fet matching

[sriegel]

i have tested all of my fets for the aleph 5 amp and have a question on matching:
which fets are the most important to match, is it:
1) all used in that channel
2)inputs and outputs each i.e. 3 inputs matched and 3 outputs matched.
3)all of them, both channels

the reason i ask is that i can get finer matching with the least amount of total devices to match, but want to know which is the best way to match my devices.
thanks for any input!
p.s. I can match all 12 devices to within 0.1 volts but can do better if i can bundle them in smaller matched batches.
scott

 

[macka]

FET Matching

Scott,

I appreciate your comments as I went througth the same exercise recently. I found lisiting all my measured devices on a spreadsheet useful so I could identify the samples within tolerance more easily.

Basically,

I would go for option aim for option (1).

INPUT PAIR
That is try and get a very close match for input pair and the current source 9610. Then try and get a very similar result from your sample for the other channel

I matched mine within 1 millivolt @3.70 +- 1 millitvolt from the sample of 50 units, so I had two pairs @3.70 +- 1 milli volt.

I did this with the test method in the A75 project and then arranged a bridge using (1) fet as a reference with two test circuits and measured the voltage differnce between them on the lowest range of my digital meter. I was easy to then meaure 10's of millivolts.

I think the imput match is most important and inter channel match is probably more inportant with the imput pairs.

This would interesting to prove in objective tests however.

OUTPUT
With the output the number from the sample within tolerance was narrower as you indicate.

So I set up a spread sheet and from 50 units got a range of 4.57 to 4.10 spread.

Luckily I got enough at @ 4.30+- 3 milli volts to use 6 + 6 for both channels for both the current source and active side.

Prior this this with the Alphe 3 I made sure each parellel pair was very close match for current sharing and then selected the nearest matching pair from the spread sheet for the active side.

This ensured that I had the closest match for both the current source and active side.

I then selected the next closest matching (2) pair for the other channel.

I hope this helps.

regards

macka

 

[sriegel]

macka,
thanks for the quick reply!
so..... if i understand you, for the output fets, it would be optimal to match the current source and active devices per channel, i.e. 3 + 3 matched as closely as possible for each channel. does that sound right? so as long as the channel fets are matched as closely as possible, there can be a slightly larger difference in comparing channel to channel average voltages?
hope i'm getting it!
thanks,
scott

 

[macka]

FET Matching

Scott,

Your understanding is correct,

Good Luck.

regards

macka

 

[Nelson Pass]

The most critical thing is to match the devices which
are in parallel so that they share current accurately.
In an Aleph 5 this means that they must be matched
in groups of three devices. Matching a wider group of
devices is nice, but not necessary.

On the input differential pair, matching the devices
gives better DC offset figures at the output, but this
can typically be adjusted out, and so it is very nice
but not essential.

The best way to pick input devices is by actually testing
them in the amplifier itself, which means socketing the
prospective devices and testing offset, distortion, and noise,
and if you want to go to the trouble, the sound We do
this on the Volksamp product, but with our other production
we used monolithic matched JFETs on the input.

 

[sriegel]

nelson and macka,
thanks sooo much for the guidance, it helps to have such an incredible resource for the DIY'er! and nelson, thanks especially for the attention you pay all of us, i get such a charge out of this project and the experience is so rewarding, although, i haven't gotten to the testing yet.
thanks again everyone!
cheers,
scott