P-channel JFETs more linear?

[lazthegaz]

I have been messing about with an audio spectrum analyser, comparing the simplest possible JFET amplifier stage (FET, unbypassed source resistor, drain connected to the power supply via a load resistor, FET always biassed to about half the supply voltage, measurements always made at the same AC output voltage). Now I noticed that when it comes to THD, P-channel devices always seem to be better, i.e. more linear. The difference is very pronounced, typically at least 10 dB and it is always the 2nd harmonic that is better (the 3rd is usually pretty much the same).

Am I overlooking something here or is this real?

[peranders]

Have you negative power supply voltage for the p-channel JFET?

[john curl]

I doubt it. Make sure that the OUTPUT of the device under test is the same with N and P. P channel devices usually have lower Gm, SO, output is lower with a given input. This makes distortion appear lower.

[lazthegaz]

First of all, yes, I do use a different polarity for the P-channel devices. These are all proper amplifying stages, with gain. Secondly, as I mentioned all these are measured at the same OUTPUT voltages.

Here are some actual measurement data. All double checked. The output voltage was always set to 4 V pp at 1 kHz.

2N5460 (p-channel). Id = 1 mA.
2nd harm = -61 dB ; 3rd harm = -71 dB below fundamantal

BF245A (n-channel). Again, Id = 1 mA
2nd harm = -50 dB ; 3rd harm = -71 dB below fundamantal

2SK170-BL (n-channel), Id = 3 mA
2nd harm = -48 dB ; 3rd = -72 dB

There is more but these are representative.

[peranders]

As Mr. Curl points out, how about the gm, must be different for these devices? Do you have the same gain for all JFET's?

[padamiecki]

I propose to a little change in a measuring: no use a source resistor, and regulate a drain resistor to at about halve of supply voltage, and shift supply voltage up to 30V. This test will be much more strict and a jfet will be forking at idssmax. Besides try analyse a complementar pairs for example j103 - k246. The Comparsion of the other types jfets might be useless. Anyway your observation is interesting!

[lazthegaz]

Thanks, Padamiecki, I am going to try that. You are right: I am not comparing like with like because of the varying amount of negative feedback the source resistor introduces.

[lazthegaz]

Here's some results at Idss, with the power supply voltage set to 30V and the drain voltage set to half the supply voltage. The input AC voltage was always adjusted so that the output was 4V pp at 1 kHz.

Device.................. 2nd / 3rd harm

2SK170................ -39 dB -70 dB
BF245B (1)........... -37 dB -86 dB
BF245B (2)........... -37 dB -82 dB
BF245A................ -40 dB -73 dB
BF244A................ -38 dB -82 dB
2N5460................ -37 dB -77 dB

One interesting thing: the overall THD hardly changes, from 1.1 to 1.4 % but the spectral distribution varies a lot, especially the 3rd harmonic. Considering the usual preference for 2nd harmonic distortion, this may be of some significance.

[padamiecki]

I guess that p-devices are not measured yet?

[lazthegaz]

The 2N5460 is a P-channel device.