Home-made tube amp ghost notes?

Hey guys, I'm trying to improve the overdriven tone of my amp. I'm loving the way it sounds for individual notes, but when more than one note is played at a time, you can definitely hear an undertone. I'm not sure whether that's just intermodal harmonics from the notes being played when heavily distorted or from a design problem in the amp. The undertone is almost completely inaudible when playing clean.

I've heard that ghost notes can come from power supplies which are not adequately decoupled. I've attached the power supply in my amp for you to see. Yes, there are a lot of caps. When I built the amp, I didn't know I could get away with less. One would think that should help, in this case.

The least decoupled node is V5, and despite that the schematic says 620R, there's a 560R resistor in there because it's all I had on hand. My math says 1/(2pi*22u*560) = 13Hz, which may not be enough. Hopefully that's all that's causing my problem.

The other nodes are:
V4 - 1/(2pi*22u*1k) = 7.2Hz
V3 - 1/(2pi*22u*1k) = 7.2Hz
V2 - 1/(2pi*22u*1.2k) = 6Hz
V1 - 1/(2pi*22u*4.7k) = 1.5Hz

The amp is also using a star ground, which I'll replace soon with a bus.
 

Attachments

  • Power Supply.png
    Power Supply.png
    64.2 KB · Views: 168
Last edited:
Ghost notes caused by inadequate power supply filtering will show up when you play a single note. They are Intermodulation between the note you play and the power supply ripple.
If you are only hearing ghost notes when you play two or more notes I suspect it is intermodulation associated with the distortion in the amp. You might be able to tell by listening carefully to the pitch of the notes. For example, if you play two notes which are a perfect 5th, the difference frequency produced by intermodulation is an octave below the lower note.
Assuming it is IM distortion, one suggestion is to reduce bass before the distortion stage(s). Many high gain amps have only partial bypassing of the first stage cathode resistor (e.g. only a 0.68uF bypass cap) to achieve this.
Another suggestion is to experiment with the biasing of your distorting stages. I have read (somewhere on the internet) that symmetric distortion (which produces only odd harmonics) does not create IM difference frequencies. (Although I don’t know the theory to support this, as yet.)
 

Tesla88

Member
2011-12-21 8:37 am
Italy
I know is a stupid question but, is your guitar in tune ? When you play more than a note you mean a chord ? If strings are not well in tune you get dissonant chords .
Also bear in mind that with overdrive you always get the sum and difference of the notes played. You get "weird notes" when you play notes with a small interval. An exeple is playing 5 fret on B string while bending at 7 fret on G string, i mean your playing an E and bendind a D to E . The less the interval the more you can hear a lower note.
If your guitar is in tune and play a power chord with bridge pick up, it must sounds good.

If chords sound weird the amp design can be the issue, most amplifiers designed for light overdrive sounds muddy when hard drived.


A complete schematic would help.
 

jgf

Member
2010-01-12 7:28 am
PA
Not related to your question, but important: The transformer primary should not be connected to safety earth! you've got four connections drawn on the IEC jack, normally there would be three, so it's not clear what's what there. Hot and neutral from the IEC jack should connect to the transformer primary, and earth should connect directly to the chassis.
 
Last edited:

steeledriver

Member
2014-12-08 10:28 pm
jgf - You are right, and I drew the schematic wrong. The primary is wired to hot and neutral, and safety earth is to the chassis. I didn't have a symbol in PSpice for the IEC jack, and I definitely drew that one wrong. I'll go fix it in my file.

Tesla88 - I made sure the the guitar is in tune, and the notes are indeed musically related to the interval I'm playing. It must be intermodulation then. And the amp was designed for light overdrive, but I've been making changes to push it harder.

Malcolm - Thanks for the reassuring post. I'll try partial bypassing to see if it helps. I'm experimenting with LED biasing instead of Rk/Ck the amp right now. All of my 12AX7s are biased at 1.7V from the red LEDs I have, which is roughly in the center. The problem existed when I had fully bypassed resistors too, and also when my second and third gain stages were hot biased to try to get asymmetric clipping - just my anecdotal two cents. I'll give this a shot and let you know if it helps!
 
Distortion causes intermods. No distortion means no intermods. You can have both, or neither.

Symmetric distortion causes odd-order intermods (and odd-order harmonics) - typically experienced as compression. Asymmetric distortion causes even-order intermods (and even-order harmonics) - typically experienced as sum and difference frequencies together with 'richer' sound due to the next octave up.
 
...Symmetric distortion causes odd-order intermods (and odd-order harmonics) - typically experienced as compression. Asymmetric distortion causes even-order intermods (and even-order harmonics) - typically experienced as sum and difference frequencies together with 'richer' sound due to the next octave up.

Many thanks for that. Could you clarify a little what odd and even order intermodulation is? Am I right in thinking that (f1+f2) and (f1-f2) are examples of even order, while (f1+f1+f2) and (f1+f2+f2) are examples of odd order?

Can you recommend a good reference that covers this?
 
Symmetric distortion causes odd-order intermods (and odd-order harmonics) - typically experienced as compression. Asymmetric distortion causes even-order intermods (and even-order harmonics) - typically experienced as sum and difference frequencies together with 'richer' sound due to the next octave up.

Sorry to quibble, but part of that is not quite right.

Symmetric distortion refers to a transfer function which can be described mathematically as an 'odd' function (or a function having 'odd symmetry') and yes that does produce only odd harmonics.

Asymmetric distortion refers to a transfer function which has neither 'odd' nor 'even' symmetry mathematically - it produces both odd and even harmonics, in general.

A mathematically 'even' function is needed to produce only even harmonics. These are not usually found in the audio chain. An example is full-wave rectification. Notice that after full-wave rectification the fundamental itself has been removed, which is correct since the fundamental is the first harmonic (i.e. an odd harmonic).

An odd function is where f(x) = -f(-x)
and an even function is where f(x) = f(-x)
 
Last edited:
Double check the bias and plate voltages in each stage to make sure you didn't make a mistake that would cause too much I.M. distortion. Some is often desirable (for a more natural sounding distortion spectrum shape), but too much can suck.

Another source of ghost notes could be Rf energy coming in from the guitar and/or the AC cord. That energy could beat against other distortion products and produce I.M. difference frequency energy that could fall in the audio band. I always put a .01uF 3kV cap across the primary of the power tranny (and the secondary, instead of across each diode, and across each large filter cap), and a passive Rf filter at the guitar input (10K R in series and a 220p to Gnd. for a -3dB at 36kHZ, assuming the guitar source Z is 10k ohms (it's almost always 6K - 10K unless it's a piezo pickup).

I have found that it is also caused by my ear-brain mechanism when playing too loud. My guitar will sound out of tune, even though I just got done tuning it with a digital tuner, and it sounds perfect at lower levels.
 
Malcolm Irving said:
Am I right in thinking that (f1+f2) and (f1-f2) are examples of even order, while (f1+f1+f2) and (f1+f2+f2) are examples of odd order?
Yes.

Sorry to quibble, but part of that is not quite right.
Sorry, I oversimplified - but in amplifiers an asymmetric waveform will almost always have second-order nonlinearity dominating so you will get fundamental plus second-order terms, plus some higher order terms at a much lower level.
 
... in amplifiers an asymmetric waveform will almost always have second-order nonlinearity dominating so you will get fundamental plus second-order terms ...

Thanks for clarifying. I agree.

If the nonlinear transfer function is exactly quadratic e.g. y = a + b x + c x(squared)
where y is the output and x is the input
with an input of two sinusoids at different frequencies f1 and f2,
the output consists of only f1, f2, 2f1, 2f2, f1+f2, and f1-f2 (plus the DC component).
i.e. no higher than second order terms.
 
Sorry about the late follow-up guys, I'm writing exams at the same time as doing this project.

I partially bypassed V1a, and while I liked the effect of cutting some extra bass on the amplifier tone as a whole, I still hear a lot of intermodulation when the amp is overdriven. I've center biased every stage in the amplifier, and yet the intermodulation sometimes sounds almost as loud as the notes I'm playing. The overtones are definitely musically related to the intervals I'm playing.

Bob Richards - I don't have any caps across the PT primary to bypass unclean mains signals. I'll give that a shot when I get home from the library.

Another issue with the sound of the distortion is that it "wanders" around a little bit, like an extra layer of light fuzz that comes and goes when you hit a big chord and let it ring out. Perhaps the two problems are related, or maybe it's not a problem and is fairly common in tube amps. I've heard the same thing in some other peoples' home made amps. My first thought is that I'm not attenuating enough between stages, and the signal is causing the grid bias of some of the overdriven preamp stages to shift around. However, I've rotated lower gain, larger headroom tubes throughout the amp (and re-biased each time to maximize headroom) such as 12AT7s and 12AU7s instead of 12AX7s, and the wandering fuzz persists.

I'll try to get some sound clips later today.
 
A full schematic would be great, but if you don’t have that, it would be good to know more details of the circuit topology.
Is this a push-pull amp? If so, what type of phase inverter do you use?
Your power supply schematic shows V1 through to V5 each having their own power supply filter/decoupling cap. That seems more than necessary, but I don’t think it can cause any problems.
There appears to be a lot of preamp gain, and if you have a cathodyne phase inverter, you should include an extra-large grid stopper on the grid of the cathodyne, to avoid some very weird distortion, see:
http://www.valvewizard.co.uk/cathodyne.html
 
I put together a schematic quickly in LTSpice - there might be a mistake or two in it, but it should be mostly correct based on my memory (Edit: one mistake - the 300V on V4b should be 285V). Since LTSpice doesn't have switch or potentiometer models, I improvised. My old PSpice schematic is a little outdated and on a hard drive in another city. There is indeed a lot of pre-amp gain in this amp, and attenuation. I've been experimenting with LED biasing, and I really like the sound.

V1/V2 are 12AX7s
V3 is a 12AU7
V4 is a 6GH8A
V5 is a 12AY7
V6/V7 are 6L6GCs

Malcolm - the closest thing to a cathodyne in this amp is V2a, in which the FX send is tapped from the cathode and the dry signal is amplified and tapped from the plate. I had thought that might be a source of distortion, so I put a 220k grid stopper in.

There are two volume controls in the preamp - one between V1a and V1b, and one after the tone stack but before V2a. Depending on how I adjust these two controls, I can get a clean signal through the entire preamp, overdrive just V1b, overdrive just V2b/V3a, or overdrive all three. No matter which mix I use, I notice the "wandering fuzz" sound - as soon as I start to go into overdrive. The IM distortion gets worse the more distorted the signal gets, which it should.

My PI doesn't have a grid stopper - and I noticed a lot of amps don't. I don't think this is my problem because even with the pre-PI MV turned WAY down and the preamp knobs cranked I notice the "wandering fuzz".
 

Attachments

  • Official Schematic v1.pdf
    85.9 KB · Views: 18
Last edited:
Nice design - I am also a fan of LED biasing.
I am wondering about the possibility of an intermittent oscillation at the onset of distortion.

It would be worth trying anode bypass caps (220p-1n) on the 12AX7 plate resistors; I find I need them on the high gain sections of a preamp - and see them on many modern commercial designs eg Marshall.

Also wondering if your FX return is grounded when not in use?
 
Thanks! Yes, the FX return is grounded when not in use.

Oh snap, I forgot to include that in the schematic. On the plate of V2a there's a 560pF silver mica cap. I have a few 390pFs and 1nF caps, so I'll try them on some of the other plates tonight.

Here's a quick sound clip. I can't run at very high volume because my roommates are studying too, but there's a sample of clean and pretty heavily distorted. I tried to play enough chords and leads to give you an idea what it sounds like at different dynamics, and a little bit where I played around on the lowest and highest strings to get really noticeable IM distortion. Around 1:36 I hit a chord hard to make that "wandering fuzz" sound happen.

Thoughts?

Edit: the sound file was too big for the forum, so I put it on SoundCloud.
https://soundcloud.com/steeledriver/tube-amp-demo
Don't mind my phone going off.
 
Last edited:
I tried bypassing plate resistors all over the amp with caps from 390pF to 22nF. While it definitely has an effect on the tone, the fizz remains. So does the IM. Could it be a lead dress issue? I've got a star ground in the amp right now, which I'm planning on converting into a bus-bar ground thanks to a very educational conversation with Mr. O'Connor. That'll probably cut five feet of ground wires out of the amp. There are cases where I have signal wires going out to and returning from a pot in very close proximity. I tried not to let any large signal wires run near small signal wires, and only cross on right angles, but I have a lot of parallel ground wires.

I've rolled different tubes around the amp and the overdriven fizz remains. I have HitRay, GE, TungSol and JJ 12AX7s, and I've even tried rolling EH 12AT7, GE 12AU7, and JAN 12AY7s around the amp and the problem persists. I don't think its related to the tubes themselves.
 

Attachments

  • IMG_20160418_173936950.jpg
    IMG_20160418_173936950.jpg
    737.4 KB · Views: 37
  • IMG_20160418_173942480.jpg
    IMG_20160418_173942480.jpg
    900.9 KB · Views: 36
Last edited: