Just saw that you already added GND VIAs in a later post.
There are still some things in my opinion that could be changed.
The input pins GND and VCC could be swapped, so you can place a low value NP0 capacitor between the VCC and GND pin and between the VREF and GND pin.
A low value NP0 capacitor in parallel to C3 and close to the input pin of the opamp could be beneficial in terms of noise cancellation, too.
If it doesn't bring anything you still have the possibility to unpopulate it later.
Instead of using a trace for VCC you could use a plane.
You could also consider placing the input and output pins on a 100mil grid to be able to place your board onto a bread board.
There are still some things in my opinion that could be changed.
The input pins GND and VCC could be swapped, so you can place a low value NP0 capacitor between the VCC and GND pin and between the VREF and GND pin.
A low value NP0 capacitor in parallel to C3 and close to the input pin of the opamp could be beneficial in terms of noise cancellation, too.
If it doesn't bring anything you still have the possibility to unpopulate it later.
Instead of using a trace for VCC you could use a plane.
You could also consider placing the input and output pins on a 100mil grid to be able to place your board onto a bread board.
Last edited:
That's generally considered a bad idea in high frequency design. Also, it is unnecessary in this case. One well known reference on PCB design that includes info on power bypass practices and why we should use certain ones over others is Henry Ott's excellent book, Electromagnetic Compatibility Engineering.
In the case of most opamps we use for these dacs (AD797 would be an exception, if one did use it), what actually helps sound quality noticeably (and does not cause opamp power impedance problems at their highest frequencies, unlike improper use of NPO caps) is to use some large value film caps in parallel where +-15v power comes into the dac board. A combination experimentally found to work well uses Wima MKS4 film caps in the following parallel values: 10 + 10 +22 +33 +33 (all values in uf). One set of parallel film caps is connected from +15v to ground and another set from -15v to ground. Not cheap, but very effective at getting rid of the last bit of opamp graininess in this and other similar opamp circuits. However, that sort of thing should perhaps be left until near the end of the project. Since it is not cheap, there isn't much point in doing it unless dac board circuitry has been very substantially improved first. Its only suggested to fix the very last of the sound quality issues.
It should also be mentioned that some people have tried using low cost DC Link Film Caps since they are much cheaper than Wima MKS4. They do kind of work, but not as well as Wima, IME. Another possible alternative that has been suggested is to use Nazar regulators, something I have not tried yet, but it would be a very cost-effective solution if it works as well as the caps. More info is available on Nazar regulator details if there is any interest.
Last edited:
Hi guys. I finally got my board from eBay but I'm concerned as i may have a different version to what is being modded on this thread. My version is 1.07
Is that an issue?
Is that an issue?
Hi Alex,
1.07 is the standard version these days and has been for quite some time. It should be fine for modding. If any specific concerns or questions, maybe you could post of pic for us to look at.
-Mark
Perhaps time for a few more words about AK4499 eval board.
At the moment I have switched to using a slightly modified TP Cronus/Amanero clock setup using Crystek 957 series 45/49MHz clocks. I found some weird problems with it at first when trying to use it with u.fl to pin header coax interconnection cabling to the eval board I2S inputs. Generally, I find that u.fl connector performance tends to deteriorate rather quickly the more times they are disconnected and reconnected. Spring tension in the contact fingers loosens up rapidly with any movement. If used, I would suggest to connect them once and don't touch them after that and or replace them as necessary. In terms of jitter and noise effects at RF, I have found that even an insufficiently tightened SMA connector can adversely affect sound quality if it increases jitter/noise on a master clock line. Therefore, at this time I am using soldered connections at the Cronus end using ribbon cable. The ribbon cable terminates in female pin header connector with gold plated contacts. The the little contact assemblies are attached to ribbon cable conductors using crimping followed by soldering. The Cronus has been mounted such that the clocks point away from the AK4499 dac chip, and side facing the dac chip mostly consists of ground plane. That seems to give the lowest distortion/jitter clocking I have found so far, as determined by subjective listening tests of cymbal and vocal harmonics coherence at higher audio frequencies. One symptom that improved with this clock setup (vs other USB boards with integrated NDK SDA clocks) is that cymbals do not have an artificially metallic tone to them, at least it is much less now and more small level details sound less like HF distortion and more like actual musical instruments at HF.
With that out of the way, other experiments have involved jumpering out the 10-ohm resistors used for the Reference Voltage supply nearby to the dac chip. Other experiments have involved trying some film caps at various power supply related points.
The bottom line with respect the the resistor and capacitor experiments is that there appears to be a design trade off that has to be reckoned with. More HF detail can be revealed/exposed, but not all of it is necessarily more enjoyable to listen to for long periods of time. The eval boards as configured by default seem to trade off some HF detail for a smoother and more pleasant sound quality, with enhanced bass and rolled off highs. Don't know how it looks on an FFT at different frequencies and or with pink noise, could be FR looks pretty flat even with some filtering effectively occurring due to power supply design. Changing resistors and caps as I have done changes the sound quite a bit, and not necessarily all for the better. Kind of sounds like even lower jitter clocking might help more, at least I have not found a point at which further improvement in clock signal quality no longer has any audible effect (particularly so when power supplies are changed to allow more HF details through.
Will have to discuss with Jam and other listeners which SQ trade offs are felt to be most subjectively preferable, for now anyway.
Also, would like to find ways to keep improving sound quality in the case where more HF detail is revealed. Perhaps time to focus more on LV digital power supply quality.
At the moment I have switched to using a slightly modified TP Cronus/Amanero clock setup using Crystek 957 series 45/49MHz clocks. I found some weird problems with it at first when trying to use it with u.fl to pin header coax interconnection cabling to the eval board I2S inputs. Generally, I find that u.fl connector performance tends to deteriorate rather quickly the more times they are disconnected and reconnected. Spring tension in the contact fingers loosens up rapidly with any movement. If used, I would suggest to connect them once and don't touch them after that and or replace them as necessary. In terms of jitter and noise effects at RF, I have found that even an insufficiently tightened SMA connector can adversely affect sound quality if it increases jitter/noise on a master clock line. Therefore, at this time I am using soldered connections at the Cronus end using ribbon cable. The ribbon cable terminates in female pin header connector with gold plated contacts. The the little contact assemblies are attached to ribbon cable conductors using crimping followed by soldering. The Cronus has been mounted such that the clocks point away from the AK4499 dac chip, and side facing the dac chip mostly consists of ground plane. That seems to give the lowest distortion/jitter clocking I have found so far, as determined by subjective listening tests of cymbal and vocal harmonics coherence at higher audio frequencies. One symptom that improved with this clock setup (vs other USB boards with integrated NDK SDA clocks) is that cymbals do not have an artificially metallic tone to them, at least it is much less now and more small level details sound less like HF distortion and more like actual musical instruments at HF.
With that out of the way, other experiments have involved jumpering out the 10-ohm resistors used for the Reference Voltage supply nearby to the dac chip. Other experiments have involved trying some film caps at various power supply related points.
The bottom line with respect the the resistor and capacitor experiments is that there appears to be a design trade off that has to be reckoned with. More HF detail can be revealed/exposed, but not all of it is necessarily more enjoyable to listen to for long periods of time. The eval boards as configured by default seem to trade off some HF detail for a smoother and more pleasant sound quality, with enhanced bass and rolled off highs. Don't know how it looks on an FFT at different frequencies and or with pink noise, could be FR looks pretty flat even with some filtering effectively occurring due to power supply design. Changing resistors and caps as I have done changes the sound quite a bit, and not necessarily all for the better. Kind of sounds like even lower jitter clocking might help more, at least I have not found a point at which further improvement in clock signal quality no longer has any audible effect (particularly so when power supplies are changed to allow more HF details through.
Will have to discuss with Jam and other listeners which SQ trade offs are felt to be most subjectively preferable, for now anyway.
Also, would like to find ways to keep improving sound quality in the case where more HF detail is revealed. Perhaps time to focus more on LV digital power supply quality.
Last edited:
Hey Mark, You might want to start a new thread on the AK4499 stuff, since you're getting into it pretty deep now.
Hey MrSlim, agreed it might be getting to be about that time since eval boards are starting to appear at Digikey, and presumably at other electronics outlets around the world.
On the other hand, I suspect some of what AKM has done with AK4499 eval board might be very applicable to consider for future prospective Sabre dac designs. Likewise, some things we have learned about Sabre dacs may turn out to applicable to AK4499.
That having been said, it might be more orderly to keep discussion of the two dacs in separate threads once more people start working specifically with AK4499 projects. So far, the amount of specific AK4499 discussion has been pretty minimal. No one has asked about programming it, hooking up an Arduino or otherwise getting the AKM control box solution working. In fact, the 3rd AK4499 eval board at Digikey still hasn't sold, most likely due to the price. If price keeps people around here from diving in, then the number of AK4499 posts may be rather low, or at least possibly slow to pick up momentum.
Maybe people are waiting for low cost Chinese AK4499 boards to appear on ebay or aliexpress. If they are made with mostly junk parts and as few of those as possible, and or if the boards don't have ground plane space for attaching mods, then those boards may be unlikely to convey the sound AKM wants people to hear, and they may be difficult to mod into condition where they could.
On the other hand, I suspect some of what AKM has done with AK4499 eval board might be very applicable to consider for future prospective Sabre dac designs. Likewise, some things we have learned about Sabre dacs may turn out to applicable to AK4499.
That having been said, it might be more orderly to keep discussion of the two dacs in separate threads once more people start working specifically with AK4499 projects. So far, the amount of specific AK4499 discussion has been pretty minimal. No one has asked about programming it, hooking up an Arduino or otherwise getting the AKM control box solution working. In fact, the 3rd AK4499 eval board at Digikey still hasn't sold, most likely due to the price. If price keeps people around here from diving in, then the number of AK4499 posts may be rather low, or at least possibly slow to pick up momentum.
Maybe people are waiting for low cost Chinese AK4499 boards to appear on ebay or aliexpress. If they are made with mostly junk parts and as few of those as possible, and or if the boards don't have ground plane space for attaching mods, then those boards may be unlikely to convey the sound AKM wants people to hear, and they may be difficult to mod into condition where they could.
Last edited:
Thanks again for all this, Mark
Yes, at a certain level some parts values and swaps or additions tune the sound, as I found out a couple of decades ago when playing with the DI/O - and back then that even started with the coax input impedance value... and ended at the analog output filter values.
So all interesting...
Also, I share your POV: the AK4499 is probably the way to go, but the eval board at over 500E won't attract DIYer... simply way too expensive. IMHO, such a project is interesting when you consider a total of raw board costs in the 150/200 region (and that's ALL the boards) and possibly as much for the DIY tweaking (power supplies, special parts, special clocks, caps etc.)
At the end we speak of a cost between 300 and 400 without any casing or real comfort functionalities : that's IMHO the limit for a hobby / toy but still good value considering you have potentialy a world class sound. Over that, you wil always wonder if you project was worth it vs used "best buys" products. Plus you could easily mess up and lose it all...
All IMHO, so let's wait for the chinese boards... and let's hope they do read your posts to make them attractive to us 🙂
Claude
Yes, at a certain level some parts values and swaps or additions tune the sound, as I found out a couple of decades ago when playing with the DI/O - and back then that even started with the coax input impedance value... and ended at the analog output filter values.
So all interesting...
Also, I share your POV: the AK4499 is probably the way to go, but the eval board at over 500E won't attract DIYer... simply way too expensive. IMHO, such a project is interesting when you consider a total of raw board costs in the 150/200 region (and that's ALL the boards) and possibly as much for the DIY tweaking (power supplies, special parts, special clocks, caps etc.)
At the end we speak of a cost between 300 and 400 without any casing or real comfort functionalities : that's IMHO the limit for a hobby / toy but still good value considering you have potentialy a world class sound. Over that, you wil always wonder if you project was worth it vs used "best buys" products. Plus you could easily mess up and lose it all...
All IMHO, so let's wait for the chinese boards... and let's hope they do read your posts to make them attractive to us 🙂
Claude
Register programming
Hi Mark,
Side note regarding the ak4499 evm, as Terry said Ouch 2! Really expensive at 600. The chip is at $86 at Digikey still 0 qty, supposed to be arriving in stock on 25 July. Just wondering what the costs of components would come to if one can make a board for say $50 using the datasheet suggested schematic for stereo only. Reckon it would still be on the high side.
Replacement es9038q2m aliexpress v.1.07 board arrived. Finished the mods. When I lift the MCU i2c pins, it just broke off, both of them! Guess its more brittle than the previous ones. Luckily, I still have the earlier MCU which already has lifted pins, successfully replaced with hot air.
Finally could connect for register program, odd thing is my reese52 relay switched the connection when turned on, so instead of seize dac for connecting, its already connected. Seize dac actually release the dac to mcu. I know I could inverse the connection but decided to leave it on so that the mcu works when relay is not powered.
My question is this, what could be the reason the DPLL reading jumped randomly, it will read 90 decimal on first read, when I tried to write it will show 0, then when I tried writing 17, it will show 64 or something else?
By the way, the dac played flawlessly when I exited the i2c capture.
Cheers,
Kay
Hi Mark,
Side note regarding the ak4499 evm, as Terry said Ouch 2! Really expensive at 600. The chip is at $86 at Digikey still 0 qty, supposed to be arriving in stock on 25 July. Just wondering what the costs of components would come to if one can make a board for say $50 using the datasheet suggested schematic for stereo only. Reckon it would still be on the high side.
Replacement es9038q2m aliexpress v.1.07 board arrived. Finished the mods. When I lift the MCU i2c pins, it just broke off, both of them! Guess its more brittle than the previous ones. Luckily, I still have the earlier MCU which already has lifted pins, successfully replaced with hot air.
Finally could connect for register program, odd thing is my reese52 relay switched the connection when turned on, so instead of seize dac for connecting, its already connected. Seize dac actually release the dac to mcu. I know I could inverse the connection but decided to leave it on so that the mcu works when relay is not powered.
My question is this, what could be the reason the DPLL reading jumped randomly, it will read 90 decimal on first read, when I tried to write it will show 0, then when I tried writing 17, it will show 64 or something else?
By the way, the dac played flawlessly when I exited the i2c capture.
Cheers,
Kay
Hi Kay,
There are two circuit boards that comprise the AK4499 eval board. Both are 4-layer, and the motherboard is rather large. Circuitry is more complex than for Sabre dacs. The Reference Voltage (5v, otherwise similar to AVCC, not Vref) uses one Jung regulator for each channel, plus a network of 10-ohm resistors in series with some big electrolytic caps. Lots of parts. There is also one NJC7805 per channel, and NJC are probably the best sounding of those type regulators. They sound better than LM317, etc., too. Then there are at least 4 low voltage regulators for digital functions.
I think you can probably see large multi-layer boards and lots of carefully chosen components makes AK4499 more complicated to diy, at least if the goal were to be to reproduce eval board sound quality (which is very good as-is, but balanced out only).
Regarding the other issue, I sent a PM, but I think if trouble reading/writing registers you might want to check the I2C signals with a scope to make sure no problems. Bits that are set for DPLL should not change on their own, and the problem is unlikely to be inside the dac chip.
-Mark
There are two circuit boards that comprise the AK4499 eval board. Both are 4-layer, and the motherboard is rather large. Circuitry is more complex than for Sabre dacs. The Reference Voltage (5v, otherwise similar to AVCC, not Vref) uses one Jung regulator for each channel, plus a network of 10-ohm resistors in series with some big electrolytic caps. Lots of parts. There is also one NJC7805 per channel, and NJC are probably the best sounding of those type regulators. They sound better than LM317, etc., too. Then there are at least 4 low voltage regulators for digital functions.
I think you can probably see large multi-layer boards and lots of carefully chosen components makes AK4499 more complicated to diy, at least if the goal were to be to reproduce eval board sound quality (which is very good as-is, but balanced out only).
Regarding the other issue, I sent a PM, but I think if trouble reading/writing registers you might want to check the I2C signals with a scope to make sure no problems. Bits that are set for DPLL should not change on their own, and the problem is unlikely to be inside the dac chip.
-Mark
Last edited:
Good morning.
For now I am going to power up the dac board with a Salas uBib for the +/-15 and an XRK971 simple cap mx board for the 5v.
I need to set up the Ubib for the correct current requirements. Can anyone help me please with what the current drawn from the board might be on the 15v rails.....with the 5v reg disconnected.
Thankyou
For now I am going to power up the dac board with a Salas uBib for the +/-15 and an XRK971 simple cap mx board for the 5v.
I need to set up the Ubib for the correct current requirements. Can anyone help me please with what the current drawn from the board might be on the 15v rails.....with the 5v reg disconnected.
Thankyou
Last edited:
Hi,
If you have the un-modded chinese v1.07 dac board the current consumption according to the description is : "Dual 15V mode: +15V about 120MA, -15V about 10MA" and with TFT display is "Dual 15V mode: +15V about 150MA, -15V about 10MA".
Kay
Hi,
The numbers Kay quoted sound about right from my recollection, after having measured them some time ago.
However, applying fancy power supplies to the basic dac board circuitry probably won't help sound quality much. There are many other things that need attention if the goal is better sound quality along the lines of what ES9038Q2M is capable of. The good news is that we have quite of bit of experience making the dac boards sound much better. The bad news is that it takes a lot of work to fix everything that really needs fixing.
Thanks for both replies. Sorry I hadn't even thought of looking at the description on ebay or Ali.
I wander how much current will be spared by not powering 7805 from the 15v rail.? I shall try to measure.
Mark I do appreciate you have helped to pioneer the mods to this board and your adamance to implementing the proper op amp avcc buffer . That is a bit beyond me for now but I shall keep learning until a point I am happy to do so.
I wander how much current will be spared by not powering 7805 from the 15v rail.? I shall try to measure.
Mark I do appreciate you have helped to pioneer the mods to this board and your adamance to implementing the proper op amp avcc buffer . That is a bit beyond me for now but I shall keep learning until a point I am happy to do so.
Last edited:
It's an easy track cut exercise to remove the 15v feed to the 7805 if you need to give it something considerably less than the 15v.
I can't give figures on the 7805 but one board I had ran very cool, another very warm with the same 12-15v voltage applied. I put this board to one side and left it just in case I needed to refer back to it as I modded the other.
Hi Kay, Mark,
I'm ready to join you guys on the MCU work, I have the Arduino in place (3.3v SCL/SDA). Tell me more about lifting these pins, I have the 1.07 board and I read (or misunderstood) that applying the jumper J1 and J2 would have the same effect as lifting the pins.
It has been said that for version 1.07 boards installing J1, J2 at the same time will halt I2C bus activity by the dac board built-in MCU. Didn't work on my 1.07 board nor on Kay's, IIUC. If it does work on your dac board then that makes it easy to take over control of the I2C bus. Best way to tell would probably be to monitor the I2C bus lines with a scope while J1, J2 are installed. If no bus activity then you are good to proceed without pin lifting or trace cutting. If you have a display, its said it will halt updates with J1,J2 installed. However, normally the MCU is talking to the dac chip constantly, and might still be with J1, J2 installed. Inputting I2S audio normally turns on the LOCK LED. If disconnecting the I2S input turns off the LED then the MCU would appear to still be polling the dac chip via I2C bus.
If I2C bus is still active you would need to disconnect the built-in MCU from the bus somehow or other. For the first dac or two, I lifted the MCU I2C pins from their solder pads and soldered on leads so I could run them through a small relay if I wanted the MCU reconnected.
On a subsequent dac I cut traces between the dac chip I2C pins and the corresponding MCU pins. Then the pullup resistors were patched into the dac chip side of the trace cuts by soldering on pieces of #30 wire wrap wire. The second method involving cutting traces was harder to do, but less risk of breaking off MCU pins while trying to lift them.
That said, breaking off MCU pins need not mean the end of the dac board. The Arduino can do it all with a little help with your friends here.
Pics of all the above described mods are available if you would like me to find the links or re-post them.
Last edited:
Thanks Mark, I've been going back over the various threads, trying not to get too distracted.....I've just added 14uf (2x4.7, 2x2.2) of PP caps I had spare from an old crossover to the +/- 15v supply. I can probably find some 10uf and a few 1uf for the 'full spread'. A very nice improvement to the sound.
I've just moved all the boards/PSUs to a metal case so on the way to making things a little more permanent.
So, for the Arduino, ideally, I'll need some sort of DPDT relay to switch between the Arduino and the MCU if the J1/J2 doesn't work as I assume that once the regs have been altered you need to switch back to MCU control.
I just need to find some time now!
Not necessarily, in some cases the MCU may change your settings back to where it had them before. Mostly, keeping the MCU there is so people can learn register programming a little at a time and always fall back to default operation if they want.