Bragi, WM8804/ES9023 DAC

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
After I designed Thor, an all discrete Lateral FET headphone amplifier I was in the need of a simple good sounding DAC with Toslink in for use with my PC.

Hence I made this 70 x 70 mm board with a WM8804 receiver, ES9023 DAC, 4 onboard 3V3, Low noise, LDO regulators, 2 for WM8804 IC, 1 for ES9023 and lastly 1 for the separate 50 MHz XO used on the ES9023 IC.

I made it to my own requirements so there is only Toslink in.

Add a transformer and RCA connectors and you're done.
 

Attachments

  • Bragi-Schematic.png
    Bragi-Schematic.png
    49.5 KB · Views: 968
  • Bragi-Layout.png
    Bragi-Layout.png
    146.7 KB · Views: 950
Last edited:
Have you already made it working/sent PCB to production?
I see several issues with routing...

Yes I have already sent it to the manufacturer and am waiting for the boards to be delivered. I am aware of the fact that the routing is not optimal. Just something I threw together in a weekend so I had something better than what my PC onboard soundcard could deliver. :)

However, I do have plans for a better board somewhere down the line and because of that any suggestions regarding layout would be most welcome. :)
 
Last edited:
I hope you'll read it as constructive feedback in positive manner :)

You can't cut ground under i2s traces. Never :) Return currents from I2S traces flow under these traces, and when it hits the cutout, it goes around the cutout, making a big loop with a cutout in between = slot antenna, which radiates with the rates and multiplies of i2s signal. Top layer helps a bit, with VIAs etc, but the cutout affects trace impedance too which could cause reflections/ringing of traces - both bad for signal integrity and EMC.

Your oscillator supply trace is too long, and it radiates enormous amounts of RFI. You need to place ferrite bead near the decoupling cap of oscillator to cut HF noise from entering the long trace.

All the supply traces are long, and they connect on the opposite side of the IC (looking from PSU). If you put the PSU under the ICs, you'll get short traces, and no I2S overlapping.

I don't have ES9023's datasheet (ESS cowardly hide them from publicity), so i'll just guess - it should have 2 supply inputs, digital and analog. You'd better use 2 separate regs, or at least don't route the supply with single trace common to both of the inputs.
If going with single common trace - detach the D_Vcc from the supply trace with ferrite.
You can use the supply trace directly for the A_Vcc.

I'm a big fan of resistors in supply lines. I bless them :)
Throw them everywhere - 5ohm before each LDO, between C26 and rectifier, between rectifier and C27...

Largish cap near the IC + resistor to LDO = really sweet thing. It both filters-out the LDO output, the frequency of current which travels in the wire between LDO and this cap drops considerably... Really sweet thing :)

By the way, you use single-rail supply, but no output caps? You'll get DC on DAC's output, unless the ES9023 has negative rail generator built-in...


Resistors - anything around 5-10ohm - depends on current demand, allowable voltage drop, and heat dissipation of resistor package...


Ah, all LDOs are transparent (short wire) in higher frequencies, loosing their filtering/regulation abilities starting at 100kHz.
 
I hope you'll read it as constructive feedback in positive manner :)

You can't cut ground under i2s traces. Never :) Return currents from I2S traces flow under these traces, and when it hits the cutout, it goes around the cutout, making a big loop with a cutout in between = slot antenna, which radiates with the rates and multiplies of i2s signal. Top layer helps a bit, with VIAs etc, but the cutout affects trace impedance too which could cause reflections/ringing of traces - both bad for signal integrity and EMC.

Your oscillator supply trace is too long, and it radiates enormous amounts of RFI. You need to place ferrite bead near the decoupling cap of oscillator to cut HF noise from entering the long trace.

All the supply traces are long, and they connect on the opposite side of the IC (looking from PSU). If you put the PSU under the ICs, you'll get short traces, and no I2S overlapping.

I don't have ES9023's datasheet (ESS cowardly hide them from publicity), so i'll just guess - it should have 2 supply inputs, digital and analog. You'd better use 2 separate regs, or at least don't route the supply with single trace common to both of the inputs.
If going with single common trace - detach the D_Vcc from the supply trace with ferrite.
You can use the supply trace directly for the A_Vcc.

I'm a big fan of resistors in supply lines. I bless them :)
Throw them everywhere - 5ohm before each LDO, between C26 and rectifier, between rectifier and C27...

Largish cap near the IC + resistor to LDO = really sweet thing. It both filters-out the LDO output, the frequency of current which travels in the wire between LDO and this cap drops considerably... Really sweet thing :)

By the way, you use single-rail supply, but no output caps? You'll get DC on DAC's output, unless the ES9023 has negative rail generator built-in...


Resistors - anything around 5-10ohm - depends on current demand, allowable voltage drop, and heat dissipation of resistor package...


Ah, all LDOs are transparent (short wire) in higher frequencies, loosing their filtering/regulation abilities starting at 100kHz.

Thanks, I get most of what you are saying and a few of the things I have already considered myself, like putting the whole PSU section below the IC's to shorten the supply trace length.

Regarding supplying the ES9023 DAC, it only has one shared analog and digital supply pin. It also has an integrated charge pump for the negative supply so no DC blocking caps are needed on the outputs. Only thing that the outputs need is a pair of 4.7nF caps to ground. I have the datasheet.

However, theres is one thing I don't get. Your "You can't cut ground under i2s traces." comment. I don't understand it. :confused:
 
Last edited:
i2s are HF transmission lines paired with the individual return grounds. Any discontinuity in symetry to ground will create reflections.
IMO the voltages fed to receiver and DAC need separation. At least a ferite bead on receiver rail.
Also... no OpAmps on the outputs? The connected load will influence the filter characteristics/efficiency. Imagine headphones directly on output. I know that there are OpAmps inside the chip but still they need a filter...
 
Last edited:
FYI, this project is dead/on hold indefinately.

Just don't have time to spend on a DAC when I'm working on my Thor headamp as well.

Also, I'm more interested in designing discrete analog circuits than digital stuff.

Have to prioritize.

Maybe I'll resurrect it someday, but for now I'm going with a Cambridge Audio Dacmagic 100 to feed my Thor headamp.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.