CS4398 analog filter

[nX3NTY]

Hi there, I'm new here

I have an X-Fi Elite Pro, it have a very nice CS4398 DAC which sounds very good. Its not a good card for music, even Creative put cheap capacitor and wrong value on the VREF and decoupling caps (which I have corrected) Digital supply to DAC has been changed to Sanyo OS-CON and X-Fi CA20K DSP supply has been changed to Rubycon MBZ

At Head-Fi forum I read that this is a voltage out DAC so basically the analog filter is just to filter out-of-band frequency and not for preamplification or current to voltage converter (correct me if wrong) and basically it can be bypassed, so I decided I just tap both channel positive analog out and just use 47uF BG NX HiQ capacitor to RCA out, to my surprise it sounded excellent and clean without any ill effect even with headphones (I haven't tested with RMAA though)

So I wonder why do bother put those opamp and all the filter if a simple capacitor will do? Is it safe on what I'm doing? My setup is Technics SU-V7 amplifier to Gale 3030 floorstander, and Philips SBC HP200 or Koss KSC-75 headphone (feed to Technics headphone amp)

[Kay]

if a simple capacitor will do

you need more, but the question is o.k.

Why don't use a filter alá crossover for a loudspeaker
with capacitors and inductors for a anti-aliasing filter
and using a buffer after that?

[dmills]

The output impedance of the converters is quite high and they are not specified for driving a capacitive load....
Further by only using one of the outputs you are leaving yourself open to all the high frequency common mode hash that the differential stage would have cancelled for you.

Finally, there will be quite a lot of ultrasonic hash present on that output (up into the Mhz region), which apart from being an EMC disaster, is also going to do nothing good to the downstream components (hopefully they will have inputs which are correctly band limited, but many systems don't).

When designing anything you should be conservative in what you generate and liberal in what you accept, the opamp low pass filter is there to band limit the output and to ensure that what is sent to the downstream component is 'conservative' (it will probably work even if downstream is poorly designed). Further it protects the DAC from static and the like (Opamps are typically more robust then small geometry mixed signal parts).

As for using a LC filter, why?? Inductors are about as far from ideal as it gets, a capacitor and a gyrator will beat an inductor for small signal low frequency use nine times out of ten.

IMHO, removing the buffer is taking things to 'as simple as possible, and then some', at least if you want to build a reliable system.

Regards, Dan.

[nX3NTY]

I doing this is because there is something wrong with the circuit when I mod the opamp.

I wonder, why is the need of low pass filter near the DAC when in the amplifier and crossover in the speaker already have it to limit the frequency in the first place?

I want to build to CS4398 datasheet analog out with, but I need few parts first. Does anyone familiar with this DAC and have some simple analog out for me to follow? Maybe I should follow Kay's idea of doing crossover style low-pass filters?

[dmills]

Originally Posted by nX3NTY

I wonder, why is the need of low pass filter near the DAC when in the amplifier and crossover in the speaker already have it to limit the frequency in the first place?

Because you cannot rely on the amp having suitable filtering!

Further, you cannot assume it is driving an amp at all, what if I have it hooked up to something that reacts badly to out of band signals (Like say something with an ADC with inadequate filtering), or an amp which runs out of loop bandwidth and goes non linear, what if the speakers have piezo horns and some numbpty has 'cost engineered' the series resistor?
With that filter present, all the above work as well as they ever do, without it you get birdies, whistles and potentially smoke.

AS I say, be conservative in what you generate, leaving that filter out is not being conservative in what you generate and is expecting downstream components to fix your mess for you.

The data sheet circuit is very, very simple.

Regards, Dan.

[nX3NTY]

Thanks for explaining, Dan. I still new in audio. I done some reading, and yes I'll be an idiot to ditch the filter because it's there for a reason. I found that even the noise in ultrasonic region, it can find its way through to audible frequency and mess up the sound and equipment like you said

To build an active filer requeres me to build linear PSU (I don't want to use PC PSU, and the internal circuit in Elite Pro is fubar) but in the meantime I don't have enough time and resources to build one

My only choice now is Kay's idea, build low pass filter, I opt for simple RC low pass filters. I found a couple of resistor and capacitor quite suitable enough for audio to pass through (cutoff frequency is 19.4kHz, good enough for redbook CD's). I found out after using this filter, it sound a lot cleaner and more listenable without much fatigue after long listening time

[Kay]

O.K.
but I'm think about such thing

[Hurtig]

Well, as others have mentioned, the DAC-chip is NOT intended for driving "heavy"/capacitve loads like a cable or the input on your amp. Therefor you need a buffer.

But more important:
The CS4398 (Like almost any other newer DAC-chips), is a differential DAC. This means that the output signal is NOT represented between Gnd and the output. It is represented between Vout+ and Vout-.
There will be some common mode signal on these, and if you only use one of them in reference to Gnd, you will NOT get the real signal out of it, even though it will play music. You just totally ruin the performance. Just like thoose people listening to old TDA1543 without buffers, believing that the "different" sound they hear is better...

That's why you should always use a differential to single ended buffer. Using an op-amp however... That's just not the best way, in spite of very nice specifications.

[Hurtig]

Originally Posted by Hurtig

Well, as others have mentioned, the DAC-chip is NOT intended for driving "heavy"/capacitve loads like a cable or the input on your amp. Therefor you need a buffer.

But more important:
The CS4398 (Like almost any other newer DAC-chips), is a differential DAC. This means that the output signal is NOT represented between Gnd and the output. It is represented between Vout+ and Vout-.
There will be some common mode signal on these, and if you only use one of them in reference to Gnd, you will NOT get the real signal out of it, even though it will play music. You just totally ruin the performance. Just like thoose people listening to old TDA1543 without buffers, believing that the "different" sound they hear is better...

That's why you should always use a differential to single ended buffer. Using an op-amp however... That's just not the best way, even though it has very nice specifications.

[nX3NTY]

Finally I got some time to buy the components and build it. Although local shop don't have the exact value of resistor, I just use the close approximate value plus all the low capacitance value only availabe in ceramics Despite that now I got what you mean Hurtig, it sounded much fuller now, and some detail has been brought up