Thanks for the reply, I may be interested in a bare board with these minimum requirements -
Coax RCA (or bare gnd/+ solder pads) to take SPDIF input from cd/streamer
Unbalanced Stereo RCA (or bare gnd/+ solder pads) for analogue output
Volume control connectors for 10k pot (or solder pads)
xv/0/xv power supply input (to connect a rectified power supply)
Any thing else would be a bonus (DC supply, Toslink, USB, balanced output, display etc.)
I am looking to put together an 8 channel dac for a DSP system so 4 boards with this performance would be good (also taking into account the Khadas was better on Coax SPDIF than USB).
One 2-channel ES9038Q2M goes for about $15 in small quantities, so $60 for four chips.
One 8-channel ES9028PRO goes for probably less than $60 in small quantities. In that case there would no need to correct for chip to chip dac output level variations. There might be some other benefits too. One low-jitter MCLK clock can be fed directly into a single dac, etc.
Also, Sharc chips, commonly used for speaker DSP, support I2S without the need for jittery optical or SPDIF connections. Galvanic isolation between the USB (or other input source) and the remaining DSP/DAC/Power-Amp electronics can be provided simply and at low cost.
Last edited:
Somebody buy just one, listen to it, then tell the others whether to buy or run?
Is there a real correlation between ultra low distortion numbers and SQ for DACs? or are we swinging back to the 70s and 80s chasing the zeros? Getting it low is good but is that the real determinant in DACs? Anyone care to comment. I recently ordered a Khadas TB to see and listen for myself.
I believe Mark has said that you can build for measurements or build for sound
I think he might agree that the standard measurements are missing some aspect(s) which determine sound quality which is better than what is normally heard from DACs or maybe the correct measurements are being done but the low levels seen in those measurements which are usually considered below audibility are not inaudible when it comes to music listening - maybe those thresholds only apply to simple test signals & not music?
Anyway, I recently purchased a Topping D10 which measures in most respects as exemplary & my listening was posted here
This shortfall in sound is only noted by comparison to DACs which deliver realism & engagement of interest - it is not noticed as a flaw if these haven't been heard before
Last edited:
I only know of 2 makers of multichannel DACs from this 1-chip sabre DACs - Exasound and Okto Research - both not very cheap. I wonder why there are no china boards available for MC application with this chips...
I2S input of above mentioned 2ch boards of course would also be appreciated - even if I think wiring is not so easy for 8 I2S channels with common clock signals (e.g. from a miniSharc board) which probably must be buffered additionally and signal lines can be very sensitive (had some issues with my wires for I2S DACs in the past) - more than SPDIF from my experience.
Level variations between chips are no problem as they will drive my 3-way + DBA set-up pair-wise L+R for each way and need to be adjusted anyway. But a common Master volume as es9028pro can do is of course a big benefit.
8-channel dacs only use one set of clock lines. Data lines are needed for each stereo pair. More for DSD since it uses one data line for each channel.
The thing about I2S is that wiring needs to be kept very short (and with sufficient ground lines if on separate boards), or additional circuitry and cost might be needed to maintain signal integrity and sound quality. Easy enough in theory if the dac chip is right next to the Sparc chip.
Last edited:
The idea is to go for both, but if there is a conflict then remember the ultimate goal is listening for entertainment and for enjoyment.
Help with DAC choise
Hi again.
Since my last post here I have been looking at options for a reliable and good sounding multi channel DAC and the Okto DAC 8 seems to be a great solution for the money. The lately updated design of it is not very DIY friendly since there is no longer an MCU to control the ESS chip. Therefore the integration of MIniSHARC is not a good option. On the other hand I looked closely on using the DAC 8 with PC software based crossovers and DSP and it looks like a new possible way forward for me. Something like Jriver with a help of Audiolense + 8 channels in Okto 8 can do the trick.
Now, I have to sort out a few things related to the possible setup with the DAC 8. I have a few SET amps for MF and HF horn drivers which obviously have unbalanced RCA inputs, while the DAC 8 has only balanced XLR outputs. Okto recommends using transformers which aren’t cheap, so I’d like to look at other possible options.
ESS Sabre on their website gives schematics for high performance balanced and unbalanced IV circuits. In both diagrams the first part of the diagram is basically the same, so I was wondering if adding on the second part to convert to unbalanced is a good idea (red marked on the attached picture)? I may use a separate box for it with a separate linear regulated power supply unit. Again, Sabre’s recommendation is AD797 Opamp, but what would be on opinion here?
As mentioned, I will need to use 2 unbalance-input amps = 4 channels = 2 Opamps. Can I combine all of them in one box and supply with common power supply or will it be much preferred to split them in two units with separate supply?
Thanks,
Anton
Hi again.
Since my last post here I have been looking at options for a reliable and good sounding multi channel DAC and the Okto DAC 8 seems to be a great solution for the money. The lately updated design of it is not very DIY friendly since there is no longer an MCU to control the ESS chip. Therefore the integration of MIniSHARC is not a good option. On the other hand I looked closely on using the DAC 8 with PC software based crossovers and DSP and it looks like a new possible way forward for me. Something like Jriver with a help of Audiolense + 8 channels in Okto 8 can do the trick.
Now, I have to sort out a few things related to the possible setup with the DAC 8. I have a few SET amps for MF and HF horn drivers which obviously have unbalanced RCA inputs, while the DAC 8 has only balanced XLR outputs. Okto recommends using transformers which aren’t cheap, so I’d like to look at other possible options.
ESS Sabre on their website gives schematics for high performance balanced and unbalanced IV circuits. In both diagrams the first part of the diagram is basically the same, so I was wondering if adding on the second part to convert to unbalanced is a good idea (red marked on the attached picture)? I may use a separate box for it with a separate linear regulated power supply unit. Again, Sabre’s recommendation is AD797 Opamp, but what would be on opinion here?
As mentioned, I will need to use 2 unbalance-input amps = 4 channels = 2 Opamps. Can I combine all of them in one box and supply with common power supply or will it be much preferred to split them in two units with separate supply?
Thanks,
Anton
The AD797 performs extremely well. I'm using them in some test circuits because they have the lowest distortion (measured with direct FFT using an APx 555) of all monolithic amplifiers that I can find. Their single gain stage design also means that the little distortion that they have is low order. For example, the LM4562 is very clean, but when it warms up, it shows increased 3rd and 4th harmonic, and you can start to see 5th. Cool the chip down and this goes away. It's also no longer available in a single.
The OPA828 is also a potentially nice single op amp, but I have not tested it yet. Oddly, it's even more expensive than the 797 in production quantities, so unless you need a FET, possibly for bias current reasons, the 797 may be a better value.
The OPA828 is also a potentially nice single op amp, but I have not tested it yet. Oddly, it's even more expensive than the 797 in production quantities, so unless you need a FET, possibly for bias current reasons, the 797 may be a better value.
Antonfazar,
keep in mind how you can measure the set-up if you go the PC DSP way. If you find a way to use the PC also as signal generator and recorder for measurement everything is fine. If you need an external Input for signal generation and microphone input Octo DAC might be difficult to combine with another PC interface for the measurement.
It is a pitty Octo gave up the miniSharc idea - would have been exactly what I was looking for..Also cannot see the reason to go for AES instead of SPDIF inputs.
keep in mind how you can measure the set-up if you go the PC DSP way. If you find a way to use the PC also as signal generator and recorder for measurement everything is fine. If you need an external Input for signal generation and microphone input Octo DAC might be difficult to combine with another PC interface for the measurement.
It is a pitty Octo gave up the miniSharc idea - would have been exactly what I was looking for..Also cannot see the reason to go for AES instead of SPDIF inputs.
Hi Anton,
You seem to have the right basic idea. R6 and R7 form part of the input filter time constant so you need to know what those are. Or perhaps you could buffer the octodac output first with two more opamps, but that gets to be a lot of opamps in the signal path.
AD797 that Monte suggests is very nice opamp, but one no longer used by ESS for output stages. AKM and ESS both now use OPA1612. The reason on the ESS side, at least, is because of the 'ESS hump' issue. Seemingly, it turns out nobody found a way to reliably solve it if using AD797 for the output stage (although AD797 still makes for a really good AVCC buffer).
Some discussion on the ESS hump and opamp selection can be found at ASR: ESS THD ‘Hump’ Investigation | Page 14 | Audio Science Review (ASR) Forum
Also, its possible if the DC offset from the I/V converter stage is already removed at octodac outputs, then AD797 might be okay for differential summing/filtering use, not sure. If the offset is removed, it might be helpful to know of it was done as described at ASR or if DC blocking caps are used, and also if there are equivalent output resistances more or less the same at R6 and R7 in the schematic you posted .
In addition, I personally would be concerned if octodac had no MCU at all. It could be controlled over I2C by the XMOS chip or some other part that might not be recognized as performing what would normally be the MCU function.
Unbalanced AES is essentially the same as SPDIF. Turning SPDIF into more-or-less balanced AES takes an inexpensive little pulse transformer. The technical difference between AES and SPDIF has to do with the way some non-audio data bits are interpreted. For most purposes and in most dacs, the differences are ignored.
One thing that does matter with SPDIF or AES is that it is a transmission line format and needs to be terminated at the end with a resistor of the correct value. You would need to know if that is done internally by the dac, or if you need to do it externally at the SPDIF/AES input connector. In the case if SPDIF, it is normally done inside the dac.
You seem to have the right basic idea. R6 and R7 form part of the input filter time constant so you need to know what those are. Or perhaps you could buffer the octodac output first with two more opamps, but that gets to be a lot of opamps in the signal path.
AD797 that Monte suggests is very nice opamp, but one no longer used by ESS for output stages. AKM and ESS both now use OPA1612. The reason on the ESS side, at least, is because of the 'ESS hump' issue. Seemingly, it turns out nobody found a way to reliably solve it if using AD797 for the output stage (although AD797 still makes for a really good AVCC buffer).
Some discussion on the ESS hump and opamp selection can be found at ASR: ESS THD ‘Hump’ Investigation | Page 14 | Audio Science Review (ASR) Forum
Also, its possible if the DC offset from the I/V converter stage is already removed at octodac outputs, then AD797 might be okay for differential summing/filtering use, not sure. If the offset is removed, it might be helpful to know of it was done as described at ASR or if DC blocking caps are used, and also if there are equivalent output resistances more or less the same at R6 and R7 in the schematic you posted .
In addition, I personally would be concerned if octodac had no MCU at all. It could be controlled over I2C by the XMOS chip or some other part that might not be recognized as performing what would normally be the MCU function.
Unbalanced AES is essentially the same as SPDIF. Turning SPDIF into more-or-less balanced AES takes an inexpensive little pulse transformer. The technical difference between AES and SPDIF has to do with the way some non-audio data bits are interpreted. For most purposes and in most dacs, the differences are ignored.
One thing that does matter with SPDIF or AES is that it is a transmission line format and needs to be terminated at the end with a resistor of the correct value. You would need to know if that is done internally by the dac, or if you need to do it externally at the SPDIF/AES input connector. In the case if SPDIF, it is normally done inside the dac.
Okto Dac8 module manual: https://www.oktoresearch.com/downloads/dac8_module_manual.pdf
Page 5 shows the I2S input connector. That's what I would want to interface to Sharc.
The dac8 Stereo model supports SPDIF or AES, at least that's what it says in the specifications section: Okto Research ...They also say it can be interfaced to RPi for networked input. Maybe that uses the I2S interface as on Dac8 Module?
If inputs are exclusively AES and not SPDIF compatible, you can't use a simple attenuation pad and or transformer (as Rane describes for going in the other direction, from AES to SPDIF) to raise SPDIF output voltage level enough for AES inputs without using active circuitry. Many devices can handle either SPDIF or AES input levels, but probably good to ask before buying. Even in the worst case, one could cobble together essentially a digital amplifier for SPDIF to AES.
Best thing would probably be to contact Okto, tell them what you want to do, and see if they can help. Might help to say you found the Dac8 Module manual and that looks interesting.
Page 5 shows the I2S input connector. That's what I would want to interface to Sharc.
The dac8 Stereo model supports SPDIF or AES, at least that's what it says in the specifications section: Okto Research ...They also say it can be interfaced to RPi for networked input. Maybe that uses the I2S interface as on Dac8 Module?
If inputs are exclusively AES and not SPDIF compatible, you can't use a simple attenuation pad and or transformer (as Rane describes for going in the other direction, from AES to SPDIF) to raise SPDIF output voltage level enough for AES inputs without using active circuitry. Many devices can handle either SPDIF or AES input levels, but probably good to ask before buying. Even in the worst case, one could cobble together essentially a digital amplifier for SPDIF to AES.
Best thing would probably be to contact Okto, tell them what you want to do, and see if they can help. Might help to say you found the Dac8 Module manual and that looks interesting.
Last edited:
Hi,
The manual looks obsolete - new board looks different now - but sure, best way is to ask Octo.
One thing I just realized and I stated above as "problem" is the ability of measurement when connecting the Octo to PC which performs the DSP. In such case you could use the Octo AES inputs to connect it with another measurement PC signal output (or with en external player, too). As the Octo is seen as 8 in and 8 out device by your PC this should work - nice!
The manual looks obsolete - new board looks different now - but sure, best way is to ask Octo.
One thing I just realized and I stated above as "problem" is the ability of measurement when connecting the Octo to PC which performs the DSP. In such case you could use the Octo AES inputs to connect it with another measurement PC signal output (or with en external player, too). As the Octo is seen as 8 in and 8 out device by your PC this should work - nice!
May I ask for a show of hands if anybody is still working on modding the ES9038Q2M boards?
Obviously, I still owe some more work trying to simplify getting best performance when using AK4137 with ES9038Q2M. However, in the best case it would involve some additional fabrication of circuitry and register programming with Arduino or other MCU. Still interest in adding more complexity to the project in return for better sound quality?
Kayata might be the only one, seems like.
Just trying to gauge the level of interest to help me think about prioritizing activities.
Obviously, I still owe some more work trying to simplify getting best performance when using AK4137 with ES9038Q2M. However, in the best case it would involve some additional fabrication of circuitry and register programming with Arduino or other MCU. Still interest in adding more complexity to the project in return for better sound quality?
Kayata might be the only one, seems like.
Just trying to gauge the level of interest to help me think about prioritizing activities.