X Aleph absolute voltage compensation scheme

[macka]

Since the original X Aleph circuit was published there has been from time to time discussion around controlling or minimising the absolute offset where R1, R4, R44 and R45 help control absolute DC offset at the outputs by sinking current to ground when the outputs sit either above or below 0V potential.

Depending on the scale of the design and final implementation the native circuit can exhibit a significant absolute offset in the order of 7 - 10 volts or more at turn on from cold and this voltage can take some time (over an hour) to settle to a residual level as the amplifier warms up

The McMillian resisters can be effected to bring down and control the absolute offset more quickly.

But neither the output resisters to ground or the McMillian resisters offer true voltage / thermal compensation

There are two problem I have discovered with the absolute offset.

Heating of the output resisters to ground and failure of these resisters.

Abnormal heating of the lower half of the output followers while the additional negative voltage flows across the Drain/Source.

Some diy builders have also sighted ill effects of over liberal use of low value McMillian resisters.

I have for a long time wondered how to resolve this issue.

Well the other night while listening to my Pass diy amps with a nice glass of red an idea popped into my head.

After studying and monitoring my own X Aleph I have observed it would seem the current source for the input Pair is causing the bias and absolute voltage shift with temperature.

With some application of lateral thinking it occurred to me that the effect of the heating of the resisters to ground could be used to thermal track the biasing of the current source

So what I have jigged up is a 10 k thermister with a negative temperature coefficient that monitors the output resister to ground temperature (the ones that get hot) as a result of the absolute voltage swing negative at turn on.

When the resisters heat up it forces the thermister to get hot and its resistance reduces proportionally in the short term. The actual thermister is strapped across biasing resister R24 for the 9610 fet current source.

As the resister cools the thermister resistance rises and we can a control system.

As the amplifier heats up overall the thermal heating and changes in temperature continue to be tracked by the thermister. The real advantage of the thermister is the management of the large absolute offset from a cold start.

The effect is a feedback system that monitors and maintains an equilibrium of the absolute voltage offset.

In my initial experiment using 75R resisters to ground and 22k McMillian resisters the offset voltage was stabilised from an inital reading of 12 volts to less than 5 volts within 1 minute and within 5 minutes its within -+0.5 volts

I will monitor the compensation scheme over the next few days and publish details. I felt it not necessary to publish a schematic because it is so simple. Just make sure the thermister is themally bonded to the output resisters to ground

The beauty of the idea is its simplicity

Ian

Patent pending : Mack's Absolute Offset Compensation c2008

[gl]

That's a great idea Ian. I don't know if will work for me though since the 100R resistors I use are the aluminum cased, chassis mounted Dale type that don't really heat up much that I've noticed. I will investigate though. It seems like too promising an idea not to try.

Cheers,
Graeme

[Zen Mod]

what's about different CCS arrangement , less temp dependable ?

I'm always using plain two bjt thingie ; nice thing about them is that each one is pulling (tempco wise ) in different direction ....
so - when temp connected , they're pretty stable .

[macka]

Thanks for the replies.

Its just an idea and as we know all sorts of things go on in diy.

I am still messing with it.

One thing I noticed is that Grey's 33R or to ground seems to damp the offset better then 100R

A lot of thngs effect the offset like the output stage bias

I wonder what Nelson's preferred value is for R to ground?

Ian

[tryonziess]

Yes it would be nice to know what a lot of Nelson's resistor values are.
The offset issue and a little lack of knowledge is what has kept me away from this project to date. Please post the results of your more in depth tinkering on this. Thanks Tad

[macka]

Probably a bit late in they day to here from the Master and Cyclotrone man .

Its nearly wine o'clock!

Ian

[macka]

Oe issue I have realised is that when the offset passed through 0.00 volt and start to track negative the thermister acts again and pushes the negative voltage more.

So I need to level the action with a series resister or perhaps a postive coefficient resister in series with the other thermister attached to the heat sinks to counter that effect.

Ian

[macka]

I found that using some what lower value MCmillian resisters of 11K and reducing the resisters to ground for 34R without the thermister was the most stable and settled quickest.

22K McMillian resisters to not reduce the offset quickly at all

I should point out my X Aleph has Jfets and I have baised the output stage for 0.320 volts across 0.5R source resisters using 16 ouput power fets per channel.

If Graeme has time I would like to compare notes with his offset experiences.

Ian

[Nelson Pass]

Originally posted by tryonziess
Yes it would be nice to know what a lot of Nelson's resistor values are.
The offset issue and a little lack of knowledge is what has kept me away from this project to date. Please post the results of your more in depth tinkering on this. Thanks Tad

There is not much point in you guys doing exactly everything the same as I do. Many
of my choices are arbitrary. If I thought your choices were problematic, I would say so.

Currently we are using using the UGS5 for the front end of the XA amplifiers. Its offset
does not drift much, so we don't have a compensation scheme for it.

The most directly effective method is probably the scheme that you see on the internal
circuit of the THS4131, where the current source for the diff pair is adjusted against the
common mode output DC. This is also the technique seen in the F1 current source amp.