• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

New FIFO buffer for RPI/SBCs

Thats really nice :)


OK so new firmware (MCU) was made and tested today. We retested 20Khz, 15Khz and 10Khz (at those frequencies the old firmware had a distortion , for example at 20Khz the thd+N was -50db!!!)

With new firmware the problem was eliminated (thd_+N is almost the same from 1Khz to 20Khz)but we will wait until tomorrow to finish all tests. No hardware change whatsoever




I would estimate that Katana will ship again Wednesday/Thursday next week .
 
I guess I could. First, hopefully things will turn out well in the end for Katana. I sure hope so because we sure seem to need a very good and accessible DAC at an affordable price.

What we did here was set up three dacs all playing the same music through a very low, SOA low distortion power amp. if there is any distortion at all it will be audible. The first dac is a Benchmark DAC-3 which is a well known reference standard. The second was a modified Chinese ES9038Q2M dac I have been describing in another thread over here: ES9038Q2M Board (So far, nobody has modded one of those dacs as much as I have because it is a lot of work, too much work for pretty much everyone, it seems.) The third dac was Katana.

Because the manual for Katana said it sounds best with 3 power supplies, and because it just says the power supplies should be 5v, 4A, I took that literally as follows: The RPi was powered by its original wall wart supply, the wire leads from the Katana board were connected to a clean linear 5v supply, and the USB connector on the microcontroller was connected to a 5v wall wart switcher. Thought I would try that first, at least.

The dacs were level matched with an SPL meter reasonably close. Two listeners compared all the dacs against each other as much as they wanted and they could adjust the volume levels up and down for comparison purposes as they wished. Both listeners ranked the dacs in this way: 1) Dac-3, 2)Modded Chinese dac, and 3)Katana.

It was noticed that Katana sounded the most different of the three, with a full, round midrange, but some unnatural distortion at higher frequencies such as cymbal sounds.

So, cdsgames and I exchanged a few emails. I described the findings and asked about possible power supply issues. Before I had a chance to exactly describe each of the power supplies I was using, I got an email back saying they had found a problem and were working on it.

That's all I know at this point.

The only thing I am really sort eager to find out when the updated firmware is available is whether the dac rankings will change and my modded dac will come in third, or maybe it will be a tie. It would probably be best for virtually everyone if I could honestly say Kantana comes in 2nd place. There is no way it can beat DAC-3, but if it could beat my modded dac it will be a very good dac, very possibly the best dac under $1,000 as cdsgames would like it to be. At $250 it would be steal, IMHO. Assuming that happens, of course.
 
Last edited:
"There is no way it can beat DAC-3"

Why not?
* What matters is the implementation of a DAC not the DAC chip itself in the first place
* USB DACs are flawed by definition
* And than that DAC-3 is stuffed with all kind of clutter
* there's more - you can very easily fool yourself with

On the first glance you doesn't seem to have squeezed most out of your
Katana setup.

E.g.
* The way you powered the PI doesn't seem OK to me
* No word about the PI SW setup
* What DAC filters are u using?
* ...

I'm pretty confident that a Katana tower plays extremely well if properly implemented and burned in!
I mean - I do prefer my (IMO rather well implemented) Kali-Piano21 combo over several commercial DACs I owned in the past.
I'd expect Katana to play in a different league - if properly implemented.
A pity that Nirvana is not ready yet!

And then your testers also mentioned Katana sounds different to the other DACs - beside the detected flaws.
That's the very first thing I told cdsgames (he'll confirm that) - different in a to me very promising way!

Just a remark: Talking about a 250$ DAC is misleading. With PS and PI you quickly run into the $500 range.
And that are $500 in the DIY arena! Perhaps there also need to be an isolator-HAT put to the BOM sooner or later
to squeeze the last bit out of the setup.
That means we're talking factor 2 to 3 devices in the commercial area that can be considered
competition. At these levels you'll find dual-mono ES9038Pro implementations.
Even in the $1000 range you can find this or that pearl. It's gonna be a tough race.


Let's see how things develop. IMO it's much too early to draw conclusions.

Enjoy.



PS: A pity that the controller board needs to be swapped for a firmware change. Would be nice if a binary could be downloaded to it somehow (via PI/GPIO or USB-C).
 
Last edited:
"There is no way it can beat DAC-3"
Why not?
* What matters is the implementation of a DAC not the DAC chip itself in the first place
* USB DACs are flawed by definition
* And than that DAC-3 is stuffed with all kind of clutter
* there's more - you can very easily fool yourself with
On the first glance you doesn't seem to have squeezed most out of your
Katana setup.
E.g.
* The way you powered the PI doesn't seem OK to me
* No word about the PI SW setup
* What DAC filters are u using?
* ...
I'm pretty confident that a Katana tower plays extremely well if properly implemented and burned in!
I mean - I do prefer my (IMO rather well implemented) Kali-Piano21 combo over several commercial DACs I owned in the past.
I'd expect Katana to play in a different league - if properly implemented.
A pity that Nirvana is not ready yet!
And then your testers also mentioned Katana sounds different to the other DACs - beside the detected flaws.
That's the very first thing I told cdsgames (he'll confirm that) - different in a to me very promising way!
Just a remark: Talking about a 250$ DAC is misleading. With PS and PI you quickly run into the $500 range.
And that are $500 in the DIY arena! Perhaps there also need to be an isolator-HAT put to the BOM sooner or later
to squeeze the last bit out of the setup.
That means we're talking factor 2 to 3 devices in the commercial area that can be considered
competition. At these levels you'll find dual-mono ES9038Pro implementations.
Even in the $1000 range you can find this or that pearl. It's gonna be a tough race.
Let's see how things develop. IMO it's much too early to draw conclusions.
PS: A pity that the controller board needs to be swapped for a firmware change. Would be nice if a binary could be downloaded to it somehow (via PI/GPIO or USB-C).

+1
I think there is nothing special about DAC-3.
I would like to try Katana with a very good PS like Paul Hynes SR4 or SR7.

Matt
 
Hi all ,


Mark has very graciously tested the setup and reported back that he thinks hi frequencies sound "dry" and some distortion was audible



Instead of dismissing his report , a flurry of activities started ..


We tested all boards and... yeap he was absolutely right.


1-9Khz THD+N was good, 10-20 was really bad. (less then -50)
Hmmm so clearly Mark was hearing a real issue.



Anyway we found the firmware problem , we recalculated and retesting. As soon as testing is complete I will give a proper answer.


Hardware is already here , Jtag is on the way.


Thx to all (Mark , Soundcheck) for helping
 
Ok proper answer




Someone in our soft team f***** up royally .


Our MCU driver is set as master and uses 128xSR in synch mode



To make you all understand the meaning of this.


Ess sabre has a great feature call DPLL. Dpll has bandwidth divided into 15 settings + OFF = 16 (in asynch mode).



DPLL takes your jittery clocks and cleans them . Great right ?
No, not al all , not for Katana clocks . You see any dpll adds jitter. Usually its much smaller than jitter from original clocks , but Katana clocks were designed to have absolute low jitter (femtoseconds and I am not talking marketing bull)


Staying on the same subject , I will refer you to Mark , he has done some experiments on the Dpll of ess9038q2m and sound quality.


He was reporting that with dpl banwith going from 4 lowest to 3 lowest , sound quality has improved. I believe he wants to go down further to 2 lowest or maybe to 1 (lowest dpll bandwith)


Now remember dPLL adds jitter. Mark is hearing jitter by changing to a lower setting on dPll (lower bandwith = lower added jitter)


Enter Synch mode with Katana. Why use dPLL at all on ess9038q2m ? Our clocks are clean so we use a little trick called SYNCH mode where dPLL is bypassed . Direct clocks no added sauce . Dpll setting = 0



Makes sense right ? Hold on, electronics are never simple. Enter DOP. DOP is not supported in synch mode.


Behind the scenes we implemented a very sophisticated driver that sw betwen asynch (dop) and synch (pcm). A hot head changed some settings though and last firmware is using synch with mclk meant for asynch (synch mclk is always 128xSF) We tested at 1Khz..but not higher.


Mark tested the Katana with a distortion of -50!!! at 10Khz and higher. We are getting now same numbers from 0-20 Khz. Apples will be compared with apples .



On DAC 3 and Katana.

The architecture its different. DAC 3 puts its trust in sample conversions and digital filters. They are using DPLL (no idea what bandwith) .Analog stage is text book implementation of a normal opamp with emphasis on great THD+N.

Katana puts its trust on clocks jitter (like Mark has discovered and like Kali has showed to many people, less jitter = incredible sound) supercaps on AVcc and dicreete opamps for sound quality.
Oh yeah THD+N on Katana is -112.75 while DAC 3 is -113. Lets call it (almost) a draw.


Will you hear the distortion diff between the DAC3 and Katana ? Impossible
Will you hear the difference in jitter betwen the 2 DACs ? Absolutely you will
Will you hear the Katana supercaps electons in AVCC vs textbook implementation of Avcc of ess Sabre ? Yes you will
Will you hear the difference betwen monolitic opamps vs discreete design in class A amplification that Katana has ? YES YOU WILL.


I have a lot of respect for Mark , he has done a lot of work helping this community (check the ess9039 thread) , but we never shared a glass of wine . He is not my buddy. His comments can be trusted . Let him test again the Katana with the correct firmware vs Dac 3 and far east designs and share his thoughts .



Meanwhile , Soundcheck , Greg and many others will do the same. Opinions, warts and all, are welcome thats what makes us strong.
 
Last edited:
cdsgames - just read your comments.

Q:
There's a "DOP" on/off switch offered by the driver through alsamixer.
I'm wondering if that one equals "async clocking" (for DOP) on/off then !?!?


sync vs async on Sabre

Just to confirm:
I've been running sync clocking from Kali towards other ES9023 HAT DACs (the clock on the DAC I removed) some time ago.
The difference was well worth that exercise. There's no going back if you drive it this way.


"opinionated critics"

Not in a destructive way I hope. :D

As long as there's constructive feedback involved I know you guys appreciate it.
e.g: Now - finally all LEDs a green ;) ( they still can't be turned off altogether though ;) )



I do have to comment one thing again.

The power wire interfaces.

Jtag for 15/-15/gnd output stage. Some kind of molex for 5V on the DAC.
No DIY friendly connector for the 5V of the controller at all - just USB-C.
Locations for these are left and right and none on the front.
On the front side we've got the PI power.
Bottom line. For those using multiple supply options, they'll potentially be looking at a nice EMI/RFI inviting wire-web with all kind of connector/cable types being involved. I think it would be very easy to get that straightened out.
 
One thing I noticed is the power up sequence in case of powering the boards with different supplies. How should one solve this issue best? Do you need delay circuits for each power supply to keep the correct sequence? Would have been nice to apply this in the DAC directly e.g. waiting until circuit that needs power first gets it then releases the next one and this one the next.. or probably you might provide such a power sequence board as add. component? Guess nobody wants 3 power switches on his DAC.
Regarding AVCC, did you compare your super cap solution with some opamp reg solution? Lot of electrons are nice but the caps must be charged and impedance of the caps at low frequency is increasing what is not best for de-coupling. How did you solve this and get good PSSR?
Can't wait for the first reviews! Great design!
 
Soundcheck , DOP is just to switch DOP off not synch vs asynch



Yeap I am aware of the power and and connections but there are multiple things to take in consideration and choices we had to make. Still I will keep in in mind.


Freezeboz, power sequence should not be a problem , we will try to expand on that on later stage. Impedance of SC at low frequency is at its lowest , for higher frequency (where SC is not that good) we have lots of film decoupling. I did not compare direct vs opamps because opamps used for 2 things:


1.Low low noise in audio band (but we achive even better with with a triple filter on Avcc , such as "ripple eater circuit , LPF and LDO.). Of course L and R are treated like different AVccs and both receive indepandant rails with separate ripple eater , LPF and LDOs
2. Low impedance in audioband. I can assure you that a SC + decoupling has better impedance than opamps and most important the loopback of electrons is only few mm..






Our design is well taught. I stand by it.


No opamps for us in the very important Avcc (its textbook implementation and works well though). Still SC are better for SQ if you ask the "few" here that bought the Boss Dac.
 
Last edited:
I must say, kudos to Markw4 & his mates!

They listened, reported back what had to be some really uncomfortable news, and were proven correct.

Also thanks to cdsgames for sharing all the details. Obviously that's completely the DIY ethos, but when it's your professional baby & your money, it can be hard to do. Much easier to vaguely blame it on sw, and carry-on like nothing happened.

Personally, I've been sitting on the fence whether I really need a $250 DAC -- I'm no where near the level of enthusiast that many on the forum are -- but how could one not want to support such a great guy.
 
...at some point the DAC becomes "transparent" given limitations in the rest of one's listening chain.

Quite true. This is why I strongly recommend that people who are modding dac boards over in the other thread also mod an LME49600 head phone amp board for high performance. That way they have something they can rely on to evaluate changes to dac SQ with mods they are making.

The HPA boards I encourage folks to mod are like this to start with: 1PCS LME49720NA+LME49600 headphone amplifier KIT | eBay

Much like the cheap ES9038Q2M DAC boards, the HPA boards are junk to begin with but can be made very accurate using things like low-distortion SMD thin film resistors, an added ground plane, proper bypassing, etc.
 
IIs there a thread where these mods are detailed?

There are some pictures here: http://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-63.html#post5402398

After taking a look if it still is of interest please post over in the other thread and we could talk more or I could answer any questions about the particulars of modding. Also, probably worth downloading LME49600 data sheet which has some helpful application info.

EDIT: It may help to know, if anyone doesn't already know, that when viewing pictures in the forum they do not always show at full size and resolution. When a picture is open, if you hover the mouse over the lower left corner and white X appears, that means higher resolution can be selected by clicking on the X.
 
Last edited: