CS8420/AD1896 and 1541

Smiffy

Member
2002-08-21 3:30 pm
UK
Hi All
I may be totally wrong here and may have not read the datasheets properly(knowing me)but if the TDA 1541 can handle a 176.4KHz max sampling rate is there anything stopping me using a CS8420 or AD1896 and upsampling to 176.4KHz or lower and feeding it to the TDA1541 without oversampling?.
Smiffy
 

Smiffy

Member
2002-08-21 3:30 pm
UK
Ta for the help
I always thought they were different.Haven't picked up a soldering iron for ages and have forgotten a lot of inf.On another note with reg 1541 which way do the versions go.Is R1 the lowest performing followed by S1 and then S2 the best?Thanks for the help.Just removed SAA7220 from Musical Fidelity Digilog.Sounds noticably better.Smiffy
 

Lenin

Member
2002-02-15 1:42 am
London
A Digilog? Pleaes post some more info!

Hi Smiffy,

Just picked up a digilog myself - been planning to finally get stuck into some digital kit for a while.

Do you happen to have a schematic? (Mine has a mk3 board)

Could you post some more info on your mods, maybe a picture too?

Thanks.
 
oversampling or upsampling is the same
This is not quite true. According to the common usage of both terms, "oversampling" is the classic interpolation with a whole-numbered factor (mostly 4x, 8x, or 16x), whereas "upsampling" means asynchronous sample rate conversion with independent input and output clock. Exactly what CS8420, AD1896, and SRC4192 do.
 
It doesn't "work in the same way" but from an information theory standpoint the result is the same. The DF1704/6 is less complicated than an ASRC because it doesn't need a rate estimator, huge tables of FIR coefficients, and so forth. But if you consider what is happening to the signal in an ASRC with Fs_out == Fs_in * N where N is an integer, it is the same thing that would happen to the signal in an Nx oversampling filter.
 

rfbrw

Member
2001-10-26 11:51 pm
.
jwb said:
It doesn't "work in the same way" but from an information theory standpoint the result is the same. The DF1704/6 is less complicated than an ASRC because it doesn't need a rate estimator, huge tables of FIR coefficients, and so forth. But if you consider what is happening to the signal in an ASRC with Fs_out == Fs_in * N where N is an integer, it is the same thing that would happen to the signal in an Nx oversampling filter.

That is tenuous to say the least. One might as well argue that a Wankel engine is the same as a square 4 because they both take in fuel/air and have a rotating bit.

ray.
 
Sorry, I couldn't tell if there was an actual rebuttal in there somewhere. I'll make it more clear: an ASRC and an "oversampling digital filter" have the same effect on the information stream when the output-to-input ratio is an integer (2,4,8 etc). They both, to my knowledge, use interpolation and decimation filters to achieve the result. That the ASRC can do so asynchronously makes it far more complex, and indeed the ASRC induces further error on the signal with non-integer ratios. But they are not fundamentally different processes.

I await your educational remarks.
 

rfbrw

Member
2001-10-26 11:51 pm
.
Given the large volume of information out there on this topic, from comments on the AD1890 by Robert Adams in the rec audio groups to the recent thread in this forum, the patent applications for the CS8420 and the AD1890/96 devices and articles in the AES journal not to mention the datasheets, you can easily educate yourself but it appears you choose not to and as such it was an act of sheer folly on my part to even contemplate challenging your assumptions. More fool me, I'll know better next time.

ray.