Overcoming interaction between control settings in FMV tone stack

[printer2]

Gnobudy, have you tried Gar's one knob tone control?

[Gnobuddy]

Gnobudy, have you tried Gar's one knob tone control?

I haven't tried that one, no. I've seen it in Merlin Blencowe's preamp book, though.

I think this one is worth a quick LTSpice simulation to see how it performs. However, my guess is that it will produce very poor boost characteristics and lopsided frequency response curves (cut and boost very different), because the signal at the cathode is so much smaller than the signal at the anode.

-Gnobuddy

[J M Fahey]

It´s an active Treble Boost/Cut and as such better than the old Cut only one.
In principle crossover Boost/Cut frequency should be about the same, so C2 should be about 10/15X C3 .
As is, it cuts above 4kHz but boosts above 28 Hz so a gain boost rather than an equalizer.

I bet C2 = .047uF (600Hz) to 0.022uF (1200 Hz) and C3 2200 to 4700pF would be more balanced.

Oh, and the pot should be AntiLog to have an even range.

[Gnobuddy]

In principle crossover Boost/Cut frequency should be about the same, so C2 should be about 10/15X C3.

This is one of the things that rings some minor alarm bells in my head. IIRC, Thevenin impedance at the cathode of the triode on its own is {(1/gm) + (Rp+rp)/mu}. Then you parallel that with the 5.6k external cathode resistor to get the impedance at the cathode of the whole circuit. I make that about 1.6 kilo ohms using the usual datasheet values (gm = 1600 umho, mu = 100, rp = 60k), along with Rp = 100k and Rk = 5.6k.

Since gm is not a particularly accurate parameter, and the impedance at the cathode depends on it, there is already some unpredictability in the bass control's corner frequency.

Since boost is generated by bypassing the 5.6k external cathode resistor, leaving the internal 1.6k cathode resistance of the triode in circuit, my quick estimate is that maximum boost is limited to (5.6+1.6)/1.6, or about 13 dB. Maximum cut, of course, is -infinity dB. So we can already tell that cut and boost curves will be heavily assymetrical, with a lot more cut available than boost.

If I have time I'll run it through LTSpice in case I've made any mistakes in my back-of-the envelope calculation. Today is a busy day, though.

-Gnobuddy

[SoulFetish]

Presence or Resonance controls actually reduce feedback as they are advanced form their zero position. (Hiwatt is an exception.)

Tubes naturally produce low order distortion that raises with signal level. This sounds great for guitar or bass. (some solid state circuits can also do this.) If you apply feedback around tubes or even local feedback at the cathodes, the distortion spectrum shifts to higher order harmonics, although overall the total distortion level is lower. This doesn't sound as good. Evidence: Compare a Music Man amp to an all tube Fender equivalent. The Music Man just sounds cold and sterile, The Fender is much warmer even if you plug straight in with the Volume set on 2. You can instantly tell the difference in a blind test.

.

LT! How you been, man? Thanks for waying in on this. I feel like it's been a while since I've heard from you.
Your point makes sense. Although, wouldn't this be true in any feedback network? The entire frequency response gets flattened and extended, because signal voltages are fed back proportionally. Also, doesn't local feedback serve to minimize distortions that are generated in the stage? Wouldn't any low order harmonics present at the input remain the same component of the overall signal? So, how bad could it be in an otherwise transparent stage right at the output of the amp?
The reason I ask is, I was kind of rooting for this to be a good idea.

[loudthud]

Getting a little off topic here,

Although, wouldn't this be true in any feedback network?

According to Nelson Pass, the feedback needs to be more than about 10dB to generate higher order harmonics. Link: https://www.passdiy.com/project/arti...n-and-feedback

The entire frequency response gets flattened and extended, because signal voltages are fed back proportionally. Also, doesn't local feedback serve to minimize distortions that are generated in the stage?

Yes, local feedback minimizes THD (emphasis on the Total), but it also shifts the spectrum if there is too much FB, >10dB.

Wouldn't any low order harmonics present at the input remain the same component of the overall signal?

Yes, this is an important point. You can use one or two non-feedback stages to warm up the signal, then FB stages can be used as long as they don't have bad clipping behavior. There are lots of solid state bass amps with a 12AX7 front end.

So, how bad could it be in an otherwise transparent stage right at the output of the amp?

Speakers don't sound as good for a guitar amp if they are driven but a low impedance. They sound better with current drive. So it could go either way.

IMHO, the lower the feedback, the better the sound. There are cases where some feedback sounds OK and doesn't completely mess up the sound like the cold clipper in a Trainwreck Express.

[PeanutNore]

I have a valve guitar amp that uses a little MOSFET as a "sourceodyne" phase inverter. It works better than a triode phase inverter in every way, except it will probably sound nasty if clipped. So I put a fixed attenuator between the (valve) stage that drives it, and the MOSFET gate. Attenuation is set so that, with the valve at full clip, the MOSFET is still happily linear. (Easy to do because MOSFETs only need a couple of volts between source and drain to keep operating perfectly, unlike a 12AX7 that wants a hundred volts or more.)

I'm excited to see that you've been able to do this succesfully - I tried making an LND150 "sourceodyne" and it worked really well for a few minutes, until I think I fried it due to inadequate protection for the MOSFET gate. I've got a tiny "clone" of a JCM800 running an ECC99 dual triode in self-split configuration, and it would get a lot more output with a PI, even one without any gain. If you feel like sharing any tips or a drawing of how you've wired it, I'd be eternally grateful.

I have had success in building a long-tailed pair out of LND150s to add some power amp simulation to a guitar pedal, but I have it designed around a 9 volt supply. I'm not sure what will happen if I try to scale it up to 350v.

[Gnobuddy]

If you feel like sharing any tips or a drawing of how you've wired it, I'd be eternally grateful.

Sure thing, pic attached. The pic shows the schematic and LTSpice simulation. The real thing has been working in my little 6AK6 amp for a couple of years now.

There's nothing clever about this circuit, really. The Supertex datasheet says the LND150 can tolerate up to +/- 20 V between gate and source. So I used a pair of back-to-back 12V zeners in series to keep the gate from going more than about 12.6V positive or negative of the source. They work in conjunction with the 47k "gate stopper".

Speaking of the gate stopper, I've tried up to 1 megohm there with no adverse effects. It's been a year or two, so my memory may not be accurate, but I don't think I even heard any problematic treble loss.

The 1.5 meg, 3.9 meg, and 680k resistors bias the gate to one-quarter of the B+ voltage, which is optimum for a "SourceODyne" and gives you maximum output signal capability. A MOSFET can saturate with the drain within a couple of volts of the source, so the source and drain signals can practically "touch in the middle" at one-half power supply voltage; maximum peak-to-peak output signal at source or drain is therefore almost half B+ (much more than you would get from a triode in the same circuit).

In my case, B+ was about 150V, so my SourceODyne can put out two signals of nearly 75V peak-to-peak each, maximum. This was far, far more than I needed to drive my output stage (a pair of 6AK6 pentodes, which only need about 20 Vpp for maximum clean output, and maybe 40 Vpp to be deep into overdrive.

Keep in mind the LND150 is quite small, so it can't dissipate much power without overheating. I try to keep it under 250 mW, which is conservative. Biased with the source at one-quarter B+, the quiescent voltage across the LND150 is about one-half B+. In my case, there is about 75V DC across the LND150, and quiescent current through it is around 0.5 mA, so the power dissipated is under 40 mW (that's 0.04 watts!). It stays cool and happy, in other words.

My B+ is pretty low because this was a little 2-watt amp built around a pair of 6AK6 output pentodes. But the LND150 should handle 350V B+ too. Biased properly (gate at 1/4 B+), it will have 175V DC between source and drain. As long as you pick the source and drain resistors so current through them is no more than, say, 1.5 mA, the LND150 should be fine. In other words, with 350V B+, make the source and drain resistors at least 56k, and you should be fine. The gate bias network can stay exactly the same (680k, 3.9meg, 1.5meg).

I have had success in building a long-tailed pair out of LND150s to add some power amp simulation to a guitar pedal

Interesting! LND150s have such a high transconductance compared to valves that I would have thought you would just get gross, buzzy, fuzz-box distortion all the time. I take it that's not what happened, though. I would be interested in seeing that schematic, if you feel like sharing!

I don't think LND150s can replace a long-tailed-pair phase splitter in a guitar amp, by the way. A long-tailed-pair built around a pair of triodes has no local feedback, so it is quite nonlinear, and has a "growl" of its own, which is part of the amp's sound. If you replace the triodes with LND150s, it will have far too much gain; if you fix the gain problem by adding unbypassed source resistors (i.e. adding local negative feedback), you get a perfectly transparent phase splitter. No "growl", and it won't sound like the real thing.

Cathodynes have 100% feedback, so they have so little distortion that they have no "tone" of their own. A "SourceODyne" is therefore a perfect replacement, as it is also perfectly transparent and has no audible distortion of its own.

-Gnobuddy

[Gnobuddy]

I built and tested the active tilt control. After a short test, I really like it. I had been concerned that there wouldn't be enough flexibility without independent treble and bass controls, but in practice I didn't find that to be the case. For the intended owner (I'm giving this amp to a friend), it is a definite advantage, as he doesn't get along well with complex gadgets.

This tilt control works in conjunction with a "de-nastifying filter" which includes an 800 Hz notch, treble roll-off above 2 kHz, and a little bass lift below about 400 Hz. All that stuff is on a separate PCB (and a separate schematic).

Pic attached...component values are the same as the ones in my real circuit, but the MPF102 model in LTSpice doesn't bias up the same way as my real MPF102's do. In real life, both MPF102s are biased to have half B+ at the drain. The BC550C emitter follower also biases to half B+, this time at the emitter.

-Gnobuddy

[PeanutNore]

Thanks! This is a lot simpler than I was expecting. I think one of my problems when I tried it before was the way that I had tried to source-bias it, like a cathode biased cathodyne. I think I forgot that the gate was referenced to something close to the source voltage rather than ground and neglected an important DC blocking cap.

The same mistake appears in the drawing that I have (below) for this LND150 LTP - between the "LEVEL" pot and R27 there should be a ~22n cap, because the gate resistor R28 isn't referenced to ground but rather the tail. I added that cap once I breadboarded it, and it works great. I'm using a simple op-amp differential amplifier to recombine the two phases for the output, since it's part of a preamp pedal.

[Gnobuddy]

Thanks! This is a lot simpler than I was expecting.

Zen circuitry.

I had tried to source-bias it, like a cathode biased cathodyne.

Gotcha. Even with the right DC blocking caps in place, this is really tricky with all FETs, because they have such wide parameter spreads. I've seen a lot of bad FET circuits on the 'Net, circuits which worked for the original builder, but which have no guarantee of working if built by anyone else.

Having a large DC voltage across the source resistor is one good way to stabilize an FET circuit against parameter variations. The large DC voltage swamps small changes in Vgs due to JFET parameter spreads. This would work for the LTP circuit too, though there is the additional problem that the two FETS in the pair might be widely mis-matched, so one might be completely turned off while the other one draws all the current.

Thanks for sharing your circuit, by the way. Did you match the MOSFETs in the LTP, or did it sound good by sheer chance?

Now you've got me wondering if I have enough JFETS of one type to find a well-matched pair among them!

I'm using a simple op-amp differential amplifier to recombine the two phases for the output, since it's part of a preamp pedal.

I've had the same idea, in a slightly different variation: make an LTP with two small-signal pentodes, run the outputs into an LND150 differential amplifier to combine the outputs like a transformer (just as you did!), divide down the output signal, and spit out a line-level signal, complete with true push-pull pentode distortion, and speaker emulation, for use with a P.A. system or clean solid-state power amp.

I'll have to try it some time.

-Gnobuddy

[PeanutNore]

I didn't match the MOSFETs, it just seems to work. It might be dumb luck, or it might be that the specific part number I've been buying for LND150s may be a binned part, but all the ones I have have been extremely consistent - as a common source amplifier on 9v with a 4k7 drain resistor and 240 ohm source resistor, the drain voltage is always within 100mV of 4.2 volts. So about 2.5% variation in current I guess. I picked 4k7 for the drain resistor based on Idss from the datasheet, and used a trimpot to find the ideal source resistance and 240 ohms was the closest standard value. Since then every one that I've tried has been dead on with those values.

[Gnobuddy]

..the drain voltage is always within 100mV of 4.2 volts. So about 2.5% variation in current I guess.

Okay, that is just extraordinary. 250% variation wouldn't be terribly unusual - Supertex says Vgs(off) can vary from -1V to -3V for the LND150. I've got a fistful of MPF102s, all from one batch, that have Vgs(off) anywhere between 1.2 V and 2 V (and that is a lot closer than the extremes of the factory spec.)

But 2.5%? I've never heard of discrete FETs with that level of matching. Heck, I doubt that level of matching is achieved with two FETS on one die inside an opamp. You, my friend, may be the luckiest man on earth, and you should run out and buy a lottery ticket right now!

-Gnobuddy

[PeanutNore]

Watch as the next one I pull from that tape sits at 7 or 8 volts in that circuit LOL

[Mike Sulzer]

To assess the interaction in the FMV tone stack, let’s look at how it works. I am not happy with the descriptions I found on line, and so this post starts with a partial analysis. The Marshall version is shown in the first schematic in the attachment.

The middle schematic shows the circuit when the bass and mid pots are set to zero. What’s left is a high pass circuit (Ct and the treble pot, with the load on the output affecting the frequency response somewhat), but the bottom of the pot is imperfectly connected to ground. The bigger you make Cc and Cb, the better the connection to ground. Replacing Cc with a wire would improve the connection to ground. The mid pot also affects this somewhat. So we have some potential for interaction when the bass pot is raised, and this complicated impedance changes.


The right hand schematic shows the circuit when the bass pot is set to 10 and the mid pot is left on zero. Now we can see that Rb and Cb make a low pass filter, and Cc couples that filtered signal to the output. Cc is just a coupling capacitor that blocks dc from the previous stage. And it makes a contribution to the unwanted interaction. IMO, replace Cc with a wire and put a proper coupling capacitor at the output of the preceding stage. This coupling capacitor should be somewhat larger than Cc. This is also safer for those who put their fingers where they should not go since the HV does not get into the tone stack. I have measured the response of the modified circuit and the only differences from the original are interaction level. I also have listened to it, and the differences do not seem not large, and I like the new circuit.


To complete the analysis, when the mid pot is increased from zero, it sets a maximum frequency on the decrease towards high frequencies due to Cb. So it kind of puts some middle back.