@Soundcheck,
More later, but try the MBLS in sync mode. See the Allo.com docs on the Kali for the correct expansion pin where the MCLK is present from the Kali, remove the 51R between the clock and the DAC chip, and run a wire from the expansion pin to the DAC chip side pad where the 51R was.
I suspect that will be a good lift and bring your MBLS closer to your HFBD+P.
Also upsampling to 384 might be a good thing then.
Greg in Mississippi
More later, but try the MBLS in sync mode. See the Allo.com docs on the Kali for the correct expansion pin where the MCLK is present from the Kali, remove the 51R between the clock and the DAC chip, and run a wire from the expansion pin to the DAC chip side pad where the 51R was.
I suspect that will be a good lift and bring your MBLS closer to your HFBD+P.
Also upsampling to 384 might be a good thing then.
Greg in Mississippi
Thx for the advise.
How about approaching it slightly different? by
1. removing power from the MB clock ( which also eliminates potential
clock interferences ) and then
2. attach the Kali MCLK in front of the 51R. This way there'd be still some damping in place.
And - I'd need to verify that -
3. there'd be only one 3.3/3.6V reg for the entire DAC required to feed just AVCC
if I'm not mistaken. That'll make the whole AVCC 3.6V mod setup less complex.
Basically I could reuse (no extra cost) my single 5V Mravcla reg and just reconfigure it for 3.6V (2nd onboard reg would also be taken out of operation ).
Another nice sideeffect would be ending up with just a single stage regulation.
How about approaching it slightly different? by
1. removing power from the MB clock ( which also eliminates potential
clock interferences ) and then
2. attach the Kali MCLK in front of the 51R. This way there'd be still some damping in place.
And - I'd need to verify that -
3. there'd be only one 3.3/3.6V reg for the entire DAC required to feed just AVCC
if I'm not mistaken. That'll make the whole AVCC 3.6V mod setup less complex.
Basically I could reuse (no extra cost) my single 5V Mravcla reg and just reconfigure it for 3.6V (2nd onboard reg would also be taken out of operation ).
Another nice sideeffect would be ending up with just a single stage regulation.
Last edited:
Soundcheck, that sounds good.
On removing the power, consider cutting the 5v trace to the clock regulator just before the input capacitor. You could then just scrape off the green solder mask and do a solder bridge to re-connect it.
Also if it sounds promising, consider cutting the clock output to 51R trace just before the resistor to remove the clock output from the circuit.
Good thought to keep the 51R in place. I did not like 0R in that position. OTOH, it might need a different value there for best termination and best sound given the longer feed from the Kali. I bet you'll find an optimal value between 22R - 100R.
Your thoughts on trying 3.6v on the AVCC should work ok. Always good to have the leads length as short as possible there.
Greg in Mississippi
P.S. I hope to get to trying sync mode on my modified Mamboberry this weekend along with trying my modified HiFiBerry DAC+ Pro in slave mode on a Kali. Of course, will report when I do.
On removing the power, consider cutting the 5v trace to the clock regulator just before the input capacitor. You could then just scrape off the green solder mask and do a solder bridge to re-connect it.
Also if it sounds promising, consider cutting the clock output to 51R trace just before the resistor to remove the clock output from the circuit.
Good thought to keep the 51R in place. I did not like 0R in that position. OTOH, it might need a different value there for best termination and best sound given the longer feed from the Kali. I bet you'll find an optimal value between 22R - 100R.
Your thoughts on trying 3.6v on the AVCC should work ok. Always good to have the leads length as short as possible there.
Greg in Mississippi
P.S. I hope to get to trying sync mode on my modified Mamboberry this weekend along with trying my modified HiFiBerry DAC+ Pro in slave mode on a Kali. Of course, will report when I do.
Last edited:
Greg.
Looking forward to hear about your sync-clock mod.
If it turns out to be worth it, I for sure will do it too.
###########
Another update.
I mentioned earlier that I've been trying a Gbit ethernet dongle based on a Realtek network chip. .
Max transfer rate went up to around 300Mbit/s. ( that's quite beneficial for my bulkoad/ramplayback setup -b 40000:660000 . )
Last night I was thinking to insert a AQ Jitterbug that's been sitting idle for quite some time.
Just attaching it to the USB port supposedly cleans up the power rail.
I then attached the ethernet-dongle to the Jitterbug.
I do think things further improved. But I need to verify it again.
Obviously I'd never recommend to buy a $50 Jitterbug to filter an external dongle.
However. For those who have a JB at home give it a try. I'd like to hear what you think about it.
The dongle should work on Moode and on pCP. pCP needs the net-usb.tcz to be loaded.
The related kernel module is called "r8152".
Looking forward to hear about your sync-clock mod.
If it turns out to be worth it, I for sure will do it too.
###########
Another update.
I mentioned earlier that I've been trying a Gbit ethernet dongle based on a Realtek network chip. .
Max transfer rate went up to around 300Mbit/s. ( that's quite beneficial for my bulkoad/ramplayback setup -b 40000:660000 . )
Last night I was thinking to insert a AQ Jitterbug that's been sitting idle for quite some time.
Just attaching it to the USB port supposedly cleans up the power rail.
I then attached the ethernet-dongle to the Jitterbug.
I do think things further improved. But I need to verify it again.
Obviously I'd never recommend to buy a $50 Jitterbug to filter an external dongle.
However. For those who have a JB at home give it a try. I'd like to hear what you think about it.
The dongle should work on Moode and on pCP. pCP needs the net-usb.tcz to be loaded.
The related kernel module is called "r8152".
I tried the sync-mode on the MB this weekend. Very nice, now sounds direct like the other sync-mode DACs I have and that ESS ASRC 'haze' is gone. Will take some adjustment of the series R in the MCLK line, too much emphasis in the highs with the 51R, too much the other way with a 100R, have a 65R or so there now, I suspect 75R-85R will be optimal, but want to leave the 65R in there for awhile and listen and let it settle in before changing it (and I need to get some Rs in those sizes too).
No time to post more now, will catch up later with pix of the sync-mode mod, results of a HFBD+P in slave mode on a Kali, and some networking upgrades including a pair of fiber media converters to provide ground isolation, in-line filtering, and linear/battery power.
Greg in Mississippi
No time to post more now, will catch up later with pix of the sync-mode mod, results of a HFBD+P in slave mode on a Kali, and some networking upgrades including a pair of fiber media converters to provide ground isolation, in-line filtering, and linear/battery power.
Greg in Mississippi
Last edited:
Vasilis,
I re-read what you and Soundcheck wrote about the MB & MBLS output filtering. I MIGHT try replacing the 10R with a 12R, but more likely will go for 0R there and re-try the 100R series resistor in the MCLK feed line for sync-mode. Might be the best overall.
A related question... what type of capacitor do you have for the 4700pf currently. Looks like either mylar or polyprop. I generally prefer acrylic (PPS), polyprop, polystyrene, and Teflon, in that order. If those are mylar, I will try polyprops, I have some Wima 4700pf here.
Greg in Mississippi
I re-read what you and Soundcheck wrote about the MB & MBLS output filtering. I MIGHT try replacing the 10R with a 12R, but more likely will go for 0R there and re-try the 100R series resistor in the MCLK feed line for sync-mode. Might be the best overall.
A related question... what type of capacitor do you have for the 4700pf currently. Looks like either mylar or polyprop. I generally prefer acrylic (PPS), polyprop, polystyrene, and Teflon, in that order. If those are mylar, I will try polyprops, I have some Wima 4700pf here.
Greg in Mississippi
Greg.
Good to hear about your success story.
Another idea.
I was thinking how to route the clocktrace better.
How about drilling a hole through the Mambo roof?
This way you'll get direct access to the 2nd header on the Kali.
I'm not sure though if the MCLK is available there.
The tracelength would be shortened by at least 2/3 I'd guess.
########
Over the last couple of days things further broke-in.
Luckily the soundquality improved.
Yesterday I tried something new.
I've been swapping the Meanwell/Mravcla combo against the iFi iPower/47k Rifa combo back and forth.
For the time being I use these as PI supply only.
First impression:
The iPower combo caused a more dynamic powerful image. I guess the Rifa cap delivers more than enough energy when needed.
However. The center seemed to be slightly overpowered. Somehow the
relation between left/right and center seemed to be unbalanced.
With the Mravcla Combo the image in front of me looked perfectly well balanced and still
exhibited a pretty deep, wide and evenly filled stage.
Certain percussion instruments rendered 3D nicely.
Somehow that overall energy was missing though. I need to look into this further.
Summary
The overall situation gets more satisfactory every day. It's pretty surprising though how much variables are still in the game - even with a Kali in place.
Enjoy
Good to hear about your success story.
Another idea.
I was thinking how to route the clocktrace better.
How about drilling a hole through the Mambo roof?
This way you'll get direct access to the 2nd header on the Kali.
I'm not sure though if the MCLK is available there.
The tracelength would be shortened by at least 2/3 I'd guess.
########
Over the last couple of days things further broke-in.
Luckily the soundquality improved.
Yesterday I tried something new.
I've been swapping the Meanwell/Mravcla combo against the iFi iPower/47k Rifa combo back and forth.
For the time being I use these as PI supply only.
First impression:
The iPower combo caused a more dynamic powerful image. I guess the Rifa cap delivers more than enough energy when needed.
However. The center seemed to be slightly overpowered. Somehow the
relation between left/right and center seemed to be unbalanced.
With the Mravcla Combo the image in front of me looked perfectly well balanced and still
exhibited a pretty deep, wide and evenly filled stage.
Certain percussion instruments rendered 3D nicely.
Somehow that overall energy was missing though. I need to look into this further.
Summary
The overall situation gets more satisfactory every day. It's pretty surprising though how much variables are still in the game - even with a Kali in place.
Enjoy
MCLK is available on pin 29 from the Kali. With the MBLS, since the header is mirrored to allow easy access to the pins, you have a good, fairly straight shot from the mirrored pin 29 over to the MCLK feed series R.
I did my mod by removing the series R, putting a small piece of Kapton tape over the pad that goes to the onboard clock output, them soldering the series R back on the pad going to the MCLK input on the ES9023. Then I ran a small wire from pin 29 over to the now unsoldered end of the series R.
Afterward changing out the series R is very easy and quick.
I don't think you can get a shorter run than that, but I might be missing something.
Greg in Mississippi
Last edited:
PFR-470: 4700pF 63V Evox Rifa PFR Polypropylene Film, Aluminium Foil Capacitor | Hifi Collective
But in some batch we have use the red wima , there is not difference ...
Vasilis...
Thanks. Given that it is already polyprop, I won't change it out for the Wima... I haven't found much diff between them either.
I'm not much of a silver mica fan... nor am I feeling the filter cap is a major impediment at this time, so I'll leave them alone.
THANKS!!
Greg in Mississippi
Thanks. Given that it is already polyprop, I won't change it out for the Wima... I haven't found much diff between them either.
I'm not much of a silver mica fan... nor am I feeling the filter cap is a major impediment at this time, so I'll leave them alone.
THANKS!!
Greg in Mississippi
"feeling" -- Hmmh.
You'd better get your priorities right. Easy wins first.
And... ...just to quote your own signature: "Everything matters!"
PS: Regarding the clock trace idea I had. I had the impression that the MCLK was also available on the JTAG header. My fault.
This is obviously not the case. Then PIN29 is the pin to access.
Last edited:
"feeling" -- Hmmh.
You'd better get your priorities right. Easy wins first.
And... ...just to quote your own signature: "Everything matters!"
<SNIP>
And in the real world, one must prioritize what to do next.
I generally focus on the things that I've heard make significant & decisive differences. In digital EQ, those are:
1. Digital input signal integrity.
2. Low ground noise.
3. Good wide-bandwidth low-noise and low output impedance power supplies.
In my experience, these are primary. Get these right and you get the best out of a particular unit. While other things such as output signal path are important (even moreso when it is an active stage and even moreso when it is an active I/V stage), IN MY EXPERIENCE they are secondary and do not make as large of a difference.
AND the next thing for me with the Mamboberry is getting the input clock signal massaged to minimize or eliminate the high-frequency emphasis I find I get with this setup (and the ES902x in general). Sync mode was a good step in the right direction, I need to tame the treble to really make it sing.
And then there is prioritizing time. I have a ridiculous number of projects going currently... with the HF emphasis tamed in the MB, I'll turn to those other projects... mainly because as GOOD as the MB has gotten to my ear, I still think there is more potential in the HFBD+P on the Kali once I feed the Kali from a better supply. AND I have 5 other significant digital projects that are either already above what I feel I can get from either the MB or HFBD+P or that will be above them with the next set of mods.
Of course we all have opinions and bias. These are mine. I modify for me. You do so for you. There is no expectation that what I do will please you and vice versa. That we can find gems in each other's work is gravy and I'm happy that we do. But that won't always happen.
Later!
Greg in Mississippi
P.S. I should add that I would have chosen the capacitors Vasilis already has in there or an equivalent Wima. Next up for me are polystyrenes, I MAY try some here, but that is down the road.
Last edited:
Vasilis,
After reading the writeup of your process in designing the Ideon Audio Ayazi DAC, I now feel a bit sheepish in poking around in your Mamboberry. I do appreciate your forebearance in our efforts to get better sound out of the beast in our setups... and your graciousness around our efforts too!
So I have a question for you... as I said above, after going to sync mode from the Kali MCLK, I like what I hear, but have a bit of high-frequency emphasis. A 62R series resistor in the MCLK line seemed too low, 100R seemed too high. I'm considering either:
1. An 85R-90R MCLK series R.
2. Going back to the 100R MCLK series R while going to 0R on the output series resistors.
But this is with a fairly ordinary link from the MCLK pin on the header to the MCLK series resistor... a length of 28g or so vinyl-insulated wire. I had originally thought to use a piece of u.fl coax with the shield connected at both ends. Do you think it is worth going to that first before playing with the MCLK series resistor value?
TIA!
Greg in Mississippi
After reading the writeup of your process in designing the Ideon Audio Ayazi DAC, I now feel a bit sheepish in poking around in your Mamboberry. I do appreciate your forebearance in our efforts to get better sound out of the beast in our setups... and your graciousness around our efforts too!
So I have a question for you... as I said above, after going to sync mode from the Kali MCLK, I like what I hear, but have a bit of high-frequency emphasis. A 62R series resistor in the MCLK line seemed too low, 100R seemed too high. I'm considering either:
1. An 85R-90R MCLK series R.
2. Going back to the 100R MCLK series R while going to 0R on the output series resistors.
But this is with a fairly ordinary link from the MCLK pin on the header to the MCLK series resistor... a length of 28g or so vinyl-insulated wire. I had originally thought to use a piece of u.fl coax with the shield connected at both ends. Do you think it is worth going to that first before playing with the MCLK series resistor value?
TIA!
Greg in Mississippi
Looking forward to hear your findings.
I'm still fiddling around with the PI power supplies. Somehow that Meanwell/Mravcla combo
became a bit annoying after a while.
I swapped the iPower/Rifa combo back in.
What I finally did now is cutting off 2/3 (around 1m) of that IMO much too long iPower cable.
That also required to dissect and move the ferrite bead.
I remounted and fixed it with heat-shrink afterwards.
I think more important is to have low noise ""plus"" plenty of fast energy
available in a pre-reg stage, rather than having a perfect voltage only.
I havn't tried though to hookup up a large cap after the Mravlca. Perhaps I'll try that later.
I'm still fiddling around with the PI power supplies. Somehow that Meanwell/Mravcla combo
became a bit annoying after a while.
I swapped the iPower/Rifa combo back in.
What I finally did now is cutting off 2/3 (around 1m) of that IMO much too long iPower cable.
That also required to dissect and move the ferrite bead.
I remounted and fixed it with heat-shrink afterwards.
I think more important is to have low noise ""plus"" plenty of fast energy
available in a pre-reg stage, rather than having a perfect voltage only.
I havn't tried though to hookup up a large cap after the Mravlca. Perhaps I'll try that later.
A good friend came by my house the other day with his custom made 9023-based DAC. We tried 2 sets of LPF caps of almost identical values, they both gave pretty different results in the treble department. Perhaps you should explore that area after all..
I got a question to the more experienced I2S folks over here.
If running the DAC in synchronous mode the datasheet clearly states the supported
MCLK frequencies per fs. That probably differs from DAC2DAC. Lets talk about the Sabre.
Where and how it is defined and which clock values are applied per fs in our case? (Please - KISS)
Sorry for my ignorance.
PS-EDIT: Just another question popped up. How about fs > 192kHz (256*192000~50MHz). It seems to me, that the clocks wouldn't support samplerates higher than 192kHz in sync mode!?!?
If running the DAC in synchronous mode the datasheet clearly states the supported
MCLK frequencies per fs. That probably differs from DAC2DAC. Lets talk about the Sabre.
Where and how it is defined and which clock values are applied per fs in our case? (Please - KISS)
Sorry for my ignorance.
PS-EDIT: Just another question popped up. How about fs > 192kHz (256*192000~50MHz). It seems to me, that the clocks wouldn't support samplerates higher than 192kHz in sync mode!?!?
Last edited:
What I discovered last night is that the Mamboberry IS NOT a Mamboberry LS. Dumb, huh? I went to jumper out the '10R' series resistor on the output and saw 2 red lines instead of brown and black as I expected... I never memorized the entire resistor color code, but I did know that was 2.2R or 22R or 220R or something like that, not a 10R. It measured 2.2R.
Vasilis said "The RPI have some emphasis in hi freq , special in very expensive systems sound a little bit hard , with 10 ohm the effect almost gone , general the 10 ohm soften a little bit the result". And I found a similar effect in the MCLK series R, with higher toning down the treble. Because I had some treble emphasis, I originally intended was to jumper out the 10R output series resistor and change my MCLK series R back to the 100R that sounded too dull before.
BUT now that I realized the output series resistors on the Mamboberry were 2.2R, the next thing to try was to put 10Rs there. I went to my stock and pulled out a couple of good Vishay/Dale CN55 10Rs and put them in.
And listened.
And was VERY HAPPY. Now the promise of the sync mode was realized, with the haze gone and only a very slight tilt towards the highs. Music was very dynamic, exciting, and detailed. Vocals were very well delineated, with multiple overdubs very clear. Also things retained composure and complexity during high level passages... I attribute that to the Uptone Audio LPS-1 supplies.
Of course this might be 'first listen effect', where it sounds fabulous the first time you listen to it. Or it might sound diff after the resistors have some time on them. I'll check again tonight.
And the other thing I believe I discovered is that the resistors in both the Mamboberry and MBLS that look like carbon composite likely are... and I tip my hat to you Soundcheck, there was a bottleneck in the output filter, just not the one we were discussing.
Greg in Mississippi
Vasilis said "The RPI have some emphasis in hi freq , special in very expensive systems sound a little bit hard , with 10 ohm the effect almost gone , general the 10 ohm soften a little bit the result". And I found a similar effect in the MCLK series R, with higher toning down the treble. Because I had some treble emphasis, I originally intended was to jumper out the 10R output series resistor and change my MCLK series R back to the 100R that sounded too dull before.
BUT now that I realized the output series resistors on the Mamboberry were 2.2R, the next thing to try was to put 10Rs there. I went to my stock and pulled out a couple of good Vishay/Dale CN55 10Rs and put them in.
And listened.
And was VERY HAPPY. Now the promise of the sync mode was realized, with the haze gone and only a very slight tilt towards the highs. Music was very dynamic, exciting, and detailed. Vocals were very well delineated, with multiple overdubs very clear. Also things retained composure and complexity during high level passages... I attribute that to the Uptone Audio LPS-1 supplies.
Of course this might be 'first listen effect', where it sounds fabulous the first time you listen to it. Or it might sound diff after the resistors have some time on them. I'll check again tonight.
And the other thing I believe I discovered is that the resistors in both the Mamboberry and MBLS that look like carbon composite likely are... and I tip my hat to you Soundcheck, there was a bottleneck in the output filter, just not the one we were discussing.
Greg in Mississippi
The ESS datasheets I have are the least informative of any I've seen. They do not contain the typical sampling frequency over clock frequency chart that ALL other audio DAC datasheets I've seen show.
That said, I suspect a 49.1520 mHz clock will support 384 on the ES9023 in sync mode. That DOES work with the ES9018 (Barrows does it that way) and I believe the ES9023 and ES9018 use the same oversampling and ASRC engines. 24.5760 MHz MIGHT work, but I doubt it.
Of course, you likely have a 24/22 Kali like the rest of us!
IF you try sync-mode and IF you can get it to work at 384, that MIGHT sound even better as it would minimize the filtering performed in the chip (though you can't turn it off via I2C options like you can on the ES9016 and ES9018).
Greg in Mississippi
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.