The modulated supply and voltage swing (patent #5,343,166)

[rtarbell]

This question is directed towards anyone and everyone, Mr. Pass included.

Attached is the modulated supply picture.

==> If I understand correctly, the positive voltage swing of the output would be about (V++) - 27V (if the output voltage goes any higher, the drain of the upper cascode would hit the voltage rail).

Similarly, the lower output voltage swing would be (V--) + 29V.

Just as a suggestion for improvement:

Eliminate the 27V and 29V sources, and replace them with a resistor divider between the output and the upper rail for the 27V source, and the output and the lower rail for the 29V source, to control the gates of the cascode transistors. Both resistor dividers would have two equal resistors with the midpoint connected to the gate of the respective cascode transistor.

In this way, the voltage difference from V++ to the output is placed on the two NFETs equally, and the voltage difference between the output and V-- is distributed between the two PFETs equally. This would allow a higher voltage swing overall. I know that we are trying tog et the voltage across the inner FETs low (to keep power consumption down).

[lumanauw]

Hi, Rtarbell,

I'm not the author of this patent, Mr. Pass is. But I think "27" and "29" is part numbers, not meant as 27V and 29V.
I guess part #27 and #29 is about 7-10V drop voltage source, if those 4 transistors are mosfets.

[peranders]

Read all about here:
http://www.pat2pdf.org

[djk]

25 is the bias voltage for the main outputs, about 8V for hex-fets.

27 and 29 are the level shift for the cascode outputs, about 10V~15V would work well.

33 is the current source.

[hermanv]

If you go to this site you will find several articles by Nelson Pass. There is one on "Cascoded amplifier design" that outlines what are the advantages of a fixed voltage vs one proportional to the signal across a given transistor. This article helps explain why a resistor divider will have quite a different effect than the fixed offset scheme.

http://www.passlabs.com/articles.htm

Cascoded designs have been around for a long time and like many other designs the devil is in the details. Still, they offer one solution to transistor distortion, so understanding why they work is very useful.

[djk]

That is not what the patent is about.

Maybe you could comment on the patent?

[GRollins]

"27" and "29" are simply labels for voltage sources. It's normal practice for a patent to use numeric labels like that. There is no limitation on the construction of the voltage source; resistors are fair game. In fact, thats almost certainly the method Nelson used, himself (in practice, probably a pot). However, specifying a voltage source in the generic sense allows for things like regulated voltages or perhaps even modulated voltages.
Note that the voltages at 27 and 29 will almost certainly be different due to the real world differences in Vgs between 17 and 19. Thus, it's conceivable that the voltages at 27 and 29 might coincidentally turn out to be 27 and 29V. Don't bet the farm on it, though.
Something to think about...what if you were to replace 11 and 13 with JFETs?

Grey

[hermanv]

Originally posted by rtarbell
(edit....)
Just as a suggestion for improvement:

Eliminate the 27V and 29V sources, and replace them with a resistor divider between the output and the upper rail for the 27V source, and the output and the lower rail for the 29V source, to control the gates of the cascode transistors. Both resistor dividers would have two equal resistors with the midpoint connected to the gate of the respective cascode transistor.

In this way, the voltage difference from V++ to the output is placed on the two NFETs equally, and the voltage difference between the output and V-- is distributed between the two PFETs equally. This would allow a higher voltage swing overall. I know that we are trying tog et the voltage across the inner FETs low (to keep power consumption down).

Originally posted by djk
That's not what the patent is about.

I was replying to the thread starter, rtarbell, not to the patent. The poster recommends making the outside transistor pair follow 1/2 the output voltage swing. While this does distribute the heat accross two pairs and does allow for larger swings, the patent drawing is of a cascode design. i.e. the Voltage accross the output drive transistor is constant with output voltage swing.

By replacing the voltage sources 27 and 29 with a resitor divider the benefits of a cascode design are lost. It is a different circuit topology and while either topology can be made to work well, rtarbell's "suggestion for improvement" seems to have missed the point.

[djk]

"Voltage accross the output drive transistor is constant with output voltage swing."

Items 11 and 13 are the outputs. The supply for them is from item 33, a CCS. It looks like the voltage across 11 and 13 does not become constant until item 33 runs out of current and it comes through items 17 and 19, at this point items 11 and 13 are cascoded by items 17 and 19.

A clever variation on class G.

Did Nelson Pass ever make a commercial version of this?

[Nelson Pass]

Not yet.