how to calculate cutoff frequency for Low pass (grid stop)

[mikepukmel]

Some articles, e.g. Aiken's site, talk about grid stop resistors and oscillation. Merlin's book says to choose the grid stops as high as possible without affecting the audio frequencies. I understand the basic idea of a low pass filter, but this one is a lot more complex (at least it seems so). The high school physic LP filter, the signal goes through the resistor and the capacitor is to ground.

For these tubes, there are several capacitors, not sure which are more important: plate to grid1, plate to grid2 etc to ... eventually to cathode. And all combinations.

Is there a straightforward calculation for the low pass cutoff that has the grid stop as a component of the calculation?



Thanks
Mike

[Gingertube]

Short Cut method to get a close approximation.
Take the plate to grid capacitance (from the datasheet).
Note that the input signal voltage is at the grid and at the anode is that input signal voltage x the gain, but inverted, so that there is gain +1 x the input signal voltage across that anode to grid capacitance. That makes that capacitance appear to be gain + 1 x its actual value.
Multiply that anode to grid capacitance value (from the data sheet) by gain + 1 to arrive at the Miller Capacitance - this is the MAJOR determining capacitance.
That capacitance AC wise is effectively from grid to signal ground. Add the grid to cathode capacitance value to that.
Then just use that total capacitance value and the grid stop resistance to get the corner frequency of the low pass filter.

Example:
12AX7 gain stage:
Cg1a = 1.7pF
Cg1k = 1.6pF
Gain say 50
Miller Capacitance = 1.7 x 51 = 86.7 pF
Add Cg1k = 86.7 + 1.6 = 88.3 pF
You can add say 5 pF for wiring strays = 93pF

With a 68K gridstop then that gives you 25.167 kHz as the corner frequency (1/2 pi R . C)

In a HIFI circuit you may see a small cap across the gridstop to boost high frequency response.

Also be aware that when you overdrive the stage the gain drops dramatically and the Miller Capacitance drops too.
That is why you will sometimes see (in a high gain design) a small cap wired from grid 1 to signal ground to swamp the Miller Capacitance and give more consistent results in overdrive.

The gridstop resistor works to damp the tuned circuit from grid lead INDUCTANCE and the resulting supression of parasitic oscillation is not really a result of the lowpass filter effect of the gridstop.

Cheers,
Ian

[SoulFetish]

One other thing to consider is which stage you are talking about in this case. For instance, if you're talking about the input of the amplifier, then a 68k resistor is going to contribute a large amount of resistor noise which will only get amplified along with your signal voltage throughout each consecutive stage.
So, there are two main things which need to be worked out: First, most of the problematic RF is going to enter the amplifier traveling along the shield of your guitar cable. So your going to want to provide a low inductance path to the chassis RIGHT at the input of your amplifier. This can be done by connecting a .01µF capacitor from the sleeve terminal on your input jack to the chassis with as short a connection as possible. The other thing is to use as small a resistor value as you can in order to achieve a balance in stability and low noise in this stage.

[trobbins]

One one other other thing thing is to appreciate that the input voltage signal may come with a substantial effective resistance (from the driving stage) that adds to the input stopper resistance of the next stage. That could noticeably increase the R in RC, and hence you may observe a lower corner frequency than anticipated.

[mikepukmel]

Cool, cool, thanks Ian Greg, and TR.

Re stage: basically how to ballpark each stage. First, since the original JTM45 schematics don't have grid stops, what a good value to start with for the output tubes, and why. Then, for my amp with reverb, that thing seems to be too close to the verge of going into oscillation. Id like to reduce the chance by putting grid stops around the reverb area (send/return).

For output tubes (6V6 and KT66)
The Fenders had, what 1K5 and some Marshalls 5K6 on output tubes.

I'm assuming because the capacitance is different and the tubes are pentodes rather than triodes, why we don't see grid stops on the order of 68K but an order of magnitude less?

Also, for Fender and Marshall, they didn't put grid stops on the second preamp stage tubes, only the inputs on the first preamp tube. For manufacturers, extra parts mean more cost, but for me, tinkerer, if it doesn't hurt the tone (and above 25KHZ seems plenty high up there for guitar) Id put them in if it helped to reduce the chance of something causing an oscillation, namely my less than stellar lead dress.

[bob p]

the reason that you see LP filtration on at the input to the amp, rather than on every preamp stage, is to provide RF rejection at the amp's input. As mentioned previously, the guitar cable can act as an RF antenna. You definitely don't want an RF antenna connected to the input of your amp, so all amps use a low-pass filter that's designed to reject high frequency RF content while allowing LF musical content to pass.

one RF rejection filter at the input is all that you need because the antenna is on theoutside of the chassis. your second preamp stage is presumably mounted inside of a grounded chassis that prevents any stray RF from entering the circuit.

[SoulFetish]

Cool, cool, thanks Ian Greg, and TR.

Re stage: basically how to ballpark each stage. First, since the original JTM45 schematics don't have grid stops, what a good value to start with for the output tubes, and why. Then, for my amp with reverb, that thing seems to be too close to the verge of going into oscillation. Id like to reduce the chance by putting grid stops around the reverb area (send/return).

For output tubes (6V6 and KT66)
The Fenders had, what 1K5 and some Marshalls 5K6 on output tubes.

I'm assuming because the capacitance is different and the tubes are pentodes rather than triodes, why we don't see grid stops on the order of 68K but an order of magnitude less?\

Well, let me answer it another way so that I don't confuse us all with conjecture trying to explain why fender and marshall did anything. Ignoring traditional component values for a moment, what you want to know is that you're right about triodes and pentodes having different capacitance at the input. Triodes are affected by miller capacitance, in that, there is nothing to shield the grid from the anode. So once those frequencies are "lost" via their inter electrode capacitance, only those lower frequencies unaffected can be amplified. Thereby, increasing the effective capacitance at the output.
Pentodes, on the other hand, were invented and are build to screen the grid from the anode. So you don't suffer the effects of miller capacitance. Even though marshall and Fender used low value resistors in output stages with Tetrodes and Pentodes, you could uses much higher values with very little loss in HF content.
Often it is beneficial in mitigating blocking distortion and instability overdriven stages.
For example, I use 51k grid stoppers on the cathode followers and EL84s in my output stage with no ill effect on high frequency content.
An EL84 has about 10.5pF capacitance from grid to all electrodes(except anode), and the anode to all (except grid) of 6.8pF. The grid/anode has a max capacitance of 0.5pF. Without the drudgery of figuring out Pentode gain, lets just say we're dealing with 25pF of input capacitance (we're just saying). With a grid stopper of 51k, we get a corner frequency roll off of 124.83kHz.

[bob p]

Mike, one thing to bear in mind is that you're looking at "primitive" circuits like the JTM45 as you're trying to get a handle on the "proper" grid stopper values to be used with a tube's miller capacitance to provide frequency shaping for a gain stage. That's the way that the "primitive" circuits used to do it, but it's not the way that today's more "advanced" high gain circuits work.

If you look at just about any commercially successful high gain amp today, they contain a large number of resistors and capacitors in the signal path, which may not make sense at first glance. These amps certainly use a lot more RC filters than the primitive circuits. Why? They're using specific RC circuits to create HP and LP filters to provide tone shaping and to improve stability under considerations of "excessive" distortion.

Selecting a grid stopper in combination with the Miller capacitance of the tube isn't the only way to accomplish this goal. Many manufacturers instead use specific RC circuits in the signal path to create tuned HP and LP filters without directly attenuating the grid signal. Many amp designs also implement diodes to deal directly with the blocking distortion problem caused by capacitor charging. As they say, "There's more than one way to skin a cat."

[mikepukmel]

My main concern about my rank amateur home built amps is reducing the chance of parasitic oscillation, rather than tone shaping. Ive read descriptive accounts, haven't found so much calculations, e.g. what frequency bands parasitic oscillations tend to occur in, and what components to add, or ?? else to do to get rid of it. Merlin suggests some experimentation might be needed, re size of grid stops.

On my deluxe build attempt, after adding reverb, I had the output tubes go wild for a few seconds, before I realized what it was doing and shut it off. The main problem was bad reverb, and general lead dress. i shortened a whole bunch of wires, mainly to the tube socket pins, twisted a bunch of pairs. Got rid of the immediate problem, but still found that if I tinker with the vol/bass/treble controls, I could get the output tubes to go wild.

Replaced the eyelet board with one that has the correct (or at least as designed) component layout for reverb and also as a benefit is narrower (front to back of the amp), so that the leads drop to the chassis faster and are neater, compared to the "golden" amps of the pre -cbs era.

I haven't fired up that Deluxe-in-progress yet still a little more to do, but what I wanted to try to figure out was that if I put a couple of grid stops in, at preamp inputs, the reverb recovery, and maybe tinkered with slightly larger stops on the output tubes (from 1500 to 5600), would it *help* to reduce the chance of bad oscillation? That was the reason for the question about calculating corner freq for the low pass that occurs with the grid stop and internal capacitance.

Next build, JTM 45 clone, in progress, the original original schematics had no grid stops except on the input. The clone/kit company provides a few extra resistors for output tube grid stops, and suggests to use them. I just read a thread on MEF (old thread so unfortunately photos are gone), a guy said he had lots of problems that sounded like oscillation. One of the reviewers said something like "Your lead dress is terrible". Apparently, the guy put some grid stops in, and fixed the problem.

As well, I'd hate to put a bunch of grid stops in and have the amp sound terrible. Many Fenders out there with 1K5's on 6V6's, and a little later Marshalls with 5K6 on the OT's, and none have grid stops on the second preamp tubes.

So, my absolute amateur first guess was that I could try 68K grid stops around the reverb section, since I never got a hint of oscillation before adding it, and also, the first input stage has 68k's and didn't kill the tone, so first try, why not 68k. Then, thought to ask what the calculations look like and see if the value is close to the audio band. And maybe upping the output tube to 5.6K and see what happens. Maybe in 3 stages:
- as is today
- add grid stops around the reverb
- up the grid stops on the output tubes.

Then, similar experiment for the JTM as it comes alive.


As a separate question, have you guys seen a particular frequency or band, when oscillation occurs, e.g. "Ive never seen it above X Hz or ... below Y HZ,

or is it more unpredictable?

[bob p]

It can be anything from LF motorboating to Megahertz.

I know that you have Merlin's books. They have an excellent chapter on stabilizing an amp with NFB. IMO it's worth carefully studying that chapter.

Do you have scope? And a frequency counter? Or are you flying totally blind?

[Gingertube]

The value of grid stop recommended for triodes from the golden age of tube design to guarantee freedom from parasitic oscillations was 8/gm. You can often get away with much less or even no gridstop but that is the value for "good" design.

For a 12AX7 triode with a gm of 1.6mA/V (use 1.2 mA/V for a well worn tube) use 8/0.0012 = 6.66K, so 6K8 to 10K is plenty.

Remember that the grid stop is there to damp a tuned circuit which is formed from grid lead inductance and parasitic capacitance. That grid lead inductance needs to be kept to a minimum so having the grid stop on a wiring baord remote from the tube socket is no good (extra inductance from the connecting wire).
The grid stop resistor body should be hard up against the tube socket pin. I often use a small 3 tag "tag strip" mounted on the tube socket mounting bolt to accommodat the gridstop(s) and then wire from there to the board.

If you have oscillations after fitting the grid stops then it is from some other cause.
Such as:
- bad lead dress with the grid circuit components/wiring too close to the anode circuit wires/components. If wires components have to cross arrange for them to cross at right angles to minimise capacitive and magnetic coupling.
- inadequate power supply bypassing (motorboating form of oscillation).
- OR just bad circuit design

As an additional "hint" for good design, anode loaded stages are susceptible to inductive loading (such as transformers), cathode loaded stages (eg cathode followers) are susceptible to capacitive loading (such as long cables). Often a "build out" resistor can be used to isolate that node (the anode or the cathode) from the resulting phase shift.

Cheers,
Ian

[bob p]

To put some more numbers on it, so that you can be quantitative rather than qualitative:

Parasitic oscillation comes from the combination of the tube's Miller C and the series L in the grid wiring to form a resonant circuit. The way that I'd approach the problem is to minimize the inductance in the grid wiring by shortening the wires as much as possible, and if that doesn't cure the problem then I'd add damping to the resonant circuit by adding grid resistance.

The Q of the resonant circuit can be calculated using the formula:

Q = (1/R) * sqrt(L/C)

Looking at the equation, series resistance is going to play a large role in determining Q, and both L and C to a lesser extent because we're looking at their square root. So to minimize the problem, decrease the amount of wire in the grid circuit (L) and increase the amount of resistance R, using as little as needed to get the desired result.

[mikepukmel]

To put some more numbers on it, so that you can be quantitative rather than qualitative:

Parasitic oscillation comes from the combination of the tube's Miller C and the series L in the grid wiring to form a resonant circuit. The way that I'd approach the problem is to minimize the inductance in the grid wiring by shortening the wires as much as possible, and if that doesn't cure the problem then I'd add damping to the resonant circuit by adding grid resistance.

The Q of the resonant circuit can be calculated using the formula:

Q = (1/R) * sqrt(L/C)

Looking at the equation, series resistance is going to play a large role in determining Q, and both L and C to a lesser extent because we're looking at their square root. So to minimize the problem, decrease the amount of wire in the grid circuit (L) and increase the amount of resistance R, using as little as needed to get the desired result.

Ahh, cool, thanks. So, getting that board down, close to the chassis, and removing like 15 x 1" of wire, then very carefully routing wiring, to get rid of another half that, is important. Once that is done, the resistance you're talking about, *could* be in the form of a grid stop? I.e. not only are grid stops helping with the specific low pass in the tube itself, but also in the wiring?

[mikepukmel]

Thanks everyone. Learning!

[Gingertube]

One one other other thing thing is to appreciate that the input voltage signal may come with a substantial effective resistance (from the driving stage) that adds to the input stopper resistance of the next stage. That could noticeably increase the R in RC, and hence you may observe a lower corner frequency than anticipated.

This is particularly relevant at the input stage. The source resistance/impedance from your guitar pickups can be quite appreciable and will be highest (to give lower corner frequency) when the guitar volume knobs are turned down to lower settings. For best treble run your guitar volume knob(s) at max.

Cheers,
Ian