I opened this thread for further discussion about my I2S input ES9023 + JG Filter Buffer boards.
Of course any general discussion about the detailed application (Cs, Rs, regs,...) of the ES9023 DAC chip is very welcome
I'd like to start with my current findings after some first tests:
After initial setup of the board according to the actual "beta"BOM I did some first measurements (of course I also listened and really enjoyed music). Like others reported before there is some noise arround 1.5MHz especially at the NEG pin which could be reduced by moving the 2.2µF MLCC directly to the pins of the DAC (on the board it is only a few mm away ). Furthermore I found some modulation from the music signal on the NEG pin! I tried different Elcaps (220µF Pana FC -> 470µF Pana FR with lower ESR -> 22µF Pana FR) and the lowest measured best (no listening test) so i'll probably stay with 22µF.
I remember some discussion about removing the charge pump cap and implementing external NEG voltage, however the discussion stopped and no one posted results. Did anyone test this and can report his findings?
I also measured GND noise at several positions on the board and found it to be quite clean. However my scope is quite old and not the most sensitive on earth (5mV).
Another point is the "clipping resistor" at VREF pin. I could not measure any significant clipping and liked the sound without the R, however this might be personal taste. I'll definately leave this option for the next design.
Hope some will join this discussion so that everyone can benefit from our findings
Best regards, Daniel
Of course any general discussion about the detailed application (Cs, Rs, regs,...) of the ES9023 DAC chip is very welcome
I'd like to start with my current findings after some first tests:
After initial setup of the board according to the actual "beta"BOM I did some first measurements (of course I also listened and really enjoyed music
). Like others reported before there is some noise arround 1.5MHz especially at the NEG pin which could be reduced by moving the 2.2µF MLCC directly to the pins of the DAC (on the board it is only a few mm away ). Furthermore I found some modulation from the music signal on the NEG pin! I tried different Elcaps (220µF Pana FC -> 470µF Pana FR with lower ESR -> 22µF Pana FR) and the lowest measured best (no listening test) so i'll probably stay with 22µF.I remember some discussion about removing the charge pump cap and implementing external NEG voltage, however the discussion stopped and no one posted results. Did anyone test this and can report his findings?
I also measured GND noise at several positions on the board and found it to be quite clean. However my scope is quite old and not the most sensitive on earth (5mV).
Another point is the "clipping resistor" at VREF pin. I could not measure any significant clipping and liked the sound without the R, however this might be personal taste. I'll definately leave this option for the next design.
Hope some will join this discussion so that everyone can benefit from our findings
Best regards, Daniel
I have been meaning to try a 50KR in series with ~200K POT wires as a variable R, on the Vref pin..., and I suppose a switch in series [close to the DAC] JUST for experimentation.
Then find some music laid down hot, certain female vocalists close to a mike can distort, for instance. I think the use of the pin could have other sonic impacts than just trying to prevent clipping -- I remember the day I took a UAE23 [MODDED with 3.6V and 100K there] to my Sister and Brother-in-law's -- as I seem to recall I left them 2x DACs, one with and one without. A jury of two preferred the R there in terms of other detail and staging... "your mileage may vary". So don't JUST think about clipping. Now, their system is unusual, that DAC signal would have been going immediately into a Yaqin CD2 cathode follower buffer.
Daniel do you know what V is on the NEG pin? I could get a regulator pretty fast and try it.
Then find some music laid down hot, certain female vocalists close to a mike can distort, for instance. I think the use of the pin could have other sonic impacts than just trying to prevent clipping -- I remember the day I took a UAE23 [MODDED with 3.6V and 100K there] to my Sister and Brother-in-law's -- as I seem to recall I left them 2x DACs, one with and one without. A jury of two preferred the R there in terms of other detail and staging... "your mileage may vary". So don't JUST think about clipping. Now, their system is unusual, that DAC signal would have been going immediately into a Yaqin CD2 cathode follower buffer.
Daniel do you know what V is on the NEG pin? I could get a regulator pretty fast and try it.
As mentioned the SMD pads are available on the PCB so everyone can use it or leave it
Errm, measure it
hint: should be similar to voltage at AVCC pin to generate 0V DC at the outputs
Hi Daniel,
I just received the boards from you today and they are great, professionally made and they look better than in pictures, thanks for making them available
I will start ordering parts and maybe next week receive them and in weekend solder them to the board; can't wait to hear the sound of the direct output compared to the J G buffer filter
great work
I just received the boards from you today and they are great, professionally made and they look better than in pictures, thanks for making them available
I will start ordering parts and maybe next week receive them and in weekend solder them to the board; can't wait to hear the sound of the direct output compared to the J G buffer filter
great work
I should add that to the degree it might matter, the above reading was with 100K to ground on ES9023 pin #6. Oppz, no, huge mistake, come to think of it, the -3.4 was across the pin#6 resistor, effectively, so, I have not measured the NEG pin YET, I was in a hurry :^| I will measure the NEG pin ASAP. I suspect the NEG will be more NEG, just a guess. Again, using a HifimeDIY UAE23 as the subject to take my readings from.
If the negative injection works to the NEG pin, I'm thinking something simple, either a negative ADJ regulator, or a -5V feeding loaded voltage divider: a POT or 2 R in series 32+68Ohms let's say? To ground. Cap middle to ground. This would only be to end up with less DAC noise, else why bother? What's one more regulator when we use them like salt? No Charge-pump noise? no CP cap{s}?
I have tried 330uF 25V Panasonic FM caps in the beta-board's various larger cap holes 5mm, they fit fine. The Test/Beta BOM should not be taken as Daniel's latest best test results so DO read what he has said, I hope all the comments have made it to the build-thread but maybe not. I know how anxious people get, even me... do not buy all the parts and assume the beta anything is where you want to be, meaning no offense to anyone, especially Daniel who was so brilliant to come up with this extension to the JG filter buffer. The BOM is a work-in-progress subject to changes. I LOVE THE ES9023 and think it's a diamond in the rough, we could easily get more out of it for performance than the designers intended or suggested.
If the negative injection works to the NEG pin, I'm thinking something simple, either a negative ADJ regulator, or a -5V feeding loaded voltage divider: a POT or 2 R in series 32+68Ohms let's say? To ground. Cap middle to ground. This would only be to end up with less DAC noise, else why bother? What's one more regulator when we use them like salt? No Charge-pump noise? no CP cap{s}?
I have tried 330uF 25V Panasonic FM caps in the beta-board's various larger cap holes 5mm, they fit fine. The Test/Beta BOM should not be taken as Daniel's latest best test results so DO read what he has said, I hope all the comments have made it to the build-thread but maybe not. I know how anxious people get, even me... do not buy all the parts and assume the beta anything is where you want to be, meaning no offense to anyone, especially Daniel who was so brilliant to come up with this extension to the JG filter buffer. The BOM is a work-in-progress subject to changes. I LOVE THE ES9023 and think it's a diamond in the rough, we could easily get more out of it for performance than the designers intended or suggested.
I have regulators coming to brute-force try and insert the -V in place of the charge-pump in the ES9023. The plan is to create -5v then use 28/72Ohm R or a 100-ohm POT pre-adjusted close... I do not know how much current the DAC would use in TOTAL, the negative rail would consume about 1/2? ignoring any digital stuff. So many regulators are used in DAC designs [including this one], it would be trivial to add one more. I could not find a fixed -3.6V regulator in the first attempt.
The Linear Technology LT3032 Series - Dual 150mA Positive/Negative Low Noise Low Dropout Linear Regulator comes in a adjustable version. It should probably be set to -3.3V / -3.4V rather than -3.6V because that seems to be what the charge pump generates.
Digikey # LT3032IDE#PBF-ND
The LT3032 package is not exactly easy to solder, but there is also the LT 1964.
The Linear Technology LT1964 is available in a more convenient package, adjustable and significantly cheaper. It is marginally noisier, but possibly still "better" than the internally generated negative supply.
Digikey # LT1964ES5-BYP#TRMPBFCT-ND
I hope this helps.
Digikey # LT3032IDE#PBF-ND
The LT3032 package is not exactly easy to solder, but there is also the LT 1964.
The Linear Technology LT1964 is available in a more convenient package, adjustable and significantly cheaper. It is marginally noisier, but possibly still "better" than the internally generated negative supply.
Digikey # LT1964ES5-BYP#TRMPBFCT-ND
I hope this helps.
Helps a LOT! First to experiment with whatever I have or will get fast. Now remember the NEG 3.4 was with 100K to ground off ES9023 PIN#6, which not many people would do. [Heavily modded HiFIMEDIY UAE23]
A regulator that is POS AND NEG in one package -- hey whata find! If this works we might come up with a way to "NUL" the DAC or offset...tune V balance around 0...hopefully not see DC at the outs, or, not much. Just thinking aloud, who knows it will work at all?
C9 on the beta board, I think polarity may need reversing? Or use Non-polar C? that is ES9023 "NEG".
A regulator that is POS AND NEG in one package -- hey whata find! If this works we might come up with a way to "NUL" the DAC or offset...tune V balance around 0...hopefully not see DC at the outs, or, not much. Just thinking aloud, who knows it will work at all?
C9 on the beta board, I think polarity may need reversing? Or use Non-polar C? that is ES9023 "NEG".
Polarity of C9 is indeed reversed
. Already mentioned it to some, but not all. So thats a good remark. Thanks I am out for about one week since we are currently refurbishing some rooms. My lab will be quite nice soon
Btw.: Mic5205 is pin compatible with LT1761 but the bypass C needs to connect BYP to GND while its between BYP and OUT for the LT1761.
I am looking forward to the test results regarding "disabled" charge pump and external NEG supply. Downside would be the board layout with the dual supply which would probably be worse with respect to proper ground layout
Best regards, Daniel
NEG is just one of those questions. I may just tack a -5 regulator on my +/- 15V regulator board and fly the wire over {from loaded -voltage divider, maybe a 100R POT set mid-range?} and tack a non-polar SMD cap on the NEG pin to ground. {and somewhere a sense wire for my meter}. This weekend is no good for me though; I have other obligations.
a NEGATIVE supply could come from PCB edge: 3 terminal screw lug instead of 2. Frankly, one other option is +/- 15 and +/- 5 on another PSU board or down one side of the existing art? The FAB we used? I'm pretty sure THEY do not slot or drill for break-away. {I know one that WILL} Existing board is 6x7CM and FAB charges same rate as 10x10. I KNOW that would be a lot of work. I am NOT arm-twisting. TO220 regulators get pretty cheap {NJM7915FA example USD $0.64Ea. 100+ 170uV noise, insulated TAB}, fast low-Vf diodes... a nul-adj with safety R? {for -3.??V}. or "balance" a +/- combo regulator? I can do it on perf-board too.
The next wave of LT1761 3.3V {for OSCILLATORs} arrived, I have 825x on hand.
I have a bag of Micrel 5205 3.6, but not enough for all beta boards.
I have To-92 -5 regulators and some multi-turn POTS coming for a NEG-pin test. I guess that means I have to dig out my scope. I HAVE NJM7905FA regulators here.
a NEGATIVE supply could come from PCB edge: 3 terminal screw lug instead of 2. Frankly, one other option is +/- 15 and +/- 5 on another PSU board or down one side of the existing art? The FAB we used? I'm pretty sure THEY do not slot or drill for break-away. {I know one that WILL} Existing board is 6x7CM and FAB charges same rate as 10x10. I KNOW that would be a lot of work. I am NOT arm-twisting. TO220 regulators get pretty cheap {NJM7915FA example USD $0.64Ea. 100+ 170uV noise, insulated TAB}, fast low-Vf diodes... a nul-adj with safety R? {for -3.??V}. or "balance" a +/- combo regulator? I can do it on perf-board too.
The next wave of LT1761 3.3V {for OSCILLATORs} arrived, I have 825x on hand.
I have a bag of Micrel 5205 3.6, but not enough for all beta boards.
I have To-92 -5 regulators and some multi-turn POTS coming for a NEG-pin test. I guess that means I have to dig out my scope. I HAVE NJM7905FA regulators here.
I've just measured the NEG (pin#11) pin on my unmodded UAE23 at 25C and it shows -3.4V idle and 3.39V at max volume with 160Ohm headphones plugged in directly. The VREG pin shows 1.31V. I did this measurement to add to enjoybiking's one as I expected the AVCC (pin#5) to be 3.3V but it seems like the unmodified UAE23 v1.2 actually uses an adjustable LT1763 and feeds 3.6V. My version, unlike enjoybiking's one, does not have the 100+K on pin #6 to ground yet.
That's FUNNY!! So if only one regulator, they are running the TE7022 at 3.6? Well, everything on the board. That star shaped intersection at center of board on back is at 3.6 then? Was the Vreg reading negative or +? Man I must be losing it? I'll retake the VREG reading soon, everything is put away for company *<:~) The NEG reading makes sense, that's good. I will not have any bench time for a bit yet -- at which time I will mock up the supplies... I have 36Vct transformers and +/- 15 boards... I can hang a -5 REG off that and try say a 100 ohm POT from that to ground [preset to mid range], tune it for -3.4 unloaded then apply to es9023 NEG. See if it varies. I will probably have a cap at the POT wiper too, nothing big.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.