---++ This is the design page for the Cryoboard noise filter circuit

Design files

PDF of schematic

* CryoNoiseFilterJune2011.pdf: CryoNoiseFilterJune2011.pdf

* CryoNoiseFilter.pdf: CryoNoiseFilter.pdf (12 May 2011)

Altium project

* CryoNoiseFilterJune2011.zip: CryoNoiseFilterJune2011.zip

* CryoNoiseFilter12May11.zip: CryoNoiseFilter12May11.zip

Gerbers required to manufacture PCB (not assembly)

* CryoNoiseFilterGerb.zip: CryoNoiseFilterGerb.zip

Spice simulation of the Drive/Sense circuit

* CryoboardFIlt2.asc: CryoboardFIlt2.asc

Updates

The following spice simulation has been used to demonstrate that a simple capacitive filter should suppress any noise going into the cryostat and that the resulting modification to the measured impedance can be calibrated out.

To understand how a simple capacitor can remove the noise. Simply consider its impedance at both 100Hz and 1Mhz. At 100Hz, a 10nF capacitance has an impedance of 159kOhms and at 1Mhz has an impedance of 15.9 Ohms. At 1Mhz the 15.9 impedance that is external to the cryostat is in parallel with the load that is inside the cryostat. All high frequency noise should be shunted away from the cryostat. With a 1k load, roughly 63x less noise will enter the cryostat and with a 100k load 63,000x less noise should enter the cryostat.

* FilterSimulation.jpg:

(Note that the ground on the current source has actually been removed in the actual simulation, having the ground at this location is an error and significantly changes the results)

Firstly lets just simulate the response of the Isense and Vsense filter, from this simulation you can see that both the first order filter and second order filter have virtually the same response up to 100hz. The difference between the two curves at 100hz is 0.6% of the average between them. What this indicates is that for the purposes of this simulation we can assume that when performing a Vsens/Isense any frequency dependance before 100Hz is not a result of these filters and will be directly because of the "noise filtering" we use.

* FilterSimulationResult1.jpg:

Our noise filter, will be just a 9.4nF capacitor to ground (in reality it will be two 4.7nF capacitors in parallel) one on the drive side and one on the sense side. At 100Hz, this effective 9.5nF capacitor will have an impedance of 169kOhms. Note that the capacitance cannot be made significantly larger since we wish virtually all current to flow through the load at frequencies lower than 100Hz. The maximum load being approximately 100kOhms (I believe 50kOhms is the largest we currently use).

In the following simulation we compare the vsense/isense in two different cases, when the load is 1kOhm (Blue trance) and 100kOhm (green trace). In both cases the curves are normalized to the actual load impedance, i.e 1 is a perfect measurement. (Note that this plot is not an indication of the noise performance)

* FilterSimulationResult2.jpg:

You can see in the figure that if we AC bias at 100Hz in this example the normalized V/I is different in the two cases. For the 1k load, the measurement at 100Hz is much closer to the actual load resistance than for the 100k load. This makes sense since the 9.4nF capacitance has an impedance of 169kOhm.

The amount by which our V/I changes with load can be calculated. This means that we should be able to calibrate out this affect.

The following simulation shows a model for the overall system on the left and a closer model to the actual system on the right. The model on the left is significantly easier to analyze and is very easily derivable from the model on the right.

* FilterSimulation2.jpg:

In the following figure the two models are compared. As you can see they are identical (remember that the model on the right also has the 1 and 2 pole filter still in place and we are assuming this has no affect). The model on the left is equally applicable for a 1kOhm load, simply make the load resistance in the model 500 Ohms.

* FilterSimulationResult3.jpg:

In order to generate the model on the left, the differential system has had a ground connection added directly down the middle of the system. This is why the 100kOhm has become 50k, why the 10nF capacitor is now 20nF (remember that two 20nF capacitors in series are equivalent to 10nF)

What this model tells us is that to calibrate out the affects of having 9.4nF capacitance present we simply have to calculate the ratio of the current actually flowing through the load compared to what was sourced.

The ratio between load current and supplied current is:

abs( Iload/Iin ) = 1/sqrt(1+(Rtot*Ctot*2*pi*f)^2) where Rtot = Rload/2 and Ctot=2.2nF+Cfilttotal*2

abs(Vload/Iin) is what we're measuring = Rload * abs( Iload/Iin )

The resulting expression for Rload is simply solved

Rest=abs(Vmeas/Iin) %The initial estimate of R

k=Ctot*pi*f;

Rload=Rest/sqrt(1-Rest^2*k^2) %The compensated value for R

as an example, if we put a 10nF in parallel with a 100k load and ac bias at 100hz simulations show me that 82% of the current flows through the load(same answer from 1/sqrt(1+(Rload*Ctot*pi*f)^2))

if Iin=10nA (this is what we're measuring) we only have 8.2nA flowing into the load the voltage across the load is 0.82mV the first estimate of the load resistance is therefore 0.82mV/10nA = 82kOhms

using the above equations:

k=(2.2nF+10nF*2)*pi*100=6.97e-6

Rload=82kOhms / sqrt( 1- 82kOhms^2*6.97e-6^2) Rload = 100 kOhms

---

This topic: CryoElectronics > WebHome > CryoNoiseFilterDesign Topic revision: r6 - 2011-06-30 - WinterlandUser

---