Watt Sucking Fireball Series

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Well, an emulated triode created by a Zen or Schade external
feedback maybe a bit compromised by external phase shift.
But this network theoretically has only one Mu, and therefore
tries to make parallelishy triodelike curves.

But a real triode is never perfect sharp cutoff of one Mu and
Gm. Multiple semi-parallel paths of widely differing Mu and Gm.
"Remote cutoff" is always the reality. Smoking-Amp pointed
this out to me some time ago. I think he was right.

Examine four "perfect" triode curves (diode curve + Mu gain)
all slightly different. But individually each having parallel diode
shaped curves, that none by itself ever "leans to the right".

Four "perfect but different" triodes averaged in parallel, now
begin to lean over more like one might expect of a "real but
imperfect" triode... Representing if you will, multiple internal
paths with differing cutoffs.

Anyways, seems you are pushing more toward the perfect
parallel model with RFeedback. A lower internally averaged
Mu of itself wouldn't necessarily achieve the same thing.
Am I now randomizing you correctly???
 
Last edited:
Let me ask the question another way...

If RFeedback is preferred for its linearity over internal
Mu in the lower end of this SRPP, then why choose a
Triode for this duty?

Why perhaps could an FET + BJT cascode in place of
this high Mu triode have not given a superior result?
RFeedback/RInput and not Mu sets gain and linearity
in both cases, does it not? Wouldn't a high Gm and
low Miller seem the more relevant figures of merit?

What other magical property could a Triode contain
(besides slow warmup and microphonics) to find good
abuse in this circuit where Mu doesn't seem to be
called upon to do anything?

Not arguing against it, just asking the thought process?
I definately possibly maybe woulda gone for Triode too,
just because... Now you know my logic, whats yours?
 
Last edited:
Uhh Boy !

Earlier in this thread I made the comment that progress in amplification devices, starting with triodes, has proven to promote more non linear transfer functions as device type and gain have changed. For example, most bipolar transistors have an Hfe or current transfer gain being low at small (Ic) collector currents, increasing to a peak as Ic increases, then decreasing toward the max device limit. When a number of such devices are used in an array composing an amplifier, overall gain within the loop or even highly degenerated in "open loop" design, result in complex impedance modulation and phase distortion over the convoluted range of collector current swings and gain variations. One of the reasons that a heavily biased class A amplifier tends to sound better is that the range of Ic deviation is small relative to the range of power being spent and the Hfe variation being less is more linear than would be the case with the same parts in a class B amp. Often is the case that "triple diffused" transistors like the MJL series are used because the Hfe is relatively flat over a wide range of Ic and the design linearity center is more easily managed as is the overall impedance consistency looking into the load. I have even gone so far as to use large dies like TIP32 and TIP31's in low current complimentary differential stages because the Hfe change at low current is such a small portion of the non linear curve that it is relatively constant in a front end application. (Q) So, what is the object of this? (ANS.) Linearity.
A triode, is in application, the most linear gain device ever produced. True, when selected to operate in its most linear region where the load line intersect through its curves is most evenly spaced and the Gm relation to voltage has the greatest margin. As mentioned also in prior note, when a linear transfer function is lightly fed back, the extremes of the operation range having nonlinear modes can be trued up and present little overall transient intermodulation product because gross corrections are not being employed. Much to the chagrin of some proponents of integrated op-amp technology, you will never see one in the signal path of anything I care to hear clearly. The WSF input gain stage and its tube equivalent have a peculiar current mode going on between the lowest triodes/FETs and the top triode/FET. The action of the summing node is counter current balance. Lowest Diff triode/FET and the lower "op-amp" triode/FET trade currents going into the uppermost cathode/source but across the uppermost cathode/source resistor. The complimentary currents trade nonlinear regions and sum to a linear transfer function. The uppermost triode must have a lower more linear u than the others producing gain while the diff triode/FET sinks current counter to the voltage at circuit output. The feedback resistor was finessed with the diff triode/FET Ib/Id to operate in the linear loadline of the uppermost triode/FET. In sum, no one answer is sufficient short of a treatise, but the variations that you refer to in cutoff slope are not approached here. Rp, for all practical purposes, is constant, Miller Cr is restrained, and linearity prevails in transfer function and in output impedance. Ya know, this forum is insufficient to portray what I really want to illustrate and still maintain a life. What needs to be said here will take nearly a year to write with a design guide...................but your craziness is inspiring!
 
Last edited:
A triode, is in application, the most linear gain device ever produced.

Helps to know WHY a triode is voltage linear, certainly isn't "just because".
Even though thats often my reason...

A triode wants to be a vacuum FET and behave like a Pentode. Only the
influence of space charge, an additional unseen control surface, is what
prevents it from being flat as sand.

According to OHSchade(BeamPowerTubesRCA1938) a screened Pentode
can be tricked to have the triode curves by application of a feedback
from plate to grid such as yours. Rather than the ordinary strapping of
screen directly to plate, as that simply re-connects internal Mu feedback.

But such an external path is sometimes difficult to DC couple, and if coupled
reactively, subject to phase shifts. Even if DC coupled, any externally wired
path that may replace or augment the function of Mu is longer and exposed
to more strays.

Does the presence of an external RFeedback path that overwhelms and
possibly even defeats the function of Mu take away from the "most linear
gain device" argument? Or is it simply averaging a "theoretically perfect"
(except for external path noise) Schade emulated triode of lower Mu in
parallel with the real one?

Pushing more toward an emulated device with parallel internal 300Bish
curves that look like multiple copies of a vacuum diode than the leaning
over set more typical of unequal internal paths (the set in your avatar).

But again, what is the cost in exposing this feedback to the longer
journey outside the gain device's internal fields? And does this partially
negate the primary advantage of the triode?

Document referenced above can be found at Pete Millet's site.
 

Attachments

  • Fig35.gif
    Fig35.gif
    21.3 KB · Views: 632
Last edited:
If isn't obvious what I'm driving at in that last rant.
You have stated that a minimal feedback is applied.
Yet this statement has ignored something important.

I see in your circuit very strong, fast, and slightly
non-linear internal triode feedback. In parallel with
slower, but more linear external resistor feedback.
But one way or the other, this circuit is linear due
to LOTS and LOTS of local feedback.

It ain't linear just because. A triode would otherwise
act just like a Pentode with the same Gm. Internal
feedback is feedback too, just more difficult to see.

I'm not entirely sure that sand linearized by similar
DC coupled local feedback wouldn't be just as good
or better. Probably boils down to how Gm is curved?

Any triode is lots of parallel triode paths of slightly
random statistical spread. Hard to fake that with just
one RFeedback wrapped around one whittled down
gain device of any type. But in your case, you seem
to be trying to skew the internal average in favor of
a resistively emulated external Mu, and narrow that
statistical spread in so doing.

Sometimes I go back and reread my previous postings,
even I don't understand them...
 
Last edited:
One further example to ponder:

A triode in real high A2 also acts like a Pentode or even
sand. Mu begins to lose influence as the space charge
thins out. Application of an external RFeedback could
bend those curves back into the normal Triode set...

I am trying to show the parallel between RFeedback
and Mu. They are really one and the same except
for the statistical spread of multiple internal paths.
 
DIY Low Complexity AMP

This is the cleaned up version of the one above. It was a least-parts-count Guitar Amp that was designed for a friend in 1992. The circuit has a couple of curious features. There is local feedback, (R21/R1)+1, around the input transistor pair to set the gain there to 26db and then there is a global feedback, R6/R5 which sets the overall gain to 26db. Even if the R6 feedback resistor is removed the amp remains relatively linear with a open loop gain of about 52db. This design would reinforce a higher even order harmonic when driven hard which makes it more useful as a guitar amp since the even order harmonics are octave related and sound "FAT" or "Warm". R16 is a 16K in series with a 10K trim pot so the output stage quiescent current can be set from 0 to 1.4 amps. I set it to about 600 milliamps. Both MosFets are driven counterphase from Gate to source by relatively low, 1Kohm, impedances allowing ridiculous bandwidth if a compensation cap is not used across R6. This input stage is bootstrapped to raise the input impedance of the global summing node at R6/R5 improving linearity. The upper MosFet gate drive is bootstrapped to present a stiff modulating voltage source from gate to source and couterphase to the lower MosFet. Bias decoupling capacitors, C5&C7 improve the 56V rail rejection lowering hum. There was of course a Fet preamp driving this stage to suite the guitarist which at 1Vp or .707Vrms, drives full output without clipping. Heat sink temp rise should be kept less than 30 degrees C above ambient. A rule of thumb is to use .1 sq ft. per watt of surface area to keep the heatsink at 20 degrees C above ambient with convection cooling. This is generally true for black anodize sinks, vertically aligned with 1.5 to 2" fin height, and 1/4" fin spacing. Both sides of the fin are counted as surface area. This amp had a plate about a half square foot of 2" fin stock.
 

Attachments

  • WattSuckingFireball1992Pub2.jpg
    WattSuckingFireball1992Pub2.jpg
    393.4 KB · Views: 583
Last edited:
That's Cool

I think that's very cool. Its self bias feature is pretty tight <40 mv but it has to operate at a high watt bias current or all hell breaks loose at higher power. Input impedance has to be relatively low so the gate capacitance doesn't swamp the frequency response. Must work on split supplies to establish bias to the bipolar balancing act; I like self biasing methods, see Ref:2N4401-3-AMP; and the "gain pivot" for the amp is the transfer function of the lower MosFet. Zen kinda in a twisted way. This type of circuit convolution is exactly the type of thing that can nudge a change in thinking habit. Wierd, but cool. Below is what's left from optimising; Madman Muntz lives; the lower boot had little or no influence. There doesn't seem to be a convenient way to use the small bipolar devices in the signal gain path, the're busy balancing the output for all practical purposes. The intermods are a bit high but the circuit twist is interesting stuff.
 

Attachments

  • ThingAmp.jpg
    ThingAmp.jpg
    243.3 KB · Views: 619
Last edited:
OK... I see it now a little clearer how the simplified fireball worked.
Seems a concertina splitter around Q2? Q1 handles the feedback,
and the rest of it is all bias... Very much a tube circuit translated
over to the sandstate, was that a Williamson???

Mine's a Zen (Schaded MOSFET triode emulator) like you spotted
already. Then the SRPP above is a crazy hybrid of Nelson's Aleph,
Waveborne's servo'd CCS, and MJK (split SRPP resistor) style Anti-
triode. I claim nothing original.

The negative bootstrap was pretty pointless, I don't recall anymore
exactly why I thought it was necessary? Probably nitpicking idiocy.

You can loosen it up a little into AB by putting Schottky in series
with each of the output resistors. I'd suggest MBR745 to start.
Otherwise it tends to glitch a little hard when asked to do the
impossible. You noticed raising bias was another workaround.
You sho hooked on fireballs...

I had a more evolved version on my drive somewhere, but havn't
had time to go digging or reinventing that wheel. Maybe later...
 
This has the same issues (like pointless negative bootstrap) you noted....
But I post it for your entertainment, since I can't seem to find the one
I was actually looking for...

The challenge here was keeping the plate (triode feedback loop)
voltage high enough that the grid didn't start leaking. So I settled
for dumping only half the output swing back into Mu... I guess you
can call that a 50% ultralinear strapping...

I remind that voltage gain from plate to cathode is roughly 1/(mu-1).
Which at this ratio is totally voltage attenuation, but current gained.
Concepts stolen from Smoking-Amp and OldEurope.

About as weird as you could possibly hope for.
 

Attachments

  • ULAsh.gif
    ULAsh.gif
    47.5 KB · Views: 478
Last edited:
And

The Low Complexity Guitar amp follows a circuit that I've held true to form from the beginning. It is as described, a Sinclair/Peterson structured amplifier. Where this is dramatically different from Zen/Aleph is that both MOSFET output conductance elements are driven the same way, gate to source, in counterphase, no static current source or slave load. R9 is the load resistor of a Q1/Q2 composite NPN transistor whose emitter Re=R1+R21 not only serves as a V/R current modulator through this composite but also as the composite's local gain feedback loop. AC voltage developed across Re is mirrored across R9 and that voltage is AC referred to the M1 source by the bootstrapping cap C4. Divider R12/R13 simply sets a quasi Vdd/2 at the output while R16/R17 bias M2's gate to set a quiescent totempole current that tends to also hold the center in check by DC feedback. Four active parts, but as is, maximally utilized for multiple functions as coud be imagined at that time. If I can find the prototype I'll show a photo of the totally freeform hairball wiring method used, no board, just components wired in 3 dimentional space. Very few parasitics. Also, Simple can be finessed to a high order of performance if the balance of multitask elements is carefully ordered. "The simplest circuit will be one whose element count equals the number of internal nodal functions to be performed" Dr. Gene Slottow 1973
 

Attachments

  • Sinclair-Peterson-Simple2.jpg
    Sinclair-Peterson-Simple2.jpg
    257.2 KB · Views: 466
Last edited:
OK... I see it now a little clearer how the simplified fireball worked.
Seems a concertina splitter around Q2? Q1 handles the feedback,
and the rest of it is all bias... Very much a tube circuit translated
over to the sandstate, was that a Williamson???..

The difference between what the Sinclair/Peterson phase splitter does and a concertina is huge! Many early OTL amps used a concertina drive to the output stage which makes the upper tube of the output stage a cathode follower with less than unity gain while the lower half of the output stage has Gm x RL Gain. I think even one of the Futtermans was done this way. Though the classic RCA OTL amp used a balanced differential splitter, the upper portion of the output was, unsymmetrically, a cathode follower. This asymmetry is why so much gain prior is used and then thrown away with a feedback loop. The Sinclair/Peterson amp allows both tubes to operate linearly with gain the same way as a pair driving a push-pull transformer. There is no imbalance in the transfer function between the output conductance elements relative to their drives. Even a properly designed Williamson is very linear while requiring meager feedback.
 

Attachments

  • RCA_OTL.jpg
    RCA_OTL.jpg
    89.4 KB · Views: 445
Last edited:
Sinclair/Peterson huh? OK, got it...

For what its worth, I saw your bootstrap and knew you were
driving the upper device to have gain, and not merely to follow.
Williamson drives both devices to have gain, so maybe not that
different, except the transformer push pull rather than totem.

The presence of a bootstrap makes S/P not considered to be
a Concertina splitter? I'd never heard that distinction before,
but accept that I have not yet read the full history.

I guess Sinclair/Peterson is quasicomplimentary rather than
SRPP? As there isn't a series regulated follower thingie going
on upstairs... I tend to misuse SRPP to describe things that
maybe sometimes don't apply.
 
Last edited:
An excellent article

I guess Sinclair/Peterson is quasicomplimentary rather than
SRPP? As there isn't a series regulated follower thingie going
on upstairs... I tend to misuse SRPP to describe things that
maybe sometimes don't apply.
There's an excellent article that made me feel better about the ambiguity of perception that I've noticed relating to SRPP references by many who assume that the concept is clear. Ref http://www.tubecad.com/may2000/ I was even baffled by some of your abbreviated references earlier in this thread that I'm sure in some audience, a majority of people would have a common understanding of, however, living in a bubble, I find it more useful in first technical exchange to define all references, "not obvious to one familiar in the state of the art". Though I've been designing with everything including tubes for 44 years, I seem to have a less than current vocabulary incorporating many of the abbreviated things I see referred to with ease by others like yourself. " Zen, Aleph, Anti Triode, SRPP, etc. "quasicomp" to me means NPN and NPN/PNP composite or the inverse PNP and NPN/PNP composite. If an amplifier uses only one polarity of device in its structure then quasicomp couldn't apply, or am I out of touch here too?
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.