Cathode bias, bypassing, frequency responce and voicing.

[Koreth]

So I've read the 1st chapter of Merlin's book available on his website, and have been playing around with the load line plotter spreadsheet(modified to add the -4 and -4.5 grid curves for the 12AX7), plugging in various values found on various Mesa and Marshall schematics and seeing what effects. It's quite fun and all very informative, even if I don't fully understand what all the pretty frequency response curves or load lines mean.

I do have some questions though. In his book Merlin talks about biasing a stage warmer or cooler, meaning closer to grid-current limiting or closer to cutoff respectively, which I can wrap my head around. However, I'm not sure what's meant by center biasing. I realize this means putting your bias point in the center of the operating range of the tube, but which center? When looking at the curves for a 12AX7 with a load line plotted, it seems there's a couple centers one could shoot for. One could shoot for the center of the voltage swing you'll get across your plate resistor. Or you could bias in the center of the available voltage swing of the grid. Due to the compression and expansion of the grid curves, these two points aren't ever going to line up.

An example: Lets take a HT of 350, Ra of 100kΩ. Voltage can go from approx. 110V at grid-current-limiting to 350V at cutoff, for a swing of 240V. The center of this range would be at 230V, 120V from either form of clipping with a bias of somewhere around 1.75 Vgk, which is what you get with the usual 1.5KΩ cathode resistor. From the anode voltage swing perspective, this is center-biased.

However, the 100kΩ/1.5kΩ arrangement isn't center-biased when you look at the voltage swing available to the grid. With 100kΩ/1.5kΩ It takes 1.75 volts to swing up to grid-current-limiting at 0Vgk. But to swing down to cutoff, 1.75V down to -3.5Vgk isn't enough. We need another volt to get down to -4.5Vgk to reach cutoff. -1.75Vgk when the range is from 0Vgk to -4.5Vgk isn't centered. To have the valve biased in the center of the available voltage swing of the grid, the 2.7kΩ resistor found on the input stage of many Marshalls gets you a lot closer, with the valve biased now around 2.25Vgk. Playing with the HT in the spreadsheet, this seems to hold true however big or small the HT is.

So which is the correct way to center-bias? Or should I just split the difference with something like a 1.8KΩ or 2.2kΩ cathode resistor? No matter how look I at it, 'center-biased' isn't, and each is going to have a different clipping characteristic.

One thing I noticed, the spreadsheet however doesn't seem to account for changes in the internal pate resistance (ra) caused by changes in HT. Maybe said changes aren't large enough to be significant, I dunno.

Now which has the greater effect on the voicing of the amplifier? Is it the frequency response shaping caused by the cathode resistor and bypass cap values, or is it the clipping and compression characteristics of the load line and the amp's bias point? If you were to take two preamps trying to cop a Marhsall sound, but with slightly different circuits. The first has the same load lines and bias points of the Marshall, but different frequency response due to changes to HT(and thus ra), Ck, and elsewhere in the preamp circuit. The second has different loadings and biasing from the Marshall design, but has somehow gotten the same frequency response. Which is going to sound more like a Marshall?

[Steve Conner]

Which is going to sound more like a Marshall?

Whichever one you plug into a 4x12 cabinet loaded with Celestions.

Frequency response affects tone, but the non-linear effects associated with tube bias points also affect tone in a different way. When designing an amp, you can tweak both to your taste.

People often refer to the former as "voicing" and the latter as dynamics, feel, touch sensitivity, whatever, although they're not completely independent of each other: the harmonics generated by non-linear effects change the the perceived frequency response, and frequency response shaping interacts with the generation of harmonics, depending on which stages you do it in.

We could spend hours and hundreds of column inches trying to explain the theory, or you could grab some tubes and a handful of components, build some circuits and hear the effects for yourself.

(Merlin can answer all the other questions )

[Wilder Amplification]

So you've discovered that it's impossible to find a bias setting that allows for bot input as well as output symmetry on a 12AX7. Pretty interesting isn't it?

The "voicing" has to do with the frequency response of the stage, which the cathode bypass cap has a direct effect on as it sets which frequencies will get more gain than others. The output coupling cap affects what frequencies get to pass onto the rest of the amp circuitry in greater amounts than others.

The clip point of the stage has more to do with dynamics/headroom control.

Yes...both of these things affect "what you hear". But the "voice" of a stage is all about frequency response whereas the dynamic character has everything to do with headroom and gain, which also has a direct effect on the "feel" of the stage.

Of course, a hotter biased stage will be easier to push into positive grid limiting, which also leads to blocking distortion and doesn't feel or sound very good.

[Joey Voltage]

I do have some questions though. In his book Merlin talks about biasing a stage warmer or cooler, meaning closer to grid-current limiting or closer to cutoff respectively, which I can wrap my head around. However, I'm not sure what's meant by center biasing. I realize this means putting your bias point in the center of the operating range of the tube, but which center? When looking at the curves for a 12AX7 with a load line plotted, it seems there's a couple centers one could shoot for. One could shoot for the center of the voltage swing you'll get across your plate resistor. Or you could bias in the center of the available voltage swing of the grid. Due to the compression and expansion of the grid curves, these two points aren't ever going to line up.

Okay, I'm going to start off by saying this: I think you are over thinking it a tad, which is good! I think you can apply the term center biased quite liberally, and anything in a particular range is perfectly acceptable, a good rule of thumb in many circles is usually biasing about 2/3rd the HT for a center bias.

An example: Lets take a HT of 350, Ra of 100kΩ. Voltage can go from approx. 110V at grid-current-limiting to 350V at cutoff, for a swing of 240V. The center of this range would be at 230V, 120V from either form of clipping with a bias of somewhere around 1.75 Vgk, which is what you get with the usual 1.5KΩ cathode resistor. From the anode voltage swing perspective, this is center-biased.

However, the 100kΩ/1.5kΩ arrangement isn't center-biased when you look at the voltage swing available to the grid. With 100kΩ/1.5kΩ It takes 1.75 volts to swing up to grid-current-limiting at 0Vgk. But to swing down to cutoff, 1.75V down to -3.5Vgk isn't enough

Okay you are kind of relying on a visual symmetry, and keep in mind that your also assuming that grid current starts to flow right at a grid voltage of 0V. It does happen a bit before, so you have to keep this in mind. The same is true for "cut off", you start to compress the signal a little bit before you actually bounce off the HT (I Think he even mentions this in the article). rom this you can see that it really doesn't take much to generate overdrive from a typical 12ax7 gain stage. Again you can use the term center biased for anything that puts you close to where a somewhat large, even, swing can be acquired before the onset of distortion occurs.... and even then It's a guitar amp. also if you take a look at your example, you will notice that it sits pretty close to 2/3rd the HT, and in an area where the grid curves are nice and widely spaced, and in a pretty linear region. When you start considering larger signals with the purpose of generating distortion, it kinda becomes a moot point

[Joey Voltage]

Whichever one you plug into a 4x12 cabinet loaded with Celestions.

I dunno, I think the inverse is true sometimes, I often think all my amps sound like V30's

[Joey Voltage]

Of course, a hotter biased stage will be easier to push into positive grid limiting, which also leads to blocking distortion and doesn't feel or sound very good.

Jon,

you got to be careful about this because it is also a bit of an absolute, or deserves expanding upon.

[tubeswell]

I have heard a number of people bandying a plate voltage idling ballpark between 1/2 and 2/3 HT - depending on the plate resistor, and where the cathode voltage is sitting and what sort of tube you're running etc (and of course whether the stage is a 'normal' inverting gain stage or a CF or a PI* etc)

* but of course the plate could be 3/4 of the HT for a cathodyne PI

@ Koreth - you could centre-biasing to mean one of two things I guess:

1) that you aim for a bias point on the DC load line that is sort of in the middle of where the grid curves are the most evenly-spaced apart. Say for a 12AX7 with a HT of 280 and a 100k plate resistor, 1.5V gives 1.5 voltage of positive voltage swing (towards grid current limiting) and more than 1.5V of negative voltage grid swing the other way (towards cutoff), but that the resulting plate voltage swing is still more symmetrical than it would be if you had a 2.5V bias point and were swinging 1.5V up and 1.5V down on the grid. Here the signal with theoretically have less harmonic distortion as long as the grid isn't driven too hard. However this will mean that you will get grid current limiting sooner as the grid voltage swing gets bigger - however this might not be a bad thing if you want to get a smooth entry and exit into grid current limiting; or

2) that you aim for a bias point on the DC load line that is in the middle of grid current limiting and cutoff, even though the plate voltage swing will get more asymmetrical as the grid voltage swing increases.

[Merlinb]

I'm not sure what's meant by center biasing. I realize this means putting your bias point in the center of the operating range of the tube, but which center?

To add to what has already been said, 'centre biasing' is not an exact term; it can mean both centre of the grid curves or centre of the available anode swing.

Often we will talk about 'ignoring the bunched up grid curves', in which case the centre point of the remaining curves and the centre of the anode swing will pretty much coincide.

But if we really were interested in squeezing out the absolute maximum headroom before flat-clipping, then you would have to go for the true centre of the grid curves, not the centre of the anode swing.

There are few absolutes in electronics design.