I was referring to these articles: https://resources.altium.com/p/how-reduce-emi-mixed-signal-systems-using-proper-pcb-ground-designs https://resources.altium.com/p/how-...ampaign=yt-sponsor&utm_content=altium-academy
https://www.analog.com/media/en/training-seminars/tutorials/MT-031.pdf
The altium tutorial states that having a star point ground might be advantages in low frequency designs
What would be a good example of digital isolation? can you provide a schematic and PCB example?
https://www.analog.com/media/en/training-seminars/tutorials/MT-031.pdf
The altium tutorial states that having a star point ground might be advantages in low frequency designs
What would be a good example of digital isolation? can you provide a schematic and PCB example?
Digital isolation means that you have digital isolator isolating VDD, GND and signals. On both sides of the isolator you are free to use any grounding schemes you like.
For example of layout with digital isolator see e.g. here: https://datasheets.maximintegrated.com/en/ds/MAX14434FWEVKIT-MAX14436FWEVKIT.pdf
For example of layout with digital isolator see e.g. here: https://datasheets.maximintegrated.com/en/ds/MAX14434FWEVKIT-MAX14436FWEVKIT.pdf
A few times when I have tried to get a star ground, I have realised that I need two of them.
The only time a ground plane is not the best way, is something like a class AB output stage, where local decoupling just isn't possible
The only time a ground plane is not the best way, is something like a class AB output stage, where local decoupling just isn't possible
Digital isolators have downsides:
Whether or not your DAC needs isolators is something you will find out when testing. So once again my recommendation is to design incrementally. First revision may not need to have isolators.
One option is to have isolators in the board design but leave them unpopulated and use jumper wires to bridge the isolator pads. Although that would in essence mean that you have a splitted ground plane.
- They are quite expensive and nowadays typically out-of-stock.
- They may introduce some jitter in I2S signals.
- To have proper isolation every signal between MCU and DAC needs to go through isolators (I2S, I2C, I/O). So most probably you need at least 5 unidirectional + 1 bidirectional isolator channels.
Whether or not your DAC needs isolators is something you will find out when testing. So once again my recommendation is to design incrementally. First revision may not need to have isolators.
One option is to have isolators in the board design but leave them unpopulated and use jumper wires to bridge the isolator pads. Although that would in essence mean that you have a splitted ground plane.
Maybe this would be a good time to touch on some other things:
1. Most ESS dac chips can be operated in synchronous or asynchronous mode. IME they sound best in synchronous mode for USB, but that requires two dac clocks instead of one. If using two clocks then there needs to be a mechanism to switch between them as needed. Usually the USB board has a signal to that can used for clock selection. A very good USB board is this one: http://jlsounds.com/i2soverusb.html ESS dac asynchronous mode operation is well suited for SPDIF family audio sources. ESS DAC chips have registers that can be set by an MCU as needed to configure the dac chip to operate best depending on audio source type.
2. It was stated before that UAC2 USB is not ideal for hi-fi if using Windows OS. A problem is that Microsoft contracted with Thesycon to design the Windows Sound Engine: https://www.thesycon.de/eng/home.shtml It basically turns out that only Thesycon has the expertise to write WASAPI Exclusive Mode drivers. UAC2 drivers are free, but WASAPI Exclusive drivers are paid. Last time I checked the minimum licensing was for 2,000 seats. It means that most diy projects using their own USB interfaces cannot offer truly hi-fi sound quality due to licensing costs for drivers. The usual solution for dac designs produced only in small numbers is to use a 3rd party dac board that includes licensed drivers. For a playback only dac, I2SoverUSB is probably the best choice. An alternate approach might be to use a Comtrue USB Bridge chip for which an ASIO driver is available (which IIRC is what IVX did with his ADC). ASIO is clearly better than UAC2, but not as reliable as WASAPI Exclusive Mode. Another option might be a USB board offered by diyinhk that can be modified with custom firmware (although its clocking is not the best, but that could possibly be hacked). The diyink board I am thinking of has an optional paid Thesycon driver available for a few dollars extra (no minimum number of seats required).
3. Using an MCU designed for audio that can do DSP, say, as opposed to using a Sharc chip, it should be fine. But UAC2 audio is IMHO unacceptable for serious hi-fi. Somehow or other that issue should be considered up front in the design process.
EDIT:
4. Regarding split ground planes, in general they are usually considered undesirable. However in the case of dacs sometimes designs are split between analog and digital under the dac chip. IIRC ESS used such a technique for ES9038QM evaluation board. There is usually a small neck where the two ground planes are connected in such designs. Some other discussion at: https://www.analog.com/media/en/ana...6/number-2/articles/staying_well_grounded.pdf
5. Most of what is written in PCB layout books and articles is focused more on high speed digital rather than for mixed signal designs such as dacs. The IEEE published book on grounding, "Grounds for Grounding," includes discussion of split ground planes for data conversion applications. https://www.amazon.com/Grounds-Grounding-Circuit-System-Handbook/dp/0471660086
6. Also most of what is written about PCB bypass design is for high speed digital. Some RF signals in data converters are basically analog more than they are digital. IMHO clocks should be considered to be analog. Same for clock buffers, clock dividers, ASRCs, etc. Don't know if everyone will agree on there being a distinction between digital and analog RF. IME the difference can be significant.
1. Most ESS dac chips can be operated in synchronous or asynchronous mode. IME they sound best in synchronous mode for USB, but that requires two dac clocks instead of one. If using two clocks then there needs to be a mechanism to switch between them as needed. Usually the USB board has a signal to that can used for clock selection. A very good USB board is this one: http://jlsounds.com/i2soverusb.html ESS dac asynchronous mode operation is well suited for SPDIF family audio sources. ESS DAC chips have registers that can be set by an MCU as needed to configure the dac chip to operate best depending on audio source type.
2. It was stated before that UAC2 USB is not ideal for hi-fi if using Windows OS. A problem is that Microsoft contracted with Thesycon to design the Windows Sound Engine: https://www.thesycon.de/eng/home.shtml It basically turns out that only Thesycon has the expertise to write WASAPI Exclusive Mode drivers. UAC2 drivers are free, but WASAPI Exclusive drivers are paid. Last time I checked the minimum licensing was for 2,000 seats. It means that most diy projects using their own USB interfaces cannot offer truly hi-fi sound quality due to licensing costs for drivers. The usual solution for dac designs produced only in small numbers is to use a 3rd party dac board that includes licensed drivers. For a playback only dac, I2SoverUSB is probably the best choice. An alternate approach might be to use a Comtrue USB Bridge chip for which an ASIO driver is available (which IIRC is what IVX did with his ADC). ASIO is clearly better than UAC2, but not as reliable as WASAPI Exclusive Mode. Another option might be a USB board offered by diyinhk that can be modified with custom firmware (although its clocking is not the best, but that could possibly be hacked). The diyink board I am thinking of has an optional paid Thesycon driver available for a few dollars extra (no minimum number of seats required).
3. Using an MCU designed for audio that can do DSP, say, as opposed to using a Sharc chip, it should be fine. But UAC2 audio is IMHO unacceptable for serious hi-fi. Somehow or other that issue should be considered up front in the design process.
EDIT:
4. Regarding split ground planes, in general they are usually considered undesirable. However in the case of dacs sometimes designs are split between analog and digital under the dac chip. IIRC ESS used such a technique for ES9038QM evaluation board. There is usually a small neck where the two ground planes are connected in such designs. Some other discussion at: https://www.analog.com/media/en/ana...6/number-2/articles/staying_well_grounded.pdf
5. Most of what is written in PCB layout books and articles is focused more on high speed digital rather than for mixed signal designs such as dacs. The IEEE published book on grounding, "Grounds for Grounding," includes discussion of split ground planes for data conversion applications. https://www.amazon.com/Grounds-Grounding-Circuit-System-Handbook/dp/0471660086
6. Also most of what is written about PCB bypass design is for high speed digital. Some RF signals in data converters are basically analog more than they are digital. IMHO clocks should be considered to be analog. Same for clock buffers, clock dividers, ASRCs, etc. Don't know if everyone will agree on there being a distinction between digital and analog RF. IME the difference can be significant.
Last edited:
@Markw4 TY for your extensive answer.
I`m not done with reading the ES9033 datasheet, but my first iteration will make use of the HW mode 8, this seems reasonable for the beginning
I`m not done with reading the ES9033 datasheet, but my first iteration will make use of the HW mode 8, this seems reasonable for the beginning
You have a point here. I think the easiest thing is to stick to solid GND on the two inner layers, at least for now and focus on proper layout and design. If this fails or is not good enough, other ideas should be considered. Propper layout and separation have to be none anyway and have the advantage of no cost and simple design.
This is actually not correct. WASAPI Exclusive also uses Windows UAC2 driver which comes originally from Thesycon. The main advantage of WASAPI Exclusive is that some of Windows audio engine controls can be bypassed. Soundwise there is not any difference between WASAPI Exclusive and Shared provided that the audio endpoint samplerate and bit depth matches the audio file so that no resampling is performed.
WASAPI Exclusive also supports Native DSD. Moreover WASAPI Exclusive will not force intrusion of 'Default Sound Device' and or 'Default Communication Device' streams into PCM, such as happens with UAC2, and even happens with ASIO when the data stream is PCM. In that case the ASIO stream can be resampled to the Windows default bit-depth and sample rate. ASIO is only fully safe for PCM when Windows Default Devices are assigned to some other audio interface device (and Windows is likely to set them back to the ASIO device without telling you). In any case, nobody wants to go into Windows Sound Device Settings and change the default sample rate and bit depth every time a song in a playlist is at a different sample rate or bit depth. If someone forgets to make that change then the sound quality will be corrupted by Windows not-very-hifi resampler.
Last edited:
UAC2 is good enough and only choice unless you want to implement an ASIO driver. ASIO4ALL and FlexASIO may be ok for 44.1kHz and 48kHz sampling rates. Windows resampler can also be avoided if the resampling is done in the player (e.g. foobar2000).
As explained, ASIO is not safe from Windows resampling when the data stream is PCM. HQ Player resampler may be okay since there are a lot of high quality reconstruction filter options, but its not free.
There have been several times Windows automatically reassigned Default Devices to my ASIO sound device. Each time that happened I ended up troubleshooting to find out what happened to the dac SQ. Sometimes I checked the hardware first before remembering to check if Windows screwed up the settings again. Wasted too much time on needlessly poor software.
Not going to find agreement with Mr. Bohrok on this issue it seems.
There have been several times Windows automatically reassigned Default Devices to my ASIO sound device. Each time that happened I ended up troubleshooting to find out what happened to the dac SQ. Sometimes I checked the hardware first before remembering to check if Windows screwed up the settings again. Wasted too much time on needlessly poor software.
Not going to find agreement with Mr. Bohrok on this issue it seems.
Last edited:
On what issue that is? You were dead wrong about the UAC2 driver and WASAPI Exclusive.
I agree. I really don't see the point of repeatedly mentioning my alias (with Mr. prefix) and whether or not there is agreement.
Mr. is a title used in formal address. Someone being treated respectfully as opposed to being treated as a familiar. Like in Spanish, Usted verse Tu. https://www.thespanishlearningclub.com/pages/blog?p=the-formal-you-in-spanish#:~:text=As a general rule, use,who you know relatively well. Nothing insulting intended by the use of formality. May I ask how you prefer to be addressed? If bohrok2610 is it, fine.
As you don't know my name (or gender) it is preferable to just use the username. And this discussion is not about you and I agreeing on something so no need to repeat that either. Just state your opinion.
Okay. You did say I sounded like your wife at one point. Maybe I should have been more open minded?
Regarding WASAPI Exclusive my understanding was that OP was intending to use the MCU for USB-I2S as well. Implementation of WASAPI Exclusive driver for some diy device is a mammoth task so the only practical option is to use UAC2 driver as is and learn the quirks of Windows audio engine. Or switch to Linux. Windows audio is in a sorry state but WASAPI exclusive is not a solution for diy at the moment.
My 'opinion' regarding WASAPI Exclusive Mode for diy'ers would be that some good USB boards are available with suitable drivers. Some low-production-volume commercial products use such USB boards. I use them for my diy projects and recommend them to other diy'ers for the reasons already outlined. Other than some cost (~$100), I see no reason to put up with the problems associated with standard Windows sound drivers. For the diy'er going through multiple dac projects, USB boards can be moved from one project to another. Therefore a one-time investment in a good USB board seems reasonable to me. Switching to linux may also be a good option for some people, I would not disagree with the opinion of some other unnamed person on that issue.
IME implementing a good USB-I2S device is very educational and much more satisfying than e.g. implementation of DACs.