Output tube snubber kits, any experience?

[SoulFetish]

Technically yes , although thatīs not the main point making it such a breakthrough.
The real trick there lies in making a powerful Class AB2 amplifier, "hidden in plain sight".
They can smash those grids with some 20mA on demand, any day of the week.
Grids can get as much as as 20 or 25V *positive* on them.

Al this goes well beyond any classical Cascode design, which usually delves only in relatively tame Class A1 .
So yes, the Music Man power stage is a Cascode.

[teemuk]

Some guitar amps that use cascode: Vox V125 Lead, Ibanez Starfield series, Marshall 9001 preamp, to name a few. Of course they are bass amps but Aguilar’s DB750 and DB659 also employ multitude of cascode stages. They are not an extreme rarity.

Of course the cascode is a poor choice if you actually want to capture characteristics of a single-ended triode gain stage since cascode’s going to have characteristics more similar to push-pull topologies. I think its commonly employed with FETs mainly because it minimizes struggles to match them. Not to mention it clips "softly" like tubes reputedly do (though it actually clips too soft and too symmetrically compared to generic SE common cathode circuit).

---

Back to real topic: I can instantly mention one “classic” tube amp with “conjunctive filter” and that is Dr Z’s Carmen Ghia. …Excuse me, (cough cough) I mean Hammond AO-35. There are some other vintage (guitar) amps where you can see the circuit used as well (though I wouldn’t consider them classics) and yes, Radiotron Handbook does explain why the circuit is used so it’s not just some snake oil.

Almost every solid-state amp will have such circuit too, as it is basically nothing but a “Zobel network” that compensates for increasing load inductance at high frequencies by adding parallel capacitance (and a current-limiting resistor). Can make marvels to overall stability and yes, by doing so also reduce the fizzyness of overdrive characteristics and crossover distortion. In some cases (such as this “snubber” –thingy) the circuit is tuned to operate at audio spectrum and plainly acts as a low-pass filter so yes, it removes even more fizzyness.

You will more likely encounter a conjuncctive filter in a tube HIFI amp than in a tube guitar amp, but yeah, most tube GUITAR amps do not even try to reach design extents of HiFi amps. They just pretty much use the simplest circuit that still works with modest reliability. But sometimes one needs to add a conjective filter to achieve that goal (e.g. Carmen Ghia).

[Pedro Vecino]

Probably the most popular amp that uses it is the Pro Junior.
http://www.thetubestore.com/lib/thet...atic-Rev-F.pdf

[bordonbert]

Just picked up on this thread in my wanderings. The 8.2Meg resistors are most likely there to balance the voltage across the two caps. If the caps have unequal leakage then the voltage across them is shared according to their leakages. I've had to use the same setup when putting 800V across a pair of larger caps in the past, (to be dropped in the sea and towed behind a boat no less). As a rule of thumb the current through the resistors is made to be about 10x the maximum leakage so it swamps the current leaking through the caps and sets the voltage levels across each. In the article it actually says "The two lower-power resistors essentially spread the voltage evenly across the two capacitors". Now whether that is necessary given the fact that you can't find a quoted level for leakage for an orange drop capacitor is open to argument but it is a nice detail to get right.

As to the JFET mu-amp based diagram posted by JMF, it definitely isn't a cascode. That example has the upper JFET stuck on as the load of the lower and the output taken from below it. A cascode utilises a second transistor to isolate the first (usually lower) from the output at its drain. Considering it as it is for BJTs, the second is a low input impedance common base stage acting as the collector load of the first and hence giving it a very low gain. In this way bandwidth is increased significantly by isolation from the Miller capacitance. The common base has a current gain of ~1 so keeps the gain of the stage as it uses the same collector load resistor as before. That circuit isn't even a mu-amp setup. It's really a SRPP with a dodgy output take off point, (or perhaps SEPP, see the reference to your own Merlin's site below).

It's always a question to me whether the diagrams you see for effects pedals using the so called mu-amp circuits are correct and as the designer intended or whether they have suffered as most seem to do from taking the output of each stage from the wrong point. If that additional upper source resistor is included, and it does a lot of good there, is there some benefit I can't see in taking the output from its lower end? If the output is taken direct from the source of the upper FET at the top of the resistor, gain is increased significantly and output impedance is dropped greatly. Now that gain increase might not be what is wanted but then surely gain could be better controlled in the lower source or, at least, the resistor becomes a bit redundant. It is no longer isolating the lower stage from the output.

See here for Merlin's analysis of the circuit in a valve context: http://www.valvewizard.co.uk/SRPP_Blencowe.pdf and here for the SS equivalent http://www.geofex.com/Article_Folder...p/modmuamp.htm

I know many here are pretty much expert in this field but I'm sure many are not and theses do make an interesting read.

[Helmholtz]

As to the JFET mu-amp based diagram posted by JMF, it definitely isn't a cascode. That example has the upper JFET stuck on as the load of the lower and the output taken from below it. A cascode utilises a second transistor to isolate the first (usually lower) from the output at its drain. Considering it as it is for BJTs, the second is a low input impedance common base stage acting as the collector load of the first and hence giving it a very low gain. In this way bandwidth is increased significantly by isolation from the Miller capacitance. The common base has a current gain of ~1 so keeps the gain of the stage as it uses the same collector load resistor as before. That circuit isn't even a mu-amp setup. It's really a SRPP with a dodgy output take off point, (or perhaps SEPP, see the reference to your own Merlin's site below).

If you are referring to the stage consisting of T3/T4 in JMF's example, this is indeed a SRPP amp ("shunt regulated push-pull", admittedly a strange and somewhat misleading name) and not a cascode or ĩ-amp. It is a real push-pull stage and the upper device is driven by the signal drop across R11. In a ĩ-amp the upper device has no drive signal and just acts as a high impedance active load to the lower one, making it a single ended stage with high voltage gain.

[bordonbert]

Both pairs T1/T2 and T3/T4 are SRPPs Helmholtz. The first pair just looks a bit odd with its top JFET reversed but the configuration is exactly the same, R3 is the equivalent of R11. It also has the output point taken off in that less than optimal place at the bottom of the upper source resistor. As I understand it that takeoff point choice, below or above the resistor, is crucial to getting the maximum efficiency out of the overall stage. In particular it lowers the output impedance quite dramatically and for the T1/T2 stage it is not buffered as it is for T3/T4. The first stage is also driving the tonestack so could do with that low impedance too.

[Helmholtz]

Both pairs T1/T2 and T3/T4 are SRPPs Helmholtz.

No doubt.

[bordonbert]

Errrr, ok. Sorry, my bad.

[Pedro Vecino]

I have recovered this topic (itīs interesting) to expose this question:
I'm with an amp with 2xEL84 push pull, fixed bias, 350V, 8K OT with feedback. Among multiple tests I have done to polish its distortion texture I have used a conjunctive filter (1500pf/2000V capacitor and a 22K resistor in series connected between the EL84 plates). It has a discreet effect but appreciable using maximum overdrive (similar to a Plexi turned up to max).
Is it possible to calculate approximately at what frequency and attenuation it acts?
Thanks a lot!

[Chuck H]

I have recovered this topic (itīs interesting) to expose this question:
I'm with an amp with 2xEL84 push pull, fixed bias, 350V, 8K OT with feedback. Among multiple tests I have done to polish its distortion texture I have used a conjunctive filter (1500pf/2000V capacitor and a 22K resistor in series connected between the EL84 plates). It has a discreet effect but appreciable using maximum overdrive (similar to a Plexi turned up to max).
Is it possible to calculate approximately at what frequency and attenuation it acts?
Thanks a lot!

Well I'm not a good resource for formulas. But to be honest I don't think there is one for your request. An amplifier in clipping condition, much less that it's tubes, and also EL84's that are prone to intermodulation distortions even at lower overdrive levels. This is just an arcane area of situation that doesn't have the benefit of logged research information. BUT...

Doing some work with SPICE programs and comparing it to what I can witness on the bench indicates that you are already dialed in about as well as could be expected. I have been using a 10k resistance in series with a 1500pf cap for an OT primary shunt. This was arrived at through experimentation and works well for my particular amp. In my research I have noticed that 22k actually works better in most applications. I can't speak to the roll off frequency from memory (if I get back into those programs I'll report). I can tell you that the higher resistance does decrease the knee frequency but also reduces the effective shunt. IIRC from my observations and comparisons you are on the cusp of an audible affectation, but not quite. That is to say that the change in the useful spectrum for a typical guitar amp speaker should be barely noticeable, if at all. What I DID notice about using the shunt filter with EL84 tubes is that spikes in the OT primary during clipping can be mitigated very effectively and the 22k resistance (parallel with the ubiquitous 8k primary for EL84's) in series with the 1500pf cap seemed to be the best values overall in simulations.

Keeping spikes that connect directly to the tubes down, even in an amp that clips hard often, is a good idea. The spikes can often find their way through the screen circuit and blow tube/s. I've learned the hard way so you don't have to

[Pedro Vecino]

Thanks Chuck. I worked on it for thirty minutes and didn't even bother to make it selectable for comparison.
I've tuned it by playing with full gain and a post-pi master volume to limit the volume. It is when the complex harmonic content appears (with max overdrive) that the effect is noticeable and it gradually disappears as it goes down again towards the clean sounds. All this, by ear, since the filter is still there. I have also used an attenuator to help me handling higher powers.
I first tried 2200pf and 10K. After 1500 and 10K. After 1500 and 15K and finally 1500 and 22K.
My idea is that something happens around 6000Hz, but I say it intuitively. I find it possible that certain speakers fail to detect it. My 4x12 Peavey cabinet with Jensen (the reference one) shows these things better I think than a 1x12 cabinet with a Creamback 65. I haven't tested enough as this speaker (it's new) needs volume to make it sound like it should.
The effect on the spikes that you comment is very interesting too.
Thanks again.

[Chuck H]

Ok...

Here's some SPICE stuff. I used a Tiny Terror circuit cranked up about two thirds for the simulations. The test points for the spikes were at the el84 plates. One plot with the 22k/1500p shunt filter and the other without. I overlaid the frequency plot for easiest comparison. The effect at 6k is less than 1dB.

[Helmholtz]

A Zobel network across the OT primary has several effects:

1) It is a moderate low pass filter, influencing open loop frequency/phase response.
If there is global NFB there will be little effect on closed loop response until power amp clips and NFB loses its power.

Exact calculation/simulation of the effect on open loop resonse seems complicated and would need to take into account:

- tubes' internal plate impedance (varies with plate current)
- OT leakage inductance
- Reflected speaker AC resistance and inductance (both vary with frequency).

2) It's a snubber, damping OT primary resonant spikes/ringing by means of the series resistor, which absorbs energy at HF.

3) It affects closed loop stability.

4) It can be used to compensate increasing speaker impedance and associated increasing speaker response at high freqencies.

5) The Zobel acts as an additional load at high frequencies.


Effects are partly interdependent and optimizing one of them may compromise other effects.
So one needs to decide on its main purpose and vary component values for best results.

[Pedro Vecino]

Thank you very much to both. No more can be expected regarding the above.
The idea of using the filter comes from the interest in giving it a touch of warmth with the gain to the maximum. Slightly less gritty and somewhat more fluid. The amp sounds very good on its own but with the gain at maximum the EQ loses much effectiveness and on the bright channel practically all the time I use the presence to the minimum. The filter also contributes to a slightly better centering.
Another thing I have observed is that it reacts differently compared to a slight snubber placed between the PI plates. It seems better integrated into the amp (I'm talking about very small but existing nuances). All this accepting that what one hears (feels) playing is different from what one hears as a mere spectator.
It is difficult to believe that that area between 5K and 10KHz, seeing the small difference between the curves, acquires so much relevance but I want to think that this is where the harmonic content in overdrive sits and that is always delicate matter.

Thanks again!

[Helmholtz]

You might also try a Vox AC30 type cut control between PI plates.