Octavers again...

[bea]

Hi there,

I am new here and first want to say hello.

I have the idea to build an octaver with tubes in a LoV setups. My power supply can provide a rail voltage of about 20 V, and first tests show that even recent tubes like the EH ECC82 start to work better than frustratingly in this voltage range.

I have currently spiced a GreenRinger-like circuit. Fine, but overdriven, its cathodyne stage will act as an octaver by itself, and and will be hard to avoid that. If there is some interest, i can show that simulation. BTW: IMO any GreenRinger-like circuit will act as an octaver by itself as soon as the phase splitter is overdriven.

Therefore i intend to check to use a differential stage to split the signal. Maybe tonight.

Another approach i do not fully understand ist the concept of the MuDoubler. To me the circuit looks like a cathodyne phase splitter followed by a differential stage which just puts the to halves back together again. How is the claimed cancellation of the fundamental achieved?

Thanks

Beate

[R.G.]

IMO any GreenRinger-like circuit will act as an octaver by itself as soon as the phase splitter is overdriven.

It will. In fact, a single transistor stage of any kind will act as a "octaver" if it's biased so that it clips on one side only. A single diode will do this too. What matters is the amount of octave. A properly adjusted full-wave-rectified signal will produce more "octave" and less of the original signal than any of these.

Another approach i do not fully understand ist the concept of the MuDoubler. To me the circuit looks like a cathodyne phase splitter followed by a differential stage which just puts the to halves back together again. How is the claimed cancellation of the fundamental achieved?

The Mu Doubler is a circuit that does one form of electronic computation.

For the moment, assume that no clipping happens. Then the phase splitter produces signal and inverted signal. These are fed to the differential stage. The differential stage produces the difference of the two signals, again inverted, so you have a multiplied signal and inverted signal, but the output is a current signal on the drains of the two differential FETs.

The two current signals at the drains are added by running them both into the same resistor. The Principle of Superposition notes that the result of any linear operations is the sum of the operations done individually, so the voltage that appears on the single drain resistor is the sum of the two drain currents.

This is simply zero for the signal as amplified and summed into this resistor. But there is some distortion in the FETs. Odd order distortion products will also cancel in the resistor. But even order distortion products will reinforce at the drains. So if the FET have 1% even order distortion, the signal and inverted signal at the drains will cancel, but you'll get 1% plus 1% = 2% even order distortion of the multiplied signal at that drain resistor. The cancellation at the drains must be good to fully suppress the 200% signal differences, but it's possible to tweak it in.

The circuit "computes" just the even-order distortion products when it's properly set up.

[bea]

Which means that the mu doubler might work better if i use tubes because they are less linear?

[bea]

Ok, here first results on a green ringer alike plus tube overdrive. rail voltage is 20V:



In the following pic You see:

input: gray
after 1st stage: red
after rectification: light blue - note the doubling of the frequency
after final cathode follower: dark blue. With different (larger) values  of the resistors, a much larger output is possible.

[bea]

So if the FET have 1% even order distortion, the signal and inverted signal at the drains will cancel, but you'll get 1% plus 1% = 2% even order distortion of the multiplied signal at that drain resistor. The cancellation at the drains must be good to fully suppress the 200% signal differences, but it's possible to tweak it in.

The circuit "computes" just the even-order distortion products when it's properly set up.

Ok, sounds really nice. The difficulty seems to be the symmetry of both the catodyn and especially that of the differential stage. I checked both approaches (catodyn and difference amp as phase splitter), and i did it with both tubes and a BF245B JFET.

The simulations (where matching of parts is obviously not a problem at all) show a systematic asymmetry, especially if the voltages are relatively low. Unfortunately the JFET was not much better than the tube. The differential amp being worse than the catodyn in this respect.

Maybe it will help to use a constant current source instead of the resistor in the differential stage? Raising the rail voltage will also help (with the tube approach it is more or less a must to go to at least 30V for that stage to work).

[Mark Hammer]

A belated welcome.  Nice to see you here, Beate.  Seen many of your posts over on MEF.

[PRR]

> here first results on a

But isn't that a AA12 semiconductor diode "doubler"? The tubes are working linear (nearly).

Before morning coffee I was confronted by this and noted that while it is a poor linear mixer it seems to be a fine "doubler" (at least with exact-match drive and triodes).

https://s15.postimg.org/40amvg50r/cathode_mixer.gif

A run with 20V supply, 10V at grids, 12AU7, also looks very good. Output level can be quite high, no need for after-boost (certainly not any fancy high-gain stage).

[bea]

But isn't that a AA12 semiconductor diode "doubler"? The tubes are working linear (nearly).

Yes, indeed. My idea was to check if it was possible to set up a GreenRinger lookalike with tubes. And of course i am open for different ideas. BTW: the two resistors after the diodes feed some portion of the "other" halfwave back to the signal and might make it cleaner.

The gain stages after the splitter have the purpose of an overdrive stage after the splitter. My musical idea is to obtain an overdriven "guitar" signal  form a Bass VI.

Before morning coffee I was confronted by this and noted that while it is a poor linear mixer it seems to be a fine "doubler" (at least with exact-match drive and triodes).

I think i'll give it a try - and i have also LoV models of the ECC83 and the 12AY7 at hand. Unfortunately no ECC81.

A run with 20V supply, 10V at grids, 12AU7, also looks very good. Output level can be quite high, no need for after-boost (certainly not any fancy high-gain stage).

Again - i think i was not clear about that - the gain stage following the splitter has a purpose of its own.

[bea]

https://s15.postimg.org/40amvg50r/cathode_mixer.gif

A run with 20V supply, 10V at grids, 12AU7, also looks very good.

Do i understand this correctly: in an actual implementation i would need a splitter which drives the two halves of the mixer?

(BTW: i am going to try to get more than 20 V from my power supply, probably setting up a voltage doubler instead of the simplistic approach to feed  a 2nd filter chain from the reservoir cap...)

[PRR]

> a splitter which drives the two halves of the mixer?

Yes.

[bea]

Well, some first results in LTSpice. The behavior is a bit strange: it performs better (i.e., gives a larger output voltage) if the anode volte does not exceed about 20 V. If i run the model with 30 V i need to chose R4=1Meg ...
Overdriving behavior is also surprising: due to the added harmonic content, the output will increase - the voltage gain will approach 0.5 instead of the 0.33.

The large values of the resistors, especially R1, might indicate that i might use a CCS instead (mhmm, for these tiny currents...).

In a real application i might set up the splitter with a BF245B, possibly at only 12.5 V due to better symmetry of the source. Moreover i can use my 2nd tube as a gain stage. (Purely practical reasons: i do not have space and power supply capacity for more than 2 tubes, and i have a few BF245B available).

Please let me note that the values have been optimized by trial and error for the given tube model, and these do not account for the variability of actual tube parameters.

[bea]

Ok, here some results for the cathode mixer. This works like the Mu Doubler, doesn't it?



At 20 V on the rail i obtain better results with a JFET as a splitter, but i am going to try different tubes as well. Unfortunately i don't have models of the ECC81/12AT7 - probably one of the most interesting tubes for this application - which are valid in the LoV range. But i might check for other medium mu triodes. The constant current source in the mixer gives more consistent octave doubling compared to a simple resistor, especially if i increase the rail voltage. And a larger output...

[PRR]

I believe the current source in the long-tail is not what you want. It gives a sharp corner on the down side, large even-order harmonics. With a resistor tail the gain falls off around the dip, the corner is rounded and smoother.

Of course it may depend what you want.

[bea]

That's already slightly overdriven. The harmonic content of the octave signal varies strongly with the input level.

The key problem with the approach is that the octave effect vanishes at small input voltages (+-50 mV, i.e. normal output for a guitar). Adding a stage before the cathodyne will raise problems with the headroom of that stage, and both spiking as well as frequency doubling will occur in that stage.

If i use a simple resistor the octave effect may vanish with larger rail voltage, even with 30 V .

Maybe i should check  versions with 2 ECC86 (dedicated LoV tube) or 2 ECC88 (designed for relatively low anode voltages)

[idiot savant]

Cool thread!

A few years ago I experimented with a green ringer sounding tube octaver using pentodes. It started as a random circuit fragment posted by Merlin on the ax84 forums.

It has a very ring mod/sitar-ish sound. But seems to be a little exploited method of frequency doubling.

I'm pretty sure I used a different gain control method, but I can't remember the exact version I built.

[PRR]

> That's already slightly overdriven.

IMO, it is real over-driven. It approaches a Precision Rectifier. Prec Rect throws huge high-order harmonics.

> The harmonic content of the octave signal varies strongly with the input level.

Starve it. Let idle current go very low. Let cathode resistor load the cathode signals. The sharp notch of the Prec Rect become non-precision soft round-overs.

> the octave effect vanishes at small input

Basic 2nd-harmonic distortion tends to rise as the *square* of level (3rd as the cube, etc). So yeah, the output will leap and vanish, exaggerating the input dynamics.

On my sim (below), gain varies 20:1 with input signal varied 30:1. (I suspect 1.0V invokes another nonliniearity, tho the wave looks fine.)

The Prec Rect does not do this (if sufficiently precision to near-zero levels). But at all levels it has a lot of high-order harmonics which may be irritating.

(Yeah, some in-between may be the soul of musical beauty.)

The "4V battery" in the cathodyne grid circuit could be replaced with an input cap and grid and cathode resistors; battery is just easier in SPICE.

> problems with the headroom

Who is in charge here? The guitar, or the designer? Headroom control is the heart of audio work.

You can pad-down *or* boost-up the input as needed.

At 20V, the output does look a bit low, say 1/2 or 1/3rd of "normal" guitar level. Add a post-boost stage. I hacked another 12AU7 unit (not shown), it seemed to be fine.

http://s12.postimg.org/qccaa7qod/Octaver_12_AU7_20_V.gif

[PRR]

> a little exploited method of frequency doubling.

Had to stare at that, and still not sure.

There is a trick, I think using un-bypased G2, to give two "push pull" outputs from one pentode. You aren't using the second output at G2. But the scheme gave a really kinky "transfer curve", you have a mighty large cathode resistor, it is possible it works on "the next bend out" from where it is a lousy phase spitter.

[idiot savant]

> a little exploited method of frequency doubling.

Had to stare at that, and still not sure.

There is a trick, I think using un-bypased G2, to give two "push pull" outputs from one pentode. You aren't using the second output at G2. But the scheme gave a really kinky "transfer curve", you have a mighty large cathode resistor, it is possible it works on "the next bend out" from where it is a lousy phase spitter.

Here's the link and discussion.

http://www.ax84.com/bbs/dm.php?thread=379279

I remember having to read it several times before it sunk in.

Pretty interesting, but it works!

[PRR]

me> a post-boost stage. I hacked...

http://s9.postimg.org/uz1tn1xsf/Octaver_12_AU7_boost_20_V.gif

[bea]

Thats mostly the circuit i started from.

The difficulty with this one arises from the nonlinear (quadratic?) expansion of the output signal. BTW: the Green Ringer might also have this property.

BTW: i just put it "upside down" - grounded the cathodes and fed the anodes through a common resistor. Mixing point at the junction of the anodes. Larger output, especially at small signals, but of course also the quadratic behavior.