Single or dual differential?

[Bricolo]

Originally posted by john curl
First, think about removing ALL the current sources and just using 1 resistor between the source pairs on the input. Second, this is actually a very stable design, because it has almost no gain, but 1/2A Hitachi devices can be problematic, because they are not well matched in this situation. The P channel looks like a triode, but the N channel looks like a pentode. Too much 2'nd harmonic. How do we fix this?

Mr Curl,


Cascoding could work good, but would'nt be easier to use other output devices? (better matched ones)

Could you explain me why it has low gain? I thought that dif pairs had a very high gain. Are dual differentials different?
So, can hitsware's schematic be used as if, without feedback?

[fdegrove]

Hi,

Trolling the solid state section again? .......................

Guilty as charged, sir....

[hitsware]

Originally posted by john curl
First, think about removing ALL the current sources and just using 1 resistor between the source pairs on the input.

Wouldn't that make it class B ? I.E. at quiessence there would be a virtual ground at the center of the resistor and grounds on the gates (assuming the gates are tied together (+to+ & -to-)

[hitsware]

Originally posted by hitsware


Wouldn't that make it class B ? I.E. at quiessence there would be a virtual ground at the center of the resistor and grounds on the gates (assuming the gates are tied together (+to+ & -to-)

Duuh ..... I keep forgetting jfets are depletion devices ... Still sort of a nebuosity(sp?) about the 'virtual gnd' between the halves, but ..... (1/2 way between + and -) = 0

[hitsware]

As an output stage ..... NOT a line amp .....

[Fred Dieckmann]

Change the resistor connecting the sources of N pair and P pair of JFETs to three delta connected resistors with the bottom resistor between the sources of the P JFET pair. Adjusting the value of this resistor will change the degeneration and gain for that half of the circuit to compensate for the N channel MOSFET's greater transconductance than the P channel MOSFET. The parallel combination of the other two resistors will set the bias current of the JFETs independently of the value of degeneration resistor for the P pair of JFETs. You also need to cascode the JFETs for supply voltage above about +/- 20 volts. R1 and R3 will set the bias for the the front end and output stage and will be something different than value shown on the schematic below.

[john curl]

Actually, this makes a pretty lousy power amp, BECAUSE there is no loop gain with a low Z load. Makes a darn good line amp, however.
I even made the Grateful Dead line driver to drive the final mix from the mixing board to the stage with a similar configuration. All else being equal, fets are better than bipolars, but sometimes bipolars are useful in this circuit.

[Fred Dieckmann]

Well, maybe not no loop gain. The Hitachi K1058/J162 pair biased at half an amp would give an output stage gain of about 5 for 4 ohms I believe. With two pair of output devices and some fairly low value drain resistors for the JFET front end to drive the MOSFET Gate to source capacitance, I bet you could get 16 or 20 dB of negative feedback. This wouldn't be a great measuring amp but might not sound too bad. I have the parts around...... maybe I will simulate it and build it if the numbers look reasonable.

[hitsware]

Originally posted by Fred Dieckmann
This wouldn't be a great measuring amp but might not sound too bad.

Yes! There is a certain quality to the common emitter(source) type configuration. Sort of a fullness in the low end.......
BUT beware (maybe a sim would catch it) of sudden 'mode shifts' for lack of a better term. Since the gain of the output stage does vary so does the feedback (if used). ALSO I think some sort of "shootthrough" (too much conduction from rail to rail through the outputs) sometimes occurs........mike

[john curl]

Fred's biasing has problems. Not in concept, but in execution. The resistor values are too large, and the differential gain is asymmetrical