Redesigning QUAD-405 DC bias loops

[MarcelvdG]

Hi Joost,

You get a first-order (6 dB/octave) response up to f = -z_io2/(2*pi) = -p2/(2*pi) = 1/(2*pi*(R4//R5)*C2). From there onwards it becomes second order (12 dB/octave). With the values in the first post of this thread,

C1 = 680 nF
R3 = 22.1 kohm
R6 = 221 kohm
C4 = 68 nF
R4 = 68 kohm
R5 = 56 kohm
C2 = 2.2 uF,

you get

1/(2*pi*(R4//R5)*C2) ~= 2.3557 Hz

With the original values, 4.7 kohm, 22 kohm and 100 uF,

1/(2*pi*(R4//R5)*C2) ~= 0.41097 Hz

or with 10 kohm, 10 kohm and 100 uF,

1/(2*pi*(R4//R5)*C2) ~= 0.31831 Hz

So yes, there is a range of frequencies where you lose the second-order slope, but you can dimension things such that this only happens far in the subsonic region. If you want to extend the second-order slope down to nearly 0 Hz, you can try one of the suggestions from section 7 of the document.

Best regards,
Marcel

 

[MarcelvdG]

The example in the document, with alpha = 0.55, is meant for people who don't care about the subsonic filtering. The Bode asymptote of the transfer then has a first-order slope up to 0.55 times the cut-off frequency, then changes to second-order and becomes flat at the cut-off frequency. The real response is a smooth line that approaches the Bode asymptote when it has the same slope over a large frequency range, but since the Bode asymptote only has a second-order slope over less than an octave, it never approaches a second-order slope well.

 

[joost456]

Hello Marcel,
Thanks for your explanation, I will look at your suggestions about keeping the second order slope. For me subsonic filtering is important because of the vinyl comeback/recovery in combination with bass ports in speakers.
Joost