Current Clamping Improvements
Another fundamental improvement that we have been able to incorporate into the Optopatch is the ability to perform TRUE current clamping. We had been too busy concentrating on the other innovations to give current clamping much thought until Dr. David Ogden (MRC Mill Hill, London) kindly pointed out to us in September 1995 that the standard patch clamp current clamping circuit has performance limitations that are often detectable in practice. Dr. Ian Forsythe (Leicester, UK) independently mentioned the same problem to us a few weeks later, which further encouraged us to investigate it, and the following paragraphs explain both the problem and its solution. It turns out that the optical feedback makes implementation of the solution even cleaner than it would be otherwise, but in order to emphasise the generality of the solution, the discussion assumes that a high-value resistor is used to pass the current.
Figure 5 - Ideal Current Clamp.
First of all, it is easy to design a true current-clamping circuit, and Fig. 5 shows all that is needed. The headstage amplifier is connected as a voltage follower, and the current is provided by a voltage source connected to a resistor. The load resistance into which the current is passed is the resistance of the electrode in series with that of the cell, and for the purposes of this discussion we shall assume that the electrode resistance is sufficiently low compared with the cell resistance that it can be ignored (in practice that may well not be true, but the problem can be dealt with in other ways, as described in the section on resistance and capacitance compensation, so for this discussion we can indeed ignore it). To approximate to a true current source, the value of the current-passing resistor should be high compared with the load resistor, so that the voltage at the amplifier input remains low compared with the current-passing drive voltage, since the current flow is actually given by the difference between these two voltages. With the optical current-passing method, this condition is met automatically, but when a resistor is used, the equivalent effect can be obtained by adding the input voltage to the current-passing command voltage, as Fig. 5 shows. So why don't other patch clamps work this way?
Figure 6 - Conventional current clamp
The answer is one of convenience, since it is probably fair to say that current-clamping was added to the standard patch clamp design as something of an afterthought. The problem here is that the electrode in Fig. 5 is connected to the non inverting input of the headstage amplifier, whereas for voltage clamping it needs to be connected to the inverting input as shown in Figs. 2-4. To prevent the need to reconfigure the input amplifier when switching between voltage clamp and current clamp, patch clamp amplifiers conventionally perform current clamping with the circuit shown in Fig. 6 (we show here a resistive headstage for generality, but the other headstage types are equivalent in this respect). It consists of the basic voltage clamp circuit of Fig. 2, but this time including the necessary frequency compensation stage as well, and with the output feeding back to the voltage command input via an integrator. A feedback capacitance, Cf, is shown in parallel with the feedback resistance, Rf, not just because some capacitance here is unavoidable, but because it is usually deliberately increased to give a well-defined high-frequency rolloff, that can then be compensated more precisely by the subsequent high-frequency boost stage as shown (see Sigworth, 1995, for a full theoretical treatment). As far as the frequency response is concerned, we can therefore notionally remove Cf and the boost circuit, and regard the integrator as being connected directly to the headstage. Just as in voltage clamp mode, the headstage output - after subtraction of the command voltage - is the voltage across Rf, which is a direct measure of the current through it. Any difference between this voltage (current) and the non inverting input of the integrator, to which the current-passing command voltage is applied, is fed back as an error signal to the command voltage input of the headstage. The effect of this circuit is therefore always to maintain the electrode at a voltage such that the required current is passed through the feedback resistor, and hence into the electrode.
An integrator is used in order to introduce significant low-pass filtering into the overall feedback loop. The filtering is necessary because the bandwidth of the signal from the headstage, even after equalisation by the boost circuit, is likely to be no more than 100KHz at best, and in practice it may be considerably less. To ensure stability of the overall feedback loop, its bandwidth needs to be constrained to a still lower frequency, of perhaps only a few KHz, whereas general amplifier circuits can easily have loop bandwidths in the MHz range. In particular, this applies to the input stage in Fig. 5, where the feedback is a direct connection, since the amplifier is just a voltage follower in this circuit. There is only the bandwidth of the input amplifier itself to consider in this case, and to push that well into the MHz range is a trivial matter.
Unfortunately, the low bandwidth of the Fig. 6 circuit causes a problem here, for the following reason. At frequencies beyond the loop bandwidth, there is essentially no feedback, so the amplifier input impedance falls to the value given by the parallel impedance of Rf and Cf (since the output of the input amplifier stage then acts as a signal ground). The effect of feedback around the circuit is to raise this impedance proportionately with the amount of feedback applied, and very substantial amounts of feedback would be required over the entire operating frequency range in order for the circuit to approach the performance of that in Fig. 5. As the frequency falls, the gain of the integrator rises, but the gain of the boost circuit falls, so the loop gain - and hence the amount of feedback - actually does not begin to increase until the corner frequency of the boost circuit is reached, i.e. the frequency at which the impedance of Cf equals that of Rf. What this actually means is that lowpass corner frequency of the integrator must be less than the combined bandwidth of the headstage and the high-frequency boost circuit if there is to be any feedback at all at frequencies above the corner frequency of the boost circuit!
There is also yet another problem to consider, which concerns the headstage's input capacitance, and in practice this problem may also be serious. It may not be immediately obvious that there is a problem, since the input capacitance neutralisation circuit (not shown in Fig. 6) continues to compensate correctly for capacitances between the input and ground, so the system's performance remains the same in this respect as for a step in the command voltage under voltage clamp. However, there is another component of the input capacitance, which appears between the inverting and non inverting inputs of the input amplifier. This capacitance does not load the input to a significant extent in voltage clamp mode, because local feedback, mainly provided by Cf at high frequencies, acts quickly to reduce the differential input voltage towards zero in response to a command voltage step, so it is effectively completely compensated. However, in current clamp, a step change in the voltage at the electrode will initially appear entirely as a differential input voltage, which will be reduced much more slowly by the current feedback loop, with a time constant determined by the integrator. Under these circumstances, the current to charge the differential input capacitance (which could easily be on the order of 10pF in a typical circuit), will initially be drawn from the electrode. The fact that the action of the integrator then discharges the capacitance again actually makes matters even worse, because this means that the electrode effectively has all the charge returned to it later on!
The combination of these effects is most certainly noticeable in practice. Consider for example, the case in which passing a steady current in whole-cell mode causes the cell to fire action potentials. Under these conditions both the retention of high input impedance up to a relatively high frequency and the preservation of a good transient performance are particularly important, and what can happen in practice if these conditions are not met is, in the words to us of one current-clamper, "...as if the patch clamp is trying to voltage clamp the action potentials."
Obviously the current-passing performance of a particular patch clamp amplifier will depend on the actual component values, which will in any case also depend on the current range selected, and without having our own version of Fig. 6 to analyse, it does not seem appropriate to quantify the problems any further here. The only general point we can note is that when a headstage is operating in the capacitive (integrating) mode, where Cf may be higher than for the resistive mode, it might require boost compensation up to a higher frequency for comparable performance, but on the other hand the frequency response is likely to be easier to correct in this mode, so in practice there may not be much if any difference. Since we are using another current-passing method altogether in the Optopatch, we prefer to be neutral on this subject!
Figure 7 - True Current Clamp
A much enhanced general implementation of current clamping is shown in Fig. 7. Swapping the connections to the inverting and non inverting inputs of the headstage amplifier when switching between modes would clearly be inconvenient, but we can obtain the same effect by inverting its output instead. Feeding the inverted output back to the command input of the headstage as shown effectively reverses the polarity of the inputs, making the headstage amplifier into a voltage follower. Although the feedback has to be applied via another amplifier stage, the signals here are all of low impedance, so it is easy to ensure high loop bandwidth, resulting in a performance that is to all intents and purposes identical to a conventional voltage follower. We can now pass current directly though what was Rf, driven by an amplifier which corrects for the input voltage in the same way as in Fig. 5.
Two possible practical problems intrude. First, the capacitor Cf no longer serves any useful purpose, so ideally it should be removed, but in practice this may involve additional switching in the headstage. However, in practice it could stay connected across Rf, and is effects could be compensated for by a low-pass filter (of characteristics that are complementary to those of the high-frequency boost circuit required for equalisation in voltage clamp mode) in the current clamp command input. Such filters can be made to have a much more ideal frequency response, giving much more precise correction, and in any case the correction is outside the main feedback loop, so does not have any influence on its performance. However, it would still be better for Cf to be relatively small, since it still lowers the uncorrected impedance of the current source. The second possible problem is that in the conventional headstage design of Fig. 6, the subtraction of the command voltage from the current output may take place in the headstage itself, which is not appropriate for the current clamp circuit of Fig. 7. The fact that we were made aware of the current-clamping problem during the design stage of the Optopatch made it much easier for us to cure it than might have been the case otherwise.
The headstage design in the Optopatch, as shown in Fig. 4, effectively consists of just the input amplifier stage, in which the connection between its output and the current-passing circuitry (equivalent to Rf and Cf in Fig. 3) is made indirectly by interconnections that go back to the main electronics, plus part of the optical drive system. This allows the circuit to be reconfigured between the standard voltage clamp and true current clamp modes without the need to perform any switching within the headstage itself. Although the configuration of the optical headstage made it particularly easy to implement true current clamping, equivalent arrangements could nevertheless be made for the other types of headstage as well.
Since this section was first written, Magistretti et al (1996) published an article in TINS (see bibliography), which also discussed this problem, and it provides useful further reading material on this subject.
