Hello,
I've ordered this build from australia and just completed this build. I get no audio on output, whether I use optical or metalic input.
Leds are working properly yellow on input with signal and green when audio is present. However, when I check output with scope I see 270kHz ~18.4V only.
This signal is present with or without any input connected.
Signal is perhaps some remainder of clock signal amplified.
I use 2 separate little transformers, one for 5V another for 2x15V. 5V transformer has additional rectifier and grounds are connected on PSU board.
What can I check to see whether DSD1796 is working or not?
What signals must be present for DSD1796 to start outputing audio?
I've ordered this build from australia and just completed this build. I get no audio on output, whether I use optical or metalic input.
Leds are working properly yellow on input with signal and green when audio is present. However, when I check output with scope I see 270kHz ~18.4V only.
This signal is present with or without any input connected.
Signal is perhaps some remainder of clock signal amplified.
I use 2 separate little transformers, one for 5V another for 2x15V. 5V transformer has additional rectifier and grounds are connected on PSU board.
What can I check to see whether DSD1796 is working or not?
What signals must be present for DSD1796 to start outputing audio?
Have you followed the initial test and turn on proceedures in the Jaycar notes, these are done with out the DAC board connected with the ribbon cable, sometimes these cables do not crimp correctly, have you buzzed them for continuity. Is the op voltage of 18.4v DC or AC, if it is DC one of you 15v is not connected, if it is AC you may have blown the DAC. Have you orientated the 3.3vdc reg correctly.
Do you have 5vdc on the DAC board?
Do you have +- 15vdc on the DAC Board?
Do you have +- 15vdc on each of the filter chips?
Are they orientated correctly?
Please give us some information to establish where you are!
Do you have 5vdc on the DAC board?
Do you have +- 15vdc on the DAC Board?
Do you have +- 15vdc on each of the filter chips?
Are they orientated correctly?
Please give us some information to establish where you are!
Yes, I've followed initial test and Input board is behaving as writen in docs.
I've buzzed ribbon cables, and the're working fine.
Voltage on each filter capacitor next to opamp is 29.8V, I've checked orientation and this seems to be okay as well.
Measuring +-15V rail didn't show any AC on it. DAC board uses 5V regulator and voltage going to DSD chip is 5V.
Strange thing is, that I have always around 285kHz ~4,43V signal present on output.
An externally hosted image should be here but it was not working when we last tested it.
I had not received 8n2 capacitors with this kit, so I've used SMD 8n2 capacitors, which are soldered on bottom. I've also changed 10uF capacitor for tantalum, which has better ESR. Output capacitor 22p was changed to 100p. These changes however, should not affect audio output.
I can see serial data comming out of DIR9001 chip to pin 5 of DSD1796, but DAC is not generating any signal.
What else can I check?
I was in contact with designer of this DAC, and after few emails the problem was found in 8n2 capacitors which I bought (because these were not provided in kit) and capacity was way off. 0.02nF or something like that.
I've bought new 8n2 SMD , and now DAC seems to perform better.
Local store does not have such values of capacitors even in ceramics.
I thought of building this DAC but decided against it
I thought of building this DAC but once I looked at the PCB and grounding, I decided against it. Nicholas, it's a great minimum-cost design but I think Silicon Chip magazine is cheaping out with their PCB constraints.
It's a single-sided PCB with no ground fill. There's also a ground loop between the PSU and ribbon cable grounds. The foot of ribbon cable on the DAC IC's digital inputs leads to RFI/EMI along with overshoot on the signals. I could go on, but I'm having trouble seeing 24-bit performance due to the physical layout.
I thought of building this DAC but once I looked at the PCB and grounding, I decided against it. Nicholas, it's a great minimum-cost design but I think Silicon Chip magazine is cheaping out with their PCB constraints.
It's a single-sided PCB with no ground fill. There's also a ground loop between the PSU and ribbon cable grounds. The foot of ribbon cable on the DAC IC's digital inputs leads to RFI/EMI along with overshoot on the signals. I could go on, but I'm having trouble seeing 24-bit performance due to the physical layout.
While ground fill can theoretically help a little bit by reducing the size of current loops, it isn't really that critical. I think the performance speaks for itself. As stated in later articles, once I had built a device to generate a proper, clean 24 bit digital sine wave source, the THD+N (1kHz, 20Hz-20kHz bandwidth) was measured at <0.0006%. That's barely any higher than I get if I connect the Audio Precision's sine wave generator output directly to its analysis input.
I would have rather designed a double sided board with SMD components and such to get slightly better results but it's actually pretty good despite the limitations. It helps that this is not a power amplifier so we're not talking about much AC current flow, so the magnetic radiation is quite low.
I don't think so unless the DSD1796's DGND is internally connected to AGND. Unfortunately the documentation doesn't specify. Of course if it is connected internally it would be possible to cut the DGND track. Even if so, I don't think it will make any real difference.
There's far more high frequency noise coming from the delta-sigma DAC architecture than is coupling in from the digital signal cable. Much of it is filtered out and that which remains doesn't seem to hurt performance.
I don't think any 24 bit DAC has 145dB of dynamic range. In reality they mostly achieve 18-20 bit performance. I certainly haven't seen any DACs in the price range of this kit which perform better. It may be possible to get something better but I think it's probably academic.
prairiemystic, there are a couple of other things I forgot to add.
Firstly, if you want to get much better performance than this design achieves, I think you will need to use a better DAC chip. They claim the DSD1796 can do THD+N of 0.0005%. Since I measured <=0.0006% with our design, we've pretty much matched that.
The SNR we got isn't quite as good as they claim (IIRC, 113dB A-weighted vs. 123dB A-weighted) so it might be possible to improve on that. Simply increasing the output level would probably help in that respect.
The better DAC chips from TI are pretty much all significantly more expensive than the DSD1796, which is why I picked that particular one. In the end, it more than met our goals (<0.002% THD+N, SNR of at least 100dB).
Why don't you take my schematics, modify them as you see fit and produce your own double sided board? You can put the DIR9001 next to the DSD1796 if you like to reduce any EMI from the serial audio stream and clocks. You're welcome to re-use the front end if you want or design your own.
If you want to do that and send me your design I can ask for permission to build an example and test it on the Audio Precision to see how it compares to the one published. I think that would be an interesting exercise. I would like to design such a board myself but don't really have time at the moment.
Firstly, if you want to get much better performance than this design achieves, I think you will need to use a better DAC chip. They claim the DSD1796 can do THD+N of 0.0005%. Since I measured <=0.0006% with our design, we've pretty much matched that.
The SNR we got isn't quite as good as they claim (IIRC, 113dB A-weighted vs. 123dB A-weighted) so it might be possible to improve on that. Simply increasing the output level would probably help in that respect.
The better DAC chips from TI are pretty much all significantly more expensive than the DSD1796, which is why I picked that particular one. In the end, it more than met our goals (<0.002% THD+N, SNR of at least 100dB).
Why don't you take my schematics, modify them as you see fit and produce your own double sided board? You can put the DIR9001 next to the DSD1796 if you like to reduce any EMI from the serial audio stream and clocks. You're welcome to re-use the front end if you want or design your own.
If you want to do that and send me your design I can ask for permission to build an example and test it on the Audio Precision to see how it compares to the one published. I think that would be an interesting exercise. I would like to design such a board myself but don't really have time at the moment.
Hi nvinen,
First, you did excellent in the magazine article and the design, which was obviously a lot of work, including the remote control and balanced output card too. The tested THD+N is very low. The fact that you performed these measurements is a welcome relief from the amateur ventures out here.
I frown upon two aspects of the project's construction and what I would've liked to see is more attention to high-frequency design techniques, instead of the single-sided PC board with no ground fill and the ribbon cable between the digital audio receiver and DAC board.
Although yes, this is not a power amplifier and AC circulating currents are quite low in a DAC design, they are an order of magnitude greater in frequency.
As an example, the DIR9001 receiver outputs the DAC master clock SCKO at 512*fs or (I guess) 24.576MHZ, which then goes through about a foot a ribbon cable to the DAC board. Looking at this one signal, which easily has harmonics extending past 100MHz, it does not behave like DC or audio. The RF circulating currents for this clock signal: from the DIR9001, thru the ribbon cable, to the DAC board and then return to the DIR9001... and you have a large loop there.
You get transmission line effects occurring on the ribbon cable signals (try o'scope each end) which causes a lot of overshoot and undershoot on square waves. Anything past Vcc+0.3V and GND-0.3V on these signals gets clamped by the DAC IC's substrate diodes which pollutes (mainly) the analog section of the DAC IC. You mention more noise coming out of the delta-sigma architecture "to 60MHz" and this is what I mean, as a contributor.
Looking at DAC IC manfacturer's evaluation boards, please note Analog Devices (AD1955), Asahi Kasei (AKM4396), ESS Technology (Sabre ES9008) use of termination resistors in their DAC designs on the PCM signals to alleviate this problem- even thought they have very short PCM/clock signal routing, maybe a few cm traces.
I think the need for low-impedance grounds is common knowledge in electronics. A ground fill helps and a (double-sided) ground layer helps more. Also looking at DAC IC manfacturer's app notes and evaluation boards PC board design, I find all are two-sided with ground fills (and the occasional 4-layer pcb) with mention of the need for a ground plane and fill. Here it's just a mod to the PCB design.
How all this affects fidelity is a huge discussion thread all of it's own. We're assuming that THD+N is the only metric for fidelity which was the case a couple decades ago with amplifier manufacturers heaping on the negative feedback to achieve lowest numbers on equipment that ended up sounding poor.
I am doing PCB layout and schematics for a few DAC chips and can include a DSD1796 card too. But for argument's sake- if a new PC board was made and achieved slightly lower THD+N, then what? I'm not competing here, it's the tiny design tweaks that have larger impacts on fidelity that I grumble about.
First, you did excellent in the magazine article and the design, which was obviously a lot of work, including the remote control and balanced output card too. The tested THD+N is very low. The fact that you performed these measurements is a welcome relief from the amateur ventures out here.
I frown upon two aspects of the project's construction and what I would've liked to see is more attention to high-frequency design techniques, instead of the single-sided PC board with no ground fill and the ribbon cable between the digital audio receiver and DAC board.
Although yes, this is not a power amplifier and AC circulating currents are quite low in a DAC design, they are an order of magnitude greater in frequency.
As an example, the DIR9001 receiver outputs the DAC master clock SCKO at 512*fs or (I guess) 24.576MHZ, which then goes through about a foot a ribbon cable to the DAC board. Looking at this one signal, which easily has harmonics extending past 100MHz, it does not behave like DC or audio. The RF circulating currents for this clock signal: from the DIR9001, thru the ribbon cable, to the DAC board and then return to the DIR9001... and you have a large loop there.
You get transmission line effects occurring on the ribbon cable signals (try o'scope each end) which causes a lot of overshoot and undershoot on square waves. Anything past Vcc+0.3V and GND-0.3V on these signals gets clamped by the DAC IC's substrate diodes which pollutes (mainly) the analog section of the DAC IC. You mention more noise coming out of the delta-sigma architecture "to 60MHz" and this is what I mean, as a contributor.
Looking at DAC IC manfacturer's evaluation boards, please note Analog Devices (AD1955), Asahi Kasei (AKM4396), ESS Technology (Sabre ES9008) use of termination resistors in their DAC designs on the PCM signals to alleviate this problem- even thought they have very short PCM/clock signal routing, maybe a few cm traces.
I think the need for low-impedance grounds is common knowledge in electronics. A ground fill helps and a (double-sided) ground layer helps more. Also looking at DAC IC manfacturer's app notes and evaluation boards PC board design, I find all are two-sided with ground fills (and the occasional 4-layer pcb) with mention of the need for a ground plane and fill. Here it's just a mod to the PCB design.
How all this affects fidelity is a huge discussion thread all of it's own. We're assuming that THD+N is the only metric for fidelity which was the case a couple decades ago with amplifier manufacturers heaping on the negative feedback to achieve lowest numbers on equipment that ended up sounding poor.
I am doing PCB layout and schematics for a few DAC chips and can include a DSD1796 card too. But for argument's sake- if a new PC board was made and achieved slightly lower THD+N, then what? I'm not competing here, it's the tiny design tweaks that have larger impacts on fidelity that I grumble about.
Hi nvinen,
I second that - it's a really nice project, thank you, and I've purchased the short form kit from Jaycar (and thrown away all the crappy components;-) so I can build it and compare to my own design of a few years ago based on Cirrus CS8420 and CS4392 (it's always nice to have an alternative source to try when you feel like a change...). Also how lucky we are to have the designer answering our questions in a public forum.
While I hesitate to add any criticism, there are a couple of aspects of the design that look as though they could be improved upon from an armchair perspective (ie without the benefit of trying and measuring), so I'd be very grateful to hear your thoughts on the following if you have a chance.
To me the R values used in the output filter are far too low, resulting in the I-V stage opamp output (either an OPA134 or an NE5534) being loaded with about 200 ohms. This seems likely to compromise the distortion performance. Furthermore, a large 2n2 cap is than placed across the output opamp with only 100 ohms isolation. According to Douglas Self's published measurements, the 5534's distortion starts increasing significantly below about 1k loading, and at 220 ohms is quite mediocre. The same goes for OPA134 only more so, with the manufacturer's curves showing 2k as the minimum load before distortion shoots up. Although the published THD specs don't indicate a problem, they also don't tell you anything about how it will sound, and an easier load for the opamp outputs should help to minimise the generation of distortion per se and hence can only improve the sound quality.
I found the design equations for this filter type in a Cirrus application note: AN48. The only trick is that they have used two caps (C2 in the App note) to ground where your cct used a single cap between + and - sides. Simply use half the capacitance for the single cap configuration and the equations are correct. I plugged the equations into a spreadsheet and set the normalised pole locations for a Bessel type (just a preference) set at Fc = 50.5kHz. After some playing with R and C values I've come up with a set of values that should load the opamps much less, and might give improved distortion measurements and perhaps sound quality. These are also far easier cap values to source in quality polypropylene types such as FKP2:
C2 = 4n7 (was 27n)
C5 = 2n2 (were 8n2)
R3 = 240R (were 180R)
R1 = R4 = 2k0 (were 220 and 200R)
I simulated this using Circuitmaker and the filter looks correct, also nicer than when using the original values as they produce a slight peak before rolloff (probably Butterworth).
Secondly, IMO the SPDIF input amplifier cct could do with a transformer to provide galvanic isolation, which I plan to do (having obtained a DA103C from Farnell). My question though is why there is a 300 ohm input resistor when you would want the input Z to be as close to 75 ohms as possible. Is there any particular reason for the use of this value? Is the input Z of the 'U04 not very high when used as such?
Thirdly re the ground loop question, most mixed signal device app notes recommend joining digital and analogue grounds underneath the actual device as opposed to at the PSU for best noise performance. I don't have enough experience with trying it other ways to say any more though. I think I'll use a separate transformer etc to generate the 5V digital supply in my unit anyway, so I will experiment with where the grounds are joined.
Lastly, I just wanted to agree with prairiemystic on the need for using transmission line principles when transmitting high speed digital signals over cable and connectors. Horowitz and Hill cover this topic nicely in their popular text. Normally a cheap and adequate method used is just a resistor (~22 - 47 ohms) in series with the outputs before the connector, but if you really want to do it properly the inputs at the other end can be terminated with zobel newtorks etc. As these are clock signals, having clean egdes is probably quite important. Unfortunately there doesn't seem to be an easy way of modifying the existing PCB to add series Rs - oh well.
By the way, I'd be very happy to offer my (once completed) unit for measurement with the new filter values if you would like to (though you've probably got better things to do!) since I don't have the means to measure distortion that low myself and I work not far from the SC office in Brookvale.
I second that - it's a really nice project, thank you, and I've purchased the short form kit from Jaycar (and thrown away all the crappy components;-) so I can build it and compare to my own design of a few years ago based on Cirrus CS8420 and CS4392 (it's always nice to have an alternative source to try when you feel like a change...). Also how lucky we are to have the designer answering our questions in a public forum.
While I hesitate to add any criticism, there are a couple of aspects of the design that look as though they could be improved upon from an armchair perspective (ie without the benefit of trying and measuring), so I'd be very grateful to hear your thoughts on the following if you have a chance.
To me the R values used in the output filter are far too low, resulting in the I-V stage opamp output (either an OPA134 or an NE5534) being loaded with about 200 ohms. This seems likely to compromise the distortion performance. Furthermore, a large 2n2 cap is than placed across the output opamp with only 100 ohms isolation. According to Douglas Self's published measurements, the 5534's distortion starts increasing significantly below about 1k loading, and at 220 ohms is quite mediocre. The same goes for OPA134 only more so, with the manufacturer's curves showing 2k as the minimum load before distortion shoots up. Although the published THD specs don't indicate a problem, they also don't tell you anything about how it will sound, and an easier load for the opamp outputs should help to minimise the generation of distortion per se and hence can only improve the sound quality.
I found the design equations for this filter type in a Cirrus application note: AN48. The only trick is that they have used two caps (C2 in the App note) to ground where your cct used a single cap between + and - sides. Simply use half the capacitance for the single cap configuration and the equations are correct. I plugged the equations into a spreadsheet and set the normalised pole locations for a Bessel type (just a preference) set at Fc = 50.5kHz. After some playing with R and C values I've come up with a set of values that should load the opamps much less, and might give improved distortion measurements and perhaps sound quality. These are also far easier cap values to source in quality polypropylene types such as FKP2:
C2 = 4n7 (was 27n)
C5 = 2n2 (were 8n2)
R3 = 240R (were 180R)
R1 = R4 = 2k0 (were 220 and 200R)
I simulated this using Circuitmaker and the filter looks correct, also nicer than when using the original values as they produce a slight peak before rolloff (probably Butterworth).
Secondly, IMO the SPDIF input amplifier cct could do with a transformer to provide galvanic isolation, which I plan to do (having obtained a DA103C from Farnell). My question though is why there is a 300 ohm input resistor when you would want the input Z to be as close to 75 ohms as possible. Is there any particular reason for the use of this value? Is the input Z of the 'U04 not very high when used as such?
Thirdly re the ground loop question, most mixed signal device app notes recommend joining digital and analogue grounds underneath the actual device as opposed to at the PSU for best noise performance. I don't have enough experience with trying it other ways to say any more though. I think I'll use a separate transformer etc to generate the 5V digital supply in my unit anyway, so I will experiment with where the grounds are joined.
Lastly, I just wanted to agree with prairiemystic on the need for using transmission line principles when transmitting high speed digital signals over cable and connectors. Horowitz and Hill cover this topic nicely in their popular text. Normally a cheap and adequate method used is just a resistor (~22 - 47 ohms) in series with the outputs before the connector, but if you really want to do it properly the inputs at the other end can be terminated with zobel newtorks etc. As these are clock signals, having clean egdes is probably quite important. Unfortunately there doesn't seem to be an easy way of modifying the existing PCB to add series Rs - oh well.
By the way, I'd be very happy to offer my (once completed) unit for measurement with the new filter values if you would like to (though you've probably got better things to do!) since I don't have the means to measure distortion that low myself and I work not far from the SC office in Brookvale.
Sc dac
owdeo
There are 3 Aussie DIYAudio members (2 in Sydney) who can verify that it sounds far better with the less draconian filtering as shown in the Burr Brown (TI) datasheet for the DSD1792, which is pin compatible . 100pf is all that is needed at the output.
The almost flat soundstage between the speakers is vastly improved.
SandyK
owdeo
There are 3 Aussie DIYAudio members (2 in Sydney) who can verify that it sounds far better with the less draconian filtering as shown in the Burr Brown (TI) datasheet for the DSD1792, which is pin compatible . 100pf is all that is needed at the output.
The almost flat soundstage between the speakers is vastly improved.
SandyK
Last edited:
Hi SandyK,
Thanks, that's interesting to hear. Did you guys have an opportunity to properly A-B test two units, one original and the other modified? Have your group used the exact values on p35 of the PCM1792A datasheet? The filter shown is only single pole, so you would have shorted a set of resistors and left out the single cap if so?
I'm building mine at the moment, though I'm holding off on the output filters in case nvinen has a chance to respond, as I'm keen to know whether SC considered this issue or whether they just did a bit of the old application note cloning...they mention in the aritcle having tried various opamps and finding the OPA134 best for THD, but I reckon this conclusion could be different if higher filter resistance values had been used. I've never really been satisfied with the sound of the '134, so will probably use 5534s in mine anyway. I ended up using two OPA627s and and an AD797 in my own DAC design, but these really were stupidly expensive...
The higher resistor values will probably reduce the SNR slightly, but I think the compromise needs to be made more in favour of lower distortion, so I can live with that.
Thanks, that's interesting to hear. Did you guys have an opportunity to properly A-B test two units, one original and the other modified? Have your group used the exact values on p35 of the PCM1792A datasheet? The filter shown is only single pole, so you would have shorted a set of resistors and left out the single cap if so?
I'm building mine at the moment, though I'm holding off on the output filters in case nvinen has a chance to respond, as I'm keen to know whether SC considered this issue or whether they just did a bit of the old application note cloning...they mention in the aritcle having tried various opamps and finding the OPA134 best for THD, but I reckon this conclusion could be different if higher filter resistance values had been used. I've never really been satisfied with the sound of the '134, so will probably use 5534s in mine anyway. I ended up using two OPA627s and and an AD797 in my own DAC design, but these really were stupidly expensive...
The higher resistor values will probably reduce the SNR slightly, but I think the compromise needs to be made more in favour of lower distortion, so I can live with that.
Sc dac
owdeo
I sent you a message through the system with a couple of links attached which shows what we have done.
If you didn't receive the message, email me and I will give you both links. We haven't had a chance to AB units in Sydney, but we all started out with the original and found it quite poor as far as soundstage and separation goes. The other Sydney DIY Audio member has listened to mine .In fact, in comparison with DACs that we already both had it was quite poor. (e.g. modded X-DAC V3 with external linear regulated PSU.) There are also another couple in our group in the U.K.and Spain who are also constructing them. After our reports, the chap in Spain went straight to the DSD1792 filtering and 100pF at the output, as well as polyprops as appropiate, especially at the DIR9001 pin 22. (4.7nF and 68nF) He hasn't done the rest of our changes yet, but says his is already clearly outperforming his CD player.
SandyK
P.S.
The analogue section filtering is now virtually identical to that used in my X-DAC V3 which uses the DSD1792 and DIR1703E.
It also benefitted from fitting polyprops at the DIR filter . Nicholas doesn't believe they matter, but it could even come down to much closer value tolerances.
owdeo
I sent you a message through the system with a couple of links attached which shows what we have done.
If you didn't receive the message, email me and I will give you both links. We haven't had a chance to AB units in Sydney, but we all started out with the original and found it quite poor as far as soundstage and separation goes. The other Sydney DIY Audio member has listened to mine .In fact, in comparison with DACs that we already both had it was quite poor. (e.g. modded X-DAC V3 with external linear regulated PSU.) There are also another couple in our group in the U.K.and Spain who are also constructing them. After our reports, the chap in Spain went straight to the DSD1792 filtering and 100pF at the output, as well as polyprops as appropiate, especially at the DIR9001 pin 22. (4.7nF and 68nF) He hasn't done the rest of our changes yet, but says his is already clearly outperforming his CD player.
SandyK
P.S.
The analogue section filtering is now virtually identical to that used in my X-DAC V3 which uses the DSD1792 and DIR1703E.
It also benefitted from fitting polyprops at the DIR filter . Nicholas doesn't believe they matter, but it could even come down to much closer value tolerances.
Last edited:
Thanks for letting me know SandyK - I haven't received your message so will send you an email.
I'm using polypropylene caps for the PLL filter too. Their benefits over polyester may not be very important, but it can't hurt and can only lead to better PLL performance, which could impact sound quality if using a jittery source such as a low to medium range DVD player.
My view on these sort of component choices is that if it has the potential to affect performance, I'll gladly spend a little extra. The difference may be marginal, but as the project I am constructing is a one off I'd rather just try and make it work as well as possible in the first place. I would never waste money on boutique "audiophile" components though.
As an electronics engineer, hobbyist and musician, I try to find a balance between the "right wing" approach that says a simple THD+N measurement is the only requirement to prove that the design is good and the "left wing" side of audiophile madness that is more akin to Scientology and where practically anything (except one's own perception) can be said to have an affect on sound quality. The right wing approach is just daft - equivalent to designing a car, building it, measuring its power, mpg etc and declaring it finished, but never actually driving it. The left wing side is nuts too though. Changing something (often more than one thing at a time) then listening again and declaring that quality "x" has improved (and furthermore concluding that said modification will always improve sound quality in any design) without recourse to at least going back to the original is self delusional. On the occasions when I've bothered to compare two equivalent designs, that I was sure sounded very different and one clearly superior, properly, using a relay switchover where I didn't know which switch position was which design, I found it impossible to reliably say which was consistently better. I will only try modifying something if I think it could have an effect in a measurable way, though I often don't have the means to measure it... I trust my ears but only so far...
I'm using polypropylene caps for the PLL filter too. Their benefits over polyester may not be very important, but it can't hurt and can only lead to better PLL performance, which could impact sound quality if using a jittery source such as a low to medium range DVD player.
My view on these sort of component choices is that if it has the potential to affect performance, I'll gladly spend a little extra. The difference may be marginal, but as the project I am constructing is a one off I'd rather just try and make it work as well as possible in the first place. I would never waste money on boutique "audiophile" components though.
As an electronics engineer, hobbyist and musician, I try to find a balance between the "right wing" approach that says a simple THD+N measurement is the only requirement to prove that the design is good and the "left wing" side of audiophile madness that is more akin to Scientology and where practically anything (except one's own perception) can be said to have an affect on sound quality. The right wing approach is just daft - equivalent to designing a car, building it, measuring its power, mpg etc and declaring it finished, but never actually driving it. The left wing side is nuts too though. Changing something (often more than one thing at a time) then listening again and declaring that quality "x" has improved (and furthermore concluding that said modification will always improve sound quality in any design) without recourse to at least going back to the original is self delusional. On the occasions when I've bothered to compare two equivalent designs, that I was sure sounded very different and one clearly superior, properly, using a relay switchover where I didn't know which switch position was which design, I found it impossible to reliably say which was consistently better. I will only try modifying something if I think it could have an effect in a measurable way, though I often don't have the means to measure it... I trust my ears but only so far...
I'm going to get a kit for one of these DACs in the next couple of weeks. What I think would be a useful alteration is to have an in-out socket after the input selector so I can use a Behringer DSP equaliser on whatever source is selected.
An update in case anyone's interested:
Having completed the DAC and doing some extended listening, I've settled on OPA604 for all three opamps. I tried these after a tip from a friend who used to work for Burr-Brown, and have to agree that they sound better than anything else I've tried. The humble NE5534 would be my second choice. The OPA134 is boring - it sounds clearly tonally wrong with orchestral brass sections and has anaemic bass.
Comparing the sound to the analogue output of my Rotel RCD965BX CD player to the DAC (modified with filter values and opamps per previous discussions) driven from its digital output, there is very little difference. The DAC is a little smoother and less conjested on complex choral works, but the Rotel has better bass and a slightly warmer overall presentation.
In short I'd say if you're looking for really high-end sound, don't bother with this one. I suspect the DAC IC is only intended for cheap DVD players and the implementation is pretty average. Having said that, the user interface and input switching is beautifully done - all credit to SC for this.
Having completed the DAC and doing some extended listening, I've settled on OPA604 for all three opamps. I tried these after a tip from a friend who used to work for Burr-Brown, and have to agree that they sound better than anything else I've tried. The humble NE5534 would be my second choice. The OPA134 is boring - it sounds clearly tonally wrong with orchestral brass sections and has anaemic bass.
Comparing the sound to the analogue output of my Rotel RCD965BX CD player to the DAC (modified with filter values and opamps per previous discussions) driven from its digital output, there is very little difference. The DAC is a little smoother and less conjested on complex choral works, but the Rotel has better bass and a slightly warmer overall presentation.
In short I'd say if you're looking for really high-end sound, don't bother with this one. I suspect the DAC IC is only intended for cheap DVD players and the implementation is pretty average. Having said that, the user interface and input switching is beautifully done - all credit to SC for this.
owdeo
You will find that this DAC is capable of incredible results with plenty of attention to the PSU areas, among other things.
Good to hear from you again. PM or email me if you would like to see what several other Aussies and Northern Hemisphere guys have done.
Regards
Alex.
Hi Alex - likewise - will email you soon.
I must confess to never having tried any exotic regulator designs in any line level projects, and I guess it's about time I did. I guess I'm basing my conclusion on the fact that my own DAC design (based on CS8420 and CS4392) sounds noticeably better (to unbiased ears as well as my own 😉 and only has bog standard LM317 PSUs. Perhaps the upsampling has a lot to do with this however...
I wonder also about current output vs voltage output DACs (the CS4392 is voltage out). You would think using an external IV opamp running on +/-15V rails would be superior to a voltage output DAC where (presumably) the IV opamp is internal and running from the shared 5V supply only. Perhaps using a discrete transimpedance amp is the key, per the recent Pass article. I don't understand the obsession with MOSFETs and JFETs though. For a start, where do you buy genuine audio jfets? Hitachi/Renensas don't seem to make them anymore and no one sells them except the designers who promote them as sounding superior with the usual generalisations. If I ever get around to trying this it'll be using bipolars if practical.
Cheers!
I must confess to never having tried any exotic regulator designs in any line level projects, and I guess it's about time I did. I guess I'm basing my conclusion on the fact that my own DAC design (based on CS8420 and CS4392) sounds noticeably better (to unbiased ears as well as my own 😉 and only has bog standard LM317 PSUs. Perhaps the upsampling has a lot to do with this however...
I wonder also about current output vs voltage output DACs (the CS4392 is voltage out). You would think using an external IV opamp running on +/-15V rails would be superior to a voltage output DAC where (presumably) the IV opamp is internal and running from the shared 5V supply only. Perhaps using a discrete transimpedance amp is the key, per the recent Pass article. I don't understand the obsession with MOSFETs and JFETs though. For a start, where do you buy genuine audio jfets? Hitachi/Renensas don't seem to make them anymore and no one sells them except the designers who promote them as sounding superior with the usual generalisations. If I ever get around to trying this it'll be using bipolars if practical.
Cheers!
IMHO a lot of it is due to implementation of the dac you're using.
With a current output you have the decision on the conversion process, whether it's opamp, discrete of valve.
Voltage output the choice has already been made.
As for promoting there own products........
Well the circuit diagrams are there, usually with discussion about various versions.
Etch your own pcbs, and try them.
With a current output you have the decision on the conversion process, whether it's opamp, discrete of valve.
Voltage output the choice has already been made.
As for promoting there own products........
Well the circuit diagrams are there, usually with discussion about various versions.
Etch your own pcbs, and try them.
Last edited:
Perhaps so.
Try them - how? I don't have or know where I can buy any genuine Hitachi jfets...that was my point.
As for etching boards, no thanks. Why bother when various places have an online service and will make 'em professionally for not much more than the material costs (not to mention mess) of DIYing. I suspect that was SC's poor (IMHO) reasoning for not doing the DAC boards as proper double-sided affairs with SMD bypassing caps etc - they were trying to cater for the odd DIY board etcher and those that think SMDs are too hard to solder by hand.
Try them - how? I don't have or know where I can buy any genuine Hitachi jfets...that was my point.
As for etching boards, no thanks. Why bother when various places have an online service and will make 'em professionally for not much more than the material costs (not to mention mess) of DIYing. I suspect that was SC's poor (IMHO) reasoning for not doing the DAC boards as proper double-sided affairs with SMD bypassing caps etc - they were trying to cater for the odd DIY board etcher and those that think SMDs are too hard to solder by hand.
- Status
- Not open for further replies.