F3 Help

[buzzforb]

I recently finished a pair of AJ monoblocks and find myself in audio heaven, but....I seem to be having trouble quitting there. I built these with a great deal of help from a friend and would now like to venture out on my own and build something all by my lonesome. Well, i have already failed because here i am asking for help. My next adventure is based on the F3, or more specifically, the LU Power Jfets. I would like to experiment with them first since they are cheaper than their SS counterparts. I have read the ZEN 8 and 9 articles, Vladimirk's power F3 thread, and EUVL's "another source follower thread". Based on what i have read, I would like to build a standard F3 and then increase power by paralleling devices. THis is where my question comes in. IN Vladimirks thread, he just paralleled more of the cascoded mosfets, but in EUVL's thread there is a passing comment about creating a power amp paralleling both Jfet and mosfet. What advantages are there to paralleling the jfets as well as the mosfets.

[buzzforb]

I have done some reading and perhaps found the answer. In paralleling devices, gain is multiplied, but capacitance is multiplied as well. THe LU Jfets already have high Crss (Miller Effect) and multiples would exxagerate this problem even more. Vladimirk added mosfets to handle heat and at the same time, avoided the capacitance multiplier. EUVL seems to suggest that his Taylor CS cuts the capacitance in half and this is how he would suggestdealing with the increased capacitance from paralleld devices. Hope i am close.

[VladimirK]

Originally Posted by buzzforb

I have done some reading and perhaps found the answer. In paralleling devices, gain is multiplied, but capacitance is multiplied as well. THe LU Jfets already have high Crss (Miller Effect) and multiples would exxagerate this problem even more. Vladimirk added mosfets to handle heat and at the same time, avoided the capacitance multiplier. EUVL seems to suggest that his Taylor CS cuts the capacitance in half and this is how he would suggestdealing with the increased capacitance from paralleld devices. Hope i am close.

Your thinking is absolutely right, big parasitic capacitances of LU1014 is the largest evil preventing us from getting in to sound nirvana. Cascoding the LU in Zen9 schematics helps a lot to diminish the negative effect of parasitic capacitances, but the need of connecting resistor to its gate moves us away from what we have achieved by cascoding. Therefore, if your preamp can work with 3-5kOhms load, better to decrease resistors values in the voltage NFB chain. Anyway, we must drive the F3 and Zen9 with low Zout preamplifier. Also pay attention to output caps. If you plan to increase output power, make 25-30mF output cap, preferrably consisting of a cluster of 1000uF Elna Silmics.

[buzzforb]

have you tried X-ing this amp or is this something you are not interested in? The preamp i have is based on the Borberly article about Jfets. It should be very capable of driving this amp.

[VladimirK]

Originally Posted by buzzforb

have you tried X-ing this amp or is this something you are not interested in? The preamp i have is based on the Borberly article about Jfets. It should be very capable of driving this amp.

I have not tried x-ing the Zen9, but I do not think this could lead to substantial change of sound.
In this schematics, you could try experimenting with NFB resistors, output capacitors, another caps, also with disconnecting of the Aleph Current Source modulation. These are the most sensitive points of the schematics.

[juma]

Originally Posted by buzzforb

....In paralleling devices, gain is multiplied, but capacitance is multiplied as well....

Actually, total effect of paralleling devices is quite different. Please read this:

Mosfet Output Stage Capacitance

[buzzforb]

Thanks Juma. That is very helpful, although I haven't quite got my head around it yet. It seems to suggest that while under certain conditions, capacitance is an issue when paralleling devices, when used in amplifiers where the gate sees higher voltages, the gM ofthe paralleled devices helps cancel out some of these issues. I will have to think on it more. Unfortunately I do not have the quality equipment needed to experiment first hand. I also am having trouble with the concept of capacitance and cascoded devices. Thanks again.

[juma]

buzzforb,
the concept is quite simple - the more transconductance you have, the less Vgs voltage you need to open the Drain-Source chanels of the MOSFETs and with less voltage you'll need less current to charge the MOSFET's capacitances, so the net effect is that the capacitance seems lower.

[buzzforb]

So I have done some reading while traveling. Cascoding the fets not only protects the Jfet's from excess heat, it also helps in reducing capacitance. Capacitance is reduced further by adding a source resistor, taking advantage of degeneration, which throws away gain in exchange for more linear operation and also higher input impedance. From what I have read, the addition of the source resistor can lead to increase noise. Is this only at radio frequency and not a concern for amplifiers? Suggestion for keeping degeneration and reducing noise include the use of inductors or even transistors. If relevant to audio frequencies, could his be a place of potential improvement or is his just unecessary, added complexity? This is probably obvious to most, but I figure hashing it out not he forum might help other newbs gain some understanding. Thanks!

[buzzforb]

Originally Posted by juma

buzzforb,
the concept is quite simple - the more transconductance you have, the less Vgs voltage you need to open the Drain-Source chanels of the MOSFETs and with less voltage you'll need less current to charge the MOSFET's capacitances, so the net effect is that the capacitance seems lower.

Thanks for the explanation Juma. This does indeed simplify the concept. I am not schooled in the physics of semiconductors. I wrote my last post before realizing you had responded. I need to read more about cascaded devices and try to figure out the whether or not t is best to parallel both devices or just one. It would seem that by paralleling just the upper fets, Uou have reduced the drive requirements and a single power jfet is sufficient as in vladimirks amp, but I am still curious as to the suggestion by EUVl to parallel both. Oh well, time for more studying.