LM1875 with designed-in excursion protection filter

[Dr.EM]

I was looking at using an LM1875 to power a small subwoofer (PC setup) and with the sub being ported, decided it'd be a good idea to use a protective high-pass filter.

It would be desirable to not need any additional signal level electronics to achieve this, but the input coupling cap alone can only create a 1st order/6dB filter. I notice also on the application schematic however a 1kOhm resistor and 22uF capacitor to ground:



and having built a circuit with it before where the suggested values for these were 1uF and 1.5kOhm, I became quite aware that it was acting as a HPF!

In this datasheet application, the values I assume to give -3dB at:

1/(2pi*1k*22uF) = 7.2343Hz

Which seems reasonable for good response to 20Hz. The input filter then gives a 3.288Hz break frequency. Again, very low giving 'flat' response to 20Hz.

I then calculated values to give a 28Hz frequency for both parts:



So what I wonder is, what kind of filter has this created? Topology wise, is it Butterworth, or LR, or other? Importantly, will 28Hz now be the -3dB point or the -6dB point, and I can only assume a 2nd order/12dB rolloff? Should both RC networks have equal break frequencies anyhow, or is that wrong? Any input appreciated

[pacificblue]

It is a 12 dB/oct filter and 28 Hz is the -6 dB point. The curve resembles that of a critically damped filter with Q=0,5.

[Dr.EM]

That's great, thanks for your response!

[macboy]

The filter in the feedback loop is effectively a shelving filter. The gain is reduced to 1 at very low frequencies (f<<fc) or DC. The gain at high frequencies (f>>fc) is 20 (20k/1k). The filter on the input is a low pass filter; the gain approaches 0 at very low frequencies or DC. So the filter will be 12 dB/oct until the feedback filter shelves, when it will become 6 dB/oct for lower frequencies.

[AndrewT]

The NFB cap will develop a small AC signal across it. This AC voltage results in an excess of avoidable LF distortion.

A Q=0.5 filter has a quite slow roll off compared to the standard Butterworth Q=sqrt(2) filter.

[Dr.EM]

Ah yes, I fully understand your point macboy, and that could be detrimental in a 'subsonic' filter application.

I don't understand your point about distortion so well AndrewT, though I see that a Q=0.5 filter is quite a shallow roll-off, which is again not ideal.

Sounds like it'd be best to employ another op-amp as a butterworth filter then I guess choose an input capacitor as I have shown to achieve a total 18dB roll-off?

[AndrewT]

Originally Posted by Dr.EM

I don't understand your point about distortion so well AndrewT,

read D. Self and others.

Signal voltage across the DC blocking cap in the NFB loop introduces one of Self's 8 distortions. It is easily avoided by ensuring that no AC voltage develops across the DC blocker.
That is achieved by ensuring the NFB LF bandwidth is significantly wider than the passive filter at the input to the amplifier where the second DC blocker is located.

If you want to add extra steepness to the high pass filter then keep the passive input filters, both of which give close to Butterworth Q=sqrt(2) and add an extra active filter that has a Q=1 to maintain the overall Q at the Butterworth value.
Both the Unity gain and Equal component value Sallen and Key filters can do this. The MFB filter can also do this, but it inverts the phase of the signal, i.e. it's an inverting stage and filter.

[Dr.EM]

I must take advantage of my Uni's library when I go back! Thanks for the advice all

[AndrewT]

D.Self has released much of his tech discussion to the web.
Read about the 8 distortions here.
http://douglas-self.com/ampins/dipa/dipa.htm

[mystunner]

it's hard to find this IC now!