Line input RF reject filter
Posted 19th February 2013 at 08:24 AM by abraxalito
Updated 26th February 2013 at 04:09 AM by abraxalito (Added zoomed passband plot)
Updated 26th February 2013 at 04:09 AM by abraxalito (Added zoomed passband plot)
I've only simmed this so far, no listening tests. The aim here is to have a passive filter which works into a line level impedance (i.e. 10k or greater) and which gives a decent amount of RF rejection without compromising SQ. I plan to incorporate it behind the RCA input jacks of my chipamp.
Design-wise, it started life as a simple 2nd order filter with single series L and shunt C. But when it came to wind the inductor I didn't have a wire diameter small enough to make the required value (44mH). So I split up the inductor into 2 * 22mH and then couldn't resist hanging a cap off the centre tap. The result has turned more into an over-damped elliptic with a higher corner frequency than the original, but with an impeccable phase response to 20kHz. The elliptic-type plot comes about because of the SRF of the two inductors (around 350kHz).
Since fas42 remarked on the phase performance, I've added the group delay plot for the passband. Almost pure time delay of 6.2uS.
glt asked about the inductors - there's now a snapshot of how these things look. About 13m of 0.19mm (copper diameter) enamelled wire wound on to the Chinese equivalent of a TDK/Epcos B65651 P18 bobbin with an AL=250 ferrite core.
<update> I got listening to the filter for the first time last night - didn't sound as good as without. But then my DAC is already very clean in terms of RF out, being NOS with a sharp LPF. I got a little bit of 50Hz hum which means probably my layout is suspect. Also I haven't adopted star earthing on the filter as its on a separate board. I need to investigate closer integration between the filter and the chip amp, not having it dangling on wires.
Design-wise, it started life as a simple 2nd order filter with single series L and shunt C. But when it came to wind the inductor I didn't have a wire diameter small enough to make the required value (44mH). So I split up the inductor into 2 * 22mH and then couldn't resist hanging a cap off the centre tap. The result has turned more into an over-damped elliptic with a higher corner frequency than the original, but with an impeccable phase response to 20kHz. The elliptic-type plot comes about because of the SRF of the two inductors (around 350kHz).
Since fas42 remarked on the phase performance, I've added the group delay plot for the passband. Almost pure time delay of 6.2uS.
glt asked about the inductors - there's now a snapshot of how these things look. About 13m of 0.19mm (copper diameter) enamelled wire wound on to the Chinese equivalent of a TDK/Epcos B65651 P18 bobbin with an AL=250 ferrite core.
<update> I got listening to the filter for the first time last night - didn't sound as good as without. But then my DAC is already very clean in terms of RF out, being NOS with a sharp LPF. I got a little bit of 50Hz hum which means probably my layout is suspect. Also I haven't adopted star earthing on the filter as its on a separate board. I need to investigate closer integration between the filter and the chip amp, not having it dangling on wires.
Total Comments 8
Comments
-
Just noting that you've got a phase shift of about 50deg at 20kHz; people who are paranoid about such matters, and I'm not one of them, will take note of that ...
FrankPosted 19th February 2013 at 11:23 PM by fas42
-
The paranoid should take note that the phase shift of almost 50deg corresponds to an almost pure time delay of 7uS. I will post up the group delay plot. 45deg at 20kHz, 6.2uS group delay almost flat.Posted 20th February 2013 at 03:29 AM by abraxalito
Updated 20th February 2013 at 03:43 AM by abraxalito -
glad i didnt finish the chip amp due to cap shortage. I guess this is a recommended addition.Posted 20th February 2013 at 11:29 PM by buzzforb
-
No, I'm not going to recommend it until I've tried it. Until then its just a curiosity.
The inspiration behind it is Joachim Gerhard's FET buffer for the ES9023 DAC which has improved the sound by at least one account I read. So this got me thinking that if people connect such DACs to this amp, maybe its performance will be compromised. The DAC I'm using doesn't have OOB output so whether it will improve the sound is unknown.Posted 21st February 2013 at 02:20 AM by abraxalito
-
Hi abraxalito,
Seems an easier filter to build than the output filter for the ESS DAC you recommended. If not too much trouble a mouser BOM is appreciated :-). Why hand make the inductors?Posted 22nd February 2013 at 01:02 AM by glt
-
Hi glt - its not designed for the same duty as the ESS filter - that one is really industrial strength because the 9018 output is extremely noisy. This one has lower attenuation. No Mouser BOM for this one because the inductors aren't commercially available, you do have to hand wind them. That's because I was looking for lower DCR than is available in any commercial ones - 28mH is rather a large value and if you go to Mouser for these they'll have DCRs close to 50ohms (mine are below 8ohms). With the high DCR comes lower SRF because the wire length is so long!Posted 22nd February 2013 at 01:59 AM by abraxalito
-
A time delay of 7µs represent an acoustic distance from the source of 2mm.
Just to put the things in perspective.Posted 22nd February 2013 at 01:29 PM by Esperado
-
Too large, it picks up more dirt than it filters out... Especially inside the current-pumping machine called amplifier...Posted 1st March 2013 at 04:42 PM by s3tup
