Active elliptic anti-imaging filter
Posted 20th March 2013 at 03:18 AM by abraxalito
Updated 20th March 2013 at 03:30 AM by abraxalito (Added FR)
Updated 20th March 2013 at 03:30 AM by abraxalito (Added FR)
Passive filters rock for SQ, no doubt about it but I'm still curious how good sounding an active DAC I/V post filter might be. So I've figured out an almost equivalent FR active version of my 7th order LC elliptic filter. This active elliptic has been designed using LTSpice's FilterCad program giving the pole/zero positions, then the Williams handbook of filter design helped me translate those numbers into a working circuit. Its using what Williams calls the VCVS 2nd order section based on a twin-T network to realize the zeroes.
My first attempt at an active elliptic filter was using gyrators but that proved very hard (practically impossible) to get stable with CFB opamps due to their HF gain peaking. VFB opamps I ruled out at the start for inadequate SQ - its not hard to make gyrators stable with them. Hence this approach which promises to work with CFBs though I'd guess I'll probably need to add series Rs between the stages in practice. Nothing built yet but thought I'd share the schematic. It comes with around 30dB of gain, unavoidably. The other caveat is that the final stage has a high-ish Q (around 8), despite this being a 'minimum Q' design. This does mean its very sensitive to the precise component values - especially the gain setting resistors in the FB loop of the opamp.
My first attempt at an active elliptic filter was using gyrators but that proved very hard (practically impossible) to get stable with CFB opamps due to their HF gain peaking. VFB opamps I ruled out at the start for inadequate SQ - its not hard to make gyrators stable with them. Hence this approach which promises to work with CFBs though I'd guess I'll probably need to add series Rs between the stages in practice. Nothing built yet but thought I'd share the schematic. It comes with around 30dB of gain, unavoidably. The other caveat is that the final stage has a high-ish Q (around 8), despite this being a 'minimum Q' design. This does mean its very sensitive to the precise component values - especially the gain setting resistors in the FB loop of the opamp.
Total Comments 12
Comments
-
Posted 20th March 2013 at 01:16 PM by rjm -
The aim is to see if it can improve on the NOS sound - to me, the lower the sample rate the better the sound. Upsampling to me is a recipe for inferior SQ - the faster we sample the better the dynamic performance of the DAC needs to be because there are more glitches and a greater proportion of the time is spent in settling.
Posted 20th March 2013 at 02:44 PM by abraxalito -
Have you considered doing an FDNR based filter? There is an example of one here https://www.analog.com/library/analog...inal_PtB_F.pdf page 67. It's only a 5th order butterworth, but it might give you some ideas. I suspect you won't have the same issues I have had with mine, because it is operating at a high frequency (I was doing mine at 200Hz). I like the idea of the opamps only being shunts. They of course are still in the signal path, but something about the fact that the signal does not actually pass through them appeals to me
Tony.Posted 21st March 2013 at 12:07 PM by wintermute -
Hi Tony - yes the FDNR was my first attempt at making this filter - its using gyrators and those can't get stable when the opamps are CFB type, due to the peaking at HF. Doing it at 200Hz I suspect normal opamps will be fine though, so long as they have plenty of GBW.
I'm not a subscriber to the 'nothing in the signal path' idea because to me no such thing as a 'path' for a signal, they always travel in circlesBut yep I can see the appeal. The FDNR approach uses twice as many opamps - was yours a lowpass filter? If so I prefer the MFB topologies, at 200Hz a passive LC LP will have a particularly huge inductor (perhaps 1H at least for a reasonable load impedance).
Posted 21st March 2013 at 01:55 PM by abraxalito -
I love elliptic filters, and I have recently been using them in active (DSP based) crossovers for my loudspeaker, 5th, 6th and 7th order. So I am happy when I see a nice filter like this!
That sure looks like an 8th order filter, not 7th (there are 4 notches and the trend at HF is constant amplitude, not decreasing at 6dB/oct as for an odd order filter). You mention 7th order in your blog post...
One drawback of a standard elliptic filter of high order (like 8) is the very high Q of the stage with its corner frequency nearest the cutoff frequency. This will cause a large peak in the group delay, so you might want to look into that.
If you can calculate the corner frequency and Q of each stage in the filter, you might try designing the filter as a cascade of lowpass notch filters...
-CharliePosted 21st March 2013 at 05:30 PM by CharlieLaub -
Yes Low pass. I had problems with oscillation, what I did find was that some implementations of the GIC's worked better than others (there are three basic topologies in that Jung article). The main problem I believe is that opamp is driving a capacitive load. I ended up putting some compensation caps in which (on the breadboard) brought it pretty much under control. I hope (still haven't done the proper layout and implementation yet) that when implemented carefully that remaining issues should be sorted.
Interestingly it was much better behaved with a lower bandwidth opamp than with a higher one... the opa2134 worked better than the opa2604.
some of my troubles https://www.diyaudio.com/forums/analo...ml#post2899524
Tony.Posted 21st March 2013 at 09:16 PM by wintermute -
@Charlie - I think I've come across some of your posts on here, some inspiration drawn from your enthusiasm for elliptics, so thanks! Yes, good points - this is in fact a 9th order filter - I was looking for response in the same realm as my passive 7th order design but was concerned about Q in the active realization because of component sensitivity. So I went for a minimum-Q form which pushed up the required order to achieve the same stop band.
I have an idea that, in conjunction with my transversal digital filter, I might be able to move some of the 'peakiness' to the digital part. Intend to look into that sometime... Another reason for building this active form (apart from smaller size than the passive) is that I wanted to understand how elliptics are made up of smaller parts. With a passive realization it seems that everything interacts together so its really hard to work out what's going on. But here I now understand each stage separately and how the total response is built up.
@Tony - I found an interesting BBC paper about stability problems with GICs and how to solve them, are you interested? If so I can probably find a link to it. According to that paper, the stability issues aren't in driving capacitive loads, rather because the two opamps in the gyrator feed back to each other and if the gain goes above unity at any frequency, it goes unstable. That's the reason I abandoned my design - I was using CFBs and they have peakiness in the 10s-100s of MHz.Posted 22nd March 2013 at 01:36 AM by abraxalito
Updated 22nd March 2013 at 01:53 AM by abraxalito -
Hi abrax, yes I would be interested to see that if you can find it! From the sims (and my limited understanding) I had come to the conclusion that there was a phase margin issue, but that could be completely wrong. I'm sure my breadboard layout wasn't helping anything
https://www.diyaudio.com/forums/attac...dsc_0473_s.jpg
Tony.Posted 22nd March 2013 at 10:03 AM by wintermute -
Posted 22nd March 2013 at 12:30 PM by abraxalito -
Posted 22nd March 2013 at 11:07 PM by wintermute -
Posted 23rd April 2013 at 02:31 AM by bear -
Posted 23rd April 2013 at 02:02 PM by abraxalito