Diy ES9018K2M I2S Dac

Recently I have some fun playing around with a fully assembled ES9018K2M I2S DAC kit which I bought from taobao early this year.

The price of this kit is so cheap (less than RM35 without shipping) that I can’t resist and bought one for my experimental project.



The performance specification of this DAC from ESS homepage shows +127dB DNR (Dynamic Range) and -120dB THD+N (0.0001% Total Harmonic Distortion + Noise).
Since I can’t find any specification of this IC online, I read some technical articles from ESS webpage to get more understanding of the Sabre range of DAC. You can find these articles at: http://www.esstech.com/index.php/en/support/downloads/
 
I will split my findings during experimenting with this DAC kit into 3 parts:
Part 1: The making of ES9018K2M DIY DAC
Part 2: Simple Modifications
Part 3: Advance Modifications

The choice of parts may not be ideal, actually I just fully utilise what I have in hand without buying new parts.

Although all the modifications are specified to this ES9018K2M DAC, the same idea can be implemented to ES9028Q2M and ES9038Q2M series DAC kit too.
If you are not too confident about how to make or modify your ES90*8 DIY DAC series, fell free to drop me a PM and I might be able to help if within my knowledge.

If you search around Lazada or Shopee you will be able to find a lot of choice, although most of them are from China seller.
Good luck and happy DIY!

Part 1: The making of ES9018K2M DIY DAC

I decide to make this DAC kit very simple in function with just 1 coaxial input. Other type of input such as Optical in, USB and Bluetooth can be implemented in later stage easily since I made all the function block in modular form.
Since ES9018K2M DAC kit only accept IIS input, I need to make use of DIR (Digital Interface Receiver) IC to convert coaxial in (SPDIF) to IIS format.
For this DIR job, I chose DIR9001 from my un-used stock. I bought it quite some time ago but never make it to project.
Although this DIR only accept 24bit 96kHz sampling frequency max, it still full-fill the task for my simple system setup.
 
The first thing to do was actually a SPDIF amplifier to convert low level AC coaxial input signal to DC 3.3V output level as required by DIR9001.
I build a low noise 3.3V discrete regulator for this noise sensitive circuit.



Next to do was DIR circuit, with its own 3.3V low noise discrete regulator too.


I found some un-used transformers which can be utilised for this project.


Upon checking the ES9018K2M kit, I found the left and right output resistors are of different types.
Left channel using MELF type (good quality in general) and right channel thick film (I guess?) type.
Change both resistor to 47ohm 1/4W carbon resistor (yellow arrow) since I do not have MELF resistor of this value.




I also change the capacitor (white arrow) from value 390p to 220p.

The reason is due to when I redraw the LPF (low pass filter) circuit and do frequency response simulation, I found the response (blue line) was not flat and has a peak at around 70kHz. Change to 220p will give it a flat frequency response (red line) without peaking.


Now everything is ready and is time to combine all modules together.
From my recycle bin found one unused chassis that can be temporary used for this project.
Take some time to combine all.




Test the completed DAC with 0dB 1kHz and 20kHz sine wave and confirm the output was normal without clipping.
Finally come the time for sound check.
 
First impression of sound check revealed that this DAC kit is actually very good sounding consider the price.
It clear my doubt whether this DAC using fake ESS chip (if there is any…) or not.
Bass response is firm, midrange is clear with very good high frequency resolution.
The overall presentation slightly on the lean side, so it may favour more to jazz and classical music.
My personal opinion is that it will comparable well with any good quality entry level CD player.
 

Now my hand become itchy and can’t wait any further to see how far this DAC kit can go.

~ to be continue ~

Thank you for sharing your work. Nice write-up.

Noticed you didn't use any isolation on the SPDIF input. Any reason?

Hi vt4c,
Thank you for your feedback.
Mind to elaborate more regarding isolation?
Are you refer to input isolation transformer?

My current connection of SPDIF input:

You're right...It's isolation transformer...A digital SPDIF transformer will isolate the noisy ground between the devices....Just a tweak to get better result. :-)

Since there is spare gate, you might also want to consider adding another inverter in between the first inverter and the DIR9001. The circuit right now inverts all the signal...Need to invert back to get it correct.

Also consider forcing the inverter to a known state, e.g. pulling the input high. This is to prevent floating gates from causing potential nasty oscillation to the i.c.

Hi, thanks for sharing your thought. Will definately try the isolation transformer later on. I did not use it just because I do not have any stock at the moment. But I did noticed that without isolation transformer, every time I connect SPDIF cable to the source, the analog output will become noisier!

The actual design already using 2 inverter gate, my simplified block diagram do not illustrate this in detail.

The unused gate already all pull low to prevent floating. Input inverter gate I left it floating due to pull high or pull low make the whole thing no oscillation at all.

Part 2: Simple Modifications
 
2.1) Capacitors
My version of ES9018K2M DAC kit come with a mixture of Panasonic FC and Elna miniature series capacitors.
Both are actually good choice and I guess that’s the reason it can give good sound even without modification.
Since I have some Elna Silmic and Nichicon FG capacitors on hand, I replace all the capacitors to both Silmic and FG type.
 
The listening impression with the capacitors change is quite positive.
The lean character is no more existence, replace with a sound that is full bodied and more extended in the bass region.
 
After this I also replace all the small ceramic capacitors (100p x 4pcs and 220p x2pcs) to polypropylene type.
The sound improved further with better definition in the bass, mid and high region.
 
2.2) Op-amps
1)      5532DD from JRC (come with DAC kit) – overall good performance
2)      OPA2604 – overall more to tube like sound, high frequency extension is not as good as 1) and with a lot of bass sound. Overall the sound is too warm/slow for my liking.
3)      MUSES from JRC – overall very good with neutral sound reproduction. Best among 1,2 and 3.
4)      OPA2132, 2134 – good with even wider sound stage than 3) but not as neutral as 3).
Finally I choose 4) due to more fun to listen to. No. 3) is very neutral but also a bit “dull” in presentation.
 
Above impression just my personal opinion. One should choose the capacitors and op-amps based on his own preference and most important, system matching based on equipment and room acoustic.
 

At this stage I believe it can beat most entry-level CD player but still not enough to worry good quality mid-range CD player.

A well thought of design.

Hi vt4c, thank you for your encouraging words!

Part 3: Advance Modifications
 
The next modification will be to replace the clock (yellow arrow, 100MHz) to a low jitter type.




I found some leftover clock module which I made many years back.
That was the time when I experimenting the effect of low jitter clock module towards sound quality of CD player with my friends.


The picture above show 16.9344MHz frequency, but I use the highest frequency I have ever made (45.1584MHz) for this project.
I modify the output of the clock from 5V to 3.3V (older generation CD player mostly using 5V).
 
After replace the original clock with this clock module, now come the time for sound check.
The first moment I play the music, I were surprise by the improvement in sound quality.
I have my friends listen to it and they confirm my findings.
The sound stage become so much stable and well define.
Placement of each instrument is so clear that we can easily tell which instrument come from which location.
Listening to “complex” music or large scale orchestra become so easy to follow and more relaxing.
 
This confirm my doubt regarding the jitter performance of the original clock was not that great (should be due to low selling price).

If you plan to change the clock module, I believe any good quality low jitter clock will give the same result as what I experienced.

~ to be continue ~

Replace IC regulators (3.3V) to low noise discrete type.

From my previous experience, I always have positive result when changing from IC regulators to low noise discrete type.






From picture above, we can find 2pcs 3.3V IC regulator which separately supply power to analog and digital section of DAC (good design!).
In order to nicely fit in the current location, I custom build two low noise discrete regulator with universal PCB.







Replace the IC regulator with discrete type.



During sound check session, both my friends and I were again surprise by the improvement in sound quality bring by low noise regulator.

Not only the depth and width of sound field become improve, it bring the overall sound presentation to more 3D and analog like (in other words, much more natural and life like). One of my friends even commented as he was listening to tube sound.

~ to be continue ~

More modification: replace the on-board discrete regulator.


Upon checking the discrete regulator on the ES9018K2M (yes, there is one), I redraw the circuit diagram and actually can’t understand the design.

The 2N5551 transistor seems to have its emitter and collector in reverse connection…
I decide to replace the whole circuit with discrete regulator too.
 
During sound check, I can notice the sound stage improve even further (subtle improvement, not as drastic as previous 2 modification), with frequency response extend more in the high and bass region.
 
At this stage of tune, I believe it can compare well with other mid-range CD player without problem.

One “dangerous” character of this DAC is that I tend to turn up the volume while listening to music without realizing already too loud…

My previous DAC project using Wolfson WM8740 with fully discrete single ended class A zero feedback circuit if compare to ES9018K2M, this modded ES9018K2M will win in almost every region.


Further modification
For this ES9018K2M kit actually I have another 3 step of modification in my mind, but I decide to pull a stop and leave it in current condition. I will reserve this three more steps for my future project.
1)      Change the IC op-amp to discrete type (with feedback) or discrete class A non-feedback type
2)      Change the power supply from single supply (+ only) to +/- supply (for discrete op-amp)
3)      Change the transformer to toroidal type.
In fact I already received my next project which is ES9028Q2M kit, but will not start in near future due to time constrain.
 
Additional notes
From the study of this ES9018K2M DAC circuit design, I found that the DAC output circuit is configured as voltage out mode. From ESS published performance spec, the best performance will only be obtained in current out mode. So in terms of measurement, this DAC kit might not perform as well as current out design type. But again, if we want to configure it as current out mode, we need to use 3pcs of op-amp as compare to 1pcs in voltage out mode.

Sometimes more simple design is better for sound quality, even though in terms of measurement it did not give you good result.

~ End ~

Well done. Thanks again for sharing your work

Hi vt4c, thanks for your feedback too.
Will update this topic again once I try out the isolation transformer.

Any chance of sharing the schematic for your discrete regulator? Is is a TL431 based shunt regulator or TL431 used as a voltage reference for a transistor pass element?

Thanks

Hi, will share more info after back from holiday.
Basically is a zener based transistor pass element with RC filtering at the voltage reference area.

Is your regulator a variation of the one used by Pedja Rogic in his AYA 2 Dac?

If I remember correctly, Pedja uses a J310 jfet with gate tied to the source resistor; configured to source current at about 6 ma to the Zener which is then  R C  filtered to the base of a discreet darlington pass transistor using a high hfe to92 bc560 transistor and BD135.

I myself have sometimes used a current source feeding a TL431 with the cathode to a discrete darlington pass transistor.

In any case looking forward to seeing your circuit.

Thank you for sharing your thought about discrete regulator.
I have not read about Pedja Rojic and his AYA 2 DAC before.
From your description I think the circuit topology is quite similar except for darlington section.
 
Below are the circuit that I used for my ES9018K2M project.

R1 was adjusted to 3~6mA,
R2 range from 10 to 100 ohm,
C1 and C2 range from 10u to 100u (C1, R2 and C2 will affect charge up time) and C3 I used value from 47u to 220u.
For some circuit I just short R2 and no using C2 for more simple layout.

R3 I adjust to 1~2mA current (330 to 680 ohm).

Q1 I used 2SC1815 Toshiba low noise type with good hfe linearity. Previously I used 2SC2240 but no more stock.
BC550 low noise transistor will be good substitution for Q1 as well.
Q2 I use 2SA965, 800mA type from Toshiba. Of course BD136 can be used as well.
For higher current application I used TTA004B or 2SB1187 for Q2 from my leftover stock.
 
For better ripple rejection especially for unregulated input, I replace R1 with constant current source.
Previously I used JFET (2SK246 or 2SK301) as CCS but also no more stock.
I replace JFET with bipolar transistor for my more recent project.
 
I once have a chance to measure the noise performance of IC vs Discrete regulator.
If you are interested please let me know.

Thanks for sharing schematic. Is the pass transistor an implementation of a sziklai pair? It looks like one to me.
What benefits do you get from the use of a sziklai pair as a pass transistor compared with a darlington pair?
 

I managed to locate the AYA Dac schematic for the discrete regulator. Pedja Rogic's pdf although may be downloaded is encrypted and the schematics cannot be extracted. The commentary explaining the discrete regulator is found there.

Please find appended the url for the schematic
http://www.decdun.me.uk/Pedja-Rogic/pedja-gc_decdun/pedja-gc_decdun/images/non-fb-sup_sch.gif

Another link for the archived pdf of the AYA Dac.
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=11&cad=rja&uact=8&ved=0ahUKEwjPsqPNhr3YAhUkTo8KHXYHDucQFghnMAo&url=https%3A%2F%2Fwww.nexthardware.com%2Fforum%2Fattachments%2Fdac-e-dsp%2F17084d1483740256-my-new-dac-pcm1704-ad844-tda1541_aya_dac_rev2.0c_a.pdf&usg=AOvVaw2G3RL3JSyMJI6ibGQrYpcG

Thanks

The pass transistor is indeed a sziklai pair. I prefer this configuration over darlington pair due to only single Vbe drop compare to 2x Vbe drop in darlington pair. If I remember correctly, sziklai pair can provide a better level of linearity than darlington pair which can be advantageous especially in audio application.

Thank you for your link of AYA DAC. Very technical and detail explanation of how the DAC was designed. A great textbook read for DAC design.

One particular item that catch my attention was the diamond buffer in AD844 output stage. I once compare the performance of discrete diamond buffer vs op-amp buffer, in my personal opinion, no op-amp buffer can beat the performance of discrete diamond buffer.

I attach power supply and low pass filter configuration of ES9018K2M DAC kit for your reference.
You might find it useful if you intent to do some modification.

If you happened to have ES9028Q2M I2S DAC kit, I attach the power supply block diagram and LPF configuration for your reference.
Both ES9018K2M and ES9028Q2M are actually quite similar in power supply and LPF design.
Happy DIYing