why Filter used for audio

[raj_shekar]

negative feedback on lm386

Ear hearing frequncies range 20-20k. but we use filter to attenuate the frequencies after the 20k also. why ?

 

[rtarbell]

One reason I can think of:

Although we can't hear frequencies above 20kHz, the speaker drivers still respond to these frequencies. We filter out the higher frequencies so that we are not wasting power by driving the speakers with frequencies that we can not hear.

 

[raj_shekar]

In sigma delta we actually use very high order filters to attenuate frequencies above Fs/2. Why it is so important

 

[FvM]

Actually most audio SD DACs don't use high order output filters, e. g. TI recommends 2nd order with about 200k cutoff. It's an advantage of SD DAC not to need high order analog filters.

 

[Audioguru]

Frequencies above 20kHz are filtered out to avoid interference from local AM radio stations.

Remember audio tape decks? They had a bias oscillator that recorded about 80kHz. Old amplifiers filtered it out.

The amplifiers have high distortion at very high frequencies. The high harmonics beat with each other and produce audible noises unless they are filtered out.

An audio expert says, "A very wideband audio amplifier is an accident waiting to happen".

 

[chegu.balaji]

An audio expert says, "A very wideband audio amplifier is an accident waiting to happen".

nice point.

 

[jnuhope]

An audio expert says, "A very wideband audio amplifier is an accident waiting to happen".

My question is how to determine the amplifier's UGBW used in the filter? For example, if the active RC filter's cutoff frequency is 800K for audio application, how much UGBW for the amplifier should be designed and why? I've seen rule of thumb ranging from 5x to 20x and am confused whose thumb should I follow.

Thx in advance!

 

[FvM]

I've seen rule of thumb ranging from 5x to 20x and am confused whose thumb should I follow.

The suggestion are rather close together, I think. No reason to get confused.

Actually, the filter characteristic is affected by limited OP GBW in any case, the question is to which extent. E. g. cutoff frequency is shifted, Q is increased. To determine what's applicable for you, a simulation with a real OP is the best option. Some filter design tools also allow to use OP GBW in calculation.

 

[Audioguru]

An LM386 is a little power amplifier IC.
When its gain is 20 then its -3dB bandwidth is 300kHz.
Its distortion is 0.2% at 1khz but rises above 2kHz.

It has a filter built-in that reduces its gain at high frequencies so that its phase-shift doesn't cause oscillation when negative feedback is used.Then its negative feedback at high frequencies is reduced which causes the distortion to increase at high frequencies.

 

[jnuhope]

How to do the stability check for active low pass filter? If one breaks the loop at the input of the opamp, then the loop response simulation gives me a zero as opposed to a pole. Any suggestions? Thx!

Added after 8 minutes:

An LM386 is a little power amplifier IC.
When its gain is 20 then its -3dB bandwidth is 300kHz.
Its distortion is 0.2% at 1khz but rises above 2kHz.

If the gain is 20 at 300Khz, then its dc gain is 20*300*1000=6e6. Am I missing something here:?:

 

[FvM]

Am I missing something here?

Yes, the fact that LM386 has a fixed gain by an internal feedback, thus we are talking of closed loop rather than open loop gain here. But the LM386 topic is only loosely related to the orignal filter & GBW question, I think.

 

[Audioguru]

If the gain is 20 at 300Khz, then its dc gain is 20*300*1000=6e6. Am I missing something here:?:

An LM386 has built-in negative feedback for a gain of 20. It has a built-in lowpass filter so at very high frequencies where its phase shift is 180 degrees then it doesn't have any gain to oscillate.
Its cutoff frequency is 300kHz where its gain is 0.707 x 20= 14.14.

The negative feedback is the same for DC as for AC so the gain at DC is also 20.

Adding a capacitor increases the mid frequency gain to 200 (46dB) then the bandwidth is reduced to about 50kHz. The DC gain remains at 20.

 

[jnuhope]

In a delta-sigma dac for audio application, the delta-sigma modulator followed by reconstruction filter (either implented in switch capacitor or current type) and the final stage is another RC filter to filter out high frequency as well as perform diff-to-single ended output.

What's the cut off frequency for recontruction filter and for RC filter? Is recontruction filter's cut off frequency at 20K and RC filter at much higher frequency (say 1Meg) to filter out AM radio interference?