[BC] Correct URL for the Symmetra-Peak

[Robert Orban]

At 11:58 AM 8/16/2006, you wrote:
>From: Robert Meuser <Robertm at broadcast.net>
>Subject: Re: [BC] Correct URL for the Symmetra-Peak
>To: "Broadcasters' Mailing List" <broadcast at radiolists.net>
>Message-ID: <44E33D11.4000500 at broadcast.net>
>Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>
>An opamp version would be easy if gyrators were used. I bet with the
>adjustability of the gyrator you could tweak more customized
>performance.  Next questions, how would this compare to the less complex
>Orban implementation in the 8100?  Are digital implementations better?
>Would the gyrators make for a better "sound" than all those inductors.

I forgot to answer these questions in my previous post.

The Kahn unit's magnitude unflatness will certainly not be audible and it 
has the same total phase rotation as the allpass filter in the 8100*. 
However, Kahn's non-minimum phase zeros are a bit higher in frequency than 
the 200 Hz zeros in the 8100.

Chances are that an active RC implementation would be better sounding than 
the inductors and transformers in the Kahn. However, I would not use 
gyrators because floating gyrators have lots of parts. If I were to 
synthesize an active RC that literally duplicated the poles and zeros in 
the Kahn units, I would use a non-lattice realization that does not use 
simulated inductance.

Digital implementations can have lower noise and distortion than any analog 
implementation but the best you can do is to make the phase response of the 
digital implementation have a mini-max error characteristic with respect to 
the Kahn phase response. There is no way to identically map the magnitude 
or phase response of an analog network to DSP, although you can approximate 
them arbitrarily closely if you allow the digital implementation to have an 
additional pure time delay, which will increase as the match improves. In 
practice, it is not necessary to do anything nearly this fancy when porting 
an allpass filter to DSP because the error is small since the poles and 
zeros are low in frequency compared to the Nyquist frequency of the DSP. 
Almost any common frequency transform will work.