JFET Symmetry 2

[dau]

Hi NG,
relating to my previous question (31 Jul 08) I've found out that the specifications vary pretty much from part to part.
I'd like to test the SOT23-JFET before soldering it to my board. Has anyone heard of or built a test fixture for SOT23?
Any ideas?

 

[Noway2]

No, I have not heard of any form of test fixture for a SOT23, but that isn't to say that some form of socket does't exit.

At the same time, I have to question the thinking behind your query. What it looks like you are wanting to do is to 'cherry pick' for devices that meet a particular parameter that can't be guaranteed. Unfortunately, this is a very poor way to design and while it may work for a one off prototype, it is almost guarenteed to be a failure in the long run.

The trick in designing with transistors is to design the circuit so that the behavior depends on the components you can control rather than an intrinsic property of the transistor itself. This is especially true when the dependant property is one that can vary a lot. For example, when desiging a simple common emitter amplifier, one sets the gain by choosing the ratio of collector and emitter resistances as opposed to relying on the intrinsic emitter resistance.

my suggestion would be to revisit your circuit to find a different approach that is not dependant on the varying characteristics of the device.

 

[geekEE]

I agree with Noway2 that it would be better to design the circuit so that it is not dependent on pre-screening parts for that application.

However, if you wanted to proceed with testing parts, Emulation Technology has ZIF sockets for SOT23 parts:

http://www.emulation.com/catalog/off-the-shelf_solutions/sockets/sot/

Glenn

 

[itsmoked]

I'm certainly in with Noway too.

Having to test transistors means the price for your product is all over the place. You need to pay someone a fortune to sit and test them and you can end up having to throw away no parts on one production run and all of them the next run! How can you budget for that? This means you are possibly paying 10x or 40x the original per piece price.

You would be better off paying much more for a specified higher grade part or just design the circuit more forgivingly and use a cheaper part.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

I ditto everyone else. The designer's task is not to merely produce a circuit that works, once in a while, but one that works over the entire span of environments, and component parameters.

Just because you have a working part at room temperature, doesn't even guarantee proper operation over temperature or over lifetime.

TTFN

FAQ731-376

 

[dau]

Hi all,
naturally you're right.It is a desperate attempt to boost the specs of the device at unacceptable costs.I must find another way to cure the problem or resign to the actual state of accuracy.
Thanks

 

[Noway2]

Without a better understanding of your circuit, it is difficult to make exact recommendations...but....

Are you certain that a jfet is the right part for the job or would you get better results from an enhancement mode (ie mosfet) which typically has a higher gain ratio and more controllable paramaters?

Regardless, it really sounds like you need to start by revisiting the product requirements, including the cost and determine if the two are capable of matching up.

 

[dau]

@Noway2
look at my post from 31 July 08 for some details.I switch 2 channels 1Vpeak dc to 10 MHz to an opamp with switchable gain.Perhaps a series connection of n-channel and p-channel Mosfets works.I don't know.Sounds complicated anyway.Also capacitance is an issue. My problem is that the prototype had accidentally well matched JFETs.Now that the layout is made and all corrections are included the final device is less accurate.