Aliasing Intermodulation Distortion and filterless DACs

[dimitri]

Sorry, the correct link is without dot at the end
www.microchip.com/download/tools/analog/flab/flb1040.zip

 

[jean-paul]

Did you actually measure the amount of generated RF at the output ? I doubt if this will destroy tweaters. The ones I use survive everything till now. Don't like the high RF output of cellular phones in short distance of my gear. More risky than a non os DAC.

Just to be clear:
Do you realise that filterless in this thread only means digital filterless ? The necessity of analog filtering at the output is not the point of discussion.

This is right. It is psychological issue, nobody will criticize it’s own offspring. Could you say, that you’d spend hundreds of hours of your invaluable time to build this DAC, and that in the long run the sound is not satisfying? It is impossible to spend $50 and make a good DAC

My opinion is that there are DIY ers that are very critical about their built stuff and some draw the line at a very high standard. A lot of them are very well capable of saying that their new project has failed soundwise. The matter of the money is questionnable.
Although priced a bit higher the M-Audio SuperDAC ( 300 us dollar ) gets great reviews. It is a oversampling unit.
Mine is coming in a few days and I am curious about the quality of this cheapo. Devices don't have to cost thousands of dollars to sound good.

please don’t refer to the 47 lab, that’s frauds, it is marketing issue how to sell $50 DAC to you for $1000 …

I think you're on thin ice now...

 

[dimitri]

Do you realise that filterless in this thread only means digital filterless ? The necessity of analog filtering at the output is not the point of discussion.
----
They are two sides of the medal. Digital filter also degrades a signal, like analog one. Digital filters are more convenient and cheaper, that their analog equivalents.
There is some papers regarding audibility of analog anti-aliasing filters with sharp cut-off and non-uniform group delay, but I don’t hear about audibility of digital filters.
----
Did you actually measure the amount of generated RF at the output?
----
Yes, I do. It can be small, but the amplifier should be tested also on the out of band signals. The amp behavior even with a small hf content can be unexpected. The quiescent current can be increased sharply, or the amplifier can jump to hf oscillation, etc. The amp should be stable like a rock – sorry but this is not an issue even in commercial units.

 

[Jocko Homo]

Will there be any "real" audio engineers out there that symphatise with the non os topology at all and have a suitable theory/practice evaluation ?

Not likely. You already have my explanation.

Why do we always have to read this kind of crap? If it is driven by stupidity why are the results good then? Everybody that hears the DAC's likes the sound so far. My speakers like them.

Because a lot of us feel that way. And not everyone likes them. Good by your standards, not mine. Better than it should, yes. But that is not what most feel is good.

Does it matter why one thinks it sounds good ? It is ultimately a subjective decision.

If you guys think it sounds good in your system, then fine. Makes all of us happy. Just stop trying convince the rest of us that we are the idiots. You asked for a possible explantation, but no...... instead it turns into another "flat earth" debate.

Jocko

 

[jean-paul]

Jocko, your post won't change the earth from flat to round

Nobody was trying to convince the rest of you that you are the idiots, it was merely the other way around.

 

[dhaen]

Better than it should, yes. But that is not what most feel is good.

So, Jocko, are you convinced yourself (by research or other), that OS is the best way to go in the long term? And that it's best to forget nonOS, despite it's "smile factor"?

 

[rfbrw]

Re
quote:
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
Does it matter why one thinks it sounds good ? It is ultimately a subjective decision.
--------------------------------------------------------------------------------

If you guys think it sounds good in your system, then fine. Makes all of us happy. Just stop trying convince the rest of us that we are the idiots. You asked for a possible explantation, but no...... instead it turns into another "flat earth" debate.

Jocko
--------------------------------------------------------------------------------


Jocko you quote me out of context. My point was that this was a futile quest given the impossibility of quantifying an objective definition of good. There are plenty of true believers out there ready to spread the word and they need no help from me.

ray.

 

[weissi]

So, in case of using only analog filtering, wich order would be the "best" compromise in terms of aliasing and phase. I can remember to read somewhere about 9th order? you'll need as minimum 4 opamp´s for that, sound´s not so good.... thought to use a 6th order active filter with opa4134´s after a TDA1541 non-oversampling....

 

[dimitri]

Dear weissi
you can use 3rd order linear phase (constant group delay) filter like in http://www-s.ti.com/sc/psheets/sbaa001/sbaa001.pdf
That filter has very low intermodulation distortion. You can simulate it by microchip free software, which I had already quoted. The design can also be passive - two inductors and one capacitor.

 

[Elso Kwak]

Third order Linear Phase Lowpass Filter

Dear weissi
you can use 3rd order linear phase (constant group delay) filter like in http://www-s.ti.com/sc/psheets/sbaa001/sbaa001.pdf
That filter has very low intermodulation distortion. You can simulate it by microchip free software, which I had already quoted. The design can also be passive - two inductors and one capacitor.

Hi Dimitri,
I trust you scale the design down to a crossover point of 10 or 15 kHz. Otherwise it will be rather useless in a NON-OS application....

 

[jean-paul]

Ha ha, as always it's in the details Elso

 

[dimitri]

Hi Elso,
in musical fidelity DAC there is the same 3-rd order filter, after measurement I got -3dB@30êHz -10dB@44kHz -18dB@60kHz.

 

[Elso Kwak]

Analog lowpass post filter

Hi Elso,
in musical fidelity DAC there is the same 3-rd order filter, after measurement I got -3dB@30êHz -10dB@44kHz -18dB@60kHz.

Hi Dimitri,
OK, these values are for a oversampling DAC (DAC with digital filter). If we assume 8 times oversampling we will need for the NON-OS DAC a -3dB point at 30kHz/8= 3.75 kHz. This is unpractically low and can not be used! I my DAC I am using a -3dB point at 10kHz with third order Bessel characteristic (linear phase, constant group delay)

 

[weissi]

Hmmm, so i shall use inductors..... teachers always mentioned that they are "not so suitable" in filters.... when it´s necessary to use them the best would be air coil´s, and i´m not angry when i had to use Opamp´s... beside that, 3th order can I get easily with only one opamp stage (1th order directly at the dac and the two others with the external opamp..)
My Dac will be directly connected to an "szeekers" type headphone amplifier, so there aren´t much components in the signal path, so I think additional opamp´s for filtering aren´t that bad...

Regards!

 

[jwb]

Re: Analog lowpass post filter

Elso, just to be sure I understand you, you have a DAC with a 3rd-order low pass filter at 10kHz? That seems .... low. You are throwing away half the usable bandwidth (or 75% of the bandwidth of a 96kHz recording).

 

[dimitri]

Hi Elso,
yes, you are right, it has YM3623, 8x oversampling HDCD PMD100,
PCM69, 4x5534. We can use six order. That filter I had mentioned used opamps only at hf, but in bessel opamp will receive full-range signal (with hf) at the input.

 

[dimitri]

Elso wrote > my DAC I am using a -3dB point at 10kHz with third order Bessel characteristic (linear phase, constant group delay)

Elso please could you publish schematic, I've recently simulate TI AB-026A (just for fun) and got different graphs (non uniform group delay) ...

... third order... Have you used CLC filter?

 

[pburke]

Halcyon, you're alone in the desert. I am keen on following this thread but nobody seems to be able to synthesize a testable hypothesis as you call it.
Maybe this is due to the fact that most of the visitors are DIY ers ? That build the thing without having to understand WHY sound is good ?

you're probably right - I am one of these folks who built it and don't know why it sounds so good, but since I am not into this hobby to figure out the "why" but to improve the "what" it really doesn't matter much to me what the text books say. With audio, a lot of things are simply not that easy to explain. There's the whole debate on cables with the camp of those who insist that it's just a wire and anything that properly conducts electricity will sound the same, and then there are those who actually tried different wires and HEARD the difference.

I heard the difference with my DDDAC1543, and it is so profoundly better than anything I have heard before, it isn't even necessary to look for words to define the nuances. It's blatently obvious to anyone who bothers to give it a brief listen.

If people think it's stupid and only iditots use non-oversampling DACs, then let me join the ranks of the stupid - at least they have the sweet sounding DACs

 

[Elso Kwak]

low-pass filter

Elso wrote > my DAC I am using a -3dB point at 10kHz with third order Bessel characteristic (linear phase, constant group delay)

Elso please could you publish schematic, I've recently simulate TI AB-026A (just for fun) and got different graphs (non uniform group delay) ...

... third order... Have you used CLC filter?

Hi Dimitri,
It was here:
http://www.diyaudio.com/forums/showthread.php?postid=367237#post367237
If you want a higher crossover like 15 or 20 kHz can be easily accomplished in the program.
I did have some problems with the new FilterPro under Windows as I could not get a gain of one for the second section.

 

[Dave]

A typical non-oversampling DAC will not show a flat frequency response across the audio band but will be around 4dB down at 20kHz. See "Principles of digital audio, Ken C. Pohlmann" pg. 91 for an explanation and reason for this.
Perhaps this HF roll off is what some people like the sound of. It would be even greater with an analog filter close to (or in, Elso) the audio band.

Has anyone tried to boost the HF output of a NON-OS Dac to make it flat across the audio band? Would it still sound good?
You of course would not want to boost the ultrasonic content as well.

Also you have eliminated a potentially poor quality digital filter and also possibly lowered the senstivity to jitter although I'm not totally sure if this last point is true.

So perhaps what is really needed is a top end oversampling design, with SM5847, PMD200 or custom DSP based digital filter and a decent set of tone controls!

I must admit that I have not tried a NON-OS DAC but I thought I'd put the above forward anyway.