Watt Sucking Fireball Series

[Roadbagger]

Originally Posted by carpenter

Can you teach me how to replicate that jfet front end? It doesn't seem right to just copy... I'm very impressed!

I don't see anything wrong with copying, it's a good place to start while I work up a reasonable tutorial beyond broad generalities. Not sure how deep to go, it just takes some composition time.

K-wood

[carpenter]

Originally Posted by Roadbagger

I don't see anything wrong with copying, it's a good place to start while I work up a reasonable tutorial beyond broad generalities. Not sure how deep to go, it just takes some composition time.

K-wood

Whatever you decide to write, will be excellent. I'm a fan of choke loaded power differentials (my latest endeavor was at the BAF this year--the big fellow that took two people to lug to the listening table...) and would enjoy learning more about employing the current/voltage controls you used in your design.

Thanks for what you've shared so far.

[Roadbagger]

[QUOTE=scott wurcer;1968184]

Originally Posted by Roadbagger

Ok go for it, hobby businesses are at least fun.

Definitely fun, however, where it may be potentially true for the Audio business, though I have doubts about that as well, it definitely is not the case with industrial and scientific analog circuit design and implementation. The complex nature of an IC opamp and it's convolution of gains and nonlinearities over input to output drive and voltage swing can never be fully anticipated and straightened by a complex compensation. Most nodes of interest are not brought out on a pin, at least for me. Generally noted is the trend to need massive bandwidth with highly fed-back circuitry to maintain transient linearity while highly linear low loop gain parts require far less feedback and have measurably clean transient response with little or no compensation. Many times a hybrid of the two are used to utilize the DC acuity of the op amp and the linearity of a discrete. But in audio, DC acuity is a moot point, and since all of interest is AC, the loop gain argument tends to fall away in favor of a high linearity, low loop gain, to near zero feedback approach. When the discrete parts are both linear and wide bandwidth throughout, the phase distortion (group delay vs freq) is normally a non perceptible. This, being a primary influence in the transient IM, becomes evident in the air and image around a reproduction of properly recorded music and voice. ( that's my "hobby" voice talking) So, though having a long history of using op-amps and discretes for everything from particle beam metrology, to really clean Bass amps, I will probably prefer a highly linear low loop gain approach to SOUND.

K-wood

[john curl]

Hi K-Wood, I'm glad to see your input here.

[Roadbagger]

Originally Posted by jacco vermeulen

Mr Rough,

do you mind telling what type of JFEts are used for the driver section ?

A version with LU1014D for the Sony power devices would be fun.

Well, a few things would be the same and a few would need to change. The differences between the 2SK60 and the LU1014D are significant in a couple of ways. The input capacitance of the LU part is 10 times the input capacitance of the Sony part, so the driving impedance would need to be significantly lower in order to maintain reasonable slew rate for fidelity. The reverse transfer capacitance Crss is also rather enormous for the LU part and it would seem to necessitate output cascoding to reduce that miller effect on bandwidth and phase distortion. the LU type parts are very different than the vertical grid channel devices built in the 70's and in the effort that I was involved with in the 90's. So, they just need a bit more gate drive energy and circuit finess over frequency to be truly transparent. And you're right, it would be an interesting experiment. One other configuration that I've done in the past used Supertex depletion mosfets which are kinder to the designer with regard to the critical capacitances.
K-wood

[Roadbagger]

Having a penchant for evolving tube circuits into silicon emulations of the same tends to make the signal path all majority carrier type devices,(N). So just to show that there is no prejudice here against minority carrier devices, (P), here is another watt sucking fireball amp made of 5 cent transistors and capable of >30watts. There is also the need to heatsink the face of every TO-92 to an aluminum plate which makes the mechanical layout interesting at best. The circuit is a self biased transistor array follower, modified to include gain by ratioing some negative feedback. Pretty clean and definitely cheap to build. The .01 caps are .01F or 10,000uF 35V each.
K-wood

[Roadbagger]

This version of the WSF is the modern variation of the original design done 40 years ago in my parent's basement where a great deal of lernin' happened. This amp uses a MP5100 50 amp germanium transistor in series with a 4 ohm resistor as the collector load. 3.75 amps is the quiescent current for this output stage. The original had a finessed centering bias and a 5000uF coupling cap to the speaker, but still with the same results. I added the DC servo stuff on the later version because the feet just wont stay still. Though not terribly efficient, it has a remarkably clear response through the full spectrum. The drive is curiously booted to the output for bandwidth.
K-wood

[Gordy]

Thanks for sharing your approach and experiences. This looks really interesting, and I am looking forward to learning more.

[scott wurcer]

Originally Posted by Roadbagger

tube circuits into silicon emulations of the same tends to make the signal path all majority carrier type devices,(N). So just to show that there is no prejudice here against minority carrier devices, (P),

Majority and minority carriers can be either N or P. JFET's are always majority carrier devices be they N or P channel.

[Roadbagger]

Originally Posted by scott wurcer

Majority and minority carriers can be either N or P. JFET's are always majority carrier devices be they N or P channel.

Not to be pedantic about it, that was a slight of hand comment for color but even the device designers will often call out in term, this: "A device in which the current is conducted by the charges dominant in the lattice is called a majority carrier device (e.g., electrons in n-type material, or holes in p-type material if the current is conducted by charges not dominant in the lattice, the resulting device is called a minority carrier device " (IRC) Though this is most often used in the description of channel devices like FET's, where current flow is conducted through a single type of semiconducting material and not passing through any junctions, the slight of hand was extending this to include "bipolar" junction devices for polarity sake. In the case of initial implied reference to WSF No.4, the entire signal path is N-channel devices and might be said, tongue in cheek, to be a majority carrier amplifier. Die sizes for "complimentary" devices are a good illustration of the difference. The P channel die compliment to it's N equivalent will be three times as large for the same doping regime to have "complimentary" characteristics. There will always be finessed parameters involving doping concentration and area to make the conductivity, capacitance and depletion efficiency match up, so areas can vary from 1.5 to 3 times in ratio when trade-offs are managed. Always will be the argument that there is no such thing as a truly complimentary device structure, and it's a good argument to engage when a Speyside single malt scotch is employed to refine the details therein, but a forum on this subject would turn into a full time job, and I have one of those. So thank you for the correction, I'll try to restrain my loose cannon for a more outrageous subject later.
K-wood