How to increase Harmonic distortion for clean sounds

[Gingertube]

Alan,
Both models are correct
- whether you view the AC equivalent circuit as:
a current generator with rp and RL in parallel across it
OR
a voltage generator with rp and RL in series
is purely a convenience thing. Both will give the same results.

If you are interested in voltages it is more convenient to work in the voltage domain and similarly if you are more interested in currents then work in the current domain.
This is like the old Norton (Current domain) or Thevenin (Voltage domain) analysis. Both are correct, just use the most convenient one for what you want to find out.

Cheers,
Ian

[Alan0354]

Alan,
Both models are correct
- whether you view the AC equivalent circuit as:
a current generator with rp and RL in parallel across it
OR
a voltage generator with rp and RL in series
is purely a convenience thing. Both will give the same results.

If you are interested in voltages it is more convenient to work in the voltage domain and similarly if you are more interested in currents then work in the current domain.
This is like the old Norton (Current domain) or Thevenin (Voltage domain) analysis. Both are correct, just use the most convenient one for what you want to find out.

Cheers,
Ian

But the result is totally opposite between series and parallel equivalent. as you indicated, if you assume rp in series with RL, then you are right to say to keep RL low to let rp dominate. But if you use parallel, then RL has to be kept high to let rp dominate.

[Gingertube]

Alan,
The clue in looking at the two models is this:
An ideal Voltage generator has zero internal impedance, if you replace the voltage generator in the model with a short circuit (Zero Ohms) the rp and RL are then effectively in parallel.
Conversely an ideal current generator has an infinite internal impedance (effectively and open circuit).

The result is not opposite at all, I think you missed the point in my post 2 up. It is not rp itself which is the source of the distortion. It is the changing rp with changing signal current throught the tube. So we need to think about the delta rp.

In the Voltage model we want delta rp to be signficant with respect to RL so that you get significant distortion out of the voltage divider formed by rp and RL.

I the Current model we want delta rp to be significant with respect to RL so that you get significant distortion from the signal current flowing through the changing (rp parallel RL).
The changing (rp parallel RL) is caused by the delta rp. To maximise this distortion we want delta rp to be significant with respect to RL for typical values of RL and rp that means minimizing RL.

I agree that this is not straight forward and difficult to visualize because of one additional fact:
For this analysis I would STRONGLY advise using the voltage source model rather than the current source model for one very good reason.

In the voltage generator model the generator output is Vg1k x mu volts.
In the current generator model the generator output is Vg1k x gm

As signal current changes then so does rp AND so does gm whereas mu stays much the same.

Using the current source model you have 2 things changing at once both the rp and the generator itself (because gm is changing too) which will always give you a headache in the analysis and make it difficult to get your head around.

If you use teh voltage source model then you only have to worry about the rp chnage.

Cheers,
Ian

[Alan0354]

I don't see how you can use voltage model as it is a tranconductance circuit. All BJT, FET and tubes are transconductance device with different output impedance. I can assure you both models for BJT and FET are current model. I never once saw a voltage model for BJT and FET. If you have a voltage model of tubes, please post it as I want to know.

Regarding to rp, I did not assume anything, I just use the plate cure and look at the slope. So the change of rp can be looked as change of current or change of plate voltage.

I think we are in agreement on all points except the series and parallel model where in your case, you want to minimize the RL, in my case, I want to maximize RL. And it cannot be both true.

[Gingertube]

Hi again Alan,
Either a Norton Equivalent Circuit (with a current source) or a Thevenin Equivalent Circuit (with a voltage source) may be used to analyse any circuit whether tube, bjt, jfet, mosfet or whatever.

I preferred the voltage source for this analysis because:
Look at the 2 graphs at the top right of the second page (labelled pg 491)
http://tdsl.duncanamps.com/pdf/vm345.pdf
These graphs show Ri (what you call rp), S what you call gm and u (mu) vs Anode Current

You can see that rp and gm vary significantly with anode current whereas mu does not.
When doing an AC equivalent circuit analysis to determine effects of rp and delta rp it is sensible to use the Thevenin (Voltage Source) equivalent circuit rather than the Norton (Current Source) equivalent circuit because the voltage source which equals Vg1k x mu does not change with anode current (mu is stable) and whereas the current source which equals vg1k x gm does change with anode current and so is likely to obscure what you are trying to see.

Maybe some examples will help.

Here is a paper talking about simulation but shows both voltage (Thevenin) and current (Norton) generator equivalent circuits.
Vacuum Tube Preamplifier Analysis and SPICE Simulation

Here is an example showing (amougst other things) a 2 stage 12AX7 amplifier:
http://www.radau5.ch/pdf_files/tube_b.pdf

Cheers,
Ian

[Alan0354]

Thanks for the link, I have to spend some time reading them.

[Gingertube]

No worries
Just about to finish work for the week and venture out into the 46 degrees C (115 degrees F). Adelaide had the dubious pleasure of being named by the UN as the hottest city in the world for today. Will catch you all next week if my brains don't fry on the way home.
Cheers,
Ian

[jazbo8]

But the result is totally opposite between series and parallel equivalent. as you indicated, if you assume rp in series with RL, then you are right to say to keep RL low to let rp dominate. But if you use parallel, then RL has to be kept high to let rp dominate.

I did not go through all of your hand-drawn notes, but the Norton and Thevenin equivalent circuits must provide exactly the same results, otherwise, something was not drawn/done properly.

[exclamationmark]

From personal experience, decreasing RL in addition to biasing in the colder region will definitely make amplification more asymmetrical. Even by drawing load lines, it's somewhat evident that a lower RL is more efficacious at generating second harmonic compared to a higher RL given the same bias point. IMO the effect is extremely subtle, at least in the quantity that a normal gain stage can produce.

[Chuck H]

I get the feeling what the OP is after isn't necessarily a class on waveform distortions, but rather the "black art" end of this craft regarding how to create musically desirable, unclipped tones. Now that can be very subjective, of course. Wave form distortion other than clipping can be taken all the way into the realm of effects like synthesizers and octave pedals. Not to sound guru-ish or cork sniffy, but I think the OP is after those qualities that certain amps have to transform an otherwise ordinary clean tone, such as might be had by a PA system or any number of common sounding amplifiers, into a more pleasing and interesting sound. Even feel (from a players perspective). I'm not saying I can ID those things that make some amps special while others sound common. And I'm not devaluing the process of analysis, but I think it would be best applied for ID or design of some known quantity. Randomly introducing any possible harmonic distortion that isn't clipping could be a fun experiment that MIGHT produce some useful result, but not necessarily. So, aside from waveform and tube function analysis...

If I have anything to offer it's that many great guitar amp clean tones are very close to clipping. That is, there IS some compression of the waveform, but not enough to clip it sharply. Add to that the possibility of associated power supply compression. The softer attack and near clipping compression tend to bring up harmonics WRT the fundamental. After that I can offer that some guitar amp EQ's can do EXTREAM things to a wave form. Anyone ever see (on a scope) what a VOX tonestack does to a TB preamp running an otherwise clean signal (extreme setting can actually flip the phase of the HF)!?! The combination of these things, the nature of the EQ, the power supply compression and near clipping sensitivity are probably responsible for many, if not most of the really interesting electric guitar clean tones. And this doesn't even take into consideration all the acoustic nuances of a particular guitar being played at some volume through an amp. That adds another dimension of sustain and resonances as it relates to the feel and audibility of harmonics. Consider some of those great AC30 tones from the 60's that don't sound clipped at all, but don't sound clean either!?! This would be an amp that possesses all of the above qualities in spades. And I don't think those extremes are good for every musical circumstance, but it illustrates the point. The great, harmonically rich clean tones are part of an overall amplifier system more than just the generation or accentuation of a given harmonic effect with one or two preamp stages. I tweaked for years in my bedroom trying to understand why my clean tone sucked. It wasn't until I started turning my amps up that I was able to narrow down the missing components and how they relate to the amps I was building.

JM2C

[Tahoebrian5]

Thanks Chuck, every little bit helps. I think the important thing is I have a concept. I can think about it all day long but I'm never going to get anywhere until I build something and see what she sounds like. In progress on researching where to source parts and board material now. Also, I took your advice and sold the Mesa.

[daz]

let me see if i'm understanding this....did some of you say that increasing the cathode resistor on preamp stages will increase second order harmonics? And if so, is that the case even with a bypass cap? I haven't ever noticed that but now you all have me wanting to experiment some more right when i have the amp sounding great already. (i gotta stop reading these damn amp forums)

[Alan0354]

Hi again Alan,
Either a Norton Equivalent Circuit (with a current source) or a Thevenin Equivalent Circuit (with a voltage source) may be used to analyse any circuit whether tube, bjt, jfet, mosfet or whatever.

I preferred the voltage source for this analysis because:
Look at the 2 graphs at the top right of the second page (labelled pg 491)
http://tdsl.duncanamps.com/pdf/vm345.pdf
These graphs show Ri (what you call rp), S what you call gm and u (mu) vs Anode Current

You can see that rp and gm vary significantly with anode current whereas mu does not.
When doing an AC equivalent circuit analysis to determine effects of rp and delta rp it is sensible to use the Thevenin (Voltage Source) equivalent circuit rather than the Norton (Current Source) equivalent circuit because the voltage source which equals Vg1k x mu does not change with anode current (mu is stable) and whereas the current source which equals vg1k x gm does change with anode current and so is likely to obscure what you are trying to see.

Maybe some examples will help.

Here is a paper talking about simulation but shows both voltage (Thevenin) and current (Norton) generator equivalent circuits.
Vacuum Tube Preamplifier Analysis and SPICE Simulation

Here is an example showing (amougst other things) a 2 stage 12AX7 amplifier:
http://www.radau5.ch/pdf_files/tube_b.pdf

Cheers,
Ian

What happened to the forum, I have not been able to get in since yesterday morning!!!

You are right for the most part. I read the article you posted and I read up Norton and Thevinin equivalent, The voltage model is really the Thevinin equivalent of the real circuit. I looked at it literally how the tube works, which is a transconductance amplifier that the input voltage is transformed into a current and use only the current model.

As for the result, you are right to a big extend, but it gets a little interesting, It is not as straight forward as I thought. As attached, I did some calculations and look at the result. Yes, theoretically, you get slightly more distortion if you lower the RL. But it seems that the best compromise is setting the RL close to the rp.

In my example, I use the initial rp=75K which is somewhere close to 12AX7 at low current. Then I assume the rp decrease to 50K dynamically during the frequency cycle and look at the percentage change in the gain of the stage within the cycle which translates to distortion. I use 3 cases

(1) RL=75K and look at the change in gain for rp=75K dynamically to rp=50k during the cycle.

(2) RL is about 33% lower and higher than 75K. Which is RL=55K and RL=100K and look at the change in gain for rp=75K dynamically to rp=50k during the cycle.

(3) RL is 50% lower and higher than 75K. Which is RL=37.5K and RL=150K and look at the change in gain for rp=75K dynamically to rp=50k during the cycle

You can see, when RL=75K, increase in gain when rp change from 75K dynamically to 50k is 120%. The increase slows down when you further decrease RL. At RL=50% of rp, it is 128%. For RL greater than rp, the gain increase from rp changing from 75K to 50K slow down. At RL=150K, the increase in gain is 112%. So it's really a compromise and from the example, setting RL equal rp is the best compromise as you still get enough gain and enough distortion. Lowering to 37.5K is not practical in real amp.

An further, it is interesting that increasing the RL to 150K gives so little distortion. Could this be the reason I saw some amp (Hiwatt and Vox AC30) uses 220K plate resistor just to make the preamp more linear? This is a very interesting exercise to me even I am wrong.




Also it is true lowering the plate current increase rp, which increase distortion as shown in your first link. You have been very helpful.

[Alan0354]

let me see if i'm understanding this....did some of you say that increasing the cathode resistor on preamp stages will increase second order harmonics? And if so, is that the case even with a bypass cap? I haven't ever noticed that but now you all have me wanting to experiment some more right when i have the amp sounding great already. (i gotta stop reading these damn amp forums)

Yes, if you look at page 2 of this link by Ian, for ECC83, look at each of the graph, the lower the current, the higher the Ri. On top, the lower the plate voltage, the higher the Ri.

http://tdsl.duncanamps.com/pdf/vm345.pdf

[Alan0354]

No worries
Just about to finish work for the week and venture out into the 46 degrees C (115 degrees F). Adelaide had the dubious pleasure of being named by the UN as the hottest city in the world for today. Will catch you all next week if my brains don't fry on the way home.
Cheers,
Ian

My God, when you mention about 46 degree before, I thought it's F and I thought that's comfortable!!!1 46 degree C, that's awful!!!!