Sorry for the delay, but I decided to build my own test jig to get a better picture of the noise spectrum.
There are two plots here. I had to go to a double shielded enclosure with the outside shield connected back to signal ground on my A/D. There was still some slight mains hum depending on where I put the box on my bench (out in the middle of the room was best). In any case it was reduced to 2nV or so. I tracked it down to the input cap. I thought I connected the outside foil to ground, but it seems the end termination and FET lead is enough. Next rev will have extra care here. The 997Hz tone is a mystery. I thought it might be the cal oscillator on my scope but it wasn’t.
The noise is plotted on a log scale with -87.2 dB = 1nV/rt-Hz. This was established by taking the best FET and adding a reference resistor in the gate, a Vishay 10k .001% oil immersed precision resistor. I then scaled the plot down to 1nV. This theoretically even corrects the measurements for temperature but this is not necessary under normal lab conditions. I like this better than inserting a tone especially when using FFT’s, everything is scaled without the tedium of keeping track of bin equivalent noise BW.
All the FET’s were measured at 1.2mA Id. The lowest plot is my reference line. It varies just a little and I took an estimate around 1kHz of -87.2 dB. The next two plots (moving up) are a 2SK372 (Idss = 24.19mA) and 2SK369 (Idss = 14.75) respectively. Note the scale, we are only looking at .5nV – 3nV or so. The last plot is a random 2SK170. This is a very nice example of GR noise that would not show up well on a Quantech since it is only 3nV at 10Hz. These noise levels should scale as the fourth root of Id (one of the things that I want to do eventually).
The second plot is all three FET’s with Rg = 10k. It looks like the slight rise at the low end on the 2SK170 is due to the GR noise. The roll off at the high end is probably the A/D. The unexpected result is that the other two plots rise at the high end. This could be induced gate noise but probably not impact ionization since we are at low Vgs. The lack of this in the 2SK170 and the fact that the 2SK372 is in a very small package leads me to believe this is a new/different process, this defiantly needs more study. For now these two FET’s look great for MC for sure.
The last plot is the BF862, pretty impressive for an RF FET!
There are two plots here. I had to go to a double shielded enclosure with the outside shield connected back to signal ground on my A/D. There was still some slight mains hum depending on where I put the box on my bench (out in the middle of the room was best). In any case it was reduced to 2nV or so. I tracked it down to the input cap. I thought I connected the outside foil to ground, but it seems the end termination and FET lead is enough. Next rev will have extra care here. The 997Hz tone is a mystery. I thought it might be the cal oscillator on my scope but it wasn’t.
The noise is plotted on a log scale with -87.2 dB = 1nV/rt-Hz. This was established by taking the best FET and adding a reference resistor in the gate, a Vishay 10k .001% oil immersed precision resistor. I then scaled the plot down to 1nV. This theoretically even corrects the measurements for temperature but this is not necessary under normal lab conditions. I like this better than inserting a tone especially when using FFT’s, everything is scaled without the tedium of keeping track of bin equivalent noise BW.
All the FET’s were measured at 1.2mA Id. The lowest plot is my reference line. It varies just a little and I took an estimate around 1kHz of -87.2 dB. The next two plots (moving up) are a 2SK372 (Idss = 24.19mA) and 2SK369 (Idss = 14.75) respectively. Note the scale, we are only looking at .5nV – 3nV or so. The last plot is a random 2SK170. This is a very nice example of GR noise that would not show up well on a Quantech since it is only 3nV at 10Hz. These noise levels should scale as the fourth root of Id (one of the things that I want to do eventually).
The second plot is all three FET’s with Rg = 10k. It looks like the slight rise at the low end on the 2SK170 is due to the GR noise. The roll off at the high end is probably the A/D. The unexpected result is that the other two plots rise at the high end. This could be induced gate noise but probably not impact ionization since we are at low Vgs. The lack of this in the 2SK170 and the fact that the 2SK372 is in a very small package leads me to believe this is a new/different process, this defiantly needs more study. For now these two FET’s look great for MC for sure.
The last plot is the BF862, pretty impressive for an RF FET!
