General > General Technical Chat

Another boring thread-Why using a curve tracer for matching JFETs is a good idea

(1/2) > >>

Robo:
Another boring thread about Why using a curve tracer for matching JFETs is a good idea ?

Before i start to be boring i will only say that simple setup containing a voltage source and multimeter will let you measure some basic parameters and for most applications it will be sufficient and more than enough to use this method . If is this the case than dont bother to read all the boring stuff and grab a cup of tee instead .

Now il start with realy boring stuff and show some examples of how limited one point or few point measurements could be and how the measured values are prety relative.

Long story short - is that i needed a matched pair of high quality low noise JFETs for preamp imput stage .
For the job 2SK170 GR was chosen . I wasnt in a mood of monkeying around and searching for genuine toshibas on ebay so iv got LSK 170A from Linear Systems.
As for measuring setup i have used 2 Precision SMUs ,- one For Vgs suply the other for Vds.

To save me some time i started with Vgs voltage =0 and swept Vds from 0 to 10V Vds in 200mV steps.
Than i was puting gathered data into spreadsheet. Now i could sort measured values into the bins and select pairs acording to Ids values.
A litle note here. All the transistors came from one batch and were paired looking at the spreadshhed so i wasnt knowing how the curves will compare .
Now i could take a better look at the curves for selcted pairs. I had no intentions to dive into uA range so 3 curves should be sufficient for comparrisement purposes .
The Vgs voltage for each curve was suplied as follows  0V, -0.1V,  -0.2V
Vds voltage was swept from 0V to 10V  with 200mV step and here are the results :

Note the Id difference in the omic region and difference in Id for different Vgs values @ 10V Vgs

For a comparrison i have putted pair matched to 0.25mA  ;D

And at the end as alittle bonus a cuves for a pair of LSK 170A and LSJ74 .

There is totaly nothing wrong with the batch or with the quality of the transistors and i wasnt picking bad matches to show all this , i tried to match the transistors based on Id values and demonstrated how limited multimeter measurement setup is,- compared to curve tracer .

And yes i was able to match some nice pairs with no problems at all, but it wouldnt be possible without a curve tracer .

Everything what i have said or showed here is nothing new and has only informational character

Good luck to all in your future projects

Kleinstein:
The curves look quite different, despite of similar current at one point.  So at least in those examples there seems to be more than 1 prameter effecting the JFETs.

The point normally is to do the matching at conditions close to the later use (e.g. same current and Vds and than measuren Vgs).
It really depends on the circuit if the P-JFETs also need to be matched to the N channel ones in some respect. Most circuits use N-channel only.

mag_therm:
To make QUCS model better follow actual test of 2n4416 in circuit,
I adjusted both V-gs-off , and  trans_conductance parameters in QUCS
example, setting V-gs-off = -2.3 V , and  trans_conductance = 1500 umho

I have not tested enough of them to see the scatter.

Robo:
They behave diferent because of the different structures of transistor and you wont get 100% mirror image from P and N channel ,- you can get prety close tho but you need to see whats going on.
It is better to use multipoint measurement and then draw the curves but measurement accuracy will depend on measuring point count and if you wanna draw more than one curve than its gonna be  very time consuming.

Robo:
It is posible to calculate transconductance but what about ohmic region ?

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod