PI Optimal headroom

[dennyg]

I’ve been deep diving phase inverters lately.  Conclusion I’m drawing is that for high gain pre-amp designs I want to maximize headroom in the PI and PA to minimize PA distortion, assuming that PA distortion on top of a square wave will super-saturate the tone, turning it to mush.  I’ve drawn up various load line models for 12ax7, 12at7 and 12ay7 and found that the 12at7 with 100K Ra’s, 27K tail and 1K Rb will provide greatest input headroom  among the three tubes, but with about half the output swing of a 12ax7.  I modded the operating points of my highest-gain design accordingly and the results were audible, commensurate with the models.  But I do have a question:  does the higher transconductance of the 12at7 contribute to it’s effectiveness in driving power tubes?  My gut tells me that the only thing a power tube cares about at the grid is voltage (i.e. Vgk), regardless of how much current and power is transferred to it.   Yes/no/maybe so?

[92Volts]

If you overdrive a power tube (grid above 0v) it will draw grid current. Because current is only drawn in one polarity, it's limited by the coupling cap, and can lead to blocking distortion (basically, DC bias shift). But this effect will somewhat interact with the PI too, so PI tubes may sound different in those cases.

In "normal" (clean) operation, you need to push power into the bias (grid reference) resistor through the coupling cap. A 100k grid reference, which I'd consider typical, is somewhat low for a 12AX7 to drive and will limit its gain and clean power swing. But if you're already modeling the coupling cap and grid resistors-- I don't think you'll find any surprises. When you have grid resistors at or below 100k, for larger tubes or multiple parallel sets, you'd start to benefit from something other than a 12AX7.

The first point I mentioned-- overdrive happens when Vgk=0 -- is an easy rule of thumb for how much signal you need. If you have -35v on the grid for bias, you only need a 35v signal. The last time I messed around with driver stage modeling, I was trying to drive a big triode amp and needed like +/-80v of signal. Even then, it was do-able (I used a 6SN7).

[dennyg]

Oh yes, good point about the low 100K input impedance which most fixed-bias power tubes require.  Makes me wonder why higher gm valves aren't used more often in phase inverters.  After all, most pre-amps are going to deliver 10V+ to a PI at moderate MV settings, and even with a modest amplification factor of 30, the power tubes are going to get plenty of signal voltage to work with.  But maybe i'm not seeing the whole picture hence the post.

[92Volts]

12AX7s are cheap and easy to find. Using all 12AX7s through the preamp and PI makes the amp more idiot-proof... even non-idiots who are careful to install the right tubes would appreciate only needing to stock one 12A*7 type as a spare.

Because negative feedback is often applied to the phase inverter, you sometimes might need lots of gain. The 12AX7 of course gives more gain in the right circuit, and continues to perform well up to heavy-ish loads. Eventually you do need a 12AT7, and they actually are common in big Fenders, bass amps, and others that need large signal swing or drive a low power tube grid resistance.

I'm partial to 6SN7s myself because I think octal tubes look cool, and its easier to solder to octal sockets

You can make most of these tubes work in most cases, so I think the simplicity of using the same tube everywhere might really be the answer.

[dennyg]

hey man, thanks for jammin' with me on the topic.  so took a look at the 6sn7 datasheet - dang, that tube has massive headroom!  the grid curves are almost parallel so you'd get minimal harmonic distortion too.   looks like you could configure this thing to take 40Vpp or so in a PI before it would start clipping. hmmm- got me wheels a turnin'

[shooter]

took a look at the 6sn7

I like the 6SN7 as a driver for SE, I believe I've got something close to 55Vacrms.  I've driven a quad of EL84's, they woop easy, also used it on KT88. 
as long as you're modeling, look at PI trannies (inter-stage), especially when overdriven. (lots fewer parts to "plug-in"  )

[dennyg]

PI Trannies? never heard of such a thing - where can i get a download on how they work, how they wire up etc?

[shooter]

dig up some old Gibson schematics like GA-5t.

I used this tranny in a plexi build driving a pair of EL34's. (It's the 1st amp that came closest to profitability  )

http://www.triodeelectronics.com/unintrui.html

[dennyg]

interesting - the GA-5T has a cathodyne PI but I found the GA-15RVT which has a PI trannie. 
Looks like the PI is driven off plate of the last stage.  So for a plexi where the tone stack goes direct to the PI, where did you drive the PI from? 

[shooter]

Here's the schematic.  Originally I had the TS recovery as a CF but didn't have enough drive, so I changed it to a gain stage that drove the PI.
the IS trannies can't handle big current, so the 2 builds I used trannie PI, I cap coupled so it's purely AC, which in itself offers some unique tonal effects

[92Volts]

hey man, thanks for jammin' with me on the topic.  so took a look at the 6sn7 datasheet - dang, that tube has massive headroom!  the grid curves are almost parallel so you'd get minimal harmonic distortion too.   looks like you could configure this thing to take 40Vpp or so in a PI before it would start clipping. hmmm- got me wheels a turnin'

All depends on the voltage! 6SN7GTA/GTB versions are rated for 450v plate, but 1500v peak (zero current) plate voltage. If you run them with a plate resistor (like 99% of preamps), your B+ can be as high as you'd ever see outside a radio transmitter.

At 530v I've modeled a LTPPI that outputs 231v peak-peak (80v RMS). In practice? I built it and it worked but I never needed that much signal. As a single-ended driver it can do 340v peak-peak (119v RMS) which might actually be needed for big triodes. Both examples driving a 100k output grid resistance.

The sky's the limit, except you're unlikely to use higher voltage for the PI/driver than you do for your power tubes. Still, it will work for any common guitar amp tube.

[jojokeo]

I’ve been deep diving phase inverters lately.  Conclusion I’m drawing is that for high gain pre-amp designs I want to maximize headroom in the PI and PA to minimize PA distortion, assuming that PA distortion on top of a square wave will super-saturate the tone, turning it to mush.  I’ve drawn up various load line models for 12ax7, 12at7 and 12ay7 and found that the 12at7 with 100K Ra’s, 27K tail and 1K Rb will provide greatest input headroom  among the three tubes, but with about half the output swing of a 12ax7.  I modded the operating points of my highest-gain design accordingly and the results were audible, commensurate with the models.  But I do have a question:  does the higher transconductance of the 12at7 contribute to it’s effectiveness in driving power tubes?  My gut tells me that the only thing a power tube cares about at the grid is voltage (i.e. Vgk), regardless of how much current and power is transferred to it.   Yes/no/maybe so?

I don't think so. Simple answer for clean ltpi for our tubes & amps is the tried and true 470r & 10k tail and use a 12AT7. You need to design for the tube you plan to use. I have designed for 12AU7 but end up using 12AT7 in many cases. I also have plugged in 12AT7 on many many amps that were designed for 12AX7.
The 12AT7 seems to be ticket for both extremes settling for a cleaner pi for the designed 12AX7 amps and gives that little extra tone kick using instead of where a 12AU7 would normally be used. YMMV? But this is my experience.

Suggested reading: http://www.aikenamps.com/index.php/designing-long-tail-pairs-the-load-line-approach

[dennyg]

Thanks Jojo - after more reading (including Aiken's paper) here's where I am regarding operating points for a 12AT7 PI vs 12AX7:
* reduce Ra values from 100K - Aiken says optimal is 2xRp which for a 12AT7 would only be 22K - Fender used 47K's I believe for 12at7's which I'm going with next.  I trust their years of experimentation over pure theory, at least as a starting point
* the higher the tail voltage, greater the balance but at expense of Vout. Randall recommends 25% of B+ or so. In my case 27K provides 20% of B+
or around 65V yielding an effective plate voltage of 265V (with B+ of 330V)
* a load line using 47K and 265V yields a center bias of around 2.5V with Rb of 1.2K (i.e. approx 650R x 2)

Let me know if I missed anything - i'm going to tweak two of my amps over the weekend with the above and report back.
BTW, I did read somewhere that higher transconductance is a benefit to driving power tubes - something about smoothing the transition into power tube distortion due to grid current?  Any thoughts on that appreciated too

[jojokeo]

Yes, too much theory and over-analysis can lead to paralysis. Many times with all the time spent you simply could have have removed & replaced a bunch of things doing the trial & error method and using your ears - which is many times the best course to take anyway? But is also nice to understand things and/or know which directions and roads to go down?

At first glance your numbers look appropriate? At least for starting points as you say.

You get higher bandwidth the lower gain the tube has and lower bandwidth with the higher gain the tube has. It is a factor a plate resistance, transconductance, etc. I don't have my notes with me on my laptop at the moment.

[shooter]

your numbers look appropriate

agreed, cranking out the data is fun, bustin out the scope n meter n guitar n speakers n n That's like drivin a jeep down a 2-track on a sunny day