Why does one excite whetstone bridges with DC?
If one used AC, all problems re offset, drift and such would be eliminated.


Hallo Erik,

yes, it seems to be a good way, but unfortunately problems are only shifted.
Let's design a bridge which is excited by 1kHz square wave. If we assume a signal bandwidth of 5Hz then the first consequence is clear: Fundamental of bridge excitation is 200 times higher! So, whole input circuitry is 200 times more sensitive to variations of capacitive load at input. This would not present a problem, if capacitive load were stable, but unfortunately it is not. It's temperature dependent and suffers from long term drift. I do not only think about stray capacitance, but also about capacitive load provided by capacitance of unavoidable EMI input filter sitting directly at input: In order to achieve low noise performance, resistance of RC input filter should not be much greater than resistance of legs of bridge, which can be assumed to be about 300Ohm. But this causes rather big filter capacitances which gives a considerable capacitive load.
Consequence is that corner frequency of EMI input filter must be choosen very high, about 200 times higher than with DC bridge excitation. Not at all a good idea, when dealing with so small signals!!

Up to now, we only spoke about the fundamental of AC bridge excitation. But, of course, there are also those damned harmonics present!
Now, one could think, that we are only interested in the fundamental and not the harmonics, because we will them filter out later. Right?
Unfornutaley, NO! If we do not filter them out BEFORE they run into first amplifier, then they are present and make trouble. Lots of trouble!
As we stated above, corner frequency of input filter must be choosen much higher than fundamental frequency. Otherwise unavoidable variations of capacitive load (temp drift, long term drift) would ruin signal integrity. So, we must accept that at least about 10kHz (11kHz excatly) will be present at input of first amplifier, and with unfiltered intensity!
So, as consequence, first amplifier which had only to deal with signal bandwidth of 5Hz with DC bridge excitation, now has to handle 10kHz! This makes a factor of about 2000!! Why is this a problem?
Because the amplifier is much less linear at frequencies of 10kHz than for almost DC signals, and this again by a factor of 2000. With same gain setting it will show an accuracy and drift performance which is about 2000 times worse!!
Worsening of accuracy and drift performance at higher frequencies is typical for all operational amplifiers. It's caused by 'open-loop frequency response', which shows a fall of open-loop gain by 20dB when increasing frequency by a decade. As operational amplifiers always use feedback to increase linearity, decrease of open-loop gain will worsen accuracy.

An example:
An OP27 shall be used for the first amplifier. At 5Hz open-loop gain is 120dB. When choosing a closed-loop gain of 60dB (=1000), then loop-gain is 120dB -60dB = 60dB. This is the gain reserve, which helps the operational amplifier to reduce intrinsic unlinearity and drift to acceptable values.
But if we want to make the OP27 to work at 1kHz, then we get an open-loop gain of only 80dB. In combintaion with unchanged closed-loop gain of 60dB, we would result with loop-gain (gain reserve) of only 20dB. This would be totally unapropriate!! So, the only way to get the circuit working, would be to heavily decrease gain setting of this amplifier. With OP27 only a closed-loop gain of 20dB (=10) would be possible.
So, we see, when using AC bridge excitation even the simple amplifying of signal isn't simple any more.

Unfortunately, it's not enough to be linear only at fundamental of 1kHz. Also the harmonics need a high linearity of amplifier. This makes the situation even more troublesome. As OP27 provides a open-loop gain of 'only' 60dB at 10kHz, we come to a situation, where this operational amplifier can't even be used for a gain higher then 1. And up to now we didn't speak about the even higher harmonics, which are also present, of course!!

Now some people might say: Why the hell am I interested in those harmonics, when I will filter them out later?
Because they cannot be filtered out totally! An amplifier stage showing certain unlinearity will not only produce harmonics which are higher than fundamental, but also intermodulation products, which can be in the range of fundamental!! How this?
If an unlinear amplifier has to handle more than one frequency, means a bunch of frequencies, as it is the case when applying a square wave, intermodulation products are produced, with frequencies of (fx - fy), (2 x fx - fy), etc. If we take fx = 5kHz and fy = 9kHz, e.g., then according to (2 x fx - fy) an intermodulation product of 1kHz is produced!
It's not only the existence of these intermodulation products, what is disadvantagous, but more the fact, that their content is highly unstable and unpredictable with temperature change, exact sensor signal, long term drift, etc. This introduces drift, unlinearity and offset, which was originally thought to be eliminated by the choose of AC bridge excitation!!

But even if we can handle these harmonics at input stage, we again get troubles, when we want to measure the signal amplitude after amplifying. Just rectifying is not a good idea, if noise level can be higher than signal level. Sampling by an ADC isn't a good idea either. The most precise ADCs (delta-sigma) are only capable to convert signals of much lower frequencies. Also, a high Q bandpass filter with center frequency of 1kHz would be needed, if low noise level must be achieved.
If noise can exceed signal, lock-in methode is advantegous. But here again fast switching edges are coupled into inputs (or outputs) of linear amplifiers. Again drift, unlinearity and offset is introduced and it's very very difficult to obtain a dynamic range greater than 80dB, which represents a resolution of only 13bit!

So, in most cases DC bridge excitation in combination with chopper-stabilized gain stage is enough to achieve very good results. High gains can easily be obtained, while only needing a very low number of chips.
AC bridge excitation, on the other hand, is a methode used in 'special' situations, often combined with lock-in methode. AC bridge excitation must always designed very carefully! Also, according circuits providing comparable accuracy do mostly need much more chips.

Kai