maxlorenz said:Dear Soundcheck,
Now that you are happy with the sound, try a low noise (low Temp. coef.) resistor instead of the Riken which are über-noisy...(I was once fond of them, though).
With that and the TeddyReg based tower PS I bet you will get more detail...or more HF noise? Who knows...
With a TVC you could connect a balanced output DAC DC coupled to the amp...even DIDAC on balanced mode
I have postponed the square wave analysis of mine...it seems RC is a must with some amps...thanks Bernhard for the reminder.
Cheers,
M
Any low noise resistor in mind? Caddock MK 132?
I mean Riken is known for its good "sound" signature.
I do agree that a balanced DAC and a balanced AMP might be the
best approach. However. You still have a TVC in the loop.
Look-up this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=119917
No more DAC, no more AMP and TVC discussions! PCM2PWM This is IMO the way to go.
Cheers
Hi Ims,
As far as I know, settling time will be problematic with clock dependent circuits-> lack of lock-in (or delay).
I power first the DAC and then the "transport". One of my DACs takes eons to lock-in
With analog circuits the only feature that I detect is a "soft start" effect. Not bad I think
Rise time could be made faster by changing R3-R4 values (please refer to PFM link) but this will affect the performance of the filter, I believe. Contact the inventor for more details.
Hi Soundcheck,
Oh no! Not another amp project
About R, I'm not expert and you have already two advices: Bernhard's and John's (15ppm). Any 25 or lower ppm resistor should make a significant difference (I did not say improvement because you may still prefer the "tube like" Riken sound )
Cheers,
M
As far as I know, settling time will be problematic with clock dependent circuits-> lack of lock-in (or delay).
I power first the DAC and then the "transport". One of my DACs takes eons to lock-in
With analog circuits the only feature that I detect is a "soft start" effect. Not bad I think
Rise time could be made faster by changing R3-R4 values (please refer to PFM link) but this will affect the performance of the filter, I believe. Contact the inventor for more details.
Hi Soundcheck,
Oh no! Not another amp project
About R, I'm not expert and you have already two advices: Bernhard's and John's (15ppm). Any 25 or lower ppm resistor should make a significant difference (I did not say improvement because you may still prefer the "tube like" Riken sound
)Cheers,
M
Thanks Mauricio,
Your experiences about using the TeddyRegs in the analog circuit of the DAC is interesting too. I will not change the R3-R4 since it will reduce filtering capability of the regulator. I agree with you about 'soft start' is fairly bargain for the enhancement using TeddyRegs.
I will try it to supply my Lesha's style DAC (I am in exploring the dual TDA1541A matters before soldering DI8 kit. It's about five months since I got the kit from John ). But I'm not in plan to change the DI8's regulators. I will use the stock regulator as John designed them. I am willing to hear what exactly John has heard from his unique design. I believe that is the ultimate design for 8xTDA1541A so far.
Many thanks for sharing.
- ims -
Your experiences about using the TeddyRegs in the analog circuit of the DAC is interesting too. I will not change the R3-R4 since it will reduce filtering capability of the regulator. I agree with you about 'soft start' is fairly bargain for the enhancement using TeddyRegs.
I will try it to supply my Lesha's style DAC (I am in exploring the dual TDA1541A matters before soldering DI8 kit. It's about five months since I got the kit from John
). But I'm not in plan to change the DI8's regulators. I will use the stock regulator as John designed them. I am willing to hear what exactly John has heard from his unique design. I believe that is the ultimate design for 8xTDA1541A so far.Many thanks for sharing.
- ims -
UTOS2 availability
Hi luvdunhill,
I plan to publish it on this thread when available.
UTOS2 will probably be simplified (cascaded digital audio receivers). The PCM2706 will now output a SPDIF signal on the DATA output (pin 9 pulled high):
USB (PCM2706) > SPDIF > High-speed opto-coupler > CS8416 (RXP0) > I2S
TOSLINK / COAX > CS8416 (RXP1) > I2S.
Since a dual opto-coupler is used, the second opto-coupler can be used for USB detection / automatic source selection.
This simplifies UTOS2 circuit, and only requires a single high-speed opto-coupler chip (for USB galvanic insulation), this also reduces total coupling capacitance.
By cascading both USB and SPDIF receiver, the CS8416 PLL will reduce PCM2706 jitter.
The CS8416 power-down circuit is no longer required as the CS8416 is now always active, this also results in faster locking.
Hi luvdunhill,
I plan to publish it on this thread when available.
UTOS2 will probably be simplified (cascaded digital audio receivers). The PCM2706 will now output a SPDIF signal on the DATA output (pin 9 pulled high):
USB (PCM2706) > SPDIF > High-speed opto-coupler > CS8416 (RXP0) > I2S
TOSLINK / COAX > CS8416 (RXP1) > I2S.
Since a dual opto-coupler is used, the second opto-coupler can be used for USB detection / automatic source selection.
This simplifies UTOS2 circuit, and only requires a single high-speed opto-coupler chip (for USB galvanic insulation), this also reduces total coupling capacitance.
By cascading both USB and SPDIF receiver, the CS8416 PLL will reduce PCM2706 jitter.
The CS8416 power-down circuit is no longer required as the CS8416 is now always active, this also results in faster locking.
Hi maxlorenz,
Teddyregs are ideal for removing ripple voltage, I tried a modified HV Teddyreg in the tubeps module (+200V), they work so well that I could remove all gyrators from the tubedif modules. I didn't use a pre-regulator, but integrated a zener reference in the teddyreg, so I only have one series regulator darlington.
I have been doing some extensive testing with TDA1543, it seems they got a "sweet spot" between 4 and 4.63V! this is within permissible operating voltage (3 ... 8V). Since you are using active I/V, it's pointless to use high supply voltages, as this is only required when using passive I/V and high ac voltages at the TDA1543 output.
You could try this quickly by using 7805 regulators, and connecting a schottky diode in series, this will give approx. 4.6V.
It seems TDA1543 gives cleanest sound at lower chip temperatures.
TDA154X chips have current steering inputs, and basically only require small I2S signal amplitude. I tested attenuators / level shifters (1.5V DC with 700mV signal amplitude), these will significantly reduce I2S switching noise being injected in the TDA1543 substrate. I used separate attenuators / level shifters for each input.
For DATA and WS inputs:
5.6K series resistor, 3.3K pull-up (VCC), and 1K pull-down (GND)
For BCK inputs:
One level shifter, 100R series resistor, 330R pull-up, 100R pull-down, each TDA1543 clock input connected through 22R.
TDA1543 works very well with passive I/V, provided ac voltage at DAC output doesn't exceed approx. 480mV @ 4.63V, and I/V resistors are NOT connected to GND, but to a clean reference voltage. The TDA1543 Vref pin should best be left unconnected as it controls the bias current, so it could very easily inject unwanted noise / interference, especially when Vref pins of multiple chips are connected in parallel.
When using differential passive I/V, the ac voltages from multiple DAC outputs can be added, creating high ac output voltages with low gain, and not exceeding 480mV ac voltage at each DAC output. When low signal amplitudes are used with passive I/V (TDA1541A), it's advisable to use differential operation to cancel common noise / interference.
I have been testing OP-amp, discrete JFET and tube amplifiers with passive I/V (TDA1543 & TDA1541A). The following I/V resistor values (at typical DAC operating voltage) ensure low THD, despite +/- 25mV output compliance:
TDA1543: 220R / 484mVpp / 171mV rms
TDA1541A: 80R / 320mVpp / 113mV rms
It seems OP-amps with high NFB create "steps" jumping from one "stable" operating point to the next, when complex signals are applied, this gives the impression of lower resolution / grain.
Note that this has little to do with THD, and probably won't show up on (standardized) distortion tests. It's just a typical dynamic response to small signal changes. It could be possible that this becomes particularly problematic when interacting with DAC stepped output signal.
Simple discrete JFET or tube amplifiers with local or low NFB, seem to pass these complex signals more fluently, giving the impression of higher resolution.
I personally start to prefer tube differential amplifiers, and it's not because of typical "warm" tube sound caused by high (even harmonic) distortion. Recent THD measurements of the modified tubedif modules revealed approx. 0.002% THD @ 1 KHz (measured with Marconi distortion meter).
Teddyregs are ideal for removing ripple voltage, I tried a modified HV Teddyreg in the tubeps module (+200V), they work so well that I could remove all gyrators from the tubedif modules. I didn't use a pre-regulator, but integrated a zener reference in the teddyreg, so I only have one series regulator darlington.
I have been doing some extensive testing with TDA1543, it seems they got a "sweet spot" between 4 and 4.63V! this is within permissible operating voltage (3 ... 8V). Since you are using active I/V, it's pointless to use high supply voltages, as this is only required when using passive I/V and high ac voltages at the TDA1543 output.
You could try this quickly by using 7805 regulators, and connecting a schottky diode in series, this will give approx. 4.6V.
It seems TDA1543 gives cleanest sound at lower chip temperatures.
TDA154X chips have current steering inputs, and basically only require small I2S signal amplitude. I tested attenuators / level shifters (1.5V DC with 700mV signal amplitude), these will significantly reduce I2S switching noise being injected in the TDA1543 substrate. I used separate attenuators / level shifters for each input.
For DATA and WS inputs:
5.6K series resistor, 3.3K pull-up (VCC), and 1K pull-down (GND)
For BCK inputs:
One level shifter, 100R series resistor, 330R pull-up, 100R pull-down, each TDA1543 clock input connected through 22R.
TDA1543 works very well with passive I/V, provided ac voltage at DAC output doesn't exceed approx. 480mV @ 4.63V, and I/V resistors are NOT connected to GND, but to a clean reference voltage. The TDA1543 Vref pin should best be left unconnected as it controls the bias current, so it could very easily inject unwanted noise / interference, especially when Vref pins of multiple chips are connected in parallel.
When using differential passive I/V, the ac voltages from multiple DAC outputs can be added, creating high ac output voltages with low gain, and not exceeding 480mV ac voltage at each DAC output. When low signal amplitudes are used with passive I/V (TDA1541A), it's advisable to use differential operation to cancel common noise / interference.
I have been testing OP-amp, discrete JFET and tube amplifiers with passive I/V (TDA1543 & TDA1541A). The following I/V resistor values (at typical DAC operating voltage) ensure low THD, despite +/- 25mV output compliance:
TDA1543: 220R / 484mVpp / 171mV rms
TDA1541A: 80R / 320mVpp / 113mV rms
It seems OP-amps with high NFB create "steps" jumping from one "stable" operating point to the next, when complex signals are applied, this gives the impression of lower resolution / grain.
Note that this has little to do with THD, and probably won't show up on (standardized) distortion tests. It's just a typical dynamic response to small signal changes. It could be possible that this becomes particularly problematic when interacting with DAC stepped output signal.
Simple discrete JFET or tube amplifiers with local or low NFB, seem to pass these complex signals more fluently, giving the impression of higher resolution.
I personally start to prefer tube differential amplifiers, and it's not because of typical "warm" tube sound caused by high (even harmonic) distortion. Recent THD measurements of the modified tubedif modules revealed approx. 0.002% THD @ 1 KHz (measured with Marconi distortion meter).
Hi folks.
I just installed these
http://www.rhopointcomponents.com/products.asp?recnumber=182
as passive I/V stage on my DDDAC towers.
What a difference a little resistor can make. The changes I wouldn't call subtle. Seems to be much faster and more detailed. E.g. Metal (cymbals etc.) is kind of shimmering now. It is obvious now that a guitar string is a nylon string, which wasn't that obvious before. Before I haven't even thought about it, Today I just heard it. -- I'll stop here. It's every time the same thing and you ask yourself how much more information you can find on this 16/44.1 material.
Of course the slight harshness need to settle a bit.Interesting how many people a raving about Riken. I should have talked to you guys earlier.
Very interesting to see how much this el cheapo 1543 can be pimped. It seems that the DAC has even more potential reading ecdesigns post from today. Looking forward to try some more tweaks.
Thanks again for giving directions.
Cheers
I just installed these
http://www.rhopointcomponents.com/products.asp?recnumber=182
as passive I/V stage on my DDDAC towers.
What a difference a little resistor can make. The changes I wouldn't call subtle.
Seems to be much faster and more detailed. E.g. Metal (cymbals etc.) is kind of shimmering now. It is obvious now that a guitar string is a nylon string, which wasn't that obvious before. Before I haven't even thought about it, Today I just heard it. -- I'll stop here. It's every time the same thing and you ask yourself how much more information you can find on this 16/44.1 material.Of course the slight harshness need to settle a bit.Interesting how many people a raving about Riken. I should have talked to you guys earlier.
Very interesting to see how much this el cheapo 1543 can be pimped. It seems that the DAC has even more potential reading ecdesigns post from today. Looking forward to try some more tweaks.
Thanks again for giving directions.
Cheers
Hi -ecdesigns-
Thanks for your excellent comments about TDA1543's function. I will try to digest them on the following months
The 8.5V PS comes from Doede's own research about sweet spot for passive I/V in his DDDAC and was meant to be used on Soundcheck's.
My DI16 uses 6V TeddyRegs for the DACs. Should I check sound with 4.6V in it?
The sound is brilliant now, with a very pleasing "omni" character...
I fully agree.
TeddyRegs should make very clean supplies. My scope only goes down to mV. Maybe you saw further...
Hi Soundcheck,
I'm glad that theory and practice went OK together
Some may think that I only rely on mystic approaches to DIY
My own 8.5V Teddyreg for my DDDAC is ready and I should try it later this week...
(LT1085 as a pre-Reg; I made a mistake on previous post)
Cheers,
M
Thanks for your excellent comments about TDA1543's function. I will try to digest them on the following months
The 8.5V PS comes from Doede's own research about sweet spot for passive I/V in his DDDAC and was meant to be used on Soundcheck's.
My DI16 uses 6V TeddyRegs for the DACs. Should I check sound with 4.6V in it?
The sound is brilliant now, with a very pleasing "omni" character...
I fully agree.
TeddyRegs should make very clean supplies. My scope only goes down to mV. Maybe you saw further...
Hi Soundcheck,
I'm glad that theory and practice went OK together
Some may think that I only rely on mystic approaches to DIY
My own 8.5V Teddyreg for my DDDAC is ready and I should try it later this week...
(LT1085 as a pre-Reg; I made a mistake on previous post)
Cheers,
M
Yes,
keep the temperature low, about 40deg
is from importance, when higher vb of 8V to 8.5V supplyed...
I designed 16x parallel dac with a sandwitch of dacs and aluminium laminates but has to be very big...
this common piggy back design we can find on the net
does not work at all...
the temperature is running so high...
I tested last summer in Belgrade when we have more than 40deg
outside, I run the dac for 2 days, and dac was stable...
cheers
keep the temperature low, about 40deg
is from importance, when higher vb of 8V to 8.5V supplyed...
I designed 16x parallel dac with a sandwitch of dacs and aluminium laminates but has to be very big...
this common piggy back design we can find on the net
does not work at all...
the temperature is running so high...
I tested last summer in Belgrade when we have more than 40deg
outside, I run the dac for 2 days, and dac was stable...
cheers
Zoran said:
If you're talking about the DDDAC Towers, I wouldn't say
that they're not working at all !?!?
I never measured the temperature, for sure they get warmer
than 40°C. However, to me my DAC just sounds great ( Of course - Not knowing how much better it could get.
)My current benchmark a "Benchmark DAC1" is not working any better. With the latest tweaks on my DAC, the DAC1 falls easily behind my current setup.
What you realize is that the tower needs 30minutes to stabilze. But then it's IMO OK.
Hi ecdesigns and others.
I'd like to come back to your Rref/Vref reflections. I am wondering if my current setup is OK.
I presume that I need to switch also Rref to a high quality R since
it is directly involved in the current delivery for Rload.
(Please correct me if I 'am wrong here)
I did some calculations just to verify my current setup.
To summarize my thoughts:
I run 24 DACs at 8.4Vps.
I got a 2k4Ohm (24*100R) Rload and 3,85V Vload.
This is giving me an Iload of 1.6mA.
Total current would be Idac + Iload = 1,15mA+1.6mA=2,75mA
The internal Gain of the DAC is 2.
That means I need 2,75mA/2=1.375mA provided through Rref.
Rref=Uref/Iref=2,2V/1,375mA=1600R
for a 24 DAC setup I'd need somewhat 1600R/24=66Ohm (available 70R could be probably used as Rref.)
Currently I got a 2k8Ohm (24*120R DDDAC default) as Rref.
No idea it that difference is impacting the result ?
But 2k8 is obviously not really close to my calculated value of 1k6.
Do you guys have an opinion about it? Did I made a mistake here?
It was mentioned to use a current source instead of Rref.
How would a current source look like for above setup?
THX for your advise
Cheers
I'd like to come back to your Rref/Vref reflections. I am wondering if my current setup is OK.
I presume that I need to switch also Rref to a high quality R since
it is directly involved in the current delivery for Rload.
(Please correct me if I 'am wrong here)
I did some calculations just to verify my current setup.
To summarize my thoughts:
I run 24 DACs at 8.4Vps.
I got a 2k4Ohm (24*100R) Rload and 3,85V Vload.
This is giving me an Iload of 1.6mA.
Total current would be Idac + Iload = 1,15mA+1.6mA=2,75mA
The internal Gain of the DAC is 2.
That means I need 2,75mA/2=1.375mA provided through Rref.
Rref=Uref/Iref=2,2V/1,375mA=1600R
for a 24 DAC setup I'd need somewhat 1600R/24=66Ohm (available 70R could be probably used as Rref.)
Currently I got a 2k8Ohm (24*120R DDDAC default) as Rref.
No idea it that difference is impacting the result ?
But 2k8 is obviously not really close to my calculated value of 1k6.
Do you guys have an opinion about it? Did I made a mistake here?
It was mentioned to use a current source instead of Rref.
How would a current source look like for above setup?
THX for your advise
Cheers
I think this is the best tutorial on how to calculate I/V and Ref resistors for TDA1543
http://www.diyaudio.com/forums/showthread.php?postid=173483#post173483
Ciao
Andrea
http://www.diyaudio.com/forums/showthread.php?postid=173483#post173483
Ciao
Andrea
Hi Andrea.
THX. This is mainly matching my above calculations.
However. It doesn't explain everything to me. (I am not really experienced in electronic design and its practical implications)
Why do I have a different Rref, according to DDDAC standard, then the one I calculated above. What's the impact of changing it? To me the impact of changing "this and that" is not really clear.
Of course I could try it.
ecdesigns was bringing up to skip Rref to avoid additional noise . And the data sheet is also saying it is optional. What's happening with Iload in this case?
Cheers
THX. This is mainly matching my above calculations.
However. It doesn't explain everything to me. (I am not really experienced in electronic design and its practical implications)
Why do I have a different Rref, according to DDDAC standard, then the one I calculated above. What's the impact of changing it? To me the impact of changing "this and that" is not really clear.
Of course I could try it.
ecdesigns was bringing up to skip Rref to avoid additional noise . And the data sheet is also saying it is optional. What's happening with Iload in this case?
Cheers
soundcheck said:
The fact that you have 3.85V bias, tells me that the combination of Rref and Rload are in optimum. no need to change it
I am working on a small add on, to totally avoid the Rref (so it can be left open). But that will take some more time... It will automatically adjust the output at 3,85 Volt, regardless how many towers you operate. It is just one small module which you connect to the PCB analog outputs and cut the Rrefs on the tower boards. So upgrading easy... Testing is going on
In the meantime, I have heard some dddac users report back, that if Rref is paralleled with a capitor, sound improves somewhat
I heard the same about using a lm317 as current source. Room to play, but play carefully, if you pump too much current with a lm317 set up you might heat up the chips too much....
doede
soundcheck said:
than the whole biasing becomes unbalanced for your set up. You won't be able to find a good bias setup anymore with optimum results. I cannot comment the noise thing, but as said earlier, it seems to have influence, but do not expect a new dac sound as result
My calculations are based on measured (FFT and listening) results for best set up with the given powersupply voltage and maximising output voltage, at lowest THD...
That is not 100% the same as using datasheet info, which is based on theory
doede
anbello said:
This is the straight forward theoretical datasheet aproach. As said above, the optimum lies a bit off center.... the 1.2 and 1.8V borders are not so black/white. you need to stay away from the upper limit more than from the bottom limit. That is why I arrive at 3.85 bias and 1.6VRMS output with 0.3% distortion (mostly d2 which is nice ...)
doede
Doede.
Good to see you over here.
THX for the hints.
For the time being I'll switch the Rref with the type of resistor I am using for Rload - mentioned 3-4 posts ago, It's working extremely well compared to Riken. Riken slows things heavily down.
What cap did the guys apply to Rref ?
Interesting - your new approach on the output. Are you considering
not to connect Rload to ground - as it was mentioned by ecdesigns -
which seems to be an interesting idea.
Cheers
Good to see you over here.
THX for the hints.
For the time being I'll switch the Rref with the type of resistor I am using for Rload - mentioned 3-4 posts ago, It's working extremely well compared to Riken. Riken slows things heavily down.
What cap did the guys apply to Rref ?
Interesting - your new approach on the output. Are you considering
not to connect Rload to ground - as it was mentioned by ecdesigns -
which seems to be an interesting idea.
Cheers
dddac said:
Did you mean 0,03 % ? as stated in the datasheet for 0 dB, and a mean 8% for low level signals...
Or does the passive I/V raise hell ?
Now if somebody claims to hear the ca. 0,001 % of a TVC with such a DAC, this is ridiculous...
I did a listening test without the 10k termination on my TVC, it gives a disturbing treble boost, only 0,6 dB on a meter, but very noticeable.