LC vs RC filter on DAC output

[GaryKrant]

Hi all:

After reading about RC and LC filtering options, I'm not sure why major manufacturers use RC filters when dealing with multibit/sigma-delta DAC outputs.

In theory, an RC filter is going to decrease the signal level and to introduce some thermal noise. In opposition, an LC filter doesn't lower the signal level and doesn't provide resistor thermal noise.

I know that LC filters are difficult to build due to the required capacitor and inductor size but, once found a design getting both the required cutoff frequency and the possibility of comercially getting the parts, why using an RC filter instead an LC filter?

Naming some practical examples, Marantz CD6000 builds a 2nd RC filter on the output of their two DACs (prior to using an active filter), and this morning I've got a Sony MiniDisc digital stage based on a CXD2564 multibit DAC building an eight-order RC filter on their outputs.

The only problem I see from LC filters is te RF sensibility but, if properly shielded, shouln't it perform better than an RC filter when dealing with an unfiltered DAC output?

Thanks.

 

[SoNic_real_one]

1. RC filter doesn't reduce the level if it is used in an active design - OpAmp filter (with a good OpAmp). In a passive design, there are not so indicated. The Sony MiniDisc is not exactly an hi-end device, they didn't worry about S/N ratio too much.
2. Precision L components are much more expensive than an OpAmp. You want the filters to be identical on all your players, no? There are hard to use in SMD line. Individually shielding is multipling the costs by an order.

Manufacturers cut costs by any means.

 

[GaryKrant]

Ups, I forgot it: yes, I'm talking about passive filtering.

So do you think is just a matter of costs? Seems convincent, as I said a passive LC filter seems, technically speaking, a better option than a passive RC filter.

And expanding the thread to an active vs. passive filtering field, I'm trying to avoid active filtering for not depending on IC power supply implementations and IC specific tonal characteristics, ie, I want to use a passive filter approach in order to be as neutral as possible. Do you think I'm a crazy fool ?

All your comments are welcome.

 

[Nazar_lv]

Only RC, LC has big "EMI fishing"

 

[davidsrsb]

An inductor that is useful at audio frequencies needs to around 10mH. Simple chokes on ferrite cores are very non-linear and have poor tolerance. You need an air gap to get linearity, which means many more turns on a big pot core and individual adjustment with a trimmer. This all adds up to expensive

Pot core chokes are very resistant to emi as the air gap is internal

 

[GaryKrant]

An inductor that is useful at audio frequencies needs to around 10mH

True, but I'm dealing with an eight times oversampled sigma-delta output... well, seems like I was trying to impress you... what I mean is that my target cutoff freq is 352,8kHz (8x44,1kHz), and thus inductor values for that filter are far lower than 10mH. In this line I modelled a 5th order passive LC filter getting component values easily available through usual electronic components sellers (Farnell or RS-Amidata).

Anyway, by your comments seems that anti EMI measures MUST be had in mind if dealing with an LC filter. I was considering to fully shield all the LPF stage, do you think it should be enough to avoid problems or better going for pot core chokes?

 

[davidsrsb]

Ok, at 350KHz you need values probably less than 1mH
5th order filters are sensitive to tolerance. You do not want left and right to be different

 

[rascal101]

An inductor that is useful at audio frequencies needs to around 10mH. Simple chokes on ferrite cores are very non-linear and have poor tolerance. You need an air gap to get linearity, which means many more turns on a big pot core and individual adjustment with a trimmer. This all adds up to expensive

Pot core chokes are very resistant to emi as the air gap is internal

I don't quite understand the reason why the L-C filter must be resistant to EMI. For this to happen, EMI level is quite high and that the noise is reaching the output stage of the CDP. Or, the EMI noise generator is very close to the inductor and is coupling to it. If the typical CDP has a linear power supply, EMI should not be a major concern. If not, a typical switching power supply should have output chokes and an EMI filter to reduce conducted and radiated emissions significantly. If the EMI noise is external, shouldn't an EMI filter at the front end of the CDP cover this?

 

[prairiemystic]

The magnetic fields inside are not zero. EMI is generated internally from the system clock and PCM stream traces on the PCB, some hum from the power transformer too. Just put a handheld AM radio near your board and hear the whistles and birdies. Also the L-C filters must not have any mutual coupling (mounted at right angles to each other) between L and R channels. I would shield the inductors.

 

[regal]

I have never been able to find a inductor that works for this. I've been told the only ones that sound good are air core speaker types, non of these are available shielded. Shielded these yourself is no easy task.

Best filter I ever used after a DAC was a tubes miller effect.

 

[davidsrsb]

The pot core types with air gap in the centre are fairly well self-shielded, certainly several orders better than a rod or "D" core types with an air gap exposed to the world

They can be mounted close to each other without cross talk

 

[regal]

The pot core types with air gap in the centre are fairly well self-shielded, certainly several orders better than a rod or "D" core types with an air gap exposed to the world

They can be mounted close to each other without cross talk

Where do you find them?,

There is a DAC I had called the Lite DAC-60 that uses a passive I/V then LCLC then SRPP. I made it a first order LC but could never find a inductor that sounded as good as the big air core speaker inductors, problem was it hummed with those air cores. I eventually got rid of the LC filter + SRPP altogether and replaced them with an anode follower with a cap from plate to grid to increase the miller effect enough to give a -3dB around 50khz, that was the best configuration after trying many different models of shielded inductors. The cap is multiplied by Mu for its filtering so it allows you to use a smaller high quality cap.

I guess what I am saying is for a i-out DAC LC filering is troublsome, not because of the design on paper but because "audio quality" shielded inductors are difficult to source. With a V-out DAC it may be more feasible because the signal is probably less likely to pick up interference.