Drive NOS AD1865/62,PCM1704/02/63,TDA1541 from FIFO: Universal I2S-PCM driver board

I'm using a Squeezebox Touch with EDO plugin (up to 192KHz digital out) as a source. The Squeezeserver running on the PC is using SoX and can be configured to oversample on the fly, if one wants this. As I'm using the SB Touch standalone connected to a 1TB HDD I oversampled offline(manually) the music I listened so far.
 
regal, Zsolt,
since seems you play mostly 44k1/16, ripped CD, like me (I play directly CDs), what is the goal of oversampling to feed a NOS DAC?
I can understand if one want to feed a last generation SABRE (but also in that case I ask myself what advantage, since the source is anyway 44k1/16bit), but when we are talking about feed AD1865, TDA1541 or so in NOS mode, I have just a little diffcult to understand the reasons.
No advantage in signal to noise ratio, since the theorical 96dB of dynamic range does not change after oversampling.
No reduction of THD, there are other way to reduce THD.
One can run the DAC faster.... I cannot demonstrate anything, but seems it's better to run old style DAC chip slower than faster in NOS mode.
The analog filtering? I don't believe, since, for example, a Zanden filter has a phase shift in audio band of 15 degrees only.
Also, in my experience, with a simple 10nF capacitor or filterless these DACs sound better than oversampling.
I'm very curious to know.

Andrea
 
Oversampling a fliterless NOS DAC moves the aliasing artifacts upwards in frequency and outside the audible bandwidth.

But, some people don't care about such things, and accept NOS as it is.

My perception is that the NOS TDA1541 sounds best when running at 48khz. Providing it with 96khz or 192khz produces "smoother" (for lack of a better description) highs, but less dynamic bass, and less dynamic mid-range.

Of course, this is all subjective, and determined by the bitrate which you tune your DAC to work at.

Probably I theorise, with fine tuning, a person could make TDA1541 sound best at any sampling frequency.

I wont comment on other chips.
 
My perception is that the NOS TDA1541 sounds best when running at 48khz. Providing it with 96khz or 192khz produces "smoother" (for lack of a better description) highs, but less dynamic bass, and less dynamic mid-range.

That perception is pretty much shared by myself (though not from listening to the 1541, rather the 1545/1387) and other guys listening to PCM1704s over on SNA.

How I explain this is that running faster gives more glitches but also moves the image frequencies out of the way. Its those >22kHz image frequencies which cause loss of smoothness when running NOS. The increase in glitches raises the subjective noise floor and squashes the perceived dynamics.
 
it can also correct for the NOS rolloff and doing it externally in the computer you can run much more powerful linear phase apodising OSF.

Introducing any kind of digital filtering when the DAC's only got 16bits for input is a recipe for lost dynamic range. Power is no guarantee of quality incidentally - linear phase comes with pre-ringing. The apodising filters I've seen have mostly been minimum phase - which have you seen which are phase-linear?
 
I'm oversampling for the same reason as others: 1. to push the images around the sampling frequency(and its multiplies) upward so to ease the job of the analog reconstruction filter; 2. to have less high frequency roll-off. Doing it 4x only because cannot stream higher otherwise would do 8x or 16x.

Doing it on PC for two reasons: 1. higher processing power (better approximations) then "available" DFs I could use: DF1704/6, SM5842 or PMD100; 2. to try digital filters I couldn't otherwise: meridian apodizing (link) . Indeed it's not phase-linear: technical analysis of the meridian apodizing filter (link) ...but everything comes with a price.

An interesting podcast to listen (if impatient go to 36'): Discussing Negative Frequencies with John Atkinson and Scott Wilkinson

All in all hearing is very subjective. As qusp said: try it you might like it.
I personally like better with 4x apodizing filter then NOS.
 
Last edited:
Introducing any kind of digital filtering when the DAC's only got 16bits for input is a recipe for lost dynamic range. Power is no guarantee of quality incidentally - linear phase comes with pre-ringing. The apodising filters I've seen have mostly been minimum phase - which have you seen which are phase-linear?

typo, minimum youre correct, too much jargon bouncing around this old head. i'm not speaking for someone else, just mentioning why someone might do oversampling with a NOS dac, not the TDA specifically. which I freely admit o knowing next to stuff all about partially because of the 16bit limitation.

it doesnt really interest me despite the amount of engineering people do to try to squeeze the last bits out of it, because I mess with all sorts of filters in the digital realm.
 
Last edited:
All in all hearing is very subjective. As qusp said: try it you might like it.
I personally like better with 4x apodizing filter then NOS.

The nice thing is flexibility here with this board But I don't think there is an apodizing oversamping filter for the PC (let me know if I am missing something)

But ultimately the big question here is..

WILL WE HAVE BOARDS FOR XMAS ? :hohoho:
 
40


TV Man x 2
merlin el mago
rtd
jameshillj
erin
Nikola Krivorov x 3
Rupor54
vitalica
analog_sa
andrea_mori x 5
dsavitsk
tagheuer
hirez69 x 2
ccliu
BDL
cddumat
regal
zoran
Jogi
Zen(zenelectro) x 2
Buzzforb
SPWON
crobbins5421
marcus1 X 2
casshan x 2
Dweeb99 X 2
jackw X 3
 
TOTAL = 42

TV Man x 2
merlin el mago
rtd
jameshillj
erin
Nikola Krivorov x 3
Rupor54
vitalica
analog_sa
andrea_mori x 5
dsavitsk
tagheuer
hirez69 x 2
ccliu
BDL
cddumat
regal
zoran
Jogi
Zen(zenelectro) x 2
Buzzforb
SPWON
crobbins5421
marcus1 X 2
casshan x 2
Dweeb99 X 2
jackw X 3
kazap X 2
 
Last edited: