Anything unusual about this Hiwatt t20 12au7 phase splitter?

[frank57]

I drew this up as best I could from a blurry photo.
Is there anything unusual about it?
It uses a 12au7 which you don't see much these days.
It's the New Hiwatt t20 made in China amp.

[sluckey]

The 12AX7 is the split load phase inverter. It has a gain of slightly less than one. The 12AU7 is just an amplifier to bump up the amplitude of the PI signals. I've never seen this arrangment used on an EL84 guitar amp, but the circuit is not unusual.It's probably just part of the Hiwatt "let's do things a little different" mojo.

[frank57]

Is there any advantage to using that phase splitter?
What other amps used it?
Here's the power section.The one thing I don't have is the preamp.
The t20 has 2 12ax7 and 1 12au7 and 2 el84.
 

[sluckey]

No advantage that I'm aware of. Just a different circuit. Some say they can produce some bad sounding harmonics.

The split load, aka concertina or cathodyne, has been used in many amplifiers, Fender, Ampeg, Sunn, Orange, Gibson, etc, etc. Look thru Doug's schematics and you'll find it everywhere. Since the gain is less than one they are usually preceded by an amplifier, but in some cases, such as your example, the amplifier appears after the split. The old big Gibsons (Titan, Mercury, Atlas, etc.) used the same circuit as your Hiwatt.

[HotBluePlates]

What's odd is that unless the supply voltage is very low, a split-load should have all the voltage needed to output big enough signals to drive cathode-bias EL84's til they puke. Then, the lack of gain from the split-load itself is made up by including a gain stage ahead of the split load, to generate a big enough input signal to the phase inverter. It should only take 11-18v of drive signal per EL84 to push them to the max (drive signal peak voltage will equal the bias voltage).

Where you typically see the arrangement you have is with big fixed-bias output stages, where there is a split-load stage followed by the differential amplifier to drive output stages with 50v+ of bias voltage. Or, where there are cathode followers in place of your diff-amp to drive the output tube grids slightly positive (class AB2 or class B2 operation).

But if your schematic is correctly drawn from the blurry photo, they're just doing things that aren't necessary. I can make a voltage divider out of a series-parallel mix of 26 resistors, but I only need 2 resistors to make something that performs the same function to the same standard. Assuming a correct drawing, they've over-complicated things, because the same net effect could have been gotten by using a single triode ahead of the split-load as a driver, rather than 2 triodes after as a diff-amp.

[firemedic]

Maybe the differential triodes after the PI are there to overdrive the power tubes without risking the ill effects of overdriving the cathodyne itself?

[HotBluePlates]

The thing is, we normally hear of such ill-effects with a split-load used to drive a power amp using 6V6's or bigger tubes. Those tubes use a bias voltage on the order of 20-35v, depending on the B+ voltage and output power desired (which then leads in to a choice between class A or class AB).

Because of the high Gm of EL84's, very little bias voltage is required and overwhelmingly they are cathode biased. Note that the AC30 schematic indicates the EL84's idle at 10v of bias, rising to 12.5v at full output.

Any triode in split-load with more than 200v of B+ can drive such an output stage ragged before it comes close to giving up. It could probably with even less voltage, but I assumed a peak to peak output double the bias voltage, with a minimum 100v left across the triode at all times, and a slight safety factor when arriving at the 200v number.

The case for split-load distorting probably makes more sense when the output tubes start pulling grid current (when the input signal is SO big it drive the grids with a peak voltage larger than the bias, causing the grids to go positive and pull current). In this case, the cathode output presents a lower source impedance than the plate output, and is more able to supply the grid current (up to a point). The plate output of the split-load then is more likely to be loaded by the output tube pulling grid current, and its gain and output would sag. Funny enough, that should result in more even harmonic distortion from the output stage due to dissimilar drive signals, but blocking distortion becomes a problem at that plate output.

So if the characterstics of the split-load are not favorable, the differential amp after now has two plate-output stages, which should be even less desirable.

A different way to side-step the issue would be a "KAppler phase inverter" which adds to the split load provide two low-impedance cathode outputs.

So I'm missing something significant (like a B+ over 400v and a fixed-bias supply), or the schematic is incorrect, or new Hiwatt just did something silly to waste a dual-triode.

[frank57]

I don't actually have the amp, but it has buckets of gain from a demo I saw.
Here are the original photos,but it's very tough to see what's happening.
http://licklibrary.ceros.com/iguitarmag/iguitar-mag-issue-5-john-petrucci/issue5/page/106

[sluckey]

That is hard to read! JPEG artifacts and 46000+ colors to represent a black and white original! I'm surprised that someone using Photoshop on a Mac didn't understand the limitations and advantages of different image types.

[HotBluePlates]

I see... I thought you had a chassis picture.

We also don't know what's up with the preamp, as the stage after the master volume is "part of the output stage" by virtue of of the feedback loop running from the OT to that stage, prior to the split-load.

It is either a brain-damaged design, or the designer is trying to throw in a lot of extra stage, hoping they'll distort and add to the mix.

The addition of the diff-amp is bad, because there are coupling caps between it and the split-load, and from the diff-amp to the output tubes. Add in the phase shift from the (probably not-excellent) chinese OT, and you're probably begging for oscillation. You could call that feedback around "5 stages" (counting the OT as a stage). Old school designers, when wrapping feedback around 2-3 stages, spent a lot of ingenuity try to eliminate coupling caps to avoid phase shifts. If you get enough phase shift and loop gain, your negative feedback turns into positive feedback. A single, perfect cap has a theoretical maximum shift of 90 degrees; reality is the shift never quite gets to 90 degrees due to ESR and lead resistance. So feedback around two stages never reaches 180 degrees of shift away from the mid-band signal, and you're safe. With 3 stages enclosed in a loop, you're guaranteed to reach 180 degrees of shift at some frequency extreme, so you have to take steps to either slug gain or eliminate sources of shift.

Maybe that's part of why they have 220k grd reference resistor everywhere, when that would just impair individual stage gain (again, another dumb idea on their part). That's not really a good way to eliminate ringing, because you're still not addressing the issue of the phase shift at very high (or low) frequencies.

If you're right about the dissimilar plate loads of the diff-amp (maybe they're 90k and 82k, or the like), that's dumb too, because it's a diff-amp, not a long-tail. With both inputs driven by (presumably) equal and opposite input signals, there's no need for plate load compensation; that is used when you drive one grid while keeping the other grounded (generally through a cap) or feeding it a feedback signal.

I think this explains why your Hiwatt was always suffering from an incurable buzz. The modern Hiwatt company is just doing stupid stuff, and setting itself up for problems. I imagine there is a similar lack of thinking given to the layout.