power DAC
One interesting aspect of this DIY R-2R DAC is that you can make the output level higher than those R-2R monolithic DACs allow. Monolithic 24-bit R-2R DAC, probably only PCM1704, should fight the almost infinitesimal thermal noise to make the a few least significant bits make sense. Their dynamic range is capped at the voltage supply of the DAC. If the discrete DAC has output range of +/- 20V or 50V, that may really be able to achieve true 24-bit resolution. I don't know how Wadia designed their 'Power DAC', but coupling this DAC with high-current I/V stage and power output stage seems to be a very interesting approach.
One interesting aspect of this DIY R-2R DAC is that you can make the output level higher than those R-2R monolithic DACs allow. Monolithic 24-bit R-2R DAC, probably only PCM1704, should fight the almost infinitesimal thermal noise to make the a few least significant bits make sense. Their dynamic range is capped at the voltage supply of the DAC. If the discrete DAC has output range of +/- 20V or 50V, that may really be able to achieve true 24-bit resolution. I don't know how Wadia designed their 'Power DAC', but coupling this DAC with high-current I/V stage and power output stage seems to be a very interesting approach.
Konnichiwa,
If you use 0.1% Tolerance resistors with a tempco of 50ppm/K and you let the DAC reach operating temperature (without oven) the long term stability is going to be pretty good and I believe PSU Noise and switch channel resistance will become the limiting factors.
Now you can get analog switches that are specified to a a few ohm variation in channel resistance which with a 2K R2R DAC makes the switches the limiting factor already, both thermally and in absolute terms.
The short term stability of a given reference filtered by suitable Film Capacitors and buffered by an Op-Amp with good DC and AC Precision (OPA627 would seem okay) should be perfect enough to exceed 20 Bit accuracy, so a semi-discrete 24-Bit DAC with suitably well specified AND THERMALLY COUPLED switches should be fine to give > 20bit effective dynamic range.
Layout however becomes critical in such circuitry and once the speeds go up significantly (say 8 X oversampling @ 192KHz for arguments sake - even though Non-Os @ 192KHz seems an excellent option) life becomes really interesting.
Given that enough monolithic converters exist and have existed since the mid 1980's that plumb the limits of 16 Bit performance and few sources exceed realistically a 16-Bit dynamic range it is understandable that few people have bothered to do so.
MSB has made quite interesting semi-discrete DAC Modules, however the results of their implementation seem no better than reasonable high quality, but not "SOTA" and received a somewhat lukewarm review by Stereophile.
http://stereophile.com/digitalsourcereviews/799/
Given what I have seen so far and that I feel that exploiting existing monlithic devices is rarely done to any significant degree I feel that for me personally at least I shall retain that approach for a while. Once the last Multibit converters have disappeared from the market and if PCM remains the main standard for quality digital audio I shall remeber my theorising here and bother to try something on.
My experiences desining high resolution "discrete" AD circuitry (and associated analogue conditioning & scaling) for a mere < 16Bit relaible performance (admitedly down to DC and over extended temperature ranges and the like as required for industrial applications) in medium scale production was enough to scream at me "stay away if you can....!".
Sayonara
Lgrau said:
If you use 0.1% Tolerance resistors with a tempco of 50ppm/K and you let the DAC reach operating temperature (without oven) the long term stability is going to be pretty good and I believe PSU Noise and switch channel resistance will become the limiting factors.
Now you can get analog switches that are specified to a a few ohm variation in channel resistance which with a 2K R2R DAC makes the switches the limiting factor already, both thermally and in absolute terms.
The short term stability of a given reference filtered by suitable Film Capacitors and buffered by an Op-Amp with good DC and AC Precision (OPA627 would seem okay) should be perfect enough to exceed 20 Bit accuracy, so a semi-discrete 24-Bit DAC with suitably well specified AND THERMALLY COUPLED switches should be fine to give > 20bit effective dynamic range.
Layout however becomes critical in such circuitry and once the speeds go up significantly (say 8 X oversampling @ 192KHz for arguments sake - even though Non-Os @ 192KHz seems an excellent option) life becomes really interesting.
Given that enough monolithic converters exist and have existed since the mid 1980's that plumb the limits of 16 Bit performance and few sources exceed realistically a 16-Bit dynamic range it is understandable that few people have bothered to do so.
MSB has made quite interesting semi-discrete DAC Modules, however the results of their implementation seem no better than reasonable high quality, but not "SOTA" and received a somewhat lukewarm review by Stereophile.
http://stereophile.com/digitalsourcereviews/799/
Given what I have seen so far and that I feel that exploiting existing monlithic devices is rarely done to any significant degree I feel that for me personally at least I shall retain that approach for a while. Once the last Multibit converters have disappeared from the market and if PCM remains the main standard for quality digital audio I shall remeber my theorising here and bother to try something on.
My experiences desining high resolution "discrete" AD circuitry (and associated analogue conditioning & scaling) for a mere < 16Bit relaible performance (admitedly down to DC and over extended temperature ranges and the like as required for industrial applications) in medium scale production was enough to scream at me "stay away if you can....!".
Sayonara
designing a discrete dac is actually, off the topic for a diyer, asuumming that quality is demanded... there is huge theory to get to MASTER before even attempting to embark on sth like this.. why? it isnt possible to achieve the performance of the most simple ic dac out there this way, because a discrete dac means more space occupied, larger signal paths, inductances and capacities to be taken care of , that in the frequencies a dac works, must be taken account of.. how is one going to achieve better signal integrity @ 20Mhz when having say 2 cm of path (+c+l) than a solid silicon square of 9 mm2?.. consider that designing a board to utilise a ic dac is a whole art (not as simple as it looks).. building a discrete dac that will play and be good enough, is 90dB over this....
Thank you all for giving you thoughts about this DIY project. Seems that there is a possiblity, that a farly good perfoming dac can be made, but the big quistion is if it is worth the effort.
You make a very good point there. I havent got the knownledge to say how much the larger signal path will influent on the design.
Keeping the high frequency path as short as possibly, should be prefered.
This is also one of the advantages of the wieghted current source dac as I se it. the path holding high frequency signals can be minimized. The i/v conversion can be done right after the switch. This limits the high frequency signal path down to a size near the physical size of the switches. 🙂
The digital logic path will of coarse be much larger, but I think it is possible to minimize the influence from the logic.. But maybee i'm a bit naive
costiss said:
You make a very good point there. I havent got the knownledge to say how much the larger signal path will influent on the design.
Keeping the high frequency path as short as possibly, should be prefered.
This is also one of the advantages of the wieghted current source dac as I se it. the path holding high frequency signals can be minimized. The i/v conversion can be done right after the switch. This limits the high frequency signal path down to a size near the physical size of the switches. 🙂
The digital logic path will of coarse be much larger, but I think it is possible to minimize the influence from the logic.. But maybee i'm a bit naive
Is it really as high as 20MHz? I am not saying that discrete DAC module is easy to design. But I think the fastest R-2R audio DAC has about 800kSample/second conversion speed - PCM1702/1704....
Probably it's easier to take dCS approach of multi-bit delta sigma with dynamic weight averaging, and DIY the DSP code to control the switch attached to resistor networks...
Probably it's easier to take dCS approach of multi-bit delta sigma with dynamic weight averaging, and DIY the DSP code to control the switch attached to resistor networks...
Konnichiwa,
Depends. The sample rate is 44KHz for CD and up to 192KHz for DVD-A. If we oversample DVD 8 Times we have a 1.5MHz Sample Rate and we need to make sure the output of our DAC settles within a small fraction of one sample cycle. If we use non-oversampled CD then we have 44.1KHz.... ;-)
So it really depends on what you want.
Sayonara
jheoaustin said:
Depends. The sample rate is 44KHz for CD and up to 192KHz for DVD-A. If we oversample DVD 8 Times we have a 1.5MHz Sample Rate and we need to make sure the output of our DAC settles within a small fraction of one sample cycle. If we use non-oversampled CD then we have 44.1KHz.... ;-)
So it really depends on what you want.
Sayonara
I don't think costiss mean that the sampling rate is 20mHz, but he's talking about the high frequency signals handled in the dac.
correct me if I'm wrong costiss.
I prefer that the DIY dac sticks to a pretty low maximum sampling frequency, i think 200Khz should be enough.
If we really want no high frequency, signals in our dac, an approach like the : Delta-current linear-interpolation DAC thread can be chosen..
It would be easy to implement in a DIY dac.
The errors due to settling time cannot be negliated. if the settling time is 5ns and the error during the settling time is ½MSB the error equals approx. 1/10000 of the MSB. if the sampling rate is 44Khz.
This error rises as the sampling rate increases.
If a colinear DAC, and a balanced output is used the error is reduced..
correct me if I'm wrong costiss.
I prefer that the DIY dac sticks to a pretty low maximum sampling frequency, i think 200Khz should be enough.
If we really want no high frequency, signals in our dac, an approach like the : Delta-current linear-interpolation DAC thread can be chosen..
It would be easy to implement in a DIY dac.
The errors due to settling time cannot be negliated. if the settling time is 5ns and the error during the settling time is ½MSB the error equals approx. 1/10000 of the MSB. if the sampling rate is 44Khz.
This error rises as the sampling rate increases.
If a colinear DAC, and a balanced output is used the error is reduced..
its high frequencies i speak of... more than this, how is one going to utilise the switches to be soft and fast enough to handle the transitions good enough w/out ringing?.. and 16 current sources of which the last has 32768 (correct me if im wrong) times the current of the first one, are not that easy to make... and because you cant have a really big out current, if you have a max i out of say 65,536 mA, your lowest current source would be 1 uA..
that means thermal noise must be taken into account .. you should at least seal the module in resin...
In my humble opinion...
that means thermal noise must be taken into account .. you should at least seal the module in resin...
In my humble opinion...
costiss said:
Thank you for pointing out some of the critical parts of the design.
Making transitions without ringing can be a big problem, The Delta-current linear-interpolation method will solve this(i think). Short path would minimize this problem, a practical test should be the best way to show how big this problem is..
Resistors with low current noise is avialible, a quick search gave this one :
http://www.vishay.com/docs/31018/cmfind.pdf
current noise 0,1uV/V.
I don't know the current noise in a cheap 1% metalfilm resistor, does anybody have an idea of the range ?
The voltage noise from the opamp controlling the J-Fet in the current source, can "easy" be belov 5uV.
I'm not shure how to calculate the noise in a j-Fet, can't find the noise specificaion for dc, only for ac. Can anyone help ?
The noise in a switch isn't specified, but i also think this noise is non critical.
I'd like to ask another sidetracking question: How high do you gurus here think the DAC reference voltage can go up practically(not DIY practical)? Can it be much higher than 3.3V or 5V?
Konnichiwa,
Well, if you where to use Valves as switched for an R2R ladder I suspect the reference Voltage could be several 100 Volt, even >> 1,000V. Is that practical? Depends on your definition of practical.... ;-)
Using solid state switches and C-Mos logic I still think you can go up to 15V.
Sayonara
jheoaustin said:
Well, if you where to use Valves as switched for an R2R ladder I suspect the reference Voltage could be several 100 Volt, even >> 1,000V. Is that practical? Depends on your definition of practical.... ;-)
Using solid state switches and C-Mos logic I still think you can go up to 15V.
Sayonara
Kuei Yang Wang said:
Warning this a little OT 😀
I just remembered an articel about a DAC with 20bit absolute precision enjoy :
http://www.linear.com.cn/pdf/an86f.pdf
This cannot be used for audio because it is to slow, but it give a good idea of how much 20bit absolut linearity is.
the digital circuit
Any interesting ideas regarding the clock distrubution and control logic layout ?
Haven't thougth much about these issues.
But I like a masteclock placed near the DAC, the cd-drive should be controlled ad a slave from this clock.
It can be done by having a VCXO at the drive, controlled by a pll, this is the most flexible solution, but I alsothink it's the most complex.
The other solution is just feeding the master clock directly to the drive(maybe divided down).
It would be nice to have a fpga to build up the control logic, but I think it also could be build up using logic devices pretty simple. I've no accces to a fpga programmer so I have no choise 🙂
The switches should be controlled very precise, any suggestion on how to acomplish that. Is there a better way than using high speed flip flops ?
Any interesting ideas regarding the clock distrubution and control logic layout ?
Haven't thougth much about these issues.
But I like a masteclock placed near the DAC, the cd-drive should be controlled ad a slave from this clock.
It can be done by having a VCXO at the drive, controlled by a pll, this is the most flexible solution, but I alsothink it's the most complex.
The other solution is just feeding the master clock directly to the drive(maybe divided down).
It would be nice to have a fpga to build up the control logic, but I think it also could be build up using logic devices pretty simple. I've no accces to a fpga programmer so I have no choise 🙂
The switches should be controlled very precise, any suggestion on how to acomplish that. Is there a better way than using high speed flip flops ?
Re: the digital circuit
Konnichiwa,
At a guess, these are in my view two seperate topics and should be separated.
I doubt it. The logic is so simple that it is better off in simple IC's. Use a suitable 3-Dimensional SMD layout and you can have everything extremely compact.
I suspect these will be the best possible solution.
I have had a few thoughts.
I would likely use one seperate module as DAC, all SMD.
This would contain one common reference for both channels and 24-Bit control logic per channel. I suspect it would be well worthwhile making the DAC differential if we do this at all.
I would apply re-clocking having a seperate SMA/SMB clock input and likely have more SMA/SMB sockets on board to cover Data L, Data R (Data L & R normally linked together), LR Clock and Bitclock (alternative would be to use optical receivers for all but clock and using optical fibre for all key signals.
Probably also SMA/SMB sockets for Supply voltages and Analogue (current) outputs.
The PCB would have to be probably minimum 3-Layer SMD, only worthwhile doing if some volume production could be achieved.
With that small DAC PCB the DIY'er would be free to design his own Input Board, to parallel several DAC's, to use primitive oversampling systems with multiple DAC's and the like. Especially for more complex systems with multiple DAC's the Optical Receiver/Transmitter route might be a really good idea, jitter from that should be killed by the reclocker.
Of course, if we make such a DAC using industrial methodes one might also want to make a DIO Board with a Digital PLL and FIFO Buffer plus suitable SMA/SMB and optical output sections. One might want to include up to 4 sets of optical outputs (only one to be filled as standard) with selectable timing on that DIO Board to allow parallel and timeshifted DAC's.
One might even get away using suitable mounting and layout of DAC and DIO to have a DAC module direct next to the DIO and sensing the optical signal through a few mm of air.
Anyway, some thoughts on all of that. A company like LC audio should really pick up that kind of job, they could then combine it with their discrete analogue stage and such....
Sayonara
Konnichiwa,
Lgrau said:
At a guess, these are in my view two seperate topics and should be separated.
Lgrau said:
I doubt it. The logic is so simple that it is better off in simple IC's. Use a suitable 3-Dimensional SMD layout and you can have everything extremely compact.
Lgrau said:
I suspect these will be the best possible solution.
I have had a few thoughts.
I would likely use one seperate module as DAC, all SMD.
This would contain one common reference for both channels and 24-Bit control logic per channel. I suspect it would be well worthwhile making the DAC differential if we do this at all.
I would apply re-clocking having a seperate SMA/SMB clock input and likely have more SMA/SMB sockets on board to cover Data L, Data R (Data L & R normally linked together), LR Clock and Bitclock (alternative would be to use optical receivers for all but clock and using optical fibre for all key signals.
Probably also SMA/SMB sockets for Supply voltages and Analogue (current) outputs.
The PCB would have to be probably minimum 3-Layer SMD, only worthwhile doing if some volume production could be achieved.
With that small DAC PCB the DIY'er would be free to design his own Input Board, to parallel several DAC's, to use primitive oversampling systems with multiple DAC's and the like. Especially for more complex systems with multiple DAC's the Optical Receiver/Transmitter route might be a really good idea, jitter from that should be killed by the reclocker.
Of course, if we make such a DAC using industrial methodes one might also want to make a DIO Board with a Digital PLL and FIFO Buffer plus suitable SMA/SMB and optical output sections. One might want to include up to 4 sets of optical outputs (only one to be filled as standard) with selectable timing on that DIO Board to allow parallel and timeshifted DAC's.
One might even get away using suitable mounting and layout of DAC and DIO to have a DAC module direct next to the DIO and sensing the optical signal through a few mm of air.
Anyway, some thoughts on all of that. A company like LC audio should really pick up that kind of job, they could then combine it with their discrete analogue stage and such....
Sayonara
Re: Re: the digital circuit
You're probably right, a discrete solution, is also the most realistic for most DIY'ers, including me.
Yes, one common reference will be the logic descision, I also think the main powersupplies should be common, to make the R/L side as equal as possible.
By making it diffential, do you mean diffential current output, or a signed magnitude converter ?
I think a diffential current output, is very important, in order to minimize the switching problems.
A signed magnitude DAC, will also help a lot regarding the switching noise/errors, but most around zero.
A 24bit DAC would be nice, but the component price must be considered before implementing a 24 bit solution. Specially if a signed magnitude approach is choosen. But it's prefered in order to have a flexible DAC.
I really likes high speed connections, especially for the critical signals like the clock, SMA/SMB looks like the optimum price/perfomance connection.
BTW here's a high speed connector link.
http://www.microwaves101.com/encyclopedia/connectors.cfm#24mm
I'm a little concerned if it's possible to get the right termination if the clock is placed away from the dac.
All non critical data's might just as well be isolated, in order to minimize ground problems etc. I also like other isolators like the a pulse trafo, or devices like the analogs isolators.
http://www.elfa.se/pdf/56/05655055.pdf
http://www.analog.com/digital_isolator.html
The supply voltages and analogue current shouldn't hold much high frequency energy, so i can't see what a SMA connector can bring here. But i like the connectors though, so maybe we just as well use the here.
I'll answer the rest later 🙂
Kuei Yang Wang said:
You're probably right, a discrete solution, is also the most realistic for most DIY'ers, including me.
Yes, one common reference will be the logic descision, I also think the main powersupplies should be common, to make the R/L side as equal as possible.
By making it diffential, do you mean diffential current output, or a signed magnitude converter ?
I think a diffential current output, is very important, in order to minimize the switching problems.
A signed magnitude DAC, will also help a lot regarding the switching noise/errors, but most around zero.
A 24bit DAC would be nice, but the component price must be considered before implementing a 24 bit solution. Specially if a signed magnitude approach is choosen. But it's prefered in order to have a flexible DAC.
I really likes high speed connections, especially for the critical signals like the clock, SMA/SMB looks like the optimum price/perfomance connection.
BTW here's a high speed connector link.
http://www.microwaves101.com/encyclopedia/connectors.cfm#24mm
I'm a little concerned if it's possible to get the right termination if the clock is placed away from the dac.
All non critical data's might just as well be isolated, in order to minimize ground problems etc. I also like other isolators like the a pulse trafo, or devices like the analogs isolators.
http://www.elfa.se/pdf/56/05655055.pdf
http://www.analog.com/digital_isolator.html
The supply voltages and analogue current shouldn't hold much high frequency energy, so i can't see what a SMA connector can bring here. But i like the connectors though, so maybe we just as well use the here.
I'll answer the rest later 🙂
Re: Re: the digital circuit
The switch, Which I would say is the most critical part, can be made better, if they are stacked, this would give the shortest signal path. But the it would be very nice with a small SMD PCB
Yes it could be a real TDA1541 killer, except for the price 🙂 I like the optical approach, if complex systems should be made. Hope one of these DIY dac sound very good at its own, because a multiple DAC solution, dosn't fit my financial state
If this project is realized, and the sound, is unbeaten by a commercial dac, I can give LC-audio my prototype, if they're interested.
This layout could also be a challenging task for PerAnders 😀
BTW.
Which solution do you like the best ? R2R or weighted current sources. As you might have noticed, I prefer the Weighted current sources DAC.
Kuei Yang Wang said:
The switch, Which I would say is the most critical part, can be made better, if they are stacked, this would give the shortest signal path. But the it would be very nice with a small SMD PCB
Yes it could be a real TDA1541 killer, except for the price 🙂 I like the optical approach, if complex systems should be made. Hope one of these DIY dac sound very good at its own, because a multiple DAC solution, dosn't fit my financial state
If this project is realized, and the sound, is unbeaten by a commercial dac, I can give LC-audio my prototype, if they're interested.
This layout could also be a challenging task for PerAnders 😀
BTW.
Which solution do you like the best ? R2R or weighted current sources. As you might have noticed, I prefer the Weighted current sources DAC.
Re: Re: Re: the digital circuit
Konnichiwa,
Looking purely at the ability to make a reliable circuit that is not excessively complex and expensive R2R easily wins.
Using weighted current sources for 24 Bits WITH AT LEAST 20Bit monotonicity would be not a task I'd feel like attempting. And even after all that effort your resistors will be ultimatly the factor that determines linearity and all.
Even doing it 16-Bit would be highly complex, needing at least 16 Op-Amp's with excellent DC precsition AND AC behaviour, the Fet's/BJT's and so on.
One could use 12pcs 74AC574 and 12pcs octal C-Mos Multiplexer/Switch/Buffer IC's (with qualified and garanteed channel resistance) plus a few glue logic bits (Serieal to parallel conversion 24/32 bit wide), the reference source and SMD R2R ladders (4 ladders) and have a 24-Bit DAC which if we apply 12V as the switched reference voltage will give more than enough differential output to give 2V RMS full scale with good linearity.
Cost should likely be no worse than say the LC Audio Zap Filter if distributed through a normal dealer network, much less if made through Per Anders connections.
Sayonara
Konnichiwa,
Lgrau said:
Looking purely at the ability to make a reliable circuit that is not excessively complex and expensive R2R easily wins.
Using weighted current sources for 24 Bits WITH AT LEAST 20Bit monotonicity would be not a task I'd feel like attempting. And even after all that effort your resistors will be ultimatly the factor that determines linearity and all.
Even doing it 16-Bit would be highly complex, needing at least 16 Op-Amp's with excellent DC precsition AND AC behaviour, the Fet's/BJT's and so on.
One could use 12pcs 74AC574 and 12pcs octal C-Mos Multiplexer/Switch/Buffer IC's (with qualified and garanteed channel resistance) plus a few glue logic bits (Serieal to parallel conversion 24/32 bit wide), the reference source and SMD R2R ladders (4 ladders) and have a 24-Bit DAC which if we apply 12V as the switched reference voltage will give more than enough differential output to give 2V RMS full scale with good linearity.
Cost should likely be no worse than say the LC Audio Zap Filter if distributed through a normal dealer network, much less if made through Per Anders connections.
Sayonara
Re: Re: Re: Re: the digital circuit
To my best knowledge it would be easiest to accomplish high monotonicity by using a weighted current source DAC, especially if the resistors are the ultimately factor in the linearity.
but you're right it's way more complex, than a R2R dac. And this is probably reason enough not to build this.
The cost isn't overwelming can't find any octal spdt switches though.
But a quad switch like the Max4780 seems to be a good switch for the porpuose:
0,3 ohm flatnes and 0,2 ohm match between channels. 7ns Break before make period.
The 1000pcs price is only $1.41
http://pdfserv.maxim-ic.com/en/ds/MAX4780-MAX4784.pdf
Kuei Yang Wang said:
To my best knowledge it would be easiest to accomplish high monotonicity by using a weighted current source DAC, especially if the resistors are the ultimately factor in the linearity.
but you're right it's way more complex, than a R2R dac. And this is probably reason enough not to build this.
The cost isn't overwelming can't find any octal spdt switches though.
But a quad switch like the Max4780 seems to be a good switch for the porpuose:
0,3 ohm flatnes and 0,2 ohm match between channels. 7ns Break before make period.
The 1000pcs price is only $1.41
http://pdfserv.maxim-ic.com/en/ds/MAX4780-MAX4784.pdf
Does anybody have an idea of the noisefloor of a good current source ?
Any good suggestion for a better current source than the one i drawed in the schematic ?
See:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=45096&pagenumber=2
Any good suggestion for a better current source than the one i drawed in the schematic ?
See:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=45096&pagenumber=2
Re: Re: Re: Re: the digital circuit
I suggested earlier in this thread, go to Lavry engineering
http://www.lavryengineering.com/index_html.html
and download manual for DA924.
This dac uses discrete resistor weighted architechture. The
dac is in a temperature controlled oven and it has a self
calibration procedure every time it starts up. All the bits are
set current-wise with precision opamps.
It will not be sign magnitude architecture because that is
patented by BB (TI) but it uses a deglitch circuit run directly from
XO (I think based around OPA627) at the DAC OP.
So even with all this attention to detail and Dan Lavry is one of
the smartest guys in the business, it measures probably not
quite as good as some of the best monolythic DACs. However, it
is completely impervious to jitter artifacts due to the FIFO,
PLL and S+H architecture.
Check stereophile for a review of the hi-end equivalent of the DA924, same architecture inside.
http://www.stereophile.com/digitalsourcereviews/804lavry/
FWIW, stereophile rate it as one of the best but there are also
other daks which are monolythic based that they also rate
as highly.
Cheers,
Terry
Kuei Yang Wang said:
I suggested earlier in this thread, go to Lavry engineering
http://www.lavryengineering.com/index_html.html
and download manual for DA924.
This dac uses discrete resistor weighted architechture. The
dac is in a temperature controlled oven and it has a self
calibration procedure every time it starts up. All the bits are
set current-wise with precision opamps.
It will not be sign magnitude architecture because that is
patented by BB (TI) but it uses a deglitch circuit run directly from
XO (I think based around OPA627) at the DAC OP.
So even with all this attention to detail and Dan Lavry is one of
the smartest guys in the business, it measures probably not
quite as good as some of the best monolythic DACs. However, it
is completely impervious to jitter artifacts due to the FIFO,
PLL and S+H architecture.
Check stereophile for a review of the hi-end equivalent of the DA924, same architecture inside.
http://www.stereophile.com/digitalsourcereviews/804lavry/
FWIW, stereophile rate it as one of the best but there are also
other daks which are monolythic based that they also rate
as highly.
Cheers,
Terry
- Status
- Not open for further replies.