conjunctive filter

[andrew]

Hello, I noticed on the Blues express A94 and on an Octal Fatness there is a conjunctive filteroff of pins 7 on the EL84. Also on the Blues Express' preamp pin 8 there is a 15k/20uf combination but conecting both ends is a 1uf film cap. what do these mods do? Has anyone used either one on their non-Blues Expresss amps?

[Darkbluemurder]

I have never tried it but I understood from Steve Ahola's report that he used it to tame highs in the Blues Express. I guess that it's used either as an alternative to global negative feedback or in amps with very little global negative feedback. I read in an interview Dr. Z gave to ToneQuest Report that he uses conjunctive filters to even the response of the power stage in one of his models. So my understanding is it does a similar thing than global negative feedback but of course may have a different sonic effect.

Apparently only very few folks have substituted global negative feedback for a conjunctive filter, otherwise I would have expected a more intense discussion on this topic. Which raises the question: is it worth trying?

[VacuumVoodoo]

Simply put conjuctive filter a.k.a. Zobel network is used to equalize/flatten the loudspeakers impedance vs. frequency characteristic. In other words it works by reducing the reactive impedance component making the loudspeaker behave more like a resistive element.
It is not a substitute for negative feedback.

[drz400]

Zobel is usually on the secondary and is good if you know the speaker you always will be using.
Corrective filter is on the primary, good if you are looking to install it into an amp. Conjuntive works very nicely for smoothing out EL84 amps.

[VacuumVoodoo]

Zobel is usually on the secondary and is good if you know the speaker you always will be using.
Conjunctive filter is on the primary, good if you are looking to install it into an amp. Conjuntive works very nicely for smoothing out EL84 amps.

Same horse by different names. Zobel was the guy who made it popular. "Conjunctive" refers to mathematical operation whereby you dimension a series RC-link as to make its reactive component (jwC)^-1 cancel/reduce loudspeakers inductive reactance (jwL). Very simplified expressions of course.
Whether its placed on transformers primary or on the speaker is a matter of preference. IMO placing it directly across speaker coil is preferred as it allows you to dimension it correctly for that particular speaker. You want the amp to see a mostly resistive load. What tubes are in the amp is not relevant here. The electrons flowing from cathode to anode have no knowledge of metaphysical meaning of their travel, they just move obeying laws of physics.

Keep up the fun.

[ODwan]

I've tried the Zobel network on two occasions. First one was to tame high end in my 2w MiniBluesExpress. Though correctly dimensioned it made the amp lifeless and stiff in reaction. It went out again and lowering the preamp voltages did the thing for me.
Second was a mod for my Airbrake, which sounded fizzy at heavy attenuation settings. A 5µF/100v bipolar in series with a 8R2/50W made it sound just fine. I had that on a switch and only used it at "bedroom" levels. At that setting the amp doesn't see any reactance from the speaker anyway.
Timo

[drz400]

[IMO placing it directly across speaker coil is preferred as it allows you to dimension it correctly for that particular speaker. .

That is exactly why I prefer it on the primary
You never know what speaker someone is going to use.
DrZ Carmen Ghia uses a primary version (No I have nothin to do with DrZ) Drz400 is a motorcycle

Actually I have heard used the primary version called "Corrective filter" I have heard others refer to the same filter as a Conjunctive filter which is a more universal term. What I meant to say was the corrective filter as explained in the RCA book

From a RCA Receiving Tube Manual:

A corrective filter can be used to improve the frequency characteristic of an output stage using a beam power tube or a pentode when inverse
feedback is not applicable. The filter consists of a resistor and a capacitor connection in series across the primary of the output transformer.
Connected in this way, the filter is in parallel with the plate load impedance reflected from the voice-coil by the output transformer. The magnitude of
this reflected impedance increases with increasing frequency in the middle and upper audio range. The impedance of the filter, however, decreases
with increasing frequency. It follows that by use of the proper values for the resistance and the capacitance in the filter, the effective load impedance
on the output tubes can be made practically constant for all frequencies in the middle and upper audio range. The result is an improvement in the
frequency characteristic of the output stage.

The resistance to be used in the filter for a push-pull stage is 1.3 times the recommended plate-to-plate load resistance; or, for a single-tube stage, is
1.3 times the recommended plate load resistance. The capacitance in the filter should have a value such that the voltage gain of the output stage at a
frequency of 1000 cycles or higher is equal to the voltage gain at 400 cycles. A method of determining the proper value of capacitance for the filter
is to make two measurements of the output voltage across the primary of the output trasformer: first, when a 400-cycle signal is applied to the input,
and second, when a 1000-cycle signal of the same voltage as the 400-cycle signal is applied to the input. The correct value of capacitance is the one
which gives equal output voltages for the two signal inputs. In practice, this values is usually found to be in the order of 0.05 microfarad

But I have never heard of a corrective primary filter being called a Zobel
Anytime I have heard "Zobel" it has been this and needs the inductance of the speaker for the math (usually unknown)

Audio amplifiers. Zobel networks are used in audio amplifiers to dampen out high frequency oscillations that might occur in the absence of loads at high frequencies. It is the commonly seen series resistor-capacitor combination located directly at the output of the driver stage, just before the output inductor (in analog power amplifiers). Typical values are 5-10 ohms in series with 0.1 microfarads. The network limits the rising impedance of a loudspeaker due to the speaker coil inductance.

[PRR]

> You want the amp to see a mostly resistive load.

No I don't. I don't care what wacky load the amp sees as long as it sounds good.

Raw pentode power amps have high output impedance, far higher than nominal load. When you drive a typical speaker from such an amp, the rising impedance gives a rising frequency response.

In an AM radio that's the last thing you want, so you drop a big cap on it.

In a guitar amp, if the speaker was tuned for nice top-end with a low-Z amp (NFB tube or SS), it may be too shrill on a naked pentode, so we use a small cap, maybe with a resistor so it doesn not get too dull.

Since it has to be non-polar and paper caps don't get cheaper at low voltages, we normally put a small-uFd on the primary. However big power amps have OT swings too large for common paper (or plastic) caps, so it's rarely done above a few Watts.

In that application it is entirely about what you want to hear. The AM radio wants to kill everything above 5KHz and lets the sum load go capacitive. The guitar amp might let an "8 ohm" speaker rise to 12 or 20 ohms before catching it with an R-C, or might load-down the treble with 5 ohms for a tame top. Adjust to taste, not to some abstract impedance goal.

Zobel's paper is general and exact. If you want to see a constant impedance across an inductive load, do a Zobel.

Audio is rarely so fussy, and often not remotely like what Zobel described. C and R-C networks on outputs get called many different names.

A few audio apps do want near-flat impedance. A particular example is the woofer behind a passive crossover. The woofer inductance will give a very different response than computed from nominal impedance. It usually leads to a "flaw": the woofer speaks in the midrange where it is not supposed to. A near-Zobel is a simple fix.

The R-C network hanging off a transistor amp has another purpose. The falling Hfe of the transistors makes the output impedance of the amplifier..... well, in brief, if you don't have the 10 ohm 0.1uFd network, most emitter-follower output stages will find a way to oscillate hypersonically.

[VacuumVoodoo]

PRR, thanks for putting this in "if it sounds good" and a wider perspective

[Roe]

a local tech suggested trying a zobel network at v1's cathode. have anyone tried it?

[iknowjohnny]

I was told a zobel and CF are basically the same sort of thing but the CF being used on the OT primary and the zobel on the secondary. True or not, I tried both and i guess i used the wrong values when i tried it on the secondary. So i really didn't hear/feel much. On the primary which has you putting them from one plate to the other, it did in fact sound very similar to what NFB and even a cut control (depending on the values used) do. But like the cut control, i felt the CF was more of a bandaid on the real problem which was to remove harsh highs and make it sound smooth. NFB seems better than those methods and for some reason doesn't seem like a bandaid to me.

[chief mushroom cloud]

what's interesting, is that the original Watkins Dominator had a .005/3.3k across each primary...and the WEM Dominator had a .002/4.7k

both sound GREAT (but I prefer the WEM design over Watkins, but the PI is different, as well)

my point, is that it can do good things, but it's a subjective sonic debate