Matching BS170 for Softrock

Recently some BS170s supplied by Tony for the Ensemble RXTX have shown rather large variations in standing current when tested according to Robby's instructions. http://www.wb5rvz.com/sdr/ensemble/10_pa.htm 
This current is indicated by measuring the voltages (around 0.05 volts, 50mV) at R48 and R49.

It seems that many BS170s have a similar Gate Threshold Voltage but the device data sheet shows the possibility of a very large range. Such variations will give large variations in the standing current of  the Softrock PA devices . This can affect it's linearity.

A discussion on the Softrock40 Yahoo Group decided that it would be a good idea to check the six BS170s supplied with the kit and select the closest matched pair. This discussion started here http://groups.yahoo.com/group/softrock40/message/55018 but continued with different Subject Headings http://groups.yahoo.com/group/softrock40/message/55292 There are lots of posts, not all relevant. 

Below is the simple circuit suggested and my notes after I had tested a few.

May 20th. NOTE. It is still not established just how much variation in the R48/49 voltages is acceptable. In http://groups.yahoo.com/group/softrock40/message/55570 Mike says 20% will not seriously affect the performance. This message is discussing the bias circuit. If the R48/49 voltages are similar but both high, or low, then the bias could cause this. A BS170 is in the bias circuit, again variations from the typical characteristics will affect the bias voltage. The selected pair in the PA may require a different voltage.  Later tests show that all three need to be matched. A mixture of different makes, or batches, can produce wide variations.
Here http://groups.yahoo.com/group/softrock40/message/55586 is W1SRB's description of his adjustment of the bias. BUT matching all three may be a better solution.
If you are uncertain please ask in the Yahoo Softrock40 group BEFORE you do anything.



I used a plug I had in my junk box for the BS170. The pink and white wires, Drain and Gate, are connected together and go to 12 volts through a 100R resistor.
The black, Source, wire goes to power negative.
Any suitable "socket" may be used or, with care, clip leads. 
The voltmeter positive goes to the Drain and Gate, the negative to the power negative.

The current through the BS170 will be around 80mA.

My suggested 0.05 volt variation seems to give roughly the same range of variation as Robby's pages specify. You should find two of the six that are much closer. The actual measurements when installed in the Softrock are not too critical, they just need to be close to each other. 
A third BS170 must be selected that is closest to the pair, use that for Q5 in the bias circuit.

Note the BS170s may be damaged by static electricity so the usual ESD precautions are recommended. 

Below, my setup.
The voltages I measured were all just above 3 volts, I do not know what may be considered acceptable limits.

This procedure will still not guarantee that they are matched but it is the best so far.
"When you match the Gate Threshold Voltage, this does not guarantee that the transconductances will also match.  In the RXTX circuit, both the Threshold Voltage and the transconductance will affect the current through R48 and R49.  Milt W8NUE" After fitting check the R48/49 voltages, hopefully they will be close.
 At the bottom of this page I have included a test setup to do this outside The Softrock but it needs to be done carefully. Comments in the Softrock group seem to indicate this test will not be necessary.



Click on picture for full size.

The BS170 matching is to get the voltages on R48/49, at the BS170's sources, similar. Robby suggests 10%. Mike Collins, KF4BQ, says even 20% is perfectly acceptable.
Similarly the allowable amount of variation from nominal 0.05V, 50mV, is debatable.
Until someone comes up with a definitive answer I suggest that only really large variations should be of concern. I think some are taking this too seriously. Far more important is that these devices should not be overdriven.
Ken VA3ABN found some BS170s (Not supplied by Tony) that had a test voltage of around 2.4. These, when fitted, had a high standing current. Q5 should be replaced with a similar tested device and   the bias will adjust so the sources are near 50mV.
All three BS170s need to be matched.

It is likely that if the three BS170s are matched this circuit will not need altering.




There is no need to use this setup below, but it could be used to check in a circuit like that of the actual PA.
Readings need to be taken quickly at a set time after applying voltage. The BS170s are sensitive to temperature. This is normally compensated in the Softrock's bias arrangement.
No current is taken from the batteries, I used some part-used ones to get the required voltage, 2V7 in my case but it is not critical, we are just comparing the device's performance.