Hi NinoSimona,
Congratulations. I see not so many people giving a chance for those "miserable" 3-leg series regulators. And I also see that those over-engineered "super" regulators mostly focus on noise performance. I didn't see any real-life measurements of their phase and impedance curves in audio band. Phase and impedance linearity make a big difference in analogue, especially in audio circuits.
There is a very interesting thread about jbau's experiments on LM317/337 series regulators:
http://www.diyaudio.com/forums/power-supplies/143539-another-look-lm317-lm337-regulators.html
He finally came up with a simple symmetric power supply schematic at post#242 (schematic is at post#248) based on LT1085/LM337 (LM317 couldn't attain same linearity as LT1085). He observed that these regulators need at least 45mA load for entering linear zone thus, on lower loads, we need load resistors depending on voltage output configuration.
I built +5v (110R load resistor) for analogue supply of DAC chip and +/-15V (300-330R load resistor) for opamp supply. I use them on my cheap CS4398 dac kit and I'm happy with the clarity, dynamics, resolution that I attained.
My AK4399 project is in progress and I'm going to build same regulators for this project.
Last edited:
Thanks terranigma.
Yes, I read this thread of jbau too. I think it was you who actually noticed me of it some posts ago. Thanks for that.
To be more specific about the 2 regulators for dual rail applications: His experiments are regarding the LT1085 (from Linear Technology) positive reg and the LM337 (from National Semiconductor) negative reg. So you need the NS LM337, which seems to differ from other brands.
You can also consider to use two of the same regulators for a dual rail application, e.g. 2xLT1085. In order to work, you must have two separate windings on the tranny, without center tap. This way phase and impedance behavior is guaranteed identical for + and - rail.
Valeri100: I think if you build two Salas shunt regulators for +5V analog, it works probably without any problems. I heard good results from Perterma, who uses the Salas shunt for analog +5V for the single AK4399 DAC version. So the difficulty lies in supplying two DACs with one regulator, as I did.
Regards,
Nino
Yes, I read this thread of jbau too. I think it was you who actually noticed me of it some posts ago. Thanks for that.
To be more specific about the 2 regulators for dual rail applications: His experiments are regarding the LT1085 (from Linear Technology) positive reg and the LM337 (from National Semiconductor) negative reg. So you need the NS LM337, which seems to differ from other brands.
You can also consider to use two of the same regulators for a dual rail application, e.g. 2xLT1085. In order to work, you must have two separate windings on the tranny, without center tap. This way phase and impedance behavior is guaranteed identical for + and - rail.
Valeri100: I think if you build two Salas shunt regulators for +5V analog, it works probably without any problems. I heard good results from Perterma, who uses the Salas shunt for analog +5V for the single AK4399 DAC version. So the difficulty lies in supplying two DACs with one regulator, as I did.
Regards,
Nino
The more expensive DAC around AK4399 instead of electrolytic capacitors 10u use such http://detail.tmall.com/item.htm?spm=a230r.1.14.30.zT8t71&id=17238156881.
Perhaps it makes sense to try and us?
Perhaps it makes sense to try and us?
You can try it. They are tantalums, so be carefull with polarity. Tantalums have in general lower ESR than electrolytics, and probably lower ESL too, because of the small size.
I think Sunsun22 was using tantalums here, maybe he can comment on it.
Caps around AKM
HERE our grate Guru have tested those caps:
"but the vendor selections is quite good.
I've tried various OSCONS (Nichicons, Fujitsu, Sanyo) and plain capacitors and the supplied Sanyo for C30X sound really good, beaten only by Black Gate STD."
you can read about and other caps around as well.
good luck guys
HERE our grate Guru have tested those caps:
"but the vendor selections is quite good.
I've tried various OSCONS (Nichicons, Fujitsu, Sanyo) and plain capacitors and the supplied Sanyo for C30X sound really good, beaten only by Black Gate STD."
you can read about and other caps around as well.
good luck guys
Last edited:
Hi Nosian
I'm glad you like them! Today I soldered a 1uF Wima across the input and ground pin, to be certain of stable operation of the regulator. It can become unstable with too much capacitive load on the output, as I read in the datasheet. I think it's slightly better now than before, but I'm not sure of it yet.
My DAC PCB did come with a MC7805ACT in the digital +5V psu for the receiver and micro controller hardware. They could have used the same for analog and digital +5V for the DACs as well, instead of the LT1085.
For everyone who wants to check out the ON-semi regulators: Replacing the LT1085/LT1086 with them is quite simple. Just remove the two resistors and the little cap on the adjustment pin. Then bridge the pads where the adjustment cap was. Modify the MC7805ACT regulator as on the picture.
Regards,
Nino
Attachments
caps and dogs
according to AKM the caps around AK4399 should be 10uF avoiding "digital noise". on my board they are 47uf!
just wonder how holly is the dog? 10 can be 47! more better?
can it be decrease to 6.8uF? why? yes you see Wima fk2s (polyester) is upp to 10uF. Bingo!
but wait! almost impossible to find! 6.8 is everywhere and 5mm pitch.
of course vi can have two 6.8 + 3.3 almost bingo!
but placing them together! under the board? maybe worth a try!
any comments please?
according to AKM the caps around AK4399 should be 10uF avoiding "digital noise". on my board they are 47uf!
just wonder how holly is the dog? 10 can be 47! more better?
can it be decrease to 6.8uF? why? yes you see Wima fk2s (polyester) is upp to 10uF. Bingo!
but wait! almost impossible to find! 6.8 is everywhere and 5mm pitch.
of course vi can have two 6.8 + 3.3 almost bingo!
but placing them together! under the board? maybe worth a try!
any comments please?
Last edited:
Cerafine on the digital side, Silmic ARS Super Gold (Super Old also)on analog side of DAC.
Those regulators look mighty tempting; great find, BTW...
The board I have isn't the same weiliang board, has a wm8741 voltage out chip, uses a mid-rail decoupling cap that's recommended to be a 10uf as well, am using some smaller sized 100uf 25v Panasonic Pureism here, bypassed by some Wima FPK1 blue caps,.001uf, under the board.
Those regulators look mighty tempting; great find, BTW...
The board I have isn't the same weiliang board, has a wm8741 voltage out chip, uses a mid-rail decoupling cap that's recommended to be a 10uf as well, am using some smaller sized 100uf 25v Panasonic Pureism here, bypassed by some Wima FPK1 blue caps,.001uf, under the board.
Last edited:
Yes, on both versions of the 2xAK4399 DAC the capacitors are 47u.
I tried several bigger caps instead of the 47u, but it didn't improve the sound at all, so for me the Roederstein caps are just fine.
You could use the recommended 10u capacity, but I don't think it will make big changes for the better. In case of the analog/digital local decoupling capacitor: Just think of its purpose: decoupling the load from the power line, so basically it must compensate for the impedance of the psu rail. A bigger capacitor has lower ESR, that's why 47u could be better than 10u electrolytic. A Wima polyester cap has low ESR, so the decoupling capacity could be made smaller possibly. Just try a 6.8u, it could be enough.
In my experience it's best practice to keep local decoupling as small as possible, and let regulator + output capacitor do the job of supplying current to the load. Else you can get the problem of 'capacitor sound', when local decoupling is doing more than to compensate for the track impedance.
I tried several bigger caps instead of the 47u, but it didn't improve the sound at all, so for me the Roederstein caps are just fine.
You could use the recommended 10u capacity, but I don't think it will make big changes for the better. In case of the analog/digital local decoupling capacitor: Just think of its purpose: decoupling the load from the power line, so basically it must compensate for the impedance of the psu rail. A bigger capacitor has lower ESR, that's why 47u could be better than 10u electrolytic. A Wima polyester cap has low ESR, so the decoupling capacity could be made smaller possibly. Just try a 6.8u, it could be enough.
In my experience it's best practice to keep local decoupling as small as possible, and let regulator + output capacitor do the job of supplying current to the load. Else you can get the problem of 'capacitor sound', when local decoupling is doing more than to compensate for the track impedance.
Last edited:
just för the records
"1. Grounding and Power Supply Decoupling
To minimize coupling by digital noise, decoupling capacitors should be connected to AVDD, VDDL/R and DVDD respectively. AVDD and VDDL/R are supplied from analog supply in system and DVDD is supplied from digital supply in system. Power lines of AVDD, VDDL/R and DVDD should be distributed separately from the point with low impedance of regulator etc. The power up sequence between AVDD, VDDL/R and DVDD is not critical. VSS1-4 must be connected to the same analog ground plane.Decoupling capacitors for high frequency should be placed as near as possible to the supply pin.
2. Voltage Reference
The differential voltage between VREFHL/R and VREFLL/R sets the analog output range. The VREFHL/R pin is normally connected to AVDD, and the VREFLL/R pin is normally connected to VSS1/2/3. VREFHL/R and VREFLL/R should be connected with a 0.1µF ceramic capacitor as near as possible to the pin to eliminate the effects of high frequency noise. No load current may be drawn from VCML/R pin. All signals, especially clocks, should be kept away from the VREFHL/R and VREFLL/R pins in order to avoid unwanted noise coupling into the AK4399."
"1. Grounding and Power Supply Decoupling
To minimize coupling by digital noise, decoupling capacitors should be connected to AVDD, VDDL/R and DVDD respectively. AVDD and VDDL/R are supplied from analog supply in system and DVDD is supplied from digital supply in system. Power lines of AVDD, VDDL/R and DVDD should be distributed separately from the point with low impedance of regulator etc. The power up sequence between AVDD, VDDL/R and DVDD is not critical. VSS1-4 must be connected to the same analog ground plane.Decoupling capacitors for high frequency should be placed as near as possible to the supply pin.
2. Voltage Reference
The differential voltage between VREFHL/R and VREFLL/R sets the analog output range. The VREFHL/R pin is normally connected to AVDD, and the VREFLL/R pin is normally connected to VSS1/2/3. VREFHL/R and VREFLL/R should be connected with a 0.1µF ceramic capacitor as near as possible to the pin to eliminate the effects of high frequency noise. No load current may be drawn from VCML/R pin. All signals, especially clocks, should be kept away from the VREFHL/R and VREFLL/R pins in order to avoid unwanted noise coupling into the AK4399."
Attachments
Last edited:
Hi Nosian,
I didn't mean to confuse you, sorry. I was referring to the local decoupling of the power lines, AVDD and DVDD, not the voltage reference.
The datasheet is using a combination of 10u electrolytic with a 100n ceramic, in my interpretation the electrolytic for psu decoupling and the ceramic for digital noise decoupling. I don't think the electrolytic will do much against digital noise, because of the high ESR. You could replace the two with a polyester capacitor of 6.8u and see if it works. It's also depended on the regulator performance and psu ripple, whether a 6.8u is enough. Adding an extra 3.3u will increase inductance.
BTW. I'm using 1u polyester instead of the 0.1u for HF decoupling of VREFHL/R and VDDL/R. They use a 1u in the official evaluation board also.
I didn't mean to confuse you, sorry. I was referring to the local decoupling of the power lines, AVDD and DVDD, not the voltage reference.
The datasheet is using a combination of 10u electrolytic with a 100n ceramic, in my interpretation the electrolytic for psu decoupling and the ceramic for digital noise decoupling. I don't think the electrolytic will do much against digital noise, because of the high ESR. You could replace the two with a polyester capacitor of 6.8u and see if it works. It's also depended on the regulator performance and psu ripple, whether a 6.8u is enough. Adding an extra 3.3u will increase inductance.
BTW. I'm using 1u polyester instead of the 0.1u for HF decoupling of VREFHL/R and VDDL/R. They use a 1u in the official evaluation board also.
Replacing the ROE 47uf capacitors
I've just received the Dual AK4399 DAC. This board has very interesting sound. I use HDAM discrete amps as operational amplifiers.
I have a question regarding the ROE 47uf capacitors replacement. If I replace ROE 47uf by Elna Cerafine 47uf/16v will it significantly improve the sound or I almost won't hear the difference?
I've just received the Dual AK4399 DAC. This board has very interesting sound. I use HDAM discrete amps as operational amplifiers.
I have a question regarding the ROE 47uf capacitors replacement. If I replace ROE 47uf by Elna Cerafine 47uf/16v will it significantly improve the sound or I almost won't hear the difference?
That depends on how you want it to sound, basically you have to try out different caps to see if they support the sound you're after. I have found that upgrading the box-film bypasses have made a large impact on the sound for the better.
Another thing to try out, since you're using a more power-hungry op amp, check the traces from the power pins to the nearest capacitor and make sure they are the same, and .1 ohm, not more. If necessary, you can place a piece of wire over the trace to make for a better path. This is a very low cost upgrade, worked well with the discrete op amps I have.
Another thing to try out, since you're using a more power-hungry op amp, check the traces from the power pins to the nearest capacitor and make sure they are the same, and .1 ohm, not more. If necessary, you can place a piece of wire over the trace to make for a better path. This is a very low cost upgrade, worked well with the discrete op amps I have.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.