Does Pre-Matching 9610's Really Work?

[roddyama]

Hi Dale, Guys,

I realized this question was probably in the wrong forum so I re-ask it here.

I went ahead and ordered some extra 9610's (from Dale)
1.)The hope is to build a second Aleph at a later date.
2.)It sticks in the back of my head that in an old thread (un-findable with "Search") Mr. Pass said (not a quote, just from memory) that it does little good to match the differiential FETs. That the Pass method is to substitute 9610's until the DC offset disapears.

Question:
Does any one have experience with matched 9610's providing good results with a low DC offset voltage, or will it still be necessary to try several devices?

Slowly, but surely, I'm getting there.
Rodd Yamas***a

 

[fcel]

My understanding is that to obtain low DC offset, the output transistors would have to be matched as closely as possible.

 

[Nelson Pass]

If you match them for Vgs, then you usually don't
have to go to any extra effort to lower the DC offset
at the output.

Actually substituting devices into a working amplifier is
really the best way to do it, as you can test for a bunch
of performance parameters. It is tedious, but worth it.

Did I mention that we turn the amp off while performing
the substitution?

 

[harvardian]

Hi Rodd,


I will probably have plenty of input 9610's so I will try to sort in batches of three of four and send out more devices. Maybe you can report back on your success in doing this. I would expect that you would still want devices that match pretty close on Vgs.

Dale

 

[roddyama]

Hi Dale,

Thanks, much appreciated.

Rodd Yamas***a

 

[Asen]

Mr. Pass

Once you said you substitute the input devices in order to obtain lower distortion, not lower offset. From your experience, do they go together? I mean the lower offset and the lower distortion, or there are no rules of that kind?

Thanks

 

[Nelson Pass]

I'm sure there's some kind of relationship, but I don't
necessarily count on it. It depends on how much time I
want to devote to an amplifier. If I have lots of time, I'll
sub devices in and out, picking the best measuring combo
for offset, noise, and distortion, or maybe any one of these
three depending on where I'm trying to go. It's a very
tedious process.

 

[JOE DIRT®]

I feel that replacing parts may because of lack of proper Engineering of the circuit but it is the cheapest route in the orginal approach....there are so many variances when designing and building that we have to decide what is best when making a circuit work...hence subbing transistors is the best approach rather than a new board design

DIRT®

 

[Nelson Pass]

It might be a design issue, but more commonly it
is the result of imperfect parts, specifically gain devices
which distort or vary with temperature, voltage, current
and so on.

 

[GRollins]

The impact of differing Vgs values will vary depending on what circuit you intend to put them into. Something like an Aleph can tolerate more mis-match than an Aleph-X because the 'back' MOSFET (i.e. the one that takes the NFB and negative phase of a balanced input) is free to swing; its output doesn't go anywhere. If you try a mis-matched input pair in an Aleph-X (or any X, for that matter) the 'other' side will try to swing in order to bring the circuit into balance. Unfortunately, that will cause the outputs that it's hooked to to go along for the ride and the differential voltage offset at the outputs will get lopsided pretty quickly.
That said, it's still a good idea to match inputs, even in a 'normal' Aleph. The front end won't have to work so hard to achieve zero offset (or at least close to it) at the output.

Grey

 

[jwb]

I've banged this drum before, but I feel that the Aleph is much easier to build when using the 2sj109 dual FET instead of paired IRF9610s. The Vgs and temperature tracking are superb.

 

[Nelson Pass]

If you try a mis-matched input pair in an Aleph-X (or any X, for that matter) the 'other' side will try to swing in order to bring the circuit into balance. Unfortunately, that will cause the outputs that it's hooked to to go along for the ride and the differential voltage offset at the outputs will get lopsided pretty quickly.

Actually, my measurements indicate that you will pretty much
get the difference in Vgs multiplied by the DC loop gain (there's
that phrase again!) of the system, in other words its behaves
similarly to a conventional Aleph. Note though, that we normally
set the DC loop gain of an Aleph X (XA, if you like) at much higher
figures than an Aleph, thus more potential for DC offset.

 

[Nelson Pass]

Also, you can get dual P channel Mosfets at similar specs.
TI had some, and I just saw one from Vishay:

http://www.vishay.com/docs/70138/70138.pdf

This is only rated at 20 volts, but very likely will take much
more without a burp.

 

[Panelhead]

I've banged this drum before, but I feel that the Aleph is much easier to build when using the 2sj109 dual FET instead of paired IRF9610s. The Vgs and temperature tracking are superb.

I am going to build an Aleph next week. At least stuff the boards. The 2SJ109 looks interesting. Where do you source them?
Looking at my usual places only MCM stocks them. Are there better suppliers?

George

 

[Nelson Pass]

Feel free to use JFETs on the input pair, but do understand that
their low transconductance will seriously raise the distortion
figures. Enjoy, but don't be surprised if the measurements
suck.

pass/ - watches the meter through the rear view mirror

 

[grataku]

Note though, that we normally set the DC loop gain of an Aleph X (XA, if you like) at much higher figures than an Aleph, thus more potential for DC offset.

How do you go about determining this elusive parameter?

 

[GRollins]

I've always had good results with ordinary matching for the various Aleph variants I've built here. When I started on the Aleph-X prototype, I had to go back though the 9610 pairs I had on hand and get a better-matched pair in order to get the two halves of the outputs to track fairly evenly; the first two pairs I tried gave me fits, driving one side of the output high and the other low.
You're saying it wasn't the match between the 9610s? Curious...as that was the only thing I changed at that point in the game and it worked out on paper.
Now I'm starting to wonder if we're talking about the same thing. Either that or I need to get more sleep (granted, a normal state of affairs for me).

Grey

 

[Nelson Pass]

Nothing elusive about loop gain - it's that gain of the amplifier
determined by the voltage divider at the negative input.

And Grey, there is no end to the non-intuitive nature of the
X topology. There are times when something doesn't work
and you have no clue why. Eventually it surfaces, and you
are surprised, because it's not op amp behaviour. I had the
patent long before a satisfactory amplifier.

 

[macka]

Just my 2 cent worth here among you experts regards the 9610 matching.

I matched my 9610's in 2 phases, 1st rough tested each on the reference supply and put them into defined sets 3.90, 4.10 VGS etc.

I then set up one 9610 from each set of fets as a reference and compared against the others by arranging a 2nd identical current sources using precision resisters.

Measuring the VGS differential between the two fets (just like a bridge) enabled much higher resolving power. Clamping on the fets on a common heatsink further stabalized the readings.

I actually borrowed the solution from my brother Mark, a senior technician for the 7 TV Network here when building the Aleph 2 as my meter would not resolve well enough on the lower range. (helps to have one real electronics man in the family (and cheap TV repairs.....humbug)

Hope this assists,

Ian

PS Mr Pass, which J fets should we be ordering/matching for the forthcomming High Low Pass crossover, or will the beautiful Pass DIY Diva be supplying a Kit of matched Jfets like the (magical) Pearl ...just curious

 

[IanHarvey]

Looking at the Aleph input stage, zero DC offset is not guaranteed by having a differential input pair with matched Vgs/Id characteristic.

For zero DC offset, if the I/P pair are matched we need the two drain currents to be the same. This is not guaranteed - the total drain current of the two devices is set by the current source Q103 (I'm looking at the Aleph-3 service manual circuit) ; the drain current of Q101 is set by Vgs(Q108..Q109) / 390R. The drain current of Q102 is the difference between the two; it's not guaranteed to be half the total.

So minimum DC offset is best guaranteed by choosing it in-circuit for particular output transistors, or (possibly better) by varying either R106 or R108 until the drain currents are matched.

Cheers
IH