John Curl's Blowtorch preamplifier

[hitsware]

>Jfet input stages have no significant advantage
>in this regard over bjt input stages emitter
>degenerated to a similar gm.

The advantage is that the junction on the
jfet is reversed biased.
The junction on the bjt is forward biased.
(sorta like a regenerative am demodulator)
Overcoming reversedness is harder than
abbetting forwardness.

[GK]

Originally posted by hitsware
>Jfet input stages have no significant advantage
>in this regard over bjt input stages emitter
>degenerated to a similar gm.

The advantage is that the junction on the
jfet is reversed biased.
The junction on the bjt is forward biased.

I am aware of that. But the b-e junction needs to be reverse biased to some degree by the RF voltage before the BJT can behave as an adequate detector.
If the BJT input stage is heavily emitter degenerated (for other reasons such as linearity, particularly as OLG< with >frequency, lower gm / <Ccomp / >slew rate / <TIM, etc) it's immunity to AM demodulation is increased significantly as a much larger RF input voltage is required to overload it.

For stray RF voltages (especially signals in the AM broadcast frequency range), that do not cause overload, the input stage itself will just amplify them pretty much linearly along with the audio input signal.

Cheers,
Glen

[john curl]

Has everyone understood that bipolar transistors tend to rectify RF, because they are already conducting? However, jfets are have a completely turned off diode in normal operation. It is a higher barrier to cross, before RF rectification can happen.

Of course, I am speculating that this is so, but I will stand corrected by higher authority, if there is a better explanation.

[PMA]

Originally posted by G.Kleinschmidt



That is not true and a complete misrepresentation:

http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm

I speak about the book, chapter 4, not this short webpage analysis. Anyway, good to see his web result:




and compare with mine. Input stage of this kind is really not appreciated.

[GK]

Originally posted by PMA


I speak about the book, chapter 4, not this short webpage analysis. Anyway, good to see his web result:

and compare with mine. Input stage of this kind is really not appreciated.

Good grief. Your blanket claim was a plain misrepresentation of what he has published and that graph that you cite is for a BJT long tail pair without emitter degeneration and with a very low tail current, as shown in figure 6 (reproduced below)

This is far from the final input stage used in the "Blameless" amplifier (you only have to glance at the rest of that webpage) and the experiment was conducted to show the effects of the percentage of Ie imbalance on the LTP linearity.

[john curl]

This is not the Doug Self thread. Why are we discussing this?

[GK]

Originally posted by john curl
This is not the Doug Self thread. Why are we discussing this?

We are discussing and making comparisons about input stages. Part of that D.Self webpage is explicitly concerned with input stages.

[john curl]

This is pointless, it has been discussed in a biased way by Doug Self, himself. He doesn't even use fets for anything that I know of.

[syn08]

Originally posted by G.Kleinschmidt



I am aware of that. But the b-e junction needs to be reverse biased to some degree by the RF voltage before the BJT can behave as an adequate detector.
If the BJT input stage is heavily emitter degenerated (for other reasons such as linearity, particularly as OLG< with >frequency, lower gm / <Ccomp / >slew rate / <TIM, etc) it's immunity to AM demodulation is increased significantly as a much larger RF input voltage is required to overload it.

For stray RF voltages (especially signals in the AM broadcast frequency range), that do not cause overload, the input stage itself will just amplify them pretty much linearly along with the audio input signal.

Cheers,
Glen

Glen,

I agree that a degenerated BJT stage is not good as an AM detector. AM detection needs a nonlinear device to generate the 0 (signal) and 2w (to be filtered) harmonics. The more linear the stage, the less effective this process is.

Now, comparing BJTs and JFETs is not really straightforward; BJTs could easily be driven in the exponential Ic-Vbe region (which extends over many decades of current) by simply by keeping the device non-degenerated and perhaps low DC biased. Of course, a strongly biased BJT with strong emitter degeneration will be very linear, and hence a really poor AM detector. But so is a JFET stage, biased around Idss with source degeneration. In these regions, under large signal conditions, Ic and Id are still slightly exponential and parabolic dependent on Vbe and Vgs, and perhaps unless you compare a very poor BJT (Having a very low exponentil m factor) with a high gain 2SK170 JFET, the bipolar will still be a more efficient AM detector.

Now, if you bias the JFET closer to Vt (that is, e.g. by increasing the source resistor degeneration in an non LTP stage), the AM detection efficiency of the JFET will increase considerably! This is completely opposite to the bipolar behaviour and the reason is that by approaching the threshold voltage the JFET enters the subthreshold conduction region. In this region Id-Vgs is as exponential as the BJT, turning the JFET in an effective AM detector.

You see, unlikely a BJT, for non LTP stages, by increasing the source degeneration of a depletion JFET you are in fact increasing its AM demodulation efficiency. From this perspective, it would be fair to compare BJTs and MOSFET AM detection efficiencies and of course the MOSFET would win. But MOSFETs can be barely used as input stages...

One to another, if you compare a BJT LTP degenerated stage with a JFET LTP degenerated stage, biased around Idss, and the input signal is relatively small, the bipolar stage will be more efficient in detecting AM signals. If the input signal increases, then the JFET can reach the subthreshold region and could become as effective in AM detection as the BJT.

[PHEONIX]

Originally posted by john curl
This is pointless, it has been discussed in a biased way by Doug Self, himself. He doesn't even use fets for anything that I know of.

Hello John

The reason he does not use a fets is due the fact that you need a matched input pair it work with the LTP, wheras with a BJT input you can use unmatched parts and it will work fine provided you use a current mirror with them . Secondly matched Jfets are harder to get these days.

Regards
Arthur