In looking at this data sheet I see a noise figure but I don't know what it means. If I replace a irf9610 running at 2 ma with a 2n5461 running at 2 ma would the jfet typicaly have less noise than the mosfet ?
The noise figure is the ratio expressed in dB between the total noise and the thermal noise of the real part of the source impedance, assuming the source to have a temperature of 290 K.
Example: noise figure of 1 dB at 1 Mohm source resistance:
The real part of 1 Mohm is 1 Mohm.
The thermal noise of 1 Mohm at 290 K is sqrt(4 k T R) ~= 126.5525 nV/sqrt(Hz), where k is Boltzmann's constant (1.38065E-23 J/K)
The total noise is 1 dB more, so 10^(1/20) * 126.5525 nV/sqrt(Hz) ~= 141.9943 nV/sqrt(Hz)
Independent noise source add quadratically, so the noise of the JFET must be:
sqrt((141.9943 nV/sqrt(Hz))^2 - (126.5525 nV/sqrt(Hz))^2) ~= 64.3959 nV/sqrt(Hz)
For a bipolar device, this would be the result of its equivalent input noise current and voltage. For JFETs, the equivalent input noise current is negligible except at extremely high source impedance, so this will be roughly its equivalent input noise voltage. One line lower in the datasheet, the equivalent input noise voltage is specified as 60 nV/sqrt(Hz), so that sort of matches.
This is all at 100 Hz. For frequencies of a few kHz, where your ears are most sensitive, the noise figure is around 0.4 dB to 0.5 dB in 1 Mohm when biased at VDS = 15 V and VGS = 0 V, see figure 9 of the datasheet. 0.4 dB to 0.5 dB related to 1 Mohm corresponds to 39.3084 nV/sqrt(Hz) to 44.2062 nV/sqrt(Hz).
I can't answer your question about the comparison to a MOSFET because I have no data about the MOSFET's noise. What I can tell you is that these figures are quite high. There are JFETs with noise levels between 0.8 and 10 nV/sqrt(Hz) at frequencies of a few kHz.
Example: noise figure of 1 dB at 1 Mohm source resistance:
The real part of 1 Mohm is 1 Mohm.
The thermal noise of 1 Mohm at 290 K is sqrt(4 k T R) ~= 126.5525 nV/sqrt(Hz), where k is Boltzmann's constant (1.38065E-23 J/K)
The total noise is 1 dB more, so 10^(1/20) * 126.5525 nV/sqrt(Hz) ~= 141.9943 nV/sqrt(Hz)
Independent noise source add quadratically, so the noise of the JFET must be:
sqrt((141.9943 nV/sqrt(Hz))^2 - (126.5525 nV/sqrt(Hz))^2) ~= 64.3959 nV/sqrt(Hz)
For a bipolar device, this would be the result of its equivalent input noise current and voltage. For JFETs, the equivalent input noise current is negligible except at extremely high source impedance, so this will be roughly its equivalent input noise voltage. One line lower in the datasheet, the equivalent input noise voltage is specified as 60 nV/sqrt(Hz), so that sort of matches.
This is all at 100 Hz. For frequencies of a few kHz, where your ears are most sensitive, the noise figure is around 0.4 dB to 0.5 dB in 1 Mohm when biased at VDS = 15 V and VGS = 0 V, see figure 9 of the datasheet. 0.4 dB to 0.5 dB related to 1 Mohm corresponds to 39.3084 nV/sqrt(Hz) to 44.2062 nV/sqrt(Hz).
I can't answer your question about the comparison to a MOSFET because I have no data about the MOSFET's noise. What I can tell you is that these figures are quite high. There are JFETs with noise levels between 0.8 and 10 nV/sqrt(Hz) at frequencies of a few kHz.
Last edited:
Thanks Marcel, great info,
If you want a reference for a MOSFET IPS opamp, take a look at the newer TI OPA1656, note the higher 1/f corner frequency, characteristic of the semi physics. I assume they used a MOSFET over a JFET because they could get the performance they wanted with a smaller die size, thus the economical price difference of that part compared to JFET types.
If you want a reference for a MOSFET IPS opamp, take a look at the newer TI OPA1656, note the higher 1/f corner frequency, characteristic of the semi physics. I assume they used a MOSFET over a JFET because they could get the performance they wanted with a smaller die size, thus the economical price difference of that part compared to JFET types.
Last edited:
Hi rsavas, I am aware of the fact that MOSFETs generally have more 1/f noise for a given area than JFETs, but I have no data for the specific MOSFET woody wrote about. As it is a power MOSFET, it probably has a fairly large gate area and hence a fairly small 1/f noise, also since it is a PMOS - but that's just an educated guess. I also don't know the source impedance woody is interested in.