So we’re talking voltage clamp correct? So if the filter on your amplifier is set to 5 kHz, then that means any current you record faster than 5 kHz is basically gone (and in reality, you’re filtering currents down to about 1.5 kHz, see my post on filtering). But lets look at what you can actually record. If you’re saying that RsCm = 60 us, then that sets up a 1-pole filter of 2.7 kHz for both the transmembrane voltage you are applying, and transmembrane current you are trying to record. So yes, your 100 kHz sampling rate is way over kill. You don’t want to actually use the Nyquist theory on your “practical” data rate (e.g. sampling at 2.7 kHz * 2 = 5.4 kHz, because you will still be sampling noise at 5 kHz or so. So I would use a bare minimum of 10 kHz… and probably use 20/25 kHz just to get nice waveforms). But yes, you would absolutely be kidding yourself if you thought that you transmembrane currents were accurate up to 100 kHz. CpRs is another filter, but in the scheme of voltage clamp, it is usually insignificant, and Cp is more likely to cause problems due to leak subtraction, or noise (though there are edge cases where Cp*Rs is a significant filter).

> Does this filtering just apply to the response to the initial voltage jump, or does it continue as the membrane voltage comes to steady state?

The “filtering” (as in, altering a signal based on it’s frequency) only applies to the initial jump. However, as I see you know from what you say later on, there is a voltage error caused by I-clamp * Rs. The combination of the filtering and the voltage error is shown in the transfer functions in figure 2.

> Also, I thought that the reason the measured current was less than it should be in figure 3 was because of the voltage error created by the Rs. If the current is 100 pA, and Rs is 40 MOhms, then the voltage error at the peak of the current is 0.1*40 = 4 mV.

I don’t think that would be the best way to think about it. As I say in parenthesis, the simulated synaptic event is current based, so there is no driving force. Think of it like another pipette in current clamp mode, forcing a 100 pA-peak synaptic event into the cell. It doesn’t not matter what potential the cell gets to: it will always inject 100 pA. I always like to think about voltage clamp in terms of a voltage divider: one resister is made up of Rs, and the other is made up of Rm and Cm. When we apply a voltage, that voltage falls across Rs (voltage error) and Rm || Cm. When that voltage is changing rapidly, Rm || Cm appears as a very low impedance, and hence most of the applied voltage lands over Rs. But more importantly, as Rm || Cm appears as a low impedance, most current flow will over Rm || Cm, and hence you will record less current.

In reality, there are lots of ways to think about this. Some ways work better for some people than others, so have a think about it from lots of different angles (i.e. think about it from the point of view of Vcmd… think about it from the point of view of Vm and think about it if we injected current directly into the cell, i.e. at the junction of Rs and Rm || Cm.

But Let me know if you’ve got any more questions or if I misunderstood you.

]]>It turns out I did not know the answer, so I’m glad I asked. I’m hoping that now you can help me clear up some confusion and connect the dots.

Does this filtering just apply to the response to the initial voltage jump, or does it continue as the membrane voltage comes to steady state? For instance, I have a 4-pole bessel filter with a cutoff frequency of 5 kHz, but if my time constant is 60 us then the cutoff frequency of the RsCm low pass filter will be 1/(2*pi*60) = 2.65 kHz. This means I wouldn’t see fast events that I thought I should be able to see based on my sampling rate (100 kHz) and my filter. Am I thinking about this correctly, or is this filtering actually controlled by the CpRs cutoff frequency, as in current clamp?

Also, I thought that the reason the measured current was less than it should be in figure 3 was because of the voltage error created by the Rs. If the current is 100 pA, and Rs is 40 MOhms, then the voltage error at the peak of the current is 0.1*40 = 4 mV. I know when I’m evoking current in sodium channels, the holding voltage error will skew my G/V curves, but maybe I’m misunderstanding this graph. You say in parentheses “the synaptic event is current based, rather than conductance, so it is the same no matter how bad the voltage clamp is”, but if you have a voltage error in your command potential, then the driving force on the ion would be different. Maybe that’s not enough to account for the difference, though.

Thanks for your help and all your patient and thorough explanations.

]]>> Does this only filter voltage, or is current filtered as well?

Yes, both voltage and current are filtered, both the exact same time constant roughly equal to Cm * Rs. Look at figure 3 and 4 (i.e. the current and voltage in response to a synaptic event). You see that some voltage is unclamped, especially the fast components. I really should have had a figure showing the voltage and current in response to a typical square step. But if you look in the Axon Guide, look for figure 3-14. It shows the important stuff.

]]>Thanks so much for your thorough explanations about this topic. They have helped me understand my amplifier on a much deeper level. I have a question about the low pass filter created by the series resistance and the membrane capacitance, specifically when working in voltage clamp mode. I believe I know the answer, but I would like to be certain.

Does this only filter voltage, or is current filtered as well?

Thanks for your help,

Dan

Sorry for the late response. No, that is not correct. read_temp.py runs all the emails and checking scripts. If you want to just check the temperature, run

cd examples

sudo python simpletest.py

Press cntrl-c to quit, and then a restart is best (I’m not sure what would happen if both simpletest.py and read_temp.py tried to access the termometer at the same time… it may crash read_temp.py).

]]>Na Currents are tricky because they’re a) fast, b) often large, and c) create a positive feedback loop on your voltage clamp error.

a) As I’ve said, in voltage clamp, Rs (together with membrane capacitance) sets up a low pass filter for your command voltage. Sodium currents can have rise times as quick as 100 microseconds. Which means 10 kHz. Even with a trivial Cm of 10 pF and a tiny Rs of 2 MOhm, the cut off frequency of your filter is 8 kHz. So you’re really only just clamping fast enough to capture the rise time of your current.

b) Sodium currents can be large, and your steady state voltage error is equal to I_clamp * Rs… So if you have a 1 nA current, and a 2 MOhm Rs, you have a 2 mV voltage error (at steady state, which you probably wont reach, but lets ignore that)… But if you have a 5MOhm Rs and a 5 nA current, now you have a 25 mV voltage error. That is bad.

c) Sodium current depolarizes you cell, which activates more sodium channels, leading to a bigger current, which worsens your voltage clamp, which leads to more current etc etc etc… If it was a potassium current, bad voltage clamp dampens excitability. It’s still not ideal but you don’t get a run-away loss of voltage clamp. Thankfully, it is very obvious (once you know what to look for) when you see recordings like this. Have a look at this paper and compare figure 7A to 7B. You see how in the activation curve in 7A, the sodium current instantly jumps to maximum size, then slowly gets smaller? This is run-away loss of voltage clamp, also called an action current. Proper voltage clamp should produce a steady activation of sodium currents with around 20 mV between when the current starts to be active and when the current reaches maximum. You can find quite a few published papers where there is a 5 mV between the current being activated and the current reaching max. You know these people don’t have voltage clamp.

]]>In response to “Also, so long as you’re not trying to voltage clamp sodium currents, for most people, it’s more important to have STABLE Rs, than it has to have LOW Rs.” Why is it more important to have a low Rs with sodium current recordings? Is it because the Na current can be very large?

Best wishes,

Molly