# JFET Common Source DC Amplifier Temperature Compensation

Discussion in 'Electronic Design' started by D from BC, Nov 19, 2007.

1. ### D from BCGuest

My app uses a JFET..

What are some ways to temperature compensate a common source JFET DC
amplifier.

Vdd=5V
|
10K
|
+------------Vout (biased for 50% Vdd)
|d
signal--->[ N JFET
|s
|
Gnd

'DC' amplifier.

Details:
1) The signal does contain a stable bias voltage.
2) Temp range: 20C to 50C
3) Vout thermal error of +/-10mVdc.
4) BW: DC to 2Mhz

I'm guessing compensation by:
1) PTC or NTC resistor
2) Use another JFET and feed in anti drift.
3) Find a special jfet (IC??) << not an op amp!

Possible ??

D from BC

2. ### Jim ThompsonGuest

What does "temperature compensate" mean? Bias point, gain, both?

...Jim Thompson

3. ### D from BCGuest

Let's say the gate voltage is set for Vout = 2.5V...
Vout can vary with temperature..Especially with no JFET source
resistor.

For a DC amplifier like this, doesn't DC bias thermal drift at the
output also look like DC gain thermal drift?

IIRC the power supply drifting with temperature can cause bias drift
alone.

D from BC

4. ### JoergGuest

That would be my avenue. There are duals such as this one, pretty low noise:
http://www.semicon.toshiba.co.jp/docs/datasheet/en/Transistor/2SK3320_en_datasheet_071101.pdf

Buying those might be a challenge though.

5. ### Jim ThompsonGuest

Except for the "sweet spot" bias condition, gain (small signal) and
output Q-point probably vary independently.

Instead of your vague "specification" what is it you are trying to
accomplish?

...Jim Thompson

6. ### Tim WilliamsGuest

Isn't the slick solution (given in AoE) to put another one in the source,
so source current (and thus Vgs) is always equal between the two?

Hmm...but that's for a source follower. Constant source current doesn't do
you much good. Well, it would work for setting DC bias nonetheless, with
the source bypassed to ground. Doesn't do anything for transconductance.

Tim

7. ### D from BCGuest

Ahhh... I think you've caught on that I dumb down problems to the
point of making it the wrong problem

Gee...I dunno if I can devague it more...
I'll try...
I need a fast DC amplifier with an input range of 0 to
-200mV(negative) with a response up to 2Mhz.
The output of the amplifier has to reach 2.5V out for -200mV in.
There is no negative supply present.

I thought a JFET would be best..but spoiled by thermal drift.....
My alternatives were:
Use resistive network + op amp
Create a negative supply + op amp
Put one of those 'beyond the rails' op amps to the test.
Negative rail IC + op amp

D from BC

8. ### Fred BartoliGuest

Jim Thompson a écrit :
Are you sure that they are independent?
For a given jfet model gm is strongly linked to the drain current (and
essentially free from IDSS).

9. ### Jim ThompsonGuest

If -200mV Input => +2.5V Output, what does 0V Input => ?? Output

...Jim Thompson

10. ### Jim ThompsonGuest

They're not "independent", but I don't think incremental gain and bias
Q-point will exactly track.

...Jim Thompson

11. ### Fred BartoliGuest

Jim Thompson a écrit :
Ah, a bit of french meaning slept in there.

Now you said they'd probably vary independently, which they don't (for
ex. http://www.alldatasheet.com/view.jsp?Searchword=2SK170 ).

But gain and bias point sure can't exactly track. For this to be so,
you'd have to have gm proportional to Id, which you can't ignore to be
the hallmark of exponential

12. ### JoergGuest

Many regular opamps can work down to about 300mV below the negative rail
but you'd have to find one that's fast enough and low enough in offset.

If you have a spare inverter somewhere you could make in inverting switcher.

13. ### D from BCGuest

It's allowed to clip..

I'd say I'm looking for bias point thermal compensation.
Not small signal gain thermal compensation.

I believe it'll be ok for the thermal compensation to move the Q point
in order to maintain Vout at the bias level of 2.5V.
The Vout bias level is not to drift with temperature.
The new Q point due to thermal compensation and the resulting new
small signal Av.... is probably ok.

For my app..
The Avdc thermal stability (at only one Q point) is more important
than the small signal Avac thermal stability..

D from BC

14. ### John LarkinGuest

The r-r opamps do work. Sounds like a lot less hassle than trying to
work around a discrete jfet.

John

15. ### Jim ThompsonGuest

Is this part of your "alternating noisy edges" problem?

Seems you're requesting assistance at too small a scale... a larger
over-view might produce a more elegant solution when we can see loads
and continuing functions ;-)

...Jim Thompson

16. ### D from BCGuest

Bummer... No app note at the bottom..

I get suspicious when parts are hard to get...It usually means I'm
doing something the wrong way..or old way..
On occasion, it can be it being a very new way..but I don't think so
in this case..

D from BC

17. ### JoergGuest

Looks like he doesn't even need output RR. Isn't there a LM324 on
steroids somewhere?

18. ### D from BCGuest

Good guess!

It's like Dr. Frankenstein putting bits and pieces of people together
to make Frankenstein..
IIRC Frankenstein wasn't an elegant solution..

I'm just chicken to post my whole project. Also, there's the pride
factor too..
And...I'm trying to have a mostly original project..

So... it might wander around like a zombie and scare people...I made
it and it works...well kinda..

D from BC

19. ### WinfieldGuest

Well, as I understand it, and based on my measurements as well,
gm does mostly track Id, more or less independent of Idss,
especially at currents well below Idss. But I don't get the Q
discussion y'all are having here. To my mind the Id operating
point is best set independently from the particular miserable
JFET's Vgs vs. Id, etc. I mean, part-to-part, sheesh! Bummer!

But anyway, whatever, I suggest that D just go ahead and use a
JFET opamp. I mean, I'm a "big" fan of JFETs and use them as I
can, for when they're best, but what an unholy pain they are!

20. ### Phil HobbsGuest

How about an ordinary single-supply op amp wired as an inverter? Ground
the noninverting input, and bob's your uncle.

Cheers,

Phil Hobbs