Circuit questions - Inverted vs non-inverted (again)

[alvinlim]

Hi,

I have been trawling through the postings, and I am pretty clear on the way the non-inverted circuit works - a lot of information can be found from the datasheets (LM3875). I have some questions on the inverted circuit (Thorsten) and I do hope that the members of the forum can shed some light for me.

1. Why do we need such high R values for the feedback resistor - Can't I just use 22K instead of 220K for the feedback resistor, and 2K instead of 22K to connect the inverting input to ground? I thought that higher resistor values will add noise to the circuit. Correspondingly, the resistor connecting the non-inverting input to ground should be reduced to (2K * 22K / (2K + 22K). Am I on the wrong track here?
2. What's the purpose of the 10K resistor connected to the inverting input? I am guessing that it represents the input impedance to the source.
3. If I am connecting the GC to a pre-amp, can I simply just drop the attenuator? Do I need to add a resistor before the 4.7yf cap to the ground? If we are assuming that the inverting input is the 'ground', will that make any difference? Can I use 1M R here?
4. What's the purpose of the 1yf cap from pin 1 to Pin 4? and also the 0.22 R output resistor - I am guessing that both are to prevent oscillation.

Thanks!

[ukram]

1. High R for feedback resistor is needed to get "normal" input impedance of 22k. Also the 4.7uF capacitor should be replaced with 47uF if using 2k, which means electrolytics. I measured noise with both resistor values, and got the same -82 dB noise level, compared to better than -90 dB with non-inverted design (reason unknown).

2. Resistor in inverting input is used to minimize offset voltage due to input current (and "ground noise"). I tested with 4.7k and 0k and got no difference in noise or offset.

[alvinlim]

this is where i get really confused. I see the need for the 22K R to set the impedance. So, why can't I change the 220K Rf to 20K and the 10K R to 1K? I still get the same gain!

yes, the amount of feedback will increase, but isn't this consistent with Kimura's philosophy?

[Nicke]

Originally posted by alvinlim
this is where i get really confused. I see the need for the 22K R to set the impedance. So, why can't I change the 220K Rf to 20K and the 10K R to 1K? I still get the same gain!

yes, the amount of feedback will increase, but isn't this consistent with Kimura's philosophy?

It should work,BUT the input impedance will be 1 kohm...

[jan.didden]

Originally posted by alvinlim
this is where i get really confused. I see the need for the 22K R to set the impedance. So, why can't I change the 220K Rf to 20K and the 10K R to 1K? I still get the same gain!

yes, the amount of feedback will increase, but isn't this consistent with Kimura's philosophy?

You're not confused, you're right on the dot.
Indeed the 220k is bad for noise (and instability, possibly, depending on layout). One option is first to divide the output signal by say 10k and 470 Ohms, which will give you the wanted gain, and then take that point to the input via 10K, plus use the 10K input resistor. Note that the "noise gain" in this case is around 40, so the feedback is less effective, but I understand this is what the chip needs to be stable?

Jan Didden