non-inverting opamp gain stage

[lowstar]

hi folks,
i´m trying to stick a non-inverting opamp gain stage at the end of my bad stone and can´t get the damn thing to work. i studied every article and thread i could find for the last 3 hours today and 4 hours yesterday to no avail. the inverting one i did 3 days ago worked just like i expected...

this is what i did:



i connected 9v to pin8, gnd to pin4, used 470K for R2 and 100K for R1. the input goes to pin 5 (i used the 2nd half of one 4558), it comes from a cap (0.1 from the output of the bad stone). R2 i put between pin 6 and 7, from the junction pin6/R2 i ran R1 to gnd. the output i took from pin 7 via a 2.2 electro (+ facing the opamp) to out.

and there is no sound coming out.

i tried connecting the R2/R1 network to vref instead of gnd (there´s a 4V rail on the bad stone), i tried connecting vref to the input with the input signal...no go. if i let R2/R1 float without connection and touch the input (non-inverting) with 4.5v, i get sound for half a second, then it fades away.

what am i doing wrong ??

i even downloaded the device magazines from the google cache (can´t acces the pages, son´t know why) and read his articles on the opamps. he draws it exactly like the pic. why doesn´t it work for me ? i looked for shorts, i measured all voltages, i tried different opamps, nada.

this is driving me nuts.

anybody ??

[lowstar]

ok, i put the resistor network r1/r2 to vref instead of ground, and hooked up the vref to the input also via a 1Meg resistor, now it works.

but could somebody explain to me why the vref has to be both on the - and + inputs ?

and the 1M resistor (let´s call it R3), what does it´s value determine ? the R2/R1 set the gain. could i use 47K/10K or even 4.7K/1K instead (the ratio stays the same), would it be better/worse for noise or characteristics ? and is the value of R3 also dependent on those of R1/R2 ?

cheers,
lowstar

[R.G.]

Here's what's going on.

An opamp will always try to force the voltage difference at the + and - inputs to be zero. When you hold the + input at some voltage, the output moves up and tries to bring the - input up. It can only do this through the voltage divider of R2 and R1 (in this case). Your resistors are set up so that the output has to move up by 470/100 = 4.7 times as much as the - input does, so the output of the opamp has to move up to 4.7 times the bias voltage to get the - input up to the right voltage. The opamp does not have enough power supply voltage to swing that high, so the output saturates against the positive rail and stays there. No audio can get through.

If you either connect the R1 lead to the reference voltage or insert a capacitor between the R1 lead and ground, now the DC voltage difference between the two inputs can be satisfied by the output only swinging up to Vref - and so it is. In this case, the DC gain is one, but the AC gain remains at one plus R2/R1 because the capacitor (or R1 connection to Vref) has no effect on the AC gain.

[lowstar]

ahhhh...thanks ! 

so the thing left puzzling me is the value of R3 and the different possible R2/R1 sizes and their effects on the audio outcome...

cheers,
lowstar

[boogietone]

Here's what's going on.

An opamp will always try to force the voltage difference at the + and - inputs to be zero. When you hold the + input at some voltage, the output moves up and tries to bring the - input up. It can only do this through the voltage divider of R2 and R1 (in this case). Your resistors are set up so that the output has to move up by 470/100 = 4.7 times as much as the - input does, so the output of the opamp has to move up to 4.7 times the bias voltage to get the - input up to the right voltage. The opamp does not have enough power supply voltage to swing that high, so the output saturates against the positive rail and stays there. No audio can get through.

If you either connect the R1 lead to the reference voltage or insert a capacitor between the R1 lead and ground, now the DC voltage difference between the two inputs can be satisfied by the output only swinging up to Vref - and so it is. In this case, the DC gain is one, but the AC gain remains at one plus R2/R1 because the capacitor (or R1 connection to Vref) has no effect on the AC gain.

Apologies for pulling up an old thread, but the last paragraph here has me curious. For a non-inverting op amp layout, there seems to be two possibilities for connecting the inverting input to "ground."

1) Connect the inverting input to the virtual ground (V=Vref) through resistor R1.
2) Connect the inverting input to actual ground (V=0) through resistor R1 and a capacitor.

I can see that both of these arrangements are functionally equivalent - allowing the DC voltage at the inverting input to "float" to whatever the non-inverting input is. In the first case, it is effectively fixed to Vref in the same manner as the non-inverting input. In the second case, it is decoupled from the actual ground DC voltage by the cap.

Is my analysis correct?

Are these circuits equivalent?

Is one "better" than the other?

Is the DC gain of the op amp equal to 1 in both cases and how to determine the gain roll off from 1+R2/R1 at high frequency to 1 at DC?

I am guessing that a third option of putting a cap between R1 and Vref would be redundant?

Thanks.

[PRR]

> Are these circuits equivalent?

Almost. Not quite.

NFB tied to Vref: there is DC gain for any difference between "Vref at the + in pin" and "Vref at the - in pin". Ideally there is no difference, not enough to care about.

If the + in pin draws significant bias current through a very large resistor, it won't be quite Vref. Say a wheezy 5532, 1uA bias current. Say 1Meg bias resistor for Vref. 1uA*1Meg is 1V. Say the NFB network inpedance is much lower, far less voltage drop. Say the NFB imposes gain of 10. Then the 1V offset at input is 10V offset at output. In 9V systems, this causes a jam-up.

If instead the NFB network is AC-grounded through a cap, we have gain of 10 for AC/Audio but gain of 1 for DC. Say Vref is 4.5V, and that 5532 abd 1Meg resistor pulls to 3.5V. The output will be at 3.5V. We lose a little headroom but it isn't jammed-up and signal will pass.

low-Z NFB loop connected to Vref injects signal current to Vref. If Vref is not zero impedance (never is), and other stages share the Vref, signal has a sneak-path. Possible oscillation. NFB through cap to ground does the same, except ground is usually much lower Z than practical Vrefs.