First things first

[jeff]

 Is there an advantage to putting the PI first followed by a gain stage instead of vice versa? I'm thinking you could get a bigger cleaner signal to the power tubes doing it this way. Is this just a waste of a tube section?

[sluckey]

Several amps have used that scheme. An LTP phase inverter will do the same with less circuitry in most cases. Some big multi output tube amps may need a driver between the PI and PA. It's just a different way to skin a cat and some may say it has a unique sound. Go for it if you got the tubes. Here are a couple amps that did...

http://www.el34world.com/charts/Schematics/files/gibson/MERCURY_I_&_II.pdf
http://www.el34world.com/charts/Schematics/files/gibson/TITAN_MEDALIST.pdf

[Mats]

Hi
Here is one more, Kalamazoo Bass 30 amp
/Mats

[overtone]

I see driver stages after the PI on quite a few bass amp designs. Another one would be the Silvertone 1483.

[kagliostro]

Is the 400W Hiwatt the same ?

Kagliostro

[sluckey]

Is the 400W Hiwatt the same ?

It's similar, but the tube between the PI and PA is a cathode follower rather than a gain stage.

[kagliostro]

Oh, a catode follower ................ YES, I see ..............

is that to give more current to the final tubes instead of more voltage given by a gain stage ?

Kagliostro

[HotBluePlates]

More likely, it is to provide a point to inject negative feedback which will not be affected by adjusting the presence control.

[PRR]

> I'm thinking you could get a bigger cleaner signal to the power tubes

Amplifying stages distort about 5% at maximum output.

The cathodyne has large negative feedback, has lower distortion until it overloads, and even though it makes two outputs each output's overload point is nearly as big as an amplfying stage's overload.

Real difference to guitarists is the GROSS OVERLOAD behavior. Not the same for each case. Very useful amplifiers have been built either way, sometimes both ways in different runs of the same model-name.

BTW, the differential pair with differential drive does NOT need or want a cathode bypass. Why?

[Tone Junkie]

Why PRR I dont know but I would like to.
Thanks Bill

[octal]

I think I mentioned this a while back on a different thread, but the cathodyne PI followed by a drive stage is sometimes known as a "Williamson" type circuit. (It was a very famous hi-fi amp circuit.)  Of the top of my head, the Marshall Major was another example of this, as was a version the Danelectro commando. The Standel amp also supposedly used a Williamson type circuit, but I don't think I've seen a schematic for that....

[HotBluePlates]

I think I mentioned this a while back on a different thread, but the cathodyne PI followed by a drive stage is sometimes known as a "Williamson" type circuit. (It was a very famous hi-fi amp circuit.) 

The Williamson amp was much more than just that one circuit aspect. The biggest overriding factor was an output transformer with detailed specifications, which then also allowed increased negative feedback wrapped from the OT secondary to the cathode of the input stage. So the overall design had a number of refinements.

BTW, the differential pair with differential drive does NOT need or want a cathode bypass. Why?

We assume two equal and opposite signals as the input to a differential stage with differential (push-pull) drive.

A cathode bypass cap is used in a typical gain stage to bypass the signal current around the cathode resistor, so that the stage bias stays constant. This also has the effect of preventing local feedback when the cap's reactance is small compared to the cathode resistance.

IF the differential inputs signals are equal and opposite, and the two tube stages are broadly similar (not a 12AX7 triode and a 12AU7 triode), then the signal currents of each triode offset each other. There is no a.c. component across the cathode resistor, therefore no change of bias, and no local feedback. So, a cathode bypass cap has no purpose.

PRR wanted you to think about the purpose of the bypass cap. I just gave you the Cliff Notes. My bad...

[jeff]

Good point, Didn't think of that when I whipped up the schematic.
  
OK, here's another one to think about: A cathode biased power section DOES use a bypass cap, those tubes have equal but opposite signals on their grids. Why?

[frus]

BTW, the differential pair with differential drive does NOT need or want a cathode bypass. Why?

because the signal on the first tube has to be reproduced across the cathode resistor, so that the other tube amplifies the same signal. The grid of the other tube is effectively grounded, the first tube actually wiggles the cathode of the second tube, the only thing the tube cares about is the difference between grid and cathode voltage, so that's how it works. If the bypass cap is present, then both the cathode and grid are effectively grounded, so the second tube wouldn't amplify any signal at all. Am I correct?

[jeff]

because the signal on the first tube has to be reproduced across the cathode resistor, so that the other tube amplifies the same signal. The grid of the other tube is effectively grounded

I think you're thinking of a long tail pair. Take another look at the first schematic. It's a splitload phase inverter followed by a gain stage for the + and - signal. The grid of one tube is taken from the plate of the PI and the other is not grounded but taken off the cathode of the PI and is therefore out of phase to the first.

Read HotBluePlates' comments again. Think of a single gain stage. Since the cathode of the tube has a resistor to ground, as the tube draws more or less current, the voltage across that resistor changes(V=IR). Normally the cathode cap will keep the bias voltage constant by charging and discharging as the current through it changes. But with the two tubes sharing one resistor and each grid getting an equal but opposite signal, as one draws more current the other draws less and the net current remains the same through the resistor so the voltage drop across the resistor remains the same. Using a cap would be pointless because the voltage is constant.

[jeff]

 This raises an interesting question. Is the operating bias voltage of a single gain stage using a cap equal to it's idle bias? As we apply a signal to the grid we're allowing more or less current to flow through the cathode resistor changing the voltage(V=IR). Putting a cap across that resistor will charge at higher voltage and discharge at lower. If a tube is idleing at 1.5V, as more current flows, will that raise the voltage charging the cap(to a voltage higher than 1.5V)?

 So, when operating would the operating bias voltage be above the idle 1.5V.

[octal]

Real difference to guitarists is the GROSS OVERLOAD behavior. Not the same for each case. Very useful amplifiers have been built either way, sometimes both ways in different runs of the same model-name.

BTW, the differential pair with differential drive does NOT need or want a cathode bypass. Why?

This is perhaps an odd question, but would bypassing the cathode impact how the drive/diff stage reacts when the output tubes swing to grid clamping?  I guess the ability of the diff pair to source current would be strictly a function of the value of the plate resistors & B+ voltage, but how about the ability of the stage to pull down the charged coupling caps to the output tubes?

Wouldn't the presence of the bypass cap increase the distortion of the stage? (By eliminating the local NFB of an unbypassed cathode resistor.) Perhaps a good thing an a guitar amp application?

Nathan

[PRR]

> A cathode biased power section DOES use a bypass cap, ... Why?

Follow what HBP wrote: "IF the differential inputs signals are equal and opposite, ..., then the signal currents ...offset each other."

Maybe not the full story. There's an input range, for amp-triodes say zero to -3V. Gain stages are usually worked within this range. POWER stages often push past the cutoff. When one 6V6 grid is driven to -35v to -40V, it is practically an open circuit, isn't equal-but-opposite, isn't "offsetting" the other 6V6.

But cathode caps on power stages are a very complicated issue. Analysis is dubious and does not tell "how it sounds". There are many ways to build, use, and abuse cathode bias power stages. The best path is to try no, 10uDf, 50uFd, and 500uFd, over all the clean and overload outputs you may want to play, and hear how it sounds.

Several popular old amp use no bypass. These may be the ones working at lower voltage (250V on 6V6) and thus such high current that cutoff isn't happening often. (It also saved a buck, and a failure-prone part.) High voltage high fidelity cathode-bias amps will probably want a very large cap, 500uFd or so. Guitar amps often have a cap which gives a short hold-up, down near 25uFd-50uFd, and surely selected by ear.

[HotBluePlates]

This raises an interesting question. Is the operating bias voltage of a single gain stage using a cap equal to it's idle bias? As we apply a signal to the grid we're allowing more or less current to flow through the cathode resistor changing the voltage(V=IR). Putting a cap across that resistor will charge at higher voltage and discharge at lower. If a tube is idleing at 1.5V, as more current flows, will that raise the voltage charging the cap(to a voltage higher than 1.5V)?

 So, when operating would the operating bias voltage be above the idle 1.5V.

Well...

If we're talking about a small-signal stage, and the stage is not being slammed to cutoff or saturation, then the idle bias should equal the operating bias. There shouldn't be a deviation, because that's the purpose of the bypass cap: to bypass the a.c. around the cathode resistor.

Even when there is no bypass cap, the effect of feedback is very fast compared to audio frequencies, so the bias appears to stay steady, even though the tube's output is being limited. The step-by-step explanation is given so we can understand what's happening. You don't really "see" it occur, you just see lower gain/output than you have when the cathode resistor is bypassed.

However... Output tube stages may have rectification effects, which can shift the operating point away from the idle bias point. That's not important to understand at this stage; it would result in small changes in actual power output and distortion vs. calculated power and distortion.

PRR is telling you right about cathode caps being a complicated thing in output stages. Let's just say that an output stage that never strays out of class A (never cuts off) comes closest to operating as calculated with cathode bias, with and without a bypass cap. Amps that just start to veer into class AB generally have some bypass cap, and PRR is right that you want to determine what works best by ear or by measuring power output. Amps that runs far into class AB (most output power) almost never use cathode bias because the big changes in idle current vs. full power current can be large enough to be incompatible with the self-correcting nature of cathode bias. In other woprds, cathode bias could limit output power in these amps.

[jeff]

(Sorry to get off track again but sometimes my quest for knowledge leads to questions off the beaten path)
When I asked the question I was thinking it had to do with when the tubes start to go AB. This brought up another question I wanted to wait for a responce before I posed.

I've seen cathode biased power sections with a switch that disconnects the cap to reduce power. Would it be (more)effective at lower level to also split the cathode resistor?

[jeff]

 Once again my train of thought has come off the track. My plan was to build an amp with a splitload PI so I could check out it's unique sound when pushed hard, but I wanted to put in switches to remove the stage before it and add one after. That way I'd still have the gain but now the PI isn't pushed as hard. I guess it'd be cool too 'cuz I could bypass both if I'm playing quietly.(I've omitted the second channel and the tone for ease of viewing)

Will the second switch work right? I'm not sure if I can just bypass the gain stages and leave the grids connected.

Do I need to disconnect the grids and ground them?

Ground the output coupling caps?

Or is it alright as drawn(leave inputs connected and outputs going to nothing)?

[38Super]

BTW, the differential pair with differential drive does NOT need or want a cathode bypass. Why?

A differential pair, such as a modern phase inverter will not function as a diff pair if'n you bypass the cathode resistor.  In other technologies, a constant current source would be used in place of the large cathode resistor.  You can think of the large cathode resistor as a poor man's current source.  In this context, at DC balance, each side of the diff pair is conducting equal plate current (each side is passing 1/2 of total cathode current).  When one side grid voltage increases above the other's, that side of diff pair conducts more current and the other side conducts less current, since they have the same cathode voltage and the two triode stages have equal (or at least similar transconductance).  This creates a differential swing in the diff pair's plate voltages.

If you "bypass" the cathode resistor, you dont have a diff pair, you have two (relatively) AC independent gain stages with common bias.



rob

[HotBluePlates]

A differential pair, such as a modern phase inverter will not function as a diff pair if'n you bypass the cathode resistor. 

Very good point! I blew right past that.

My plan was to build an amp with a splitload PI so I could check out it's unique sound when pushed hard, but I wanted to put in switches to remove the stage before it and add one after. ... I guess it'd be cool too 'cuz I could bypass both if I'm playing quietly.

The solution is more difficult than the problem. If you needed less volume/drive, it is very much easier to incorporate a switching scheme on a volume control or other voltage divider. I know it doesn't seem as cool or organic. It will however probably work better and be very much easier and cheaper to accomplish.

Other thoughts: Because the split-load has such a large resistance in series with the cathode, it also has huge negative feedback. So, like a cathode follower, there is no gain from the input to either output (call it a gain of 1; it will really be a bit less).

The offsetting result of no gain due to feedback is that the input can now accept an input signal as big as a reasonable output sgignal without serious distortion. I'd argue that you don't need to worry about slamming the input with too much signal. Rather you need to worry about how big of an output signal you need to drive the output tubes, and whether you can make a signal that big from your available B+, given that you need 2 output signals plus extra voltage across the tube.

Don't design sections in isolation. Know the details of your output stage first, including bias, which then drives how much input signal is needed. Only after you know that, determine if you can use a split-load by balancing the needed output against the available B+.

[gldtp99]

Interesting reading---i recently rebuilt a long-neglected Silvertone 1483 that has this type of PI/driver circuit----this 1483 sits in my workshop near my '64 Gibson Titan that also has a similar PI/driver section.
But neither one is exactly the same as another similar take on the subject in two Gibson Atlas heads i have.
Thanks for the info because i've been wondering why these amps are built this way----and i've been thinking about building a channel switcher with this PI/driver set-up for a future project.
But i've been building a few amps with Paraphase PI's lately so i'm more involved with the Paraphase right now.
A friend dropped off two Marshall Jubilee heads for re-tube/service and played an agressive cascaded stage build with a Paraphase PI and cathode biased 6L6GC's---he liked the amp but not for his current 80's hairmetal band----he ended up playing it for a half hour with a big smile on his face, though..................................gldtp99

[rzenc]

My plan was to build an amp with a splitload PI so I could check out it's unique sound when pushed hard,

Merlin has a very nice article about cathodyne P.I. Also, he gives good advise regarding distortion byproducts.

http://www.freewebs.com/valvewizard/cathodyne.html

Hope this helps
Best Regards

Rzenc

[jeff]

OK, so there's a difference between these?
In one you have the current going through the resistor giving the bias voltage and the cap elimination the NFB
In the other you have twice the current going through half the resistance giving you the same bias voltage and as one tube draws more the other draws less eliminating the NBF

[jeff]

I was thinking you'd get a bigger signal before cutoff by putting the gain stage after,because you have more of the B+ across the tube while the voltage across the PI is the B+ minus the voltage across the 56K resistor to ground.

What's the maximun signal you can get before cutoff between these two?

If one has more headroom than the other, I would think, depending on which is first, it would make a difference in the relationship between when your preamp/PI and your power tubes would start to distort.

[HotBluePlates]

You probably can get a larger total output signal with a driver stage after the split-load inverter. The real question is do you need it?

Look hard at actual amplifier circuits that use a split-load. Pay special attention to the amount of bias on the output tubes, as that dictates the peak input voltage needed to drive the output stage.

You see the split-load inverter in the smallest push-pull amps (Fender push-pull amps with 6V6's) and in the biggest (Ampeg SVT, Marshall Major). In most cases, there is negative feedback from the OT secondary to the gain stage ahead of the split-load inverter. I have a habit of looking at that gain stage as part of the inverter, because it makes up much of the gain a split-load doesn't already have, and is a feedback injection point (which allows the feedback to have an impact on the inverter).

So gain stage before or after the split-load? It seems to me that amp makers have always used a stage ahead of the inverter, and in some cases another after.

Again, the necessity of the extra gain is determined by how big a signal is needed by the output stage. In a properly designed amplifier, the output stage runs out of clean headroom before anything else, so the amount of drive to the phase inverter is already settled by the designer. The exception is a master volume amp that seeks to purposely distort the preamp or phase inverter first.

[jeff]

You probably can get a larger total output signal with a driver stage after the split-load inverter. The real question is do you need it?

Right that's the part I'm missing. Thanks.