Back to basics

[PBE6]

A few dumb questions:

1. I've noticed that with a lot of my circuits (but not all) whenever I have an opamp buffer with a large Vref resistor I get a significant deviation from the expected Vcc/2 voltage divider value. The DC path is blocked from going back out by the input cap and from going anywhere else by the opamp itself. I know some opamps can have a tiny bias current, but the TL072 is listed as having a maximum of 7 nA which would only contribute a voltage drop of 0.07 V with a 10M bias resistor, nowhere near the 2 V drop I see occasionally. What's going on there?

2. I recently made a Sallen-Key high pass filter that didn't work. I suspect something wrong with the veroboard as all the connections appear correct, but it got me thinking - the caps on the input side should block any DC bias from getting to the inputs, so the DC bias should be applied through the input resistor, correct? It seems straight forward, but I'm second guessing everything at present.

2b. When putting a capacitor to ground as part of a low pass filter after an opamp gain stage, does it matter if the capacitor is connected to Vref or ground, whether or not coupling capacitors are used after the opamp? I think "no" since there should be no DC flowing, but again, I'm second guessing everything again.

[WangoFett]

Hey PBE6, here are a few potentially dumb answers.

#1 drives me mad every single time.  After trying random things and scratching my head for half an hour I remember that the impedance on my DMM is dragging the voltage down.  So I try to remember to measure the output of the op-amp instead.  Might not be your issue.  Try also measuring from 9V (or whatever your supply voltage is) to the test point.

#2 Sallen-Key HPFs seem to have a resistor from + input to ground.  Using a virtual ground instead could add a DC bias.

#2b For DC and in a DIY stompbox world it doesn't matter.  For higher frequencies it depends on the impedance of your Vref.  If your Vref has caps to ground then effectively they are in series with the cap in question so make sure they are large enough to not mess with your filters.  If your Vref is the output of an op-amp then it will look like a very small resistor to a constant voltage, so very much like a cap to ground (depends on what op-amp you use of course).

[PRR]

> buffer with a large Vref resistor I get a significant deviation from the expected Vcc/2

As Dave says: figure your meter impedance! VTVMs and DMMs are often ~~10Megs, which will be a 5% shift from the un-metered voltage of two 1Meg resistors. Passive needle-meters can be as low as 5K, which will be a HUGE shift. Some DMMs run 2Meg, 1Meg, maybe less. (Specialized power-worker meters tend to low impedances to discount stray leakage on circuits which should be carrying Real Power.)

Measure the OUTput of the opamp. This should be the "same" voltage as Vref (+/-5mV), but the opamp tolerates loading a LOT better than a hi-Z Vref node. If the opamp output is happy, it is perhaps not worth poking Vref directly.

[Transmogrifox]

Another thing you could do is
1) Measure Vref
2) Measure between Vref and the op amp input.  If you know the impedance of your meter and you know the parallel bias resistor (Rbias) you can calculate and confirm the bias current.

If you are uncertain of your meter's input impedance, then put your meter in series with a 10-Meg resistor and apply a known voltage.  Simple voltage-divider network created with your meter will give you enough information to calculate the unknown variable (meter input R).

If the current through the bias resistor ends up being more than you expect, check the datasheet for the capacitors you are using.  For example, parallel leakage currents on electrolytic caps will start to make a significant difference when you're talking about 10M.

After that I would be guessing high frequency oscillation is in play.  Of course, your meter can cause the oscillation to appear when normally it doesn't. 

So then perhaps I have worked myself back around to PRR's more simple suggestion:

If the opamp output is happy, it is perhaps not worth poking Vref directly.

[PBE6]

Great!! Thanks everyone, thought I was going crazy there for a minute.