Hi all,
I have made this topic for my upcoming project, ES9016 multichannel DAC. There are not too much multichannel DACs on the market, but I need it to drive my active loudspeaker system correctly.
This chip is reasonably priced and has a simple layout (only 48pin) but still have good enough parameters to build an outstanding DAC I think....
A picture attached shows the actually stand of my pcb. I am in the idea collecting and implementig phase... I you have any advice, experience, please share it! Thanks...
I have made this topic for my upcoming project, ES9016 multichannel DAC. There are not too much multichannel DACs on the market, but I need it to drive my active loudspeaker system correctly.
This chip is reasonably priced and has a simple layout (only 48pin) but still have good enough parameters to build an outstanding DAC I think....
A picture attached shows the actually stand of my pcb. I am in the idea collecting and implementig phase... I you have any advice, experience, please share it! Thanks...
Attachments
Here some schematics to see my power supply concept...
I desing 2 panels with 10x10cm. The top panel is the DAC pcb, the bottom panel will be the output filter pcb with I/V converters an filters, 8ch balanced and unbalanced outputs.
(There will be an option to use of an other filter pcb if only 2 channels are needed in parallel stereo mode, and perhaps with headphone amplifier integrated.)
My design is free, it means I have NO marketing considerations, make it only as a hobby. I share all my CAD-files if somebody needs it!
I desing 2 panels with 10x10cm. The top panel is the DAC pcb, the bottom panel will be the output filter pcb with I/V converters an filters, 8ch balanced and unbalanced outputs.
(There will be an option to use of an other filter pcb if only 2 channels are needed in parallel stereo mode, and perhaps with headphone amplifier integrated.)
My design is free, it means I have NO marketing considerations, make it only as a hobby. I share all my CAD-files if somebody needs it!
Attachments
Last edited:
This is indeed a useful project.
What will be the input interface?
What clocking options do you plan to use.
What do you plan to use for programming/controlling the DAC?
I am subscribed and interested.
Are you experienced in building such layouts?
The concept is easy, but getting high performance layout done is notoriously hard.
The last design i saw from an experienced guy, ended up with a host of distorition and stability issues.
Best wishes.
What will be the input interface?
What clocking options do you plan to use.
What do you plan to use for programming/controlling the DAC?
I am subscribed and interested.
Are you experienced in building such layouts?
The concept is easy, but getting high performance layout done is notoriously hard.
The last design i saw from an experienced guy, ended up with a host of distorition and stability issues.
Best wishes.
Last edited:
The input interface will be MiniDSP USBStreamer or nanoSHARC (DSP), if the 8ch plugin will be available.
Clocking option is a Crystek oscillator (e.g. CCHD-950, 100M) or it is possible to use the MiniDSP master clock (for testing it will be ok, but can be upgraded later too).
Control interface will be the Arduino nano (or similar). It is cheap and simple. I have to study up in this theme...
I have built a simple PCM1794 DAC so far, but really experienced can I not say. That is why I am here, collecting advices from (I hope really) experienced guys...
And a remark: I am not on the run, so if it takes a longer time (a year or so) to complete all the things correctly it does not really matter. Important is that it should be all ok as it is possible...
Clocking option is a Crystek oscillator (e.g. CCHD-950, 100M) or it is possible to use the MiniDSP master clock (for testing it will be ok, but can be upgraded later too).
Control interface will be the Arduino nano (or similar). It is cheap and simple. I have to study up in this theme...
I have built a simple PCM1794 DAC so far, but really experienced can I not say. That is why I am here, collecting advices from (I hope really) experienced guys...
And a remark: I am not on the run, so if it takes a longer time (a year or so) to complete all the things correctly it does not really matter. Important is that it should be all ok as it is possible...
Last edited:
If its an open project, other experts might pitch in with guidance.
2 layer seems limited for such a project. A 4-layer PCB gives flexibility needed for best single integrity practices.
Have you tried KiCad? Another open source full featured PCB design tool used by many members.
9038 currently appears to be the DAC with best specs, but is expensive.
Have you considered latest AKM devices? Comparable performance at cheaper rates.
Also consider
1. Compatibilty with Isolated/reclocked I2S input interface from device such as Ian's multichannel FIFO
2 layer seems limited for such a project. A 4-layer PCB gives flexibility needed for best single integrity practices.
Have you tried KiCad? Another open source full featured PCB design tool used by many members.
9038 currently appears to be the DAC with best specs, but is expensive.
Have you considered latest AKM devices? Comparable performance at cheaper rates.
Also consider
1. Compatibilty with Isolated/reclocked I2S input interface from device such as Ian's multichannel FIFO
Nice. IV stereo circuit located on four quadrants for a total of 8 channels seems like a nice layout.
I saw such a layout with a multichannel es9028 design.
I have no experience with 4 layers, and it makes the production expensiver too. So I stay by 2. It is in fact limited but possible to do...
I do not know KiCAD, but I took a look at that now, and I must say: not for my taste. It is very unhandy (for me at least)! On the other hand, I have all my datebase and libraries in DipTrace. This program is very easy to use and has all the function what I need. So I will stay by this.
I do not know KiCAD, but I took a look at that now, and I must say: not for my taste. It is very unhandy (for me at least)! On the other hand, I have all my datebase and libraries in DipTrace. This program is very easy to use and has all the function what I need. So I will stay by this.
The problem you'll have with two layers is managing efficient power and clock routing while still keeping input and output cleanly routed. With 2 layers you will hopping between them quite a bit, and the ground planes on both sides will be quite broken.
For simple DACs with one or two supply pins two layers is quite doable, but when the supply pins start moving to the all four ends of the chip you really need to think about a 4 layer board with at least one outer plane as an unbroken shield, specially on the digital side.
In any case, the things to watch are very good supply decoupling and low inductance in the power supply. Most issues with digital DIY boards are because of sub-optimal power routing (for example, the DIYinHK boards which are totally hopeless in this respect). This leads to very poor conversion performance.
I would also recommend against using any through hole parts as they tend to increase the power supply inductance by unacceptable amounts. If you must, at least use leadless polymer capacitors. The only way to properly handle all of this is a four layer board with one supply plane and one ground plane, with the signal in the inner layers and SMT/SMD components and connectors. If you do want to stay with two layers, optimise the clock and supply and prioritise them over the analog.
Think about design tricks like routing signal vias under the package, and moving the local regulators and bypass caps right next to the supply pins (1-2mm max). If possible separate left and right regulators and place them as close as possible. Keep only analog signal routing on the lower layer so it stays unbroken as far as possible, and use lots of vias to stitch the layers together.
I can't see the PCB design but it seems you have focused much more attention on the analog part, with shorter signal traces and the double decker I/V. I would focus on the digital part as that is where the bulk of the issues crop up. For current output DACs the output stage is less critical than the converter supply performance, which will not be good the way you have it now.
Good to see designs with this chip - I wish ESS weren't so uptight with the datasheets and the supply of chips, which would make them more interesting to the DIY community. But I don't think it's of interest to them anymore. I have a Buffalo I, an Audioquest Dragonfly, a Benchmark DAC 2 and a soon-to-be-delivered Gustard X20 Pro. Guess the common factor? But I have no interest in designing something with with their chips.
For simple DACs with one or two supply pins two layers is quite doable, but when the supply pins start moving to the all four ends of the chip you really need to think about a 4 layer board with at least one outer plane as an unbroken shield, specially on the digital side.
In any case, the things to watch are very good supply decoupling and low inductance in the power supply. Most issues with digital DIY boards are because of sub-optimal power routing (for example, the DIYinHK boards which are totally hopeless in this respect). This leads to very poor conversion performance.
I would also recommend against using any through hole parts as they tend to increase the power supply inductance by unacceptable amounts. If you must, at least use leadless polymer capacitors. The only way to properly handle all of this is a four layer board with one supply plane and one ground plane, with the signal in the inner layers and SMT/SMD components and connectors. If you do want to stay with two layers, optimise the clock and supply and prioritise them over the analog.
Think about design tricks like routing signal vias under the package, and moving the local regulators and bypass caps right next to the supply pins (1-2mm max). If possible separate left and right regulators and place them as close as possible. Keep only analog signal routing on the lower layer so it stays unbroken as far as possible, and use lots of vias to stitch the layers together.
I can't see the PCB design but it seems you have focused much more attention on the analog part, with shorter signal traces and the double decker I/V. I would focus on the digital part as that is where the bulk of the issues crop up. For current output DACs the output stage is less critical than the converter supply performance, which will not be good the way you have it now.
Good to see designs with this chip - I wish ESS weren't so uptight with the datasheets and the supply of chips, which would make them more interesting to the DIY community. But I don't think it's of interest to them anymore. I have a Buffalo I, an Audioquest Dragonfly, a Benchmark DAC 2 and a soon-to-be-delivered Gustard X20 Pro. Guess the common factor? But I have no interest in designing something with with their chips.
A lot of ideas are coming from this site: Bouwtopic AFDAC 2 - ESS Sabre ES9016 - forum.zelfbouwaudio.nl
This Gustard X20 https://www.shenzhenaudio.com/gusta...028-xmos-hifi-dac-384khz-dsd-dop-decoder.html
seems to be a really good built DAC for that price, but my points are:
- I do not want to give out near 1000$ for a 2ch device
- I need a 8ch device (I have an active LS system)
- I will NOT create the best of the best in the noise und THD dB competition
- my goal is to design a very good DIY DAC with reasonable price (good imaging, overall tonal balance and macro/microdynamic)!
- and: I like understanding and designing things...
seems to be a really good built DAC for that price, but my points are:
- I do not want to give out near 1000$ for a 2ch device
- I need a 8ch device (I have an active LS system)
- I will NOT create the best of the best in the noise und THD dB competition
- my goal is to design a very good DIY DAC with reasonable price (good imaging, overall tonal balance and macro/microdynamic)!
- and: I like understanding and designing things...
Last edited:
Agree to all the above. My Benchmark died and the design team I work for needed a standby source. Hence the Gustard. It's not built anything like the Benchmark, nor is the sound close, even running dual mono and with a newer generation of chip. Or at least, not in the first hour of running.
When laying out traces on two layers, you will usually sacrifice inductance (trace length/width) on one set of connections. There are four signals in a DAC - the power, the clock, the input and the output (yes, power is a signal as it carries varying currents at different frequencies).
The point is to ensure the digital section has access to very low impedance power supplies. The regulator -> bulk decoupling -> high frequency decoupling should all be within a very short distance of each other. Use TH parts *only* if they help you avoid vias and you can avoid interrupting the ground plane.
I can't see your files (I use Eagle myself), so you can upload images and others can collaborate.
When laying out traces on two layers, you will usually sacrifice inductance (trace length/width) on one set of connections. There are four signals in a DAC - the power, the clock, the input and the output (yes, power is a signal as it carries varying currents at different frequencies).
The point is to ensure the digital section has access to very low impedance power supplies. The regulator -> bulk decoupling -> high frequency decoupling should all be within a very short distance of each other. Use TH parts *only* if they help you avoid vias and you can avoid interrupting the ground plane.
I can't see your files (I use Eagle myself), so you can upload images and others can collaborate.
The AK4458VN is the closest chip AKM has to to the ES9016 in performance. This is also a 32bit chip so the digital volume has the potential to work as well as the sabre one. Probably cheaper too as AK4458VN is less than US$6 from Digikey.
The filter and single ended convertor is also much simpler than what is needed to get increased performance from the sabre chips.
If you are sticking with the ES9016 take a look at the Sabre evaluation board manual as that has a lot of design suggestions for layout and parts selection.
If you are willing to change your regulators Linear technology has the LT3042 which is very low noise and some other regulators in different packages that use a current based error amplifier so the noise is not gained up when you use them at higher voltages for the op-amps.
I have recently built AK4490, ES9018, ES9018k2m and ES9016 DAC's all for different systems. When I listen to them as stereo DAC's via a good headphone setup I hear very little difference between them, they all sound good.
What I do notice is that I can tell much more of a difference in the filters that come with the AK4490 and the ES9018k2m. ES9018/9016 has just fast and slow and I find it almost impossible to tell which is which.
I already have more than enough DAC's but a board with an AK4458 and a CS3318 analog volume chip would make a great front end for an active system that should be fairly cheap to put together.
The filter and single ended convertor is also much simpler than what is needed to get increased performance from the sabre chips.
If you are sticking with the ES9016 take a look at the Sabre evaluation board manual as that has a lot of design suggestions for layout and parts selection.
If you are willing to change your regulators Linear technology has the LT3042 which is very low noise and some other regulators in different packages that use a current based error amplifier so the noise is not gained up when you use them at higher voltages for the op-amps.
I have recently built AK4490, ES9018, ES9018k2m and ES9016 DAC's all for different systems. When I listen to them as stereo DAC's via a good headphone setup I hear very little difference between them, they all sound good.
What I do notice is that I can tell much more of a difference in the filters that come with the AK4490 and the ES9018k2m. ES9018/9016 has just fast and slow and I find it almost impossible to tell which is which.
I already have more than enough DAC's but a board with an AK4458 and a CS3318 analog volume chip would make a great front end for an active system that should be fairly cheap to put together.
DipTrace is free to download and use up to 300pins and 2 layers. I think it is easier to view it so like pictures with limited qualities.
I will look around for smd electrolitic-caps, and make the supply traces wider and shorter as possible. Do you meen so, or?
Last edited:
I did it already, thanks. But the evaluation board is not a right construction. It can perhaps produce good measurement values, but that's all...
Long traces, only small smd capacitors overall (in the output filters too) and a filter stage what I do not like to build...
These posts gave me some ideas:
https://e2e.ti.com/support/data_converters/audio_converters/f/64/t/248466
Attachments
Last edited:
I wasn't suggesting that you copy the entire board as the placement of connectors is meant for evaluation rather than constant use. SMD parts can allow very close spacing which helps to reduce resistance and inductance that can be difficult with other components. ESS will have chosen these parts because they are cheap, easy to manufacture and still produce an excellent measured result. I am sure you can get good results with other parts too.
The only way I know of to confirm if your layout is working is by measuring. If you can achieve the same performance as ESS then you know you have a good layout.
Apart from getting good measurements how do you plan to know if you have got the layout right?
The primary reason I don't try and produce my own boards like this is that I don't have the test equipment to verify the results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.