Question on DAC reconstruction filter

[sachi]

After recently comparing my DAC to other higher end DACs, i have noticed that there is still room for improvement with my own.

I am trying to look at what can be done to modify the sound (for better or worse only after listening would i know). Looking at the current configuration of the reconstruction filter on my DAC it seems it is using a passive first order LP filter followed by a 2nd order Sallen key filter.



I was wondering what changes i can do to this configuration.

Is it possible to just have a simple 2nd order passive filter here by doing this
I could possibly reduce the corner frequency further and/or are there other options i could try on it.

[Calvin]

Hi,

it depends on the clocking of the DAC. The higher the clock frequency the lower the order of the filter can be. And for 192kHz You might even omit with a filter alltogether just use a 1st-order passive network (technically some may say that it´s not sufficient, but sonically I prefer the no-filter versions).
With current output DACs the biggest improvement could be a simple discrete I/V-converter stage.

jauu
Calvin

[GaryKrant]

Hi Sachi:

I supose the manufacturer has accurately calculated the cutoff freq needed for that DAC so modifiying that value, in my opinion, is not going to give you more quality.

However, building a higher order passive filter should help to fit the cutoff with more precission. In any case, I think that increasing the filter order in one unit is not, again, to give you a consistent gain. You can check it simulating a 1st and 2nd order filter in many of the filter calculators available in Internet.

I really think the key to improve the preformance is, by order:

- In the used IC: AD797N seems a very good dude, but maybe some user should give you some feedback about replacing it by another model.
- In the quality of the power supply feeding the IC
- In the quality of the filter passive components: resistors might be film (for thermal stability) and of the closest tolerance possible and capacitors might be ceramic (I'm sure they are) with C0G thermal factor to get the best linearity possible.
- In the decoupling capacitors (quality and mount position, not size) in the IC power line. About qualities... uff, the ethernal debate: Elna Silmics, Rubycons Z, Black Gates,... About mounting position the closest to the IC pins.

Hope my feedback was useful.

[sachi]

Thanks Gary...the DAC already has very good quality components in the passive filter, decoupling and resistor positions. Power supply is also quite decent.

I am looking to simplify the filter down ..will run a few simulations when time permits on some possibilities and see what comes off of it.
A few details on the DAC, unit is an Assemblage DAC 2.7
DAC chip PCM1704 - 8xOS.

[wintermute]

If you want to get rid of the sallen key and replace it with a filter that is like a passive, but is really active then take a look at this app note from burr brown. http://focus.ti.com/lit/an/sbaa001/sbaa001.pdf

I've been doing a lot of research on creating LP filters using FDNR's of late. At first I struggled to work out how to get the "normalised" values for the passive filter which is transformed to the FNDR filter, until I realised I just had to use an online filter calculator and tell it the impedance was 1 Ohm. This of course assumes that the Q of the circuit is one of the standard ones such as Butterworth, Bessel, chebichev LR etc. I use this calculator --> Crossover Design Chart and Inductance vs. Frequency Calculator(Low-pass)

As the circuit you have shown does not show the load's impedance you may also need a buffer between the fdnr and the load (in fact you may need one before it as well).

This document --> Crossover Design Chart and Inductance vs. Frequency Calculator(Low-pass) also has some excellent info on FDNR's as well as a design excercise.

Note I know nothing about DAC's but I think that a reconstruction filter is an anti-alias filter, so these two docs should be of use

Tony.

[wintermute]

sorry that third link should have been http://www.analog.com/library/analog...inal_PtB_F.pdf

Tony.