Emory Cook and Gus Jose's "The Ultimate Amplifier" (Long Post Warning)
Pio1980
AK Member
Electrons.
My bench supply only adjusts to +400V, so I can't duplicate those conditions without moving my current project. I have a system that's variable to roughly +1,500V using a variac, but it takes up almost half my bench space. I wonder how much the raw supply itself might have been moving during the oscillations? If the tubes were visibly flashing, the frequency might have been too low for the supply to anchor it well. And on that note, your test might be more relevant to the use of a typical amplifier power supply, which wouldn't be regulated like a bench supply. Anyway, I'll try to run a test similar to yours when I have more space-time available.
Jack
Ok, I've been slowly working on this project and I'll have a real update post soon. Hopefully I'll be able to place the order for my first new chassis this Sunday and get started on actually building.
Before finalizing everything, I wanted to gather some input on the cathode bias resistor setup for the 5881s. The schematic calls for a single 90 ohm resistor shared between all four tubes. Is there any justification for this other than simplicity? My understanding is that shared cathode bias resistors can lead to current imbalances and don't really offer any benefits. However, I want to keep the design as close to the original schematic as possible. Also, it'll be easier to chassis mount a single 90 ohm resistor than to incorporate four separate resistors of adequate size. So, if people don't think the shared cathode bias resistor is a real issue, I would prefer to leave it.
On the assembled amp I purchased, a tech added a 10 ohm cathode resistor for each tube in series with the 90 ohm wirewound. Figured I'd mention this in case it's relevant to the analysis.
Before finalizing everything, I wanted to gather some input on the cathode bias resistor setup for the 5881s. The schematic calls for a single 90 ohm resistor shared between all four tubes. Is there any justification for this other than simplicity? My understanding is that shared cathode bias resistors can lead to current imbalances and don't really offer any benefits. However, I want to keep the design as close to the original schematic as possible. Also, it'll be easier to chassis mount a single 90 ohm resistor than to incorporate four separate resistors of adequate size. So, if people don't think the shared cathode bias resistor is a real issue, I would prefer to leave it.
On the assembled amp I purchased, a tech added a 10 ohm cathode resistor for each tube in series with the 90 ohm wirewound. Figured I'd mention this in case it's relevant to the analysis.
I don't see a problem using a common cathode resistor, generally speaking. Fixed bias would allow idling the tubes at lower dissipation.
The 10 ohm resistors allow you to measure individual bias currents, but what's the point if bias isn't adjustable?
Jack
The 10 ohm resistors allow you to measure individual bias currents, but what's the point if bias isn't adjustable?
Jack
I believe there is something about it providing a degree of balance between the tubes, figuring each "side" of the output transformer is only conducting about half the time. The single shared resistor would give a more average voltage than it would otherwise. I know I've read technical particulars about why a resistor per-channel is better than a per-tube one. There are some additional reasons if we're talking stereo too.
being able to monitor even without being able to adjust could be useful to verify you don't have one grossly mismatched tube that is hogging too much current. Basically just a verification that they're all behaving reasonably well.
being able to monitor even without being able to adjust could be useful to verify you don't have one grossly mismatched tube that is hogging too much current. Basically just a verification that they're all behaving reasonably well.
Interesting, I hadn't heard that before. Sounds like I'll leave the shared resistor then.
trainbuftony
Electron Herder
Read the article on Dave's website about the st35 power supply optimization.
It explains why the cathode resistor should be one resistor rather than 2.
Personally I opted for 2, and a switch which shorts them together to make one. Switch is spring loaded, so you hold the switch to open the contacts and check bias. Let go, and you're back to having one resistor.
It explains why the cathode resistor should be one resistor rather than 2.
Personally I opted for 2, and a switch which shorts them together to make one. Switch is spring loaded, so you hold the switch to open the contacts and check bias. Let go, and you're back to having one resistor.
Here are some pictures of the chassis design as it currently stands. Aside from taking basic layout inspiration from the original, I've redone all the layout and wiring based on the schematic. I'm mounting more stuff to turret boards than the original (to minimize messy point-to-point), and I think my design will minimize wiring length and look neater underneath. My chassis is bigger than the original, so I have more room to work with. Hopefully my redesign doesn't make the amp unstable, but I'm prepared for the possibility.
I'm currently drafting scale diagrams of each turret board to verify everything will fit. Once I'm done with that, I'm going to place the order. The chassis will be expensive, but this is a year-long project and the cost in time/tools to do something this complex myself is not worthwhile. Also, Protocase includes powdercoating in their price, which is a big plus since I want these to look professional.
I'm currently drafting scale diagrams of each turret board to verify everything will fit. Once I'm done with that, I'm going to place the order. The chassis will be expensive, but this is a year-long project and the cost in time/tools to do something this complex myself is not worthwhile. Also, Protocase includes powdercoating in their price, which is a big plus since I want these to look professional.
thats a slightly different condition though. Stereo vs mono. Within a mono amp you wouldn't have the second channel to act as a regulator. Its a good read and explains why it works like it does in a stereo amp though.
trainbuftony
Electron Herder
Forgive my simplistic understanding but I'm under the impression that when the tubes in a channel share a cathode resistor it is a form of localized feedback which needs to be present in that design in order for the ratios to work out correctly on the other feedback connections in that design.
I could be totally missing the mark, but my take away over the years is that cathode bias pp amps should share a cathode resistor and that individuals are only useful on fixed bias amps.
My example is sort of a cross between the two because its EFB equipped, but my understanding of it is that the design relies on a sharing current across that resistor.
Not trying to derail your thread with discussion of my amp....
I could be totally missing the mark, but my take away over the years is that cathode bias pp amps should share a cathode resistor and that individuals are only useful on fixed bias amps.
My example is sort of a cross between the two because its EFB equipped, but my understanding of it is that the design relies on a sharing current across that resistor.
Not trying to derail your thread with discussion of my amp....
The stock ST-35 and similar amplifiers exemplify the downside of self-bias. The push-pull pairs conduct more current at higher power levels, causing the voltage drop (bias voltage) across the cathode resistor to increase. The result is a shift toward Class B, and if the tubes are not biased properly, crossover distortion will appear, increasing distortion and limiting useful output power. In practice, this means a self-biased stage must be biased "hotter" at idle than a fixed-bias stage. The advantage of fixed-bias from this standpoint can had regardless of whether bias is applied to the grid or cathode.
Jack
Last edited:
The common cathode connection in a typical AB1 push-pull amplifier is bypassed. The cathodes exhibit no signal AC voltage and cannot contribute feedback. However, at lower signal levels where tube conduction is balanced (equal and opposite), much of the signal current is inherently cancelled. The result is more effective bypassing than with separate resistors and bypass caps.
Jack
Anyone have thoughts on deleting the first 12AU7 and feeding the grids of the 12AX7 directly with a balanced signal (no phase inverter)? I haven't done any detailed calculations yet, but my source is a balanced DAC that can do a few volts out, and this amplifier's odd design seems especially suited to this type of fully balanced approach. Losing the ability to use an unbalanced source doesn't bother me.
Going to lose a decent bit of gain. Mostly in that first triode, but grid to grid at the inverter you're a little shy of 2. Probably easily a total of 20 between the two sections. Thats a not insignificant amount of voltage you'd need to make up for in the DAC.
You could re-wire things so the two halves of the 12au7 are just acting as a straight gain stage, feeding the grids from a balanced source though.
You could re-wire things so the two halves of the 12au7 are just acting as a straight gain stage, feeding the grids from a balanced source though.
Ok interesting, I will have to think about it and read some more Morgan Jones. I'm considering a balanced DAC that can supply up to 5v RMS. Wish I knew the input sensitivity -- if it was something like 200 mv I would be fine losing the gain from the first triode and the PI.