(new design) A/D and D/A convertor buffers

[SonsOfThunder]

Butta':

Not to get picky on you, but I think you should clarify that this would be (discrete) filter design, not just a buffer. IIRC, NYDave did some stuff on just a buffer a while back...or was that you?

HTH!

 

[daArry]

u peeped this page butta? http://www.borbelyaudio.com/allfetdac.asp

 

[NewYorkDave]

[quote author="buttachunk"]the A/D input buffer; according to the AKM datasheet for AK5394A-- gain of -14.5db, high pass filter .7hz, low pass filter 320khz.[/quote]

Kinda vague. By "-14.5dB" do they mean a LOSS of 14.5dB or an inverting gain of 14.5dB? And it's 14.5dB relative to what level? What slope for the HP and LP filters? Current or voltage output?

the D/A output buffer; generally a unity gain low pass filter.

What corner frequency and slope? Current or voltage input?

If there are more complete specs in the datasheet, you could cut 'n paste 'em here instead of making us hunt around for the datasheet. :wink:

 

[SonsOfThunder]

[quote author="NYD"]Kinda vague. By "-14.5dB" do they mean a LOSS of 14.5dB[/quote] Yeah, cause you are limited to 0.7V on the input of every AD that I know of...

[quote author="buttachunk"]"why do we need complex low pass on the output since the DA convertor chip already does that ?"[/quote] That's a good question, possibly due to the various sampling freqs?

 

[mcs]

[quote author="SonsOfThunder"]Yeah, cause you are limited to 0.7V on the input of every AD that I know of...[/quote]
No, that varies a lot. On the two ADC chips I have used so far it is 1Vrms (TI PCM1800) and 4Vrms (AD1871). On some of the new TI ADCs it's 4.24Vrms... So the gain has to be variable if a general purpose circuit is the target.

Best regards,

Mikkel C. Simonsen

 

[mcs]

[quote author="buttachunk"]also, on the DA convertors, the designs i posted were 3 pole filters. but in thinking about Marik's mod, i began to wonder "why do we need complex low pass on the output since the DA convertor chip already does that ?"... [/quote]
One of the latest high-end fads is DACs with no oversampling and no output filtering of any kind. According to many tests the result are great...

Best regards,

Mikkel C. Simonsen

 

[Svart]

What about the Crystal ADCs and DACs? I've peeped at those quite a bit for something I would like to build and I wonder about a buffer for those too..

 

[Ian MacGregor]

[quote author="buttachunk"]they are not neccessary for anything below the highest maximum reproducable frequency, which in the case of 96k converters is 192khz, and for 192k convertors is 384khz....[/quote]

Isn't is the other way around?? Sampling rate of 192khz gives a max freq response of 96khz?? Nyquist??


Ian

 

[daArry]

Hey:

Pulled some info from the PCM4104 datasheet:

--

Each differential output is typically capable of providing
6.15V full-scale (differential) into a 600Ω output load. The
output pins are internally biased to the common-mode (or
bipolar zero) voltage, which is nominally VCC/2. The output
section of each D/A converter channel includes a
single-pole, switched capacitor low-pass filter circuit. The
switched capacitor filter response tracks with the sampling
frequency of the D/A converter and provides attenuation of
the out-of-band noise produced by the delta-sigma
modulator. An external two-pole continuous time filter is
recommended to further reduce the out-of-band noise
energy and to band limit the output spectrum to
frequencies suitable for audio reproduction. Refer to the
Applications Information section of this data sheet for
recommended output filter circuits.

An external output filter is recommended for each
differential output pair. The external output filter further
reduces the out-of-band noise energy produced by the
delta-sigma modulator, while providing band limiting
suitable for audio reproduction. A 2nd-order Butterworth
low-pass filter circuit with a −3dB corner frequency from
50kHz to 180kHz is recommended.

--

I posted the output (using the opa1632) in a thread earlier today - would like to know if summin else could be done tho :thumb:

 

[mcs]

[quote author="daArry"]Pulled some info from the PCM4104 datasheet:[/quote]
Have you looked at the PCM4204 datasheet also?

The recommended input buffer looks very strange to me. It has a gain of about 0.25, so you need a 24Vpp input signal!?! It also doesn't include the input protection that's needed according to other parts of the datasheet... I have asked Rochey about it, but I haven't heard from him yet.

The PCM4204 looks perfect for HDRec use BTW - you can upgrade from 4 to 8 channels just by adding an extra chip!

Best regards,

Mikkel C. Simonsen

 

[daArry]

Have you looked at the PCM4204 datasheet also?

Not in depth, but ye - this was the chip I was thinking for the a/d side since rochey mentioned it, but I really needed the d/a sorted first so spent most time on that...

The PCM4204 looks perfect for HDRec use BTW - you can upgrade from 4 to 8 channels just by adding an extra chip!

cool, same with the 4104 dac - that's why I thought ideal for an 8 chan d/a adat brd... :thumb:

 

[CJ]

here's an 88kc buttaworth!

 

[PRR]

> By "-14.5dB" do they mean a LOSS of 14.5dB

> The recommended input buffer looks very strange to me. It has a gain of about 0.25, so you need a 24Vpp input signal!?!

Yes, yes.

Remember we are coming from many-volt analog systems to under-5V digital comparators. When honking a long telephone line through an urban area, you need real power at significant voltage. But the 1-inch path from buffer-out to the gizmo inside the ADC doesn't need overpowering, and low-volt digitals work more-perfectly on analog inputs than high-volt comparators.

The -12dB/-14dB pad (or loss-buffer) brings our +24dBm/600Ω line-level peaks down to the 0.7VRMS (or so) levels that the actual ADC can swallow.

You want a buffer, not just a passive pad, because the input of the ADC is full of switching-crap and depends on a VERY low impedance drive to keep the analog input clean. If all sources drove 600Ω with grace and aplomb, a 220-150-220Ω U-pad might be OK; since modern practice dictates 10K inputs, we need an active buffer.

 

[mcs]

[quote author="PRR"]> The recommended input buffer looks very strange to me. It has a gain of about 0.25, so you need a 24Vpp input signal!?!

The -12dB/-14dB pad (or loss-buffer) brings our +24dBm/600Ω line-level peaks down to the 0.7VRMS (or so) levels that the actual ADC can swallow.[/quote]
I understand that, but the PCM4204 I talked about doesn't have a 0.7Vrms input - it has a 6Vpp input (which would be 2*2.12Vrms if I'm doing the maths right). So you actually need to amplify a standard "+4" signal for full ... (I don't remember the word I'm looking for). Then why do they suggest a circuit with 0.27 times gain, when you actually need a gain larger than 1?

Best regards,

Mikkel C. Simonsen