Warning!!! For AK4396 version...
Ouch!
I think I've found why the AK4396 version sounds tonally unbalanced...
I was reading the datasheets of AK4393 and AK4396 and the sample output stage (which incidentally is identical to the one of this board...) and they've the same topology but different resistors and caps values...
So for the two DACs a different LPF is required!!!
Then I've checked the one of AK4395 (identical to AK4393 one) and AK4399 (different values from AK4393/95 and AK4396).
So this kit should have different LPF's resistors and caps for the two chips...
Obviously I've found this information the day after I've ordered PRPs for two boards... (if someone with AK4393 version is interested PM me...I'll make a nice price )
I've attached the two output stages and, since to me they're a Sallen-Key filter, two graphs that shows the difference (made with this calculator).
The AK4396 outstage is clearly more rolled-off...
Ouch!
I think I've found why the AK4396 version sounds tonally unbalanced...
I was reading the datasheets of AK4393 and AK4396 and the sample output stage (which incidentally is identical to the one of this board...) and they've the same topology but different resistors and caps values...
So for the two DACs a different LPF is required!!!
Then I've checked the one of AK4395 (identical to AK4393 one) and AK4399 (different values from AK4393/95 and AK4396).
So this kit should have different LPF's resistors and caps for the two chips...
Obviously I've found this information the day after I've ordered PRPs for two boards... (if someone with AK4393 version is interested PM me...I'll make a nice price
)I've attached the two output stages and, since to me they're a Sallen-Key filter, two graphs that shows the difference (made with this calculator).
The AK4396 outstage is clearly more rolled-off...
Attachments
Last edited:
Low ultrasonic noise
.
http://www.diyaudio.com/forums/digi...5-4396-ultrasonic-noise-spectrum-graphic.html
That is because the level of remaining ultrasonic noise with the AK4396 is so low that it really doesn't need any analog filtering. I find the best sound by far with the AK4395 or 96 comes from running the analog direct out through high quality caps to block the dc. No filters, output Rs or buffers. Now I don't have to worry about which opamp will sound best or if it is properly decoupled as I don't have any in my signal chain.
.
http://www.diyaudio.com/forums/digi...5-4396-ultrasonic-noise-spectrum-graphic.html
^ I heard this somewhere else...
4396 seems to be better chip than 4393 so it is better to stay focused on it.
@Sandler... Ehat exactly do you mean by direct out mod... there's gotta be something to be filtered there... or not ?
@Clave ... it doesnt matter what the figures say...if it sounds good .....
Cheers
4396 seems to be better chip than 4393 so it is better to stay focused on it.
@Sandler... Ehat exactly do you mean by direct out mod... there's gotta be something to be filtered there... or not ?
@Clave ... it doesnt matter what the figures say...if it sounds good .....
Cheers
I don't think so.
IMHO the LPF filter is there not only to deal with ultrasonic noise but also to adjust the frequency response of the output if the DAC, by design, don't handle it completely.
As a matter of fact different sources (PCM or DSD for instance) needs different filters to achieve the (quite) same final frequency response.
And since different AKM DACs are coupled with different LPFs to achieve the same frequency response it can be deduced that the DAC output frequency response is different....
it doesnt matter what the figures say...if it sounds good .....
To my ears it matters... sounds good but unbalanced...
Very interesting info on the filter differences between the 93 and 96 - thanks for doing the homework.
I'd strongly advise anyone to ditch the filtered active output stage anyway. I've used a number of different 600ohm output transformers on the 93 in my src2496, and all have been a massive improvement on opamps. I briefly tried the cap only output using the positive dac output only, but to my ears this wasn't close to the quality of transformers. I suspect this is down to two things: (1) i read that these dacs like to have even loads on positive and negative outputs (2) transformers may add a pleasant colouration.
I have the 93 kit at home unbuilt and the 96 kit on the way. Got a pair of tamura td-1 transformers for the 96 and some bbc ones for the 93. Thinking about all solid polymer caps for one dac and all tantalum for the other. Might even use linear psu on one and a really nice medical smps on the other. I'm on holiday next week so looking forward to playing with all the spare parts i have lying around. Great fun
I'd strongly advise anyone to ditch the filtered active output stage anyway. I've used a number of different 600ohm output transformers on the 93 in my src2496, and all have been a massive improvement on opamps. I briefly tried the cap only output using the positive dac output only, but to my ears this wasn't close to the quality of transformers. I suspect this is down to two things: (1) i read that these dacs like to have even loads on positive and negative outputs (2) transformers may add a pleasant colouration.
I have the 93 kit at home unbuilt and the 96 kit on the way. Got a pair of tamura td-1 transformers for the 96 and some bbc ones for the 93. Thinking about all solid polymer caps for one dac and all tantalum for the other. Might even use linear psu on one and a really nice medical smps on the other. I'm on holiday next week so looking forward to playing with all the spare parts i have lying around. Great fun
Corner frequencies
I think that you will find the corner frequencies of the filters are so far out of the audio band that there will be no measurable difference to the in band response, or possibly even a flatter extension at the top, without the filter. It is just for the ultrasoninc noise of which there is very little with the AK4396.
Direct out with trannies or transformers
Running a dac chip direct from the analog pins has been discussed here since 2006. Just send the signal through a high quality cap to block the dc, or through a good transformer to eliminate the dc by balancing it on both legs. I find the sound to have the best fine detail through caps. A more pleasant and euphonic sound can come with transformers.
.
http://www.diyaudio.com/forums/digi...est-solution-output-stage-20.html#post2286802
^ I heard this somewhere else...
4396 seems to be better chip than 4393 so it is better to stay focused on it.
@Sandler... Ehat exactly do you mean by direct out mod... there's gotta be something to be filtered there... or not ?
@Clave ... it doesnt matter what the figures say...if it sounds good .....
Cheers
Running a dac chip direct from the analog pins has been discussed here since 2006. Just send the signal through a high quality cap to block the dc, or through a good transformer to eliminate the dc by balancing it on both legs. I find the sound to have the best fine detail through caps. A more pleasant and euphonic sound can come with transformers.
.
http://www.diyaudio.com/forums/digi...est-solution-output-stage-20.html#post2286802
Why AKM bother to specify different filters for different DACs?
We agree about ultrasonic noise, it's just a fact, but since there's no noise to filter why the AK4396 filter is more rolled-off?
Well I am not an expert but I was told even though not in audible range , extra frequencies can put a strain on the amplifier if it is a DC amp like Firstwatt F5 which I have...
seems like a easy solution...will give that a try just for fun and see what happens.
cheers
Don't struggle that much with the filters etc... Just listen the direct-out with blocking cap. Let your ears decide it..Then, if you feel opamps and filters are necessary again, then you can find for yourself a lot of brands/models of opamps and resistors/caps to struggle with.
Last edited:
I suggest to only replace resistors on signal path. Wattage is not issue here. AKM recommending carbon film type resistors for output stage in application notes. I also noticed that carbon film resistors has better naturalness and resolution with these dacs. It's highly recommended. I suggest you to use good quality of resistors like Kiwame, AB.
Last edited:
This night I've compared the output of my stock Samsung BD-P2500 BR Blayer (which also uses AK4396 and his proper output stage from datasheet) with mine MINI2496 and , apart refinement due to mods, there are some differences:
So, at the moment, seems that the AK4393 outstage used in the MINI2496 don't fits perfectly, but it's not a big difference, with AK4396 louder output and different on chip filtering.
I do agree, anyway the stock resistors are 1/4 Watt
I can't find this info on datasheets, can you please point me to the right document? Thanks
Kiwames (or KOA SPR) don't fit, they're too big. Probably remaining in KOA CF1/4 are good substitues.
KOA CF1/4 sounds good, I've tried them in the MyRef, but PRPs are way better despite they're metal films.
PRPs sounds full like carbon films.
Regarding Allen Bradley they're also too big but mainly they're carbon composition, not carbon film and thus they've much more noise and will drift...
- The BD-P2500 has slightly lower volume
- Highs frequencies volume is slightly different, not a big difference but the BD-P2500 was for sure more balanced
So, at the moment, seems that the AK4393 outstage used in the MINI2496 don't fits perfectly, but it's not a big difference, with AK4396 louder output and different on chip filtering.
I do agree, anyway the stock resistors are 1/4 Watt
I can't find this info on datasheets, can you please point me to the right document? Thanks
Kiwames (or KOA SPR) don't fit, they're too big. Probably remaining in KOA CF1/4 are good substitues.
KOA CF1/4 sounds good, I've tried them in the MyRef, but PRPs are way better despite they're metal films.
PRPs sounds full like carbon films.
Regarding Allen Bradley they're also too big but mainly they're carbon composition, not carbon film and thus they've much more noise and will drift...
Last edited:
I can't find this info on datasheets, can you please point me to the right document? Thanks
Ok, I'm a bit lazy:
AKM Basic Layout Tips Part 1 & 2 :
Very useful documents for chip layout. I'm totally agree on Part2's "Resistors" section.
Last edited:
Thanks Clave & Terranigma for the ideas that you suggest to us.
I have a AK4396 kit not assembled, so what is the best choice for testing it ?
1- direct coupling with output capacitor as Terranigma suggest us (please wich type, value and location in the schema)
2- modify the LP filter as Clave suggest us (value from your image, resistor type (https://www.buerklin.com/datenblaetter/E072400_TD.pdf) could be right? )
3- output transformer (please specs and cost)
I'm waiting your precious help
Thanks
I have a AK4396 kit not assembled, so what is the best choice for testing it ?
1- direct coupling with output capacitor as Terranigma suggest us (please wich type, value and location in the schema)
2- modify the LP filter as Clave suggest us (value from your image, resistor type (https://www.buerklin.com/datenblaetter/E072400_TD.pdf) could be right? )
3- output transformer (please specs and cost)
I'm waiting your precious help
Thanks
I have measured all the stock resistors. The stock resistor measurements deviate from the expected values between 1 to 3 percent (most of the cases 3 percent). I have no electronics knowledge but if we think with straight logic, these deviations must effect the overall sound quality. I wonder if I am thinking right or not.
You're welcome
I suggest you to mount it as is but socket the four DAC's output resistors with a SIL socket.
In this way you can, in every moment, pull out the resistors and plug in wires to trasformer or output caps while leaving the IV/LPF functional when resistors are plugged in socket.
If you buy the socket from the link buy also one packet of solid elcos for AK4396 AVDD, and CS8416 VA and VD decoupling...
Since you still have to assemble it I would buy and mount the proper resistors and caps from AK4396...
I was just looking at this layout and thinking, The power reg is a long way from the cs8416.
This would not be ideal from what the datasheet suggests.
"VA, Pin 6, Analog Power (Input) - Analog power supply. Nominally +3.3 V. This supply should have as little noise
as possible since noise on this pin will directly affect the jitter performance of the recovered clock"
Allan