Actually I have 3 of these cm6631 - one US$30 in my AMB y2 DAC by my bedside for headphone use, another US$30 in the dual AD1955 DAC (actually still working on this DAC and various isolation... too many projects !) and so this last one is for the ES9018, which can use the clocks from the usb board; so it makes sense to get the one with the clocks already available rather than hacking it like bigpanda et al did... they are more skilled than me... and it has a convenient 5V input, and two of the low noise regs, not just one, so in theory the clocks get cleaner power. However, the gnd is still connected so I'll have to remove the usb socket, so not completely hack-free.
BTW, if anyone wants any Tenor TE7022L devices like Teralink X2 etc... they are for sale ;-) I have a very nice ocxo 12mhz too ...
The best way to clean up the mclk etc is to use IanCanada's fifo. For what it is, it isn't expensive and will give far better results than any hack I can manage. However, I already have an fpga asrc to 384Khz with Crystek CCHD-957 in the dual ad1955 DAC, so for this es9018 dac, I want to see if the es9018 can beat the dual ad1955 without any expensive extras or hacks. The dual ad1955 cost around US$200 so it's become my benchmark for an affordable high quality DAC.
Anyway, I will test the usb board's clock outputs tomorrow and see if they are both on at the same time.
cheers
Tom
curved traces are really of dubious benefit for digital, for the analogue outputs yes, maybe MCLK, but its used for high speed analogue like RF primarily. I asked Acko about curved traces, hes an RF and digital systems engineer for MIL etc by trade, he uses curved traces on all analogue traces, but always angled for digital for shortest length
if you guys are really wanting to keep this cheap, you should really be group buying the common parts
and it seems they are.
No way I'm opening up a dac to switch a jumper every time, so either I use a PC resampler, or change to asynch. I assume I have to re-write the register to change to asynch ? I can't do this easily either so it seems the solution is the sox resampler.
No register change needed (there are no values to change) for synch/asynch. The DAC doesn't really know (or care). I think asynch is the way to go or synch if you stick to 44.1 K sample rate. But since this is diy you can try different options.
Put a socket in the clock pads so different clocks can be used (sort of like Ian's FIFO)
Put a socket in the clock pads so different clocks can be used (sort of like Ian's FIFO)
Thanks GLT.
I looked at the datasheet but couldn't find an answer. Actually, that datasheet seems quite spartan compared to most other DAC makers, but in fairness, it's not as bad as the one for CM6631.
CM6631_Datasheet_v0.8_°Ù¶ÈÎÄ¿â
Anyway, I'll use the 45mhz clock for sync on 44.1/88.2 and it'll asynch on everything else. It's not a perfect situation but hey ho.
The socket is a must and I'd hoped that v2 of the PCB would have made this easier to do with some through holes. Another change for v3....
I looked at the datasheet but couldn't find an answer. Actually, that datasheet seems quite spartan compared to most other DAC makers, but in fairness, it's not as bad as the one for CM6631.
CM6631_Datasheet_v0.8_°Ù¶ÈÎÄ¿â
Anyway, I'll use the 45mhz clock for sync on 44.1/88.2 and it'll asynch on everything else. It's not a perfect situation but hey ho.
The socket is a must and I'd hoped that v2 of the PCB would have made this easier to do with some through holes. Another change for v3....
Last edited:
Here's a photo of my efforts to fit an adp151 in the space around FB2, as the 3.3V digital regulator. I bent up the gnd pin of the reg before soldering, and I made a hole through to the ground plane rather than run a length of wire to a nearby ground. There's another photo to see where the hole comes out.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
I started to fit components today.
I think v3 should have 2 more changes :
-change the capacitor pads/holes for capacitors that decouple power supply for the the op amps. I understand you've tried to make this an easy build, but the spacings for the through hole caps around the op amps are too wide for suitable value caps. So, instead of putting pads for 0805 with the through holes before the pad, how about putting the pad and hole next to each other/side-by-side, so the though holes are 3mm apart ? It should also save space.
- change the oscillator pads and input holes so that the holes are dip8 spacing apart. This will allow people to fit a socket more easily if they choose to.
Cheers
I think v3 should have 2 more changes :
-change the capacitor pads/holes for capacitors that decouple power supply for the the op amps. I understand you've tried to make this an easy build, but the spacings for the through hole caps around the op amps are too wide for suitable value caps. So, instead of putting pads for 0805 with the through holes before the pad, how about putting the pad and hole next to each other/side-by-side, so the though holes are 3mm apart ? It should also save space.
- change the oscillator pads and input holes so that the holes are dip8 spacing apart. This will allow people to fit a socket more easily if they choose to.
Cheers
I suggest just putting the power supply rail inputs directly next to each opamp. since you dont have a regulator on the board anyway. that way you could avoid splitting the whole ground plane down the middle twice under the analogue current and voltage outs, the biggest failing on the PCB IMO
no need to worry
, a complete solid ground plane 4 layer board is coming soon for those afford to pay more this 2 layer pcb is target to the simplest and cheapest in mind to achieve the highest c/p ratio.Actually, there are many split IV DAC design on the market, the ground plane is much more fracture.
There are also many application note told analog and digital ground should be split, but star ground is always use in tube design. Actually, ferrite bead can also add noise to the power supply line but do reduce EMI. There are also many discrete component opamp in the market. There are good and bad when something is add to the design, it's just diyer's taste. but I always think the simplest solution is the best, many of my previous design is copied
. Actually, v1 pcb is the simplest welcome to copy, please keep DIYINHK logo and add the modifiy by yourname line is ok
umm my suggestion allowed the lack of the silly splits without 4 layers... you actually dont need to use 4 layers to do this properly. since you have no onboard regulators they are there for no reason ruining stereo crosstalk in the process. your comment about splitting analogue and digital planes is meaningless, you havent even done that, youve just put random and completely destructive slices through the analogue return path. plus the ES9008 app note you copied most of the design from (well the schematic anyway) specifically mentions that being of no benefit. It also specifically says not to split the ground plane under the current outputs.
I dont think there is any danger of me copying your work...
I dont think there is any danger of me copying your work...
Last edited:
The current version 2 has a 4mm-wide area at the edge of the board so the ground plane on that side is not completely split. It's on the opposite side to the op amp power entry point.
I think 4 planes is not necessary, but I'd like to see better use of the ground planes with the 2 layer.
Creating separate L/R power inputs will allow a larger area of ground in the centre of the board (instead of the small connecting area at one side). If you do this and add more ground plane vias between the two sides of the board then the grounding should be improved. Take a look at this board from AMB's site - see all the vias ?
I think 4 planes is not necessary, but I'd like to see better use of the ground planes with the 2 layer.
Creating separate L/R power inputs will allow a larger area of ground in the centre of the board (instead of the small connecting area at one side). If you do this and add more ground plane vias between the two sides of the board then the grounding should be improved. Take a look at this board from AMB's site - see all the vias ?
You are welcome!
Please keep the DIYINHK logo and add the modifiy by yourname line!
I am happy to see someone done customization on my design
Do you have any DAC PCB design before? I am happy if you can share with us
You are welcome!
Please keep the DIYINHK logo and add the modifiy by yourname line!
I am happy to see someone done customization on my design
Do you have any DAC PCB design before? I am happy if you can share with us
他不尊重你的設計。他不希望將其複製。
Are you sure ? I've picked up on considerable levels of disdain, even contempt. Sorry Qusp, but it's getting old, fast.
Previously it was said that the power lines run under the analogue output lines. To be clear, they run under the op amps themselves and there are 8 signal resistors running perpendicular to the power lines on the other side of the board. Perpendicular is not a problem, and there are ways to mitigate even this for the perfectionists out there. There are no signal traces crossing the lines, and certainly none paralleling.
Last edited:
I made a suggestion you seem to be agreeing with, I made it without being insulting, I got sarcasm in return. I wonder how many of the suggestions i've made will be on the new PCB, I see one or 2 i've made already on the update. i've got nothing but sarcasm in return from the beginning, i'm not sure how much help is due at this point, but here I go
There will be a new set of problems with a 4 layer prototype that you will pay for, it will be expensive. I agree with you, it would be better to fix the 2 layer design. 4 layers is great, but thinking it will solve all of your problems when the problems on 2 layers do not seem to be understood is foolhardy and the customers are paying for it.
the problem is not just the signal and IV resistors (the most critical resistors on the board) crossing over a massive power trace. Being perpendicular helps avoid crosstalk, but if it can be avoided it should be and its not the issue under discussion (it can be fixed by supplying the opamps from inputs next to each one, regs arent onboard, why have the inputs be destructive?) the problem is the return path is completely blocked but for a tiny gap on one side... twice....
a signal current and its return currents like to flow near each other, if a signal trace is routed on top of a 2 layer PCB with a ground plane underneath, the return likes to flow underneath the signal path and that includes the resistors. the return impedance is as important as the signal impedance, they are peas in a pod. thinking of signal as somehow separate from ground is incorrect and the sooner you get past that, the better.
then there is this misunderstanding that the opamps are analogue and the dac is digital, if that wasnt just an excuse...
this is all from the es9008 app note page the schematic was mostly lifted from
do you think it might be important to avoid cutting the ground plane/blocking the return currents?
Last edited: