Hi,

Yes, is a very good option the one you say. Just take in account that you may want to downsample your LFP  to work easier with the signal, so you don't need 32 khz for fPSPs, just 2000 hz - maybe 1600 hz is better in your case- , unless you want to look at multiunit signal in the LFP. In that case is worthy you work with high sampling rates, such as 32 khz.

Hope this helps

Hi Marcel,

Thanks for your answer. You mean "work easier with the signal", only because the data would be very large with 32 kHz sampling rates? And also, what do you think about the amplification? For LFP, people usually use lower amplification. It is OK to use higher amplification for both LFP and spike recording, right? I think I haven't understood the amplication well.   And with LFP, I am more interested in the oscillation.

Yes, when you acquire LFP with a high sampling rate you usually downsample your signaI to work more easily with a reduced number of data samples. That means to use 1 sample each of 10, for example.

I would say is better to use the right amplification for your LFP, given that you are interested in oscillatory activity. Then you have to check which is the correct amplification you need to see the real amplitude of your LFP signal in your screen. I mean there's a unique amplification that gives the real signal, and this you have to find. Normally, for oscillations, if you have your units in you recording software expressed in uV (microV), then you have to amplify 1000X.

Also, 1000X is a good amplification for see nice spikes.

Another thing is that you will need to split the signals in LFP (0.1-500hz) and Single unit (600-6000hz) signals by filtering.

Hi Marcel,

I need to make sure of it. I am a little confused with " I mean there's a unique amplification that gives the real signal, and this you have to find".  Why the amplification is unique and how to find it?

And the amplitude of the spikes I recorded is between 50 - 250 uV. So you mean amplification 1000* are both good for spike recording and LFP? When I recorded in anesthetized rats with tetrodes, I tried amplification 1000* and 500*, in both situation, I could see good spikes.

Also, if I understand correctly, I need two amplifiers with diferent bandpass filter and sampling rates, right?

Thanks!

Sorry for the criptic answer. I will try to explain it more carefully.

Imagine that you record with amplification 1000x and you get a signal of 500 uV of amplitude. Imagine then that  you switch to amplification 5000x,  you will have a signal of 2500 uV of amplitude. Which is the real measure of your signal? This is what you have to find, because there will be only one configuration that will show the real amplitude of your signal. This is important for your LFP signal, because all power spectrum measures will be related to the amplitude of the signal. This is not the case for spikes, where you normally check the waveform shape and duration or the time position with respect to other waveforms (phase), but not absolute amplitude. 

You can use a multichannel  amplifier to get all your signals bandpassed 0,1 - 8000 hz. Then you split offline the signal in LFP (lowpass 0.1-500hz) and spikes ( bandpass 600-6000hz). If you can sort spikes with that amplitude of 50 uV then you don't need to amplify more. You don't need to record with different samplng rates, you just downsample your LFP signal to work easier with analysis. 

Last answer was pretty clear. I just wanna add a little comment about downsampling. The process is called decimation, which includes low-pass filtering and a downsamplig. This is critical to avoid aliasing. Best regards.

Thanks, Marcel and Rodrigo! It is really good to have more ideas from you. The downsampling is important for me.

As has been said, yes you can, though the pyramidal layer might not be the best place to record LFPs. You can do all the processing off-line. Storage space comes very cheap these days so there is no need to throw away data while recording.

Remember to measure the frequency and phase responses of your equipment for the full 0.1 Hz to 8 kHz bandwidth. When I worked in the Mosers' lab we were unable to record LFP with tetrodes and had to use a separate electrode for LFP. I later discovered there was an interaction between the impedance in the tetrode and the input resistance of the AC-coupled headstage which resulted in a 12 Hz high-pass filtering of the signal. This suppressed the 8 Hz hippocampal theta from the tetrodes in most cases. As the equipment we used to verify the recording apparatus did not work below 20 Hz I did not see the problem while I worked in their lab. When working with a similar setup in Johan F. Storm's lab I had the offending resistors replaced and got excellent LFP from the tetrodes.

Thanks, Sturla！I use intan recording board and their headstage, so last week I contacted them. I was told I could record LFP in the way as all of you told me.  If I had good LFP with our system, I will let you know in several weeks!

The amplification you need is dependent on the ADC resolution. For example in the Mosers' lab we used 20000x amplification because the ADCs were only 8 bits. With a 16 bit ADC the resolution is 256 times better, which means amplification can be reduced accordingly. So when I used 3000x amplification in Johan F. Storm's lab the spike waveforms were actually measured an order of magnitude more precise than in the Moser's lab, although the amplification was reduced by an order of magnitude. An 8 bit ADC does not have sufficient resolution to measure spikes and LFP simultaneously. A 12 bit ADC is borderline. 16, 24 or 32 bit ADCs can easily do this. [The ADCs I used in the Mosers' lab were actually 16 bit, but the software decided to throw away the lower 8 bits, storing the signal at 8 bit per sample.]

Check the dynamic range of the ADC board (typically about 3 V) and its resolution (bits per sample) and then calculate what measurement precision you get with your level of amplification. The trick when recording spikes and LFP simultaneously is to keep the amplification so low that the LFP does not saturate the ADC. This requires lower amplification than what you would use to record spikes without LFP (by a factor of 10 or so) and thus higher ADC resolution is required to get the same precision on the measured spike waveforms.