ES9038Q2M Board

Member
Joined 2008
Paid Member
But, for some of these DIY projects we don't have a lot of expensive gear so we have to use the best methods we can come up with under the circumstances. Fooling one's self can be pretty hard to avoid of no measurement equipment and no reference DAC for comparison either. In that case and if interested in accuracy probably the best we can do is follow good advice of the manufacturer, measurements by people who are equipped to do them, and follow good engineering practices such as proper layout, use of ground planes where indicated, proper power pin bypassing, and so on, etc. That's how I like to approach it anyway. :)

In that case, shouldn't you be using an ES9028Pro-based board for modding since that's what your reference DAC uses? I have seen no indication so far that the 9028q2m is the same chip.
 
In that case, shouldn't you be using an ES9028Pro-based board for modding since that's what your reference DAC uses? I have seen no indication so far that the 9028q2m is the same chip.

Doesn't seem like it should be a problem. What the reference DAC uses internally is of no concern. It could be a black box. For now, DAC-3 is still SOA in terms of measurements and SQ. As the years go by the bars may be raised, but its fine for now. More than fine, really.

Also, according to ESS Q2M is pretty close to the same a PRO chip. It can play the same formats, operate in the same modes as PRO (for 2 channels, that is). It has 2dB worse distortion specs (-120dB vs -122dB), and mostly differs in terms of noise specs. Q2M can do -122dB and PRO can do as low as -128dB (IIRC), according to ESS. Low noise is from some combination of PRO being able to parallel channels for noise reduction and perhaps lower internal source resistance for the DAC outputs (which you might expect from the higher AVCC currents).
 
Last edited:
Is it just putting a schmitt-trigger chip into signal path between clock and DAC?
I compared Crystek with NDK - NDK has rise/fall time of 6ns and crystek CCHD 575 2ns. What would be a "low value" achievably with a Schmitt Trigger circuit?

It depends. At high frequencies clocks start to look a lot more analog than digital, at least in some ways. A couple of the bigger factors include slew rate into capacitance (how fast can a device charge up a particular load capacitance), and however much the voltage needs to swing before a high or low level is considered to be arrived at, which would be whatever the load device needs.

Roughly speaking in this situation a 2ns spec is good. However, at 100MHz, one cycle takes 10ns. ESS would like to see 4.5ns high and 4.5ns low. That leaves 0.5ns for each edge transition. By that reasoning Crystek should fail to meet the spec. Various possibilities exist. Maybe Crystek beats their spec into the whatever the Q2M load is. Maybe at some voltage level Q2M considers the edge transition complete when the level has changed by only 200mv or some small number, I don't know. Maybe some other possibility. Hard to say without some work and some test equipment.

Seems like for a one-off project the Crystek part makes a lot of sense, at least given all the other stuff that needs work to get the best out of these DACs. Someone could always go back and run some clock buffer experiments later to satisfy curiosity and maybe to try to give back something to the DIY community for the next builders that come along.
 
Last edited:
Member
Joined 2008
Paid Member
Is it just putting a schmitt-trigger chip into signal path between clock and DAC?

Possibly. This is why I referred to cdsgames because their input would be helpful. They have been working on their ess dac for about 8 months, so if they recommend a schmitt trigger maybe there's a reason. My Hifimediy ES9038Pro dac also appears to have a schmitt trigger between clock and dac. And the Quanghao/Andrea dac *also* refers to a 'high speed buffer' between clock and dac. And the Quanghao uses Crystek clocks. So is there something going on there we are not considering if they've all chosen that route?
 
@freezebox, Not sure what you are asking about re rise/fall and frequency.

4.5ns is a minimum hold time ESS would like to see for a high or a low. However, to say Crystek has 3ns is not enough information to specify everything that might matter. Clocks are are not perfectly square-shaped as a function of time. Nor are they perfectly trapezoidal-shaped. Those things are approximations to what a real analog clock looks like.

Depending on what is important for a particular application, various things can be specified. ESS doesn't specify much, so its hard to know what really matters for them without doing some tests and measurements. Please see diagrams below for some things that are sometimes specified for pulses and clock-type signals.
 

Attachments

  • Pulse.jpg
    Pulse.jpg
    42.3 KB · Views: 422
  • Pulse2.gif
    Pulse2.gif
    7.4 KB · Views: 426
  • Pulse3.jpg
    Pulse3.jpg
    8.8 KB · Views: 422
Last edited:
And the Quanghao uses Crystek clocks. So is there something going on there we are not considering if they've all chosen that route?

Hard to say. Depends if they all really know what they are doing and have the test equipment to do it right. To know what a clock looks like would take an oscilloscope, I think we all understand that. However, scopes have bandwidth limitations and input capacitance, etc., which can produce distortion in what is seen on the screen. A 1GHz or faster vertical bandwidth scope with low-capacitance active probes would help. Probably best to monitor DAC performance when it is connected to see if trying to measure is changing what is being measured.

Also possibly handy to know is that these DACs can be programmed to output MCLK on a GPIO pin. MCLK is derived from XI, which is the real clock or crystal signal. But MCLK might worth looking at with a good scope while doing buffer tests, particularly if the divisor is set to 1. Think I will leave off there.
 
Last edited:
@freezebox, Not sure what you are asking about re rise/fall and frequency.

4.5ns is a minimum hold time ESS would like to see for a high or a low. .


above you write for a 100 MHz clock one period is 10ns. So to get 4,5ns hold time the rise/fall time must be 0,5ns each. If I use e.g. a 50 MHz clock, the period is 20ns, so rise and fall time could be 11ns in sum to still have 4,5ns hold time? So the necessary "rise/fall speed" is pending on clock frequency?
 
So the necessary "rise/fall speed" is pending on clock frequency?

Maybe. What is allowable may vary. If they only specify minimum high or low times and some duty cycle limits, then that could allow for the possibility of slower transitions at lower frequencies. Assuming they have actually specified everything that is important.

If they happen to Schmitt trigger the clock/crystal input inside the DAC chip it may not matter too much so long as the rise and fall characteristics of the clock are extremely repeatable and stable from cycle to cycle. If the slopes or edge transition start/stop times vary from cycle to cycle at all then that would presumably introduce jitter.

Also, allowing any noise pick up during the rise and fall times could also increase jitter.
 
Last edited:
Babolcs,
the circuit in the post 1294 is what I finally use. It automatically sets the DAC into current mode. Nothing else to do to achieve that.. The left part of the circuit is on the bottom side on a seperate board while the right part was applied on the top side by modifying the existing circuit (replacing all components). What you dont see on the pictures are the vertical 6K8 resistors where the outputs of bottom board are connected to the circuit on the top side..Also be aware of the exact polarity of the opamp inputs. Pls ask if you have any further questions.


Hi Freezebox, All

I have done the I/V stage. Although not 100% copy of your circuit. Since I'm lazy I used the LP filter stage which was on the panel, and added the I/V stage according you schematics. Thanks it works fine.
In the I/V stage I used the LME49720. and since I didn't have at home more I used in the LP filter stage OPA2134 - will be change to LME49720 too.
(some advices can be found here also: https://www.by-rutgers.nl/IV-converter.html)



So the current status summary of all MODs I've done:
1) I/V stage

as described above


2) AVCC_L,R

I powered them with an OPA based solution. Described here OPA based Ultra Low noise VREF for DAC more detailed in post 70.

3) Power supplies

- All the panel's power supply is separated with LP5907 and this is powered with a LM317+TL431 modul
- the controller remained on the 7805 (separated)
This 2 were described in this thread in the post 1306

- The I/V stage with the LP filter and the AVCC OPA based REF are powered from a SALAS shunts v1.3

4) CAPs
- I changed the important power CAPs to Nichicon MUSE types.

5) Upsample
- I ordered the upconversion unit, suggested by Mark based on the SRC4392
(this not yet arrived yet)


I do not intend to replace the clock - according some posts here it can further enhance the voice, but maybe not too much comparing with the above described MODs.

So thanks again for your inspirations and helps. Since I'm satisfied with the voice I really stop here the moddings on this panel. I will try to find a box and will put my ugly panel into it. :).

See some more pictures. Not nice, but sounds OK.

babolcs
 

Attachments

  • IMG_1294.JPG
    IMG_1294.JPG
    925.6 KB · Views: 718
  • IMG_1295.JPG
    IMG_1295.JPG
    896.5 KB · Views: 689
  • IMG_1296.JPG
    IMG_1296.JPG
    915.6 KB · Views: 650
Well, I couldn't resist, so one is on order. For my next project I think I will go through one of the new DAC boards and see what I find that needs modding. I will let you know more when it gets here and I can take a good, close look at it.
Do you get a delivery date estimate yet? If it improvs SQ over the one you are doing the mod with its I/V output, I definitely want to order one.
 
Do you get a delivery date estimate yet?

Not yet. It's one of those things where they set the status to shipped, then it sits there saying 'awaiting pickup.' Sometimes that can go on for a couple of weeks before it starts to move. You know.

I doubt the new board is going to be better than my modded board in the beginning. (The modded board already has an IV output.) The question is probably more like how much work would it take to get it there. Hopefully, it will take a lot less work than the first on did. But who knows, we will just have to wait as see.
 
Not yet. It's one of those things where they set the status to shipped, then it sits there saying 'awaiting pickup.' Sometimes that can go on for a couple of weeks before it starts to move. You know.

I doubt the new board is going to be better than my modded board in the beginning. (The modded board already has an IV output.) The question is probably more like how much work would it take to get it there. Hopefully, it will take a lot less work than the first on did. But who knows, we will just have to wait as see.
In fact, I fully expect the new board to be less than your mod. I want to know if it has SQ improvement over the unmodded old board. If yes, how would you quantify that?

The I/V mod is doable. But the clock mod is another level.
 
Don't know about the clock mod being on another level. Could be there are easier ways and harder ways to do it. The way I did it was probably more towards the harder side. But soldering 4 little wires to the four pads and then soldering a clock a positioned a little bit above the board to the 4 wires should fairly easy. Vibration could be a concern if being held only by wires, but once it is working one could maybe put a dab of hot melt glue, or of electronics silicone adhesive between it and the board to hold it in place and keep it from moving around.

Note: Important to be aware that electronics silicone adhesive is not the same stuff as is usually found in the local hardware store glue section, but it is probably possible to find it at Amazon or one of the electronics suppliers such as Mouser or Digikey.

Regarding quantifying the new board, maybe there will be an opportunity to take some measurements. Obviously, I could compare it to DAC-3 and give a subjective assessment. Another way of looking would be to wait and see how many mods I end up doing on it until I am happy with the SQ. In a way that might be the most telling at least to some people who might be interested to know what they could be in for if they want to make it better.
 
Last edited:
As a part of some testing with all the various DAC boards and digital audio connection options available I was going to try running all the 5v supplies from one 5v power supply. Had an LT3042 board that was no doubt overkill but should work at least to do some testing.

Turns out it didn't work with the AK4137 SRC board which hung at boot when powered from the LT3042 board. Turns out the 5v was coming up too slow for the SRC board microcontroller to boot properly. Needed to do something to get it working so decided to try modding the LT3042 board for fast-start operation.

That requires connecting 2 resistors to pin 6, which was found to be connected to pins 1 and 2 by a trace running under the chip. No easy way to get at the trace so decided to try lifting pin 6. First removed big filter cap on board to create better access to the chip. Must have tried a little too hard when lifting the pin with an Exacto knife because it popped up rather suddenly and pin 7 loosened slightly. Have to be more gentle next time. Anyway, re-soldered pin 7 and connected 2 resistors to pin 6. Fast start now works very quickly, and AK4137 board boots right up with no problems. Pics below.
 

Attachments

  • LT3042 Fast Start mod.jpg
    LT3042 Fast Start mod.jpg
    271.2 KB · Views: 616
  • LT3042 Lift.jpg
    LT3042 Lift.jpg
    748.3 KB · Views: 601
  • LT3042 Soldered.jpg
    LT3042 Soldered.jpg
    713.2 KB · Views: 312
Last edited:
The LT3042 as configured is for 5v, not 3.3v. Also, 1.5A or whatever amps seems like even more overkill to use for AVCC when it comes to an ES9038Q2M.

I do have a way I am working on to try LT3045 for AVCC, but I need to get 22uf Cset caps first. Not sure what caps I want to use for that, haven't even got that far. Also have some LiFePO4 batteries around here I could try. In addition, I have AD797 opamps to try. However, before butchering up my DAC board with probably unnecessary AVCC options, there are some other things I would like to do first.

Say, since you have a PRO board now have you measured AVCC current draw at various sample rates yet? If you want to know how you might best mod your new board's AVCC power that might be a good idea.

Also, not sure if you should start a new thread for it but I would like to see some good, close-up, in-focus, hi-res pics of every part of the board. I would like to read part numbers and visually follow traces. How about that?
 
Last edited: