounds very good, but I don't like the rail voltages and I'm fiddling around with an additional stage that would give lower output impedance and enable me to remove the output caps (they get expensive if you don't want to hear them). The ZAPFilter clone could be promising, I'm trying to simplify the circuit without degrading its performance / functionality.
Requirements for me are:
- Best possible sound.
- Balanced operation with two DAC chips per channel. Balanced for me means making sure as much noise and distortion as possible appears common mode at the outputs to get canceled out at the input stage of the amplifier. This often requires more than just building two identical output stages and hooking them up to the same power supply.
- Low output impedance.
- "Normal" voltage rails, so the DAC and I/V-board can be used in existing CD / DVD players. "Normal" for me means something like +/-15 V.
- If possible, no caps in the signal path.
- Let the DAC work into an input it feels comfortable with: "virtual short to ground"
- Low noise. One of the few things you can measure and relate to what you hear.
- THD, TIM, whatever: Don't care. Can't measure it anyway. If my ears like the sound, I'm all happy.
I'm willing to drop any number of the above requirements if there are good reasons to do so - except the first. They represent what I'm aiming at if I didn't have to compromise.
I'm working on it and I'll post my results for you to comment on.
Lucas_G, good idea to compare what we have until now and take the best ideas from each design to make an ultimate I/V-stage! I think the Hoerwege DAC features a classic opamp I/V converter. If anyone needs a translation German -> English, let me know.
-----------------
Going off-topic now: I put together an airplane with some friends, and today we received an approval by the German FAA. If weather permits, we'll take it on its maiden flight tomorrow morning.
- classic opamp I/V-converter
- Pass D1 output stage and
countless variations thereof
in simulation software and
real components
- Jocko Homo's "easy-to-build-
I/V-stage" and countless
variations thereof in real
components (Jocko's original
circuit wouldn't run in simulation,
only the real-world circuit
worked for me)
- LCAudio ZAPFilter mk II clone
with lots of modifications
(simulation only)
- passive (resistor only)
"I/V-stage"
I tried them with cheap DAC chips, I wouldn't want to use my pricey PCM1704's on test setups.
I haven't yet found something that I think would be a good solution for my DAC. The Pass D1 output stage sounds very good, but I don't like the rail voltages and I'm fiddling around with an additional stage that would give lower output impedance and enable me to remove the output caps (they get expensive if you don't want to hear them). The ZAPFilter clone could be promising, I'm trying to simplify the circuit without degrading its performance / functionality.
Requirements for me are:
- Best possible sound.
- Balanced operation with two DAC chips per channel. Balanced for me means making sure as much noise and distortion as possible appears common mode at the outputs to get canceled out at the input stage of the amplifier. This often requires more than just building two identical output stages and hooking them up to the same power supply.
- Low output impedance.
- "Normal" voltage rails, so the DAC and I/V-board can be used in existing CD / DVD players. "Normal" for me means something like +/-15 V.
- If possible, no caps in the signal path.
- Let the DAC work into an input it feels comfortable with: "virtual short to ground"
- Low noise. One of the few things you can measure and relate to what you hear.
- THD, TIM, whatever: Don't care. Can't measure it anyway. If my ears like the sound, I'm all happy.
I'm willing to drop any number of the above requirements if there are good reasons to do so - except the first. They represent what I'm aiming at if I didn't have to compromise.
I'm working on it and I'll post my results for you to comment on.
Lucas_G, good idea to compare what we have until now and take the best ideas from each design to make an ultimate I/V-stage! I think the Hoerwege DAC features a classic opamp I/V converter. If anyone needs a translation German -> English, let me know.
-----------------
Going off-topic now: I put together an airplane with some friends, and today we received an approval by the German FAA. If weather permits, we'll take it on its maiden flight tomorrow morning.
In my belief Uli’s and Nelson’s I/V stages are the best. Lucas’s links to I/V stages that use op-amps are all inferior, because of the feedback needed by the opamp and the large number of parts the opamp puts in the signal path. Say, you're not the same Lucas who designed the electrical system in my type-3 Jaguar
. (Just kidding)
Uli’s and Nelson’s I/V stages both very good well designed designs with only slight differences. Here are the tradeoffs the Nelson Pass D1 is known to work, we know this because Nelson has already built a few hundred. The Uli designed has as yet not been tested in real life. The Uli designed does not need DC blocking caps on the output the d-1 does. The Uli designed requires hard to find mat02. The d-1 uses parts that are all in digi-key. The Uli design may have lower noise then the d-1, however I think the difference is inaudible. Which one is better? I don’t know.
I think if uli would be kind enough to add his DC nulling trick to the d-1 so we could remove the DC blocking caps we would have a winning combination.
. (Just kidding)Uli’s and Nelson’s I/V stages both very good well designed designs with only slight differences. Here are the tradeoffs the Nelson Pass D1 is known to work, we know this because Nelson has already built a few hundred. The Uli designed has as yet not been tested in real life. The Uli designed does not need DC blocking caps on the output the d-1 does. The Uli designed requires hard to find mat02. The d-1 uses parts that are all in digi-key. The Uli design may have lower noise then the d-1, however I think the difference is inaudible. Which one is better? I don’t know.
I think if uli would be kind enough to add his DC nulling trick to the d-1 so we could remove the DC blocking caps we would have a winning combination.
:LBHajdu said:In my belief Uli’s and Nelson’s I/V stages are the best. Lucas’s links to I/V stages that use op-amps are all inferior, because of the feedback needed by the opamp and the large number of parts the opamp puts in the signal path. Say, you're not the same Lucas who designed the electrical system in my type-3 Jaguar
I wish I were... Then I would be rich enough to hire me the bodyguards to protect me from anguished Jaguar-drivers like you!
:LBHajdu said:
Indeed!
Could this lead to another 800 piece PCB order...?
I think few realize what an improvement this could be for the overall soundquality of their system...
Good commercial DACs would cost you many thousand dollars
Regards,
Lucas.
parts issue
Lucas + Leve,
I didn´t know that it´s such an issue to get MAT´s in the US!
In Europe they are quite common for hi-end DIY projects.
according to AnalogDevices (former PMI) they cost 6,- USD
if you buy 100+
I´m on rework of my DAC, it´s basically a hypertweaked
'elektor' design, with different psu and 'my' analog output
which is a mixture of NP´s D1 and some bipolar stories.
As the software for the used GAL is available for about 15,-USD
I will draw the complete circuit (CS8414, DF1704 or -6, 4x PCM1704) and maybe some PCB work.
Maybe AD is willing to tell you where to buy MAT´s, maybe the
guys from AudioXpress (www.audioxpress.com) can tell you.
Uli
Lucas + Leve,
I didn´t know that it´s such an issue to get MAT´s in the US!
In Europe they are quite common for hi-end DIY projects.
according to AnalogDevices (former PMI) they cost 6,- USD
if you buy 100+
I´m on rework of my DAC, it´s basically a hypertweaked
'elektor' design, with different psu and 'my' analog output
which is a mixture of NP´s D1 and some bipolar stories.
As the software for the used GAL is available for about 15,-USD
I will draw the complete circuit (CS8414, DF1704 or -6, 4x PCM1704) and maybe some PCB work.
Maybe AD is willing to tell you where to buy MAT´s, maybe the
guys from AudioXpress (www.audioxpress.com) can tell you.
Uli
Nibolo,
The D1 does not care about oversampling, nonoversampling, 16bit, 24bit. All the D1 cares about is that the dac outputs current (as the TDA1541A does) and that that current is somewhere in the neighborhood of 2mA. The TDA1541A outputs 4mA I don’t know if that’s too much or not, but it’s a lot easer to accommodate the circuit for too much current then for too little. However if your building both halves of the d1 you will need two TDA1541A’s per channel in inverted phase thats 4 dacs for a 2 ch system.
On a somewhat different subject, could someone report as to if they hear any difference between 16bit and 24bit dacs. The only reason I am interested in 24bit is future proofing and some DVDs like the extended addition of “The lord of rings” comes with stereo recordings on the DVD. I don’t know if these recording are in 16 or 24bit.
uli,
I look forward to your circuit. What’s a GAL?
The D1 does not care about oversampling, nonoversampling, 16bit, 24bit. All the D1 cares about is that the dac outputs current (as the TDA1541A does) and that that current is somewhere in the neighborhood of 2mA. The TDA1541A outputs 4mA I don’t know if that’s too much or not, but it’s a lot easer to accommodate the circuit for too much current then for too little. However if your building both halves of the d1 you will need two TDA1541A’s per channel in inverted phase thats 4 dacs for a 2 ch system.
On a somewhat different subject, could someone report as to if they hear any difference between 16bit and 24bit dacs. The only reason I am interested in 24bit is future proofing and some DVDs like the extended addition of “The lord of rings” comes with stereo recordings on the DVD. I don’t know if these recording are in 16 or 24bit.
uli,
I look forward to your circuit. What’s a GAL?
GAL
GAL is short for Gate Array Logic I think, its something
instead of a graveyard of former 74LS ...
You design a graveyard of logical gates (74LS) and then
a smart digital guy shows up and says:
Youre a dinosaur man, today we programm all that ****
in a GAL, PAL, PLA some weird microcontroler...
The complete circuit has just this 'controler', the Crystal
receiver and the TEXAS DF1704 and 4x PCM1704.
My last ones were from BURRBROWN, hope the Texas ones
sound equal!
(and some peripherals of coarse )
I decided not to use the DF1706, I split the circuit into
4 PCB´s: Left, Right, Digital in, PSU. The digiboard will
be easily exchangeable (CS8414, DF1704, GAL-22V10)
Uli
GAL is short for Gate Array Logic I think, its something
instead of a graveyard of former 74LS ...
You design a graveyard of logical gates (74LS) and then
a smart digital guy shows up and says:
Youre a dinosaur man, today we programm all that ****
in a GAL, PAL, PLA some weird microcontroler...
The complete circuit has just this 'controler', the Crystal
receiver and the TEXAS DF1704 and 4x PCM1704.
My last ones were from BURRBROWN, hope the Texas ones
sound equal!
(and some peripherals of coarse
)I decided not to use the DF1706, I split the circuit into
4 PCB´s: Left, Right, Digital in, PSU. The digiboard will
be easily exchangeable (CS8414, DF1704, GAL-22V10)
Uli
Hi all,
here is one more less respectful (in terms of a D1 project) person.
I am currently building a DAC with the D1 output stage. It uses a DIR1703 as a receiver, Analog Devices AD1896A upsampler to 96kHz, DF1706 as digital filter and to seperate the channels, and four PCM1704 DACs for symetrical output.
Now I have to think about the lower output current of the PCM1704 (1.2mA max) compared to the 2mA of the PCM63. What to change to get full level output again?
I odered a proto PCB for this last week, so if all works out well, I can tell something about it in about for weeks.
Peter
here is one more less respectful (in terms of a D1 project) person.
I am currently building a DAC with the D1 output stage. It uses a DIR1703 as a receiver, Analog Devices AD1896A upsampler to 96kHz, DF1706 as digital filter and to seperate the channels, and four PCM1704 DACs for symetrical output.
Now I have to think about the lower output current of the PCM1704 (1.2mA max) compared to the 2mA of the PCM63. What to change to get full level output again?
I odered a proto PCB for this last week, so if all works out well, I can tell something about it in about for weeks.
Peter
Re: GAL
I also decided not to use such things as AD1896 as the CS8414
and DF1704 are designed for 96kHz. The DF1704 puts out 768kHz
(8 times oversampling) same as DF1706 (192kHz 4times OS)
As there are no devices on the market for 192kHz (and no software) IMHO it makes no sense to fill up the enclosure with
expensive stuff, where increase in quality is at least questionable.
BTW the DIR1703 is primarily designed to support DeltaSigma
Converters, and they are too questionable. I believe the only
alternate choice for the PCM1704U-K are the "Ultra Analog"
converters but somewhere must be a point where to stop
putting money into that unit. This extramoney I will put into better
cables speakers....
PS: when changing the resistor from 1k5 to 2k5 the parallelcap
must be changed to 1.5nF (exactly 1.62) to stay at the same
-3dB frequency.
Uli
uli said:
The complete circuit has just this 'controler', the Crystal
receiver and the TEXAS DF1704 and 4x PCM1704.
My last ones were from BURRBROWN, hope the Texas ones
sound equal!
(and some peripherals of coarse
I decided not to use the DF1706, I split the circuit into
4 PCB´s: Left, Right, Digital in, PSU. The digiboard will
be easily exchangeable (CS8414, DF1704, GAL-22V10)
Uli
I also decided not to use such things as AD1896 as the CS8414
and DF1704 are designed for 96kHz. The DF1704 puts out 768kHz
(8 times oversampling) same as DF1706 (192kHz 4times OS)
As there are no devices on the market for 192kHz (and no software) IMHO it makes no sense to fill up the enclosure with
expensive stuff, where increase in quality is at least questionable.
BTW the DIR1703 is primarily designed to support DeltaSigma
Converters, and they are too questionable. I believe the only
alternate choice for the PCM1704U-K are the "Ultra Analog"
converters but somewhere must be a point where to stop
putting money into that unit. This extramoney I will put into better
cables speakers....
PS: when changing the resistor from 1k5 to 2k5 the parallelcap
must be changed to 1.5nF (exactly 1.62) to stay at the same
-3dB frequency.
Uli
The AD1896 can work fine at 96kHz output Fs, the same is for the DF1706. Differences between DF1704 and DF1706 are the 3.3V opertating voltage of the 1706, missing internal pull-up Rs on the 1706 and that the 1706 CAN do 4-times oversampling from 192kHz. But it can still do the 8x OS from 96kHz. And the 1706 is available from TIs sample program
Reason for me to take the AD1896 were the slightly better specs. The only way to use the AD1896 with 192kHz is to use it in slave mode, means that it is clocked from another device.
I do not know if there is a problem with the DIR1703 with this design. I am not aware of diferences to the crystal receiver chips, but would like to know.
I was told that the DIR1703 has good jitter suppression.
-peter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.