4 gain stage comparison

[tubenit]

I found this (reportedly) Jim Kelly preamp design interesting.  It has a James tone stack after the 2nd gain stage, some unusual component values,  solid state diodes before the 2nd gain stage all into a concertina phase invertor.
I have NO idea if the JK preamp is drawn correctly or not?

Anyone know what the diodes do?  Note the direction of how they are drawn.  And what about the 5.6M resistor?  Is that some type of local feedback?  IF not, what purpose does it have?

Compare the 4 gain stages into a concertina phase invertor (JK amp) to the 4 gain stages into LTPI.

Interesting amp!  I have no desire to build it but I am intrigued with the design.  Reportedly, there is a quad of 6V6's running at 480+ volts on the plates. Wow!

With respect, Tubenit

[PRR]

The 5.6Meg AND the 1.5Meg set cathodyne grid DC bias to 0.21 of supply.

Perfect bias for a perfect cathodyne is 0.25.

With the im-perfect 12AX7 (significant internal resistance) we might figure a bit less than 0.25, maybe 0.22 or 0.20. There's no point in shaving the baloney too-too exactly; 5.6+1.5 was handy and plenty close enuff.

This scheme needs clean power at B+ point (C).... uh, mmmm, well, actually it has about the same supply crap rejection as any other cathodyne+driver.

V1B grid clips/clamps at 1.36V positive but can swing negative infinity. V1A can probably toss 25V-50V through a full-up volume pot to V1B grid. If +/-25V is clamped at +1.36V, the average shifts to about negative 23V. This will cut-off V1B hard, "blocking".

A million+ amplifiers work fine that way.

This "JK" clamps V1B grid at negative 1.2V, roughly symmetric to the +1.36V grid clamp. V1B will stay in happy bias up to outrageously huge levels from V1A. Happy but distorting like a new Corvette shoved through a 2 foot pipe.

[tubenit]

As I am understanding it .......... the first four stages of the JK amp are giving a reasonably "clean" signal with the power amp creating more of the distortion

VS.

The first four stages of the Dumblish amp creating overdrive into a relatively clean power amp

Is that correct? 

With respect, Tubenit

[HotBluePlates]

As I am understanding it .......... the first four stages of the JK amp are giving a reasonably "clean" signal ...

EDIT: The explanation below is wrong; the diodes would need opposite orientation to do what I describe below. See post further down.

Look again at the diode clamp on the JK's V1B.

PRR said, "V1B grid clips/clamps at 1.36V positive but can swing negative infinity." This refers to the 1.36v of bias at V1B, and pretends the diodes aren't there. This is the point where the incoming signal's positive peak drive the grid-to-cathode voltage up to 0v, so grid-current at V1B and the tube clamping (blocking) the incoming signal.

The two diodes clamp the incoming signal in one direction. You should read "clamp" as equivalent to "clip" and don't allow the incoming signal to exceed that voltage, but only on one side of the waveform. So there's assymetric distortion at V1B's grid if you input a big enough signal at the input jack and turn up the volume control enough.

Assuming local feedback due to V1A's cathode resistor cuts gain down to 25, the clamp/clip happens at a combined input jack signal & volume control setting of 1.2v/25 = 0.048v. By that I mean if the volume is full up and a 48mV peak signal is applied to the input jack, the amp distorts at V1B's grid.

This may seem like it's no better/different than the natural blocking/clamping at/before 1.36v peak due to V1B bias. However, the diodes clamp rather than the clamping being due to grid current, so the resulting distortion shouldn't be "farty" or sound like typical blocking distortion (due to a delayed recovery from grid clamping due to grid current if no diodes were present).

If other circuit voltages were known, it might be possible to determine if the diode clamp will prevent grid blocking in later stages.

Separately, I'd call this a "3 gain stage amp" and not count the stage before the split-load inverter, since it is within the power amp feedback loop and contributes to the swing needed to drive the output tubes.

[VMS]

Schematic:

http://www.jimkelleyamplifiers.com/schem.htm

There used to be more that maybe can be found with wayback machine.

Also TUT5 has a chapter of this amp.

[2deaf]

However, the diodes clamp rather than the clamping being due to grid current, . . .

If "clamp" is meaning the same as "clip", doesn't the grid current continue to clip the positive signal while the diodes clip the negative signal?

[2deaf]

The schematic that VMS referenced shows two diodes on the grid of V3a.  These diodes are crucial because there is no use in making a nice, symmetrical clip coming off of V1b if you are just going to have the same problems further down the line.

[Willabe]

Side note;

I read an interview with Bonnie Raitt where she was raving about using a Jim Kelly amp in the studio that she had borrowed from Jackson Brown. 

[kagliostro]

http://diyourself.ru/play/add-diode-clipping-distortion-to-your-guitar-amp.html

under Step 4: Inserting the circuit the second image seems your amp circuit

However if the circuit is the same or not I think the purpose of the diodes (clipping distortion) is the same

Ciao Jeff

Franco

[HotBluePlates]

However, the diodes clamp rather than the clamping being due to grid current, . . .

If "clamp" is meaning the same as "clip", doesn't the grid current continue to clip the positive signal while the diodes clip the negative signal?

Yes, thanks for the correction!! The diodes would have to be cathode-to-ground and have a voltage reference to prevent grid current the way I described it.

As 2deaf notes, grid clips the signal in the positive direction while the diodes clip it in the negative direction. The signal input would need to be ~1-1.36v (peak) in the positive direction and 1.2v (peak) in the negative direction to resulting in clipping both sides of the wave.

So these diodes don't prevent grid blocking the way I described; the volume control, if turned down, might limit some blocking at this stage's grid. Other resistors may help limit blocking at later stages.

Schematic:

http://www.jimkelleyamplifiers.com/schem.htm

And we have a similar schematic in the Hoffman Library here.

[Willabe]

http://el34world.com/charts/Schematics/files/_Miscellaneous/jimkelly_facs.pdf

[2deaf]

The number of diodes used isn't overly critical and two diodes may not be optimal.  I used four diodes in series with a 7.5K resistor on the grid of the second stage and again on the grid of the third stage of a three stage pre-amp a couple of decades ago.  I apparently thought that was optimal at the time.

An interesting thing happens when you limit the negative swing on the signal and the average that PRR mentioned moves up.  It now takes less signal at the input to achieve grid current clipping at the second stage.  Or said another way, the same input signal will clip more at the second stage.  The hiss with high gain pre-amps comes from the first stage and its associated parts, so a higher clip-to-signal ratio can allow you to reduce the hiss by reducing the amplification needed for the same amount of overdrive.

The amp that I built had way less hiss for the bite, but it also had that diode sound to it.  Nowhere near as bad as up-down diode clipping somewhere along the signal path, but enough to keep me from pursuing it.

BTW, the schematic that VMS referenced and Kagliostro pointed out has a bias on the PI that looks non-functional to me.

[PRR]

I have assumed the diodes "point up". I had to re-check that, make sure I wasn't thinking upside down. Of course we never know about our sources; as 2deaf says, obviously mistaken details slip in.

> The number of diodes used isn't overly critical and two diodes may not be optimal.

Yes.

> An interesting thing happens when you limit the negative swing on the signal and the average... moves up.  It now takes less signal at the input to achieve grid current clipping at the second stage.  Or said another way, the same input signal will clip more at the second stage.

Yes.

The difference is quite clear at high/medium level, visually. What it does for "sound" is murkier.

Thinking about it is a poor excuse for blowing the big 27 cents and *trying* it. You don't really need any added terminal strips or PCB pads. Twist several diodes end-to-end, tack one or more from grid to ground. You can try 1 or 7, up or down, with some-K added resistor, faster than you can type.