Non-feedback parametric linearisation in FETs

[x-pro]

This is my small Christmas present for those who like zero NFB single-ended class A amplification

I've worked in my spare time (for last year and half) on a home project to make a very linear single-ended transistor amplifier without NFB and one of my ideas I've described in a small article on my web site here:

http://www.ant-audio.co.uk/Theory/Parametric_Linearisation.htm

The performance of this circuit, despite the simplicity of the concept, is quite impressive - here is one example:

For a voltage amplifier stage using this idea I've measured distortion levels vs RMS current in the load resistor. Idle DC current was 4.5 mA, for 0.15 mA RMS current THD were 0.0025%, for 0.5 mA - 0.03%, for 1.15mA RMS (that is a swing from 3 to 6 mA (!) ) - only 0.16% .

It is possible to make this approach work even without the source resistor, thought it requires very careful transistors selection. THe circuit of my headphone amplifier (A.N.T. Audio "Amber") includes a voltage amplifier stage based on this idea.

Merry Christmas and Happy New Year!

Alex Nikitin

 

[jan.didden]

Interesting, Alex. Is there info on the ANT headphone amp on the 'net? Does the 'T' stand for "Technology"?

Jan Didden

Edit: found it. Beautifully executed, very professional looking. Where do you get those enclosures?

 

[x-pro]

Interesting, Alex. Is there info on the ANT headphone amp on the 'net? Does the 'T' stand for "Technology"?

Jan Didden

Jan,

look at my web page: www.ant-audio.co.uk . There is also a review of it on head-fi :

http://www4.head-fi.org/forums/showthread.php?t=88528

And well guessed - A.N.T. does stand for Alex Nikitin Technology

Alex

P.S. - enclosure is a standard Hammond case.

 

[darkfenriz]

damn
i cannot see pictures in your link
seems intersting...
could you attach these figures?

cheers

 

[x-pro]

damn
i cannot see pictures in your link
seems intersting...
could you attach these figures?

cheers

Here is the link to a PDF file:

http://www.ant-audio.co.uk/Theory/Parametric_Linearisation.pdf

Cheers

Alex

 

[tschrama]

Hi X-pro,

Best new years wishes for you and thx for the PDF.. looks very interesting .. enough to keep me spice-ing for a week ... at least....;-),


cheers,
thijs

 

[Bricolo]

what's the difference with a cascode?

 

[x-pro]

what's the difference with a cascode?

Hmm - 8-10 times lower distortion? However it is only true for a specific combination of the "upper" and "lower" transistor parameters and for the right value of the drain current.

Alex

 

[Hennie]

Very interesting indeed. To avoid opamps some I/V converters for DACS use valves. They rely on something like a 50 ohm resistor to ground followed by a valve voltage amplification stage. Would you say this topology is suitable for this purpose? That would be great, because few CD players have space for valve stages.

 

[x-pro]

Very interesting indeed. To avoid opamps some I/V converters for DACS use valves. They rely on something like a 50 ohm resistor to ground followed by a valve voltage amplification stage. Would you say this topology is suitable for this purpose? That would be great, because few CD players have space for valve stages.

This topology can be used in many applications, including this one . For 50 Ohm and, say, FS 1 mA RMS output current of the DAC, the maximum input voltage would be 50 mV - near the "sweet spot" for this circuit, with a possibility of a very decent performance - with the distortions about 0.002% (mostly 2-nd order) and SNR about 100 dB.

Alex

 

[Hennie]

Thanks for the reply. I think that it is something that definitely merits a closer look.

 

[mikeks]

Thanks Alex...

 

[smoking-amp]

Hi Alex,

Interesting circuit. Isn't the top Fet still operating in the pinch-off region with high output impedance?

Don

 

[Cortez]

This method works with BJT's too ?

 

[x-pro]

Hi Alex,

Interesting circuit. Isn't the top Fet still operating in the pinch-off region with high output impedance?

Don

Hi Don,

The top FET is operating with high output impedance, and this is a good thing - as it allows for high gain if necessary, and if you need a lower output impedance you can just connect a smaller resistor as a load.

Alex

 

[x-pro]

This method works with BJT's too ?

No. BJT's are current contolled devices and this method won't work with them. In theory the "upper" FET could be replaced by a valve, however I didn't try this (yet) .

Alex

 

[Bricolo]

and with mosfets?

 

[x-pro]

and with mosfets?

Why not? Perhaps not with all MOSFETs equally well and there are possible problems to be addressed with power FETs - i.e. input capacitance would be high with such a low drain voltage, biasing arrangement would not be trivial etc. However I plan to try this idea for a SE MOSFET power amplifier when I'll have time. As you may see from my first post, it is possible to have low distortion even when the drain current amplitude is very high (in relation to the idle DC current) , so this approach would be useful in a SE power amplifier.

Alex

 

[Ultima Thule]

Very Nice Alex!

Have you used same devices for both upper and lower FET's?

Which device is benefiting most of your approach, the upper or lower FET, or maybe this approach can't be divided in to which device is benefiting of it?

Could this be done with lower device consisting of a FET and upper device of a BJT, and even vice verse lower BJT/upper FET?


Michael

 

[x-pro]

Very Nice Alex!

Thank You!

Have you used same devices for both upper and lower FET's?

It is possible, however I've found that with a right selection of two different devices the linear part of the characteristic could be made longer

Which device is benefiting most of your approach, the upper or lower FET, or maybe this approach can't be divided in to which device is benefiting of it?

In this circuit the "lower" FET is the beneficiary, the "upper" FET is working hard to provide this benefit .

Could this be done with lower device consisting of a FET and upper device of a BJT, and even vice verse lower BJT/upper FET?

Michael

Not really, as you need a specific fit of two curves to linearise the transfer characteristic of the "lower" device. With FETs, both JFETs and MOSFETs it works, however BJTs are completely different beasts

Alex