DIY ES9038pro Board

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi All,


I'm looking to build my first DAC, all from scratch. After some recent success with reflow soldering of some SMT I'm feeling up to the challenge of finer pitch parts. I searched around and decided to go with ES9038 because it includes a SPDIF receiver. I'm standing on the shoulders of giants, having studied PCBs and schematics from the web as well as getting some great help form forum member isiora. Aside from his Clearpath DAC (http://www.diyaudio.com/forums/digital-line-level/307221-clearpath-ess-dac-project.html) this is another reference design close to what I'm going for:

ES9018 USB DAC - Hi-Resolution System 

Attached are some screen shots of my work-in-process PCB... not yet complete but getting to where are the big chunks and dodgy routing are close to done.


Notes:
  • It's a 4 plane board with the signal on the top layer, power supply runs on the bottom, a ground plane, and a 5vDC power plane (from which all the other rails will be regulated and filtered). Sabre recommends a single ground plane.
  • The white "+" and "-" symbols are vias to power and ground respectively.
  • The ES9038pro chip is right of center, a 64 pin qfn package. I included a thermal pad and vias beneath.
  • The right of that is a Crystek oscillator. Does anyone know if these have power supply bypassing internally?
  • To the right of the oscillator, a header for a micro-controller. I'm not isolating it due to space constraints (trying to meet a 2.5x3.8" size limit for pricing reasons).
  • Above and below the oscillator are the I/V stages. I'm using the MSOP version of the OPA1632 due to it's higher heat dissipation. Lot's of current coming from this DAC.
  • At the top and bottom edges of the board near the middle are regulators for the low-noise 3.3v analog supplies. I'm using ADM7154 fixed voltage regulators.
  • To the left of the ES9038 are 4 LT1963 fixed regulators providing digital and less sensitive analog supplies.
  • In the upper left are the inputs. Currently I only want to cram a transformer-isolated SPDIF and a toslink connector. If a miracle happened I might try for an AES3 but I'm not holding my breath - it's a first draft. I will try to get an isolated I2S header in below the toslink, for use with future USB-I2S modules.
  • I'll bring power to the board via a header in the lower middle. Ground, 5vDC for the plane, +/-X for the opamps.


So, I don't know, most of the ES9038 threads I can find are people showing off finished projects and beautiful PCBs. I'm very much still working out if this is a crazy project. I'd like some feedback or criticism though I know that can be tough without a real schematic. Anyway, I wanted to share and I'll update as I go along.


Brian
 

Attachments

  • ES9038_All.png
    ES9038_All.png
    92.4 KB · Views: 2,547
  • ES9038_Top.png
    ES9038_Top.png
    66.1 KB · Views: 2,519
  • ES9038_Bot.png
    ES9038_Bot.png
    58.9 KB · Views: 2,372
  • ES9038_SS.png
    ES9038_SS.png
    62 KB · Views: 2,406
following with interest.

Which IV design are you using? I heard there are challenges related to the handling of large current capability of ES9038.

I am really hoping for some multichannel solutions based on ES9038.

Hi diyaudnut,


I'm planning to use a standard I/V configuration but with an op amp that should be able to handle the current and power dissipation. Typical setups use 2 inverting op amps with a virtual ground for the DAC to drive and an RC feedback to low pass the stage. I'm a fan of the OPA1632 fully differential op amp which has an MSOP package that can handle the dissipation. I linked a screen shot of my simulation circuit below. In truth, this is not a tested design/application, so part of me wants to just put headers on the board and deal with I/V after I know I got the rest right.


Brian
 

Attachments

  • Screen Shot 2017-07-26 at 5.57.44 PM.png
    Screen Shot 2017-07-26 at 5.57.44 PM.png
    111 KB · Views: 2,307
Keeping the power supplies to the I/V opamps clean is going to be a major challenge ISTM. You'll want to look at the output impedance of the regs feeding your OPA1632s and estimate the noise generated on the rails by their classAB output stages.

Because of the relatively low PSRR of the OPA1632? I plan to regulate and filter off-board, and filter locally at the op amp (2-10uF tantalum, .1uF MLCC). I have no idea how to calculate the noise.

This is the biggest risk of the project, seeing as I have no existing I/V stage to reference I am just kind of hoping that this will work out OK. I am very much open to suggestions for an I/V stage.
 
Because of the relatively low PSRR of the OPA1632?

I've not inspected the OPA1632's DS for PSRR as I don't pay much attention to PSRR plots in DSs as they don't normally specify the test conditions (loading on the output, presence of signal to displace the OPS from the mid-point).

The challenge comes from the relatively large currents needed to be sourced to return to the DAC.

I plan to regulate and filter off-board, and filter locally at the op amp (2-10uF tantalum, .1uF MLCC). I have no idea how to calculate the noise.
Regulating off-board brings additional impedance (wires, connectors) to the supplies and this impedance raises the correlated supply noise.

This is the biggest risk of the project, seeing as I have no existing I/V stage to reference I am just kind of hoping that this will work out OK. I am very much open to suggestions for an I/V stage.
In the absence of published DSs for the 9038 I'm rather at a loss to make specific suggestions. What I do suggest though, assuming you're after the best subjective sound quality is that you avoid using classAB stages in your I/V converters. Ideally all signal circuits should be run purely in classA. A high transconductance MOSFET might work well (in common gate mode) but its just a guess in the absence of detailed technical info.
 
The ES9038 datasheet does have some schematic for output stage, but not sure if thats an IV stage.
Does the ES9018 IV schematic published by ESS work for this chip?

They offer new stereo and multichannel eval boards for ES9038. Both have opamp output stages.
Further the latest Oppo UDP-205 also has this chip with opamp output stages.

But i agree that these designs might not have been done with highest fidelity in mind.
 
The ES9038's datasheet does have an I/V "schematic", but it's just a "high level schematic", with no actual parts values.

The 9018's datasheet had a "real" schematic, but it's not applicable to the 9038 due to the latter's higher current output.

The Oppo has a rather bad sounding I/V stage, I hear.
 
The challenge comes from the relatively large currents needed to be sourced to return to the DAC.

Got it. Good point.

The 9018's datasheet had a "real" schematic, but it's not applicable to the 9038 due to the latter's higher current output.

I had planned to use this topology but adjust the the resistor/cap in the loop to better match the higher current. I think a 58 ohms feedback resistor would yield ~+4dbu of output at full scale. The question is can the opamp handle it and/or am I walking into some other problem.

would discrete opamps be a better option?

Any other discrete IV designs?

How about some class A transcond amp from Pass?

Can Tom's Mod86 be adapted for this? 8 Channel DAC with integrated amplification.

So many options. I'm seriously considering omitting the IV stage on this first build so I can play with different types. A discrete option could certainly dissipate enough power but most discrete stage I find have higher THD than op amp designs. I'm a big fan of discrete opamps (I've used the JH990 a few times). I'm simulating some tube stuff as well.
 
...and There's the possibility of a passive I/V stage. I sent an email to Souter asking about the applicability of their 1465 1:10 DAC transformer. They list the ES9018 so I'm asking specifically if this transformer can handle the 60mA peak to peak of the 9038. I'll report back on what they say. To get a ~3.5V peak to peak signal (+4dbu) the DAC outputs would see about 6 ohms.

I don't think I have the guts to include an I/V on the board right now without having an example to, uh, copy from. I'm reworking the board to include headers only for now.


Brian
 
Based on my decision to exclude an IV stage for now I went back and reworked the layout a bit. Attached is the most complete version I have yet. I'll have to rig it up with a passive I/V for testing if I get that far.

Is there a smaller crystal I can consider using? The Crystek CCHD-950 is probably overkill since the DAC will be used in slave mode exclusively.
 

Attachments

  • ES9038 DAC .png
    ES9038 DAC .png
    102.4 KB · Views: 969
XO

Based on my decision to exclude an IV stage for now I went back and reworked the layout a bit. Attached is the most complete version I have yet. I'll have to rig it up with a passive I/V for testing if I get that far.

Is there a smaller crystal I can consider using? The Crystek CCHD-950 is probably overkill since the DAC will be used in slave mode exclusively.

The Crystek CCHD-575 is available in 100M, and the data sheet shows better performance than that of the 950 (phase noise at 10 Hz). Save ya some board space.
 
I heard that the 9038 can use lower speed clocks, as it has some kind of onboard multiplier, for those who have seen the data sheet, is that correct? If so it might be advantageous to run a standard speed XO, as they typically have lower phase noise.

The minimum MCLK is specified as "128fsr" for either synchronous or asynchronous mode, so for 384K you'd need a clock of at least 49.152MHz.

The chips do support an MCLK divider (and not multiplier) but that is meant to be an energy saving feature.
 
I should have mentioned: I plan to use a 100Mhz oscillator and use the NCO value to set it as needed. The clock gear is only a divide by 2, 4, and 8 function, it's the 32-bit NCO that gets you to a specific FSR multiple. I'm honestly not sure if I even need to concern myself with the specific clock frequency if I'm using spdif. The data sheet is unclear but I'll work it out.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.