Hi,
Okay, not to wanting to speak out of turn, but while we all wait with baited breath for K+H to overcome the paralysis stage following parenthesis, there is another way:
Saffier's Sabre DAC in Granite Case | Diy HiFi Supply
This is "Zero Feedback" and "Discrete", though it does use tubes and coupling capacitors and hence is in the view of K+H totally and irrevocably flawed.
Ciao T
Okay, not to wanting to speak out of turn, but while we all wait with baited breath for K+H to overcome the paralysis stage following parenthesis, there is another way:
Saffier's Sabre DAC in Granite Case | Diy HiFi Supply
This is "Zero Feedback" and "Discrete", though it does use tubes and coupling capacitors and hence is in the view of K+H totally and irrevocably flawed.
Ciao T
Hi,
Not my project BTW, if you look around you can probably find a shop that can make you such a granite case. I am currently trying to decide between glass and light coloured stone for my speakers.
The OPA1632 is not quite a traditional Op-Amp.
It's fully differential folded cascode circuit is very close to how I would I design a discrete circuit (except I'd use J-Fets on the input and single ended Sziklai Pairs with J-Fet as first transistor as output, but that is personal preference).
The only thing I dislike a lot with OPA1632 is that Burr Brown omitted bringing the four junctions out that would allow setting the gain without recourse to looped feedback. If this was present I would use this chip a lot.
On the positive side, the DC gain is fairly low for amplifiers designed to use feedback, so feedback factors are quite low and the open-loop bandwidth is somewhere between 20-90KHz based on interpolating from the datasheet, compared to often single Hz figure for normal Op-Amp's. Also, while 0,00005% THD for unity gain clearly make use of the 60-80dB NFB, inherent distortion is still quite low at something like 0.05% THD open loop.
All this means it does not sound much like "normal" Op-Amp's. It is in fact one of the "Op-Amp's on my list that are quite nice for audio applications.
Shame, as said, BB did not give us a chance to use this IC open loop, setting the gain with local degeneration.
Ciao T
Not my project BTW, if you look around you can probably find a shop that can make you such a granite case. I am currently trying to decide between glass and light coloured stone for my speakers.
even though it uses the dreaded op-amps in the output stageopa1632's it still sounds pretty good imo
The OPA1632 is not quite a traditional Op-Amp.
It's fully differential folded cascode circuit is very close to how I would I design a discrete circuit (except I'd use J-Fets on the input and single ended Sziklai Pairs with J-Fet as first transistor as output, but that is personal preference).
The only thing I dislike a lot with OPA1632 is that Burr Brown omitted bringing the four junctions out that would allow setting the gain without recourse to looped feedback. If this was present I would use this chip a lot.
On the positive side, the DC gain is fairly low for amplifiers designed to use feedback, so feedback factors are quite low and the open-loop bandwidth is somewhere between 20-90KHz based on interpolating from the datasheet, compared to often single Hz figure for normal Op-Amp's. Also, while 0,00005% THD for unity gain clearly make use of the 60-80dB NFB, inherent distortion is still quite low at something like 0.05% THD open loop.
All this means it does not sound much like "normal" Op-Amp's. It is in fact one of the "Op-Amp's on my list that are quite nice for audio applications.
Shame, as said, BB did not give us a chance to use this IC open loop, setting the gain with local degeneration.
Ciao T
The ESS DAC has now been playing in my system, but unfortunately something mysterious went wrong with our analog stage, resulting in some strange distortion in one channel, so a bit of time will pass by locating and fixing the problem.
But no big deal, it sounded just like a burned transistor.
I did get a listen to the original board though, which is based upon op-amps from Analog Devices, the AD797 which is Analogs top of the line flagship chip.
But no big deal, it sounded just like a burned transistor.
I did get a listen to the original board though, which is based upon op-amps from Analog Devices, the AD797 which is Analogs top of the line flagship chip.
Thanks T, very interesting post as usual
I'd imagine stone would be perfect for speaker enclosures, won't be much fun when it comes to moving them about though
If its decided that the Sabre dac is used in this project I really hope it comes with or at least has an easy option to use quality regulation, going by recent results I wouldn't go with anything less than decent shunts with these dacs.
BTW these are the only newer type dacs I've liked so far, the reference for me always used to be the classic TDA1541A
Cheers,
Leo
Hi,
If you want Followers....
First, consider running open loop, with extra resistors on the supplies to pull enough current to lower the emitter impedance, for example 100mA drawn per emitter follower (I know it's wasteful and non-green - but this about SQ) will have around 0.3 Ohm Emitter impedance and in the case of most current"hybrid" voltage output DAC's will only need two or three such sections.
Secondly, compensate the Vbe thermal drift.
Third, make the reference voltage not from any shunt, but simply with a passive network from the pre-regulated (LM317) voltage and use lowish value resistors and very large capacitors on the base, one for each transistor.
Given the current we are drawing in this case I'd probably instead make a supply where the DC path is low impedance but the AC is dealt with by a "AC Current Source" and then a closed loop Sziklai Pair Transistor to act as AC only shunt. The result is that DC side things are left to regulators that are good at it and for AC we make basically a rather interesting discrete linear regulator.
Voltage losses can be under 1V and current losses in the 5 - 20mA range. Lower cutoff and LF noise depends mainly on the values of the capacitors used, noise can be low as that of the Transistor in the AC Shunt.
Really tricky people may even make the whole thing zero feedback, feed-forward only. This is what I use on clock oscillators.
For hints look up Wenzel Associates "Finesse Noise Shunt" and then think about bootstrapping the series resistor...
Ciao T
If you want Followers....
First, consider running open loop, with extra resistors on the supplies to pull enough current to lower the emitter impedance, for example 100mA drawn per emitter follower (I know it's wasteful and non-green - but this about SQ) will have around 0.3 Ohm Emitter impedance and in the case of most current"hybrid" voltage output DAC's will only need two or three such sections.
Secondly, compensate the Vbe thermal drift.
Third, make the reference voltage not from any shunt, but simply with a passive network from the pre-regulated (LM317) voltage and use lowish value resistors and very large capacitors on the base, one for each transistor.
Given the current we are drawing in this case I'd probably instead make a supply where the DC path is low impedance but the AC is dealt with by a "AC Current Source" and then a closed loop Sziklai Pair Transistor to act as AC only shunt. The result is that DC side things are left to regulators that are good at it and for AC we make basically a rather interesting discrete linear regulator.
Voltage losses can be under 1V and current losses in the 5 - 20mA range. Lower cutoff and LF noise depends mainly on the values of the capacitors used, noise can be low as that of the Transistor in the AC Shunt.
Really tricky people may even make the whole thing zero feedback, feed-forward only. This is what I use on clock oscillators.
For hints look up Wenzel Associates "Finesse Noise Shunt" and then think about bootstrapping the series resistor...
Ciao T
I know the "Finesse Noise Shunt" from Wenzel. I never did use it, mainly because it involves a series resistor, resulting in high Zout a frequencies below the "cut-off" of the clean up circuit. But we may consider using this circuit. For the clock-circuit it may be very interesting!
How abou de-coupling on this circuit? Will you need another (small) series resistor between the clean-up circit and the de-coupling caps?? Else the clean-up will "fight" the de-coupling caps.
How abou de-coupling on this circuit? Will you need another (small) series resistor between the clean-up circit and the de-coupling caps?? Else the clean-up will "fight" the de-coupling caps.
O man!!! To many Einstein's here. I've read over 30 pages and you are going nowhere with this project.
I just want to build myself a very simple and "musical" dac.
Please, can anyone of you to provide me a clear schematics and a pcb schematics. I don't know to design the pcb schematics. Please help me with the pcb schematics. Otherwise I know a lot of dac schematics but without the pcb.
Thanks!!!
I just want to build myself a very simple and "musical" dac.
Please, can anyone of you to provide me a clear schematics and a pcb schematics. I don't know to design the pcb schematics. Please help me with the pcb schematics. Otherwise I know a lot of dac schematics but without the pcb.
Thanks!!!
The question is, if you just want to build some random DAC or something special. If you are into some random DAC, just pick something else.
This project is dedicated to the R&D of a DAC that has that something extra. And that takes some time. Sorry! But we will NOT compromise on performance. There are plenty of designs like that out there.....
This project is dedicated to the R&D of a DAC that has that something extra. And that takes some time. Sorry! But we will NOT compromise on performance. There are plenty of designs like that out there.....
Hi,
Depends upon the definition of "high".
Anyway, this circuit is actually open-loop and does not do anything to the resistors effective impedance.
The impedance after the Wenzel "Finess" Circuit is the series resistor plus any capacitors to ground. As I find that all else being equal having a real capacitor control impedance usually is better than having an active device I don't mind.
For "constant current draw" circuits (like clocks - okay, they have spikes at the clock frequency, but local descoupling caps should take care of that anyway as it is in 10's of MHz) the resistance does not matter.
But the circuit does not HAVE to be open loop.
Nor does the resistor have to be a resistor only, it could be boot-strapped. In this case the closed loop circuit can have VERY low impedance AND very high noise rejection, for use with loads that are not constant current.
And as the circuit is still "AC Only" it can be made so that any noisy local zenners or references are avoided and we do not add ton's of noise back, after we just struggled to remove it.
Ciao T
Depends upon the definition of "high".
Anyway, this circuit is actually open-loop and does not do anything to the resistors effective impedance.
The impedance after the Wenzel "Finess" Circuit is the series resistor plus any capacitors to ground. As I find that all else being equal having a real capacitor control impedance usually is better than having an active device I don't mind.
For "constant current draw" circuits (like clocks - okay, they have spikes at the clock frequency, but local descoupling caps should take care of that anyway as it is in 10's of MHz) the resistance does not matter.
But the circuit does not HAVE to be open loop.
Nor does the resistor have to be a resistor only, it could be boot-strapped. In this case the closed loop circuit can have VERY low impedance AND very high noise rejection, for use with loads that are not constant current.
And as the circuit is still "AC Only" it can be made so that any noisy local zenners or references are avoided and we do not add ton's of noise back, after we just struggled to remove it.
Ciao T
It is very ok this dac. I know is special. Just send to me your own schematics and pcb schematics. It's ok if this dac have an opamp analog stage. I just discovered the magic of OPA2604. It's just ok!
Thanks!
Thanks a lot for the recommendation. By the way Thorsten, I am very happy to see you back on the forum. I did a lots of reading and learned a lot from your site a long time ago, regarding speaker building. I still recall an excellent Watt Puppy DIY construction you did.
Thanks Pano! As you know I have Lundahls and wanted to experiment with something that is 1:1, vs my 1:4.
Interesting....!! How do you bootstrap the series resistor? I may be blind on this... please enlighten me....
This Wenzel "clean up circuit has been in my mind for years, but never really got any further.
Sorry, but if you don't care about the sonic performance, but just want some random DAC ASAP, this surely isn't a project for you. Pick the first finished project you find, and that may make you happy.
In contrast to many others, we will not launch this project as soon as we have something that will play. The project will be launched when we are satisfied with the performance. That's how High End audio should be.... But unfortunately VERY few designers will invest the time needed. Normally$$$ are more important. And that's why most audiogear is rushed through R&D.
power supply impedance and decoupling capacitors
Regarding the power supply impedance and micro-resonances of decoupling capacitors, I find it puzzling that film caps are preferred over ceramic ones. Are people confusing power supply decoupling with the audio signal path? They are two different beasts.
I've always used two-three decoupling caps. Bulk capacitance is from low ESR electrolytics or tantalums, midband from film types and HF from ceramics.
If you take an ASRC or even a 25MHz oscillator module, the IC's current has a strong 25MHz (clock) component, with a 3rd or 5th harmonic to that you end up needing decoupling to 100MHz. Film caps crap out in the 1-10MHz region.
The WM8805 PLL has a freq. multiplier needing a 90-100MHz frequency (!) and I'm doubtful of a 0.22uF at that frequency. I note some chip maker use a 0.1uF and 0.01uF to decouple digital logic. At the very least a ferrite bead or inductor in series with power to this chip.
Regarding the power supply impedance and micro-resonances of decoupling capacitors, I find it puzzling that film caps are preferred over ceramic ones. Are people confusing power supply decoupling with the audio signal path? They are two different beasts.
I've always used two-three decoupling caps. Bulk capacitance is from low ESR electrolytics or tantalums, midband from film types and HF from ceramics.
If you take an ASRC or even a 25MHz oscillator module, the IC's current has a strong 25MHz (clock) component, with a 3rd or 5th harmonic to that you end up needing decoupling to 100MHz. Film caps crap out in the 1-10MHz region.
The WM8805 PLL has a freq. multiplier needing a 90-100MHz frequency (!) and I'm doubtful of a 0.22uF at that frequency. I note some chip maker use a 0.1uF and 0.01uF to decouple digital logic. At the very least a ferrite bead or inductor in series with power to this chip.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.