32-bit DAC

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jump to Latest
hifimaker

hifimaker

Member
Joined 2003
hifimaker.com
georgehifi said:
AK4397 data sheet says minimum load resistance for the analog output is 1kohm, does this mean it's a voltage output dac?

Cheers George
Click to expand...

Hi George, Yes it's a voltage out DAC.

PigletsDad said:


They need to get a new version of the laws of physics out.

Let us ignore thermal noise entirely; there is a fundamental limit from quantum mechanics, given by Planck's constant. This about 6.6x10^-34 Js. If we take a a 0.6microsecond time period (from the sampling rate), the fundamental noise uncertainty energy in a sample is 10^-27J, so the noise power is about 10^-21W

If we set the reference power level at say 1mW (0dBm) this is -180dB. Applying the basic 6dB per bit rule, we find that the maxiumum resolution, for this sampling rate and reference level, is 30bits.

So the claimed figures can be shown to be pure and simple marketing BS.
Click to expand...


Fully agree that 32 bit will prpbably not have any meaningful technical improvement. AKM did clain that the additional bandwidth does help better represent the transients of the music.

I'm looking forward to see if there is a subjective improvement over the AKM4396.

-David
 
gfiandy

gfiandy

Member
Joined 2003
Hi,

To those wondering what the use of 32bit math in a DAC is. Normally the most obvious benifit comes for the fact that it puts the quantising floor much further away from the useful data. It is normal pratice in these systems not to dither, so you get lots of corelated quantisation noise. 24bits puts this just about far enough away from a 16bit noise floor but not for the new high res formats that are 24bit to start with.

However there is as I am sure you are all aware much more to making a good DAC than just the numbers on the package or even the chip itself the implementation is normally the limiting factor.

Regards,
Andy
 
A

arild

Member
Joined 2006
AK4397

I am currently using one of Peter Daniels TDA1543 DAC's and i think it sounds pretty good, however i understand this is old technology and the AK4397 is supposed to be a huge improvement http://www.asahi-kasei.co.jp/akm/en/product/ak4397/ak4397.html
The AK4397 is getting some attention from companies Like APL-HIFI they are using this DAC in their latest CD player NWO 3.0GO that they will launch next week at the CES show in las Vegas, here is the information i am able to find so far http://www.dagogo.com/Spotlight/APL-Hi-Fi.html
As you can read from the article APL uses 20qty 4387 per side with a pair of ECC99 tubes and custom lundahl transformers i believe the CD player retails around $30k. i guess my point is i would be very interested if someone where to populate some PCB's for a DAC based on the AK4397 chip.
 
A

abzug

Disabled Account
Joined 2006
PigletsDad said:
Let us ignore thermal noise entirely; there is a fundamental limit from quantum mechanics, given by Planck's constant.
Click to expand...
Why ignore it? It's much larger in magnitude than the quantum one you present, and from a calculation someone posted on another forum, then 24 bit quantization is sufficient.

If we set the reference power level at say 1mW (0dBm)
Click to expand...
Where does this come from? The reference threshold should be set at threshold of pain at the ear.

Finally, the calculation must assume the bandwidth of the noise, and the fact that signals which do not have their power spread over as wide a band as the noise can be heard even if they are several decibels below the noise floor.
 
PigletsDad

PigletsDad

Member
Joined 2007
battletonphoenix.co.uk/
abzug said:

Why ignore it? It's much larger in magnitude than the quantum one you present, and from a calculation someone posted on another forum, then 24 bit quantization is sufficient.


Where does this come from? The reference threshold should be set at threshold of pain at the ear.

Finally, the calculation must assume the bandwidth of the noise, and the fact that signals which do not have their power spread over as wide a band as the noise can be heard even if they are several decibels below the noise floor.
Click to expand...

I was just taking the most extreme assumptions (e.g. cryogenic electronics, like those used for radio telescopes and strategic IR sensors) I could to show the utter absurdity of 66bits as a claim.

You are quite right, a more realistic calculation of dynamic range WRT the properties of human hearing suggests that 24bits is not only sufficient, but probably plenty!
 
S

stefanobilliani

Member
Joined 2002
PigletsDad said:


You are quite right, a more realistic calculation of dynamic range WRT the properties of human hearing suggests that 24bits is not only sufficient, but probably plenty!
Click to expand...


The problems that occours down to the listener are fruict of the way the CD /DVD is mastered . There , in the mastering , there is need to large number of bit for the processing/mixing of the signals . An the errors in truncating are deadly fatal . There is no turning back .
 
V

vasyachkin

Disabled Account
Joined 2008
as far as DACs go 32 bit is rather useless

but in interfaces (SPDIF etc) it would be very welcome

i just ran into a problem clipping a 24bit signal yesterday ...

i guess any new 32bit product is good because it makes it more likely for other products to switch from 24bit to 32 in the future

i would very much like for 32bit to become the standard because most chips already run at 32 bit but interfaces still run at 24 which is stup1d imho.
 
T

Telstar

Member
Joined 2007
Re: AK4397

arild said:
As you can read from the article APL uses 20qty 4387 per side with a pair of ECC99 tubes and custom lundahl transformers i believe the CD player retails around $30k. i guess my point is i would be very interested if someone where to populate some PCB's for a DAC based on the AK4397 chip.
Click to expand...

Alex is also making a relatively affordable DAC which uses the same chips and tubes, for under $4000.
 
A

abzug

Disabled Account
Joined 2006
vasyachkin said:
interfaces still run at 24 which is stup1d imho.
Click to expand...
Why is it stupid, considering it's a waste of bandwidth?
Take a given connection, say USB, Ethernet, or HDMI. These are standard and have fixed maximum transfer rates. You can send more channels of 24 bit audio than 32 bit audio of the same sample rate.
 
V

vasyachkin

Disabled Account
Joined 2008
abzug said:

Why is it stupid, considering it's a waste of bandwidth?
Take a given connection, say USB, Ethernet, or HDMI. These are standard and have fixed maximum transfer rates. You can send more channels of 24 bit audio than 32 bit audio of the same sample rate.
Click to expand...

i only need 2 channels ...

plus to worry about bandwidth of an audio signal is a little like worrying about impact on global warming of a flea fart

maybe bandwidth will come into play if you run like 100 channels ... but if you're running a studio then you should be able to afford that bandwith
 
V

vasyachkin

Disabled Account
Joined 2008
abzug said:
What's stupid is introducing more standards than are necessary because someone is having trouble doing something as trivial as truncating to lower word lengths.
Click to expand...

that depends. if the new standard was fully backward compatible and would not appreciably increase hardware cost then why not ?

at this point i think hardware cost is pennies anyway and if you don't understand why truncating bits cannot solve all problems, well, too bad
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Top Bottom