Fender's change from 12ax7 to 12at7 on the PI?

[DuaneOh]

From the brown to the black era of fender amps they changed the PI tube from 12ax7 to 12at7. Why? My only guess is that they wanted to reduce the gain of that stage.

Then in the silver era they went to 47k anode resistors, to further reduce gain? That lowered the output impedance of the PI but they also lowered the input impedance to the power tubes with 68k grid leaks, does this better control possible blocking distortion?

I know that there were other resistor changes that accompanied the tube changes. I'm just trying to understand the reason for these changes and why I would use a 12at7 in the PI.

[PRR]

I think you are using selective data.

The shift to 47k and 12AT7 comes with bigger and more output tubes delivering higher power, no?

[tubeswell]

12AT7 is able to source more current and (therefore, when used as a PI/driver), this counteracts the (bigger) loss of (signal) current through output tube grid leak resistors in amps that employ beefier output tubes (that need lower grid leak resistance to keep in a ‘safe’ operating zone).

[DuaneOh]

I think you are using selective data.

The shift to 47k and 12AT7 comes with bigger and more output tubes delivering higher power, no?

Yes, I'm trying to understand why a 5F8-A can drive 4 5881's with a 12ax7 but an AA796 Twin needs a 12at7 with 47k anode resistors? Or a different amp, Marshall 1959 can drive 4 EL34's with a 12ax7. I believe that they are all class AB1, is the current to drive these valves that different?

12AT7 is able to source more current and (therefore, when used as a PI/driver), this counteracts the (bigger) loss of (signal) current through output tube grid leak resistors in amps that employ beefier output tubes (that need lower grid leak resistance to keep in a ‘safe’ operating zone).

looking at the AB763 Twin, it uses 220k grid leaks and the 82k / 100k anode resistors for the 6l6's, why did they change the design for the AA796?

[HotBluePlates]

... an AA796 Twin ... 12at7 with 47k anode resistors? ...

looking at the AB763 Twin, it uses 220k grid leaks and the 82k / 100k anode resistors for the 6l6's, why did they change the design for the AA796?

Use your own logic:  there are 68kΩ grid leaks running down to the Bias Balance pot, rather than the 220kΩ that existed previously.  Having a plate load around 1/2 (or less) of the following grid leak resistance is generally a good plan.

47kΩ might be lower resistance than a 12AX7's internal plate resistance (at a reasonable operating point).  12AT7 might have been attractive as a balance between lower internal resistance & reasonably high amplification factor.
___________________________________

My guess for the move to smaller grid-leak & lower plate load resistance is to help recover from brief overload & grid-blocking on what is otherwise a clean amp.

12AT7 is able to source more current ...

IMO "current" gets overstated when talking about phase inverters.

Given a 400v supply and 100kΩ plate load, maximum current is 400v/100kΩ = 4mA.  Use a pair of KT88s to make up your long-tail inverter, and the current will still never get as high as 4mA.

     For AB769 Twin Reverb, 380v supply - 120v cathode = 260v, so 260v / 47kΩ = 5.5mA maximum.
     That's reasonably in a 12AT7's wheelhouse & a bit much for a 12AX7.  Actual tube current will be less than 5.5mA, because around 1/3 of supply voltage needs to remain across the tube (so more like 3.7mA max).

But let's look at what current an overdriven 6L6 might pull.

     The 6L6GC data sheet shows the grid current we might expect during positive peaks.  At only +4v the grid pulls at least 5mA and possibly as much as 10mA.  This means the grid went from ∞Ω to as low as 400Ω.  It's not possible that a 12AT7 with a 47kΩ plate load will drive that gracefully.

     Viewed a different way, the 6L6 bias is around -40v.  Ignoring the tube, the total load is 47kΩ in parallel with 68kΩ, or roughly 28kΩ.  Generating 40v across 28kΩ only take 1.43mA.  Ignoring the total load on the phase inverter is 400-800Ω, we could estimate driving the 6L6 +4v beyond the onset of clipping will require 1.43mA + 5mA ---> over 6mA.

     We already saw the 12AT7 & 47kΩ plate load probably can't manage over 4mA, so the phase inverter output will be clamped by 6L6 grid current.

          Real Amps that drive grid current use something like 6L6 or EL34 into a transformer into output tube grids.

[pdf64]

Mitigating farty power amp overdrive was probably part of the goal in reducing LTP gain and max available signal output voltage swing. Shorter time constants and lower signal gain / voltage leads to reduced bias excursion when overdriven.

All these amps you mention operate in class AB1, no current drive per se is required, it’s just a voltage signal, as far as the output valve control grids are concerned.
Hence my thinking is that the key issue motivating the changes is the limiting value for grid circuit resistance, eg 6L6 it’s 100k in fixed bias.
When the valve gets very hot, grid current can increase, which across the grid leak will tend to create a voltage in opposition to the bias. So the valve gets even hotter, grid current increases, and the valve spirals into self destruction.
In cathode bias / lower total maximum dissipation, the grid leak resistance limit can be relaxed.

[DuaneOh]

Thanks for everybody's replies.

To make sure that I understand: Changing to 47k and 68k in the AA796 Twin was done to reduce bias excursion and grid-blocking (same thing?), which causes blocking distortion (farting). This change also reduces gain so the amp won't distort as early and as much.

Going from 220k to 68k on the grid leak would increase the high pass Hz with the .1u coupling cap. Wouldn't it have been easier to reduce the coupling cap like Marshall did from the JTM45 to the 1987? I know that Marshall and Fender are very different amps, but to solve the blocking distortion issue wouldn't it make sense to make the smallest change possible in the very popular black amps?

I'm not clear on why Fender changed the 12ax7 to a 12at7, for instance, from the 6g3 to the AB763 Deluxe? I see the tail of the LTP and feedback is slightly different, I'd guess to accommodate the change, but the anode and grid leak resistors are the same. Was it done to reduce gain on the LTP?

Just trying to understand the reason for these changes. Thanks again for your help.

[PRR]

....a 5F8-A can drive 4 5881's....

Can it? The 5F8A greatly exceeds the "maximum" 0.1Meg Rgk for 6L6 or 5881 in fixed bias. Maybe these amps were killing tubes, or just too fussy about their tubes. When Leo (an accountant) asked why the tubes were dying, RCA/GE tole him to respect the limits and everybody could be happier.

Two at 0.1Meg is 50k per side. The 12AX7 is not happy pulling 50k, see the amplifier table. Both gain and max swing are way down. Later there was a similar Mu tube built double-wide for more current, but even so the usual "grunty 12AX7" is the 12AT7.

WHY would we fret too much about driver gain? The driver needs some gain, sure, but mostly gain more/less is done in the low-level stages before the driver. IF we wanted "less" we usually add some frill: fancier tone-stack etc.

[sluckey]

Fender's design engineers made changes they considered to be upgrades at the time. Some changes may have been to compensate for less than optimum wiring practices. And some may have been driven by the bean counters. Some of the changes may not make sense today. The only way to know for sure "why" would require getting into the head of the engineers. So, that leaves us to speculate and that may leave you questioning the speculation.

Many of us tend to view the tweeds, browns, and blacks of the '50s and '60 as being the best and most desirable amps Fender ever made. And all the changes made since then took away from the magic. But consider this, changing from tweeds to browns and then blacks were major redesigns, not just a simple tube change or resistor change. I never hear anyone wondering why those major changes happened. Usually money is a big factor.

[shooter]

would require getting into the head of the engineers

spending time with them was an adventure, getting into their minds is a bridge to far 

[jamaio]

would require getting into the head of the engineers

spending time with them was an adventure, getting into their minds is a bridge to far 

I am not and engineer but I work with several Electrical Engineers, from my experience they over complicate everything. My job is to implement their designs and make them work in the real world. Several times I have removed some of the complications and simplified things to make it work. There is no need for a $1000. design that can be done with a $5. part, the $1000. design would have required major work and several hours spent reconfiguring everything, the $5. part took 5 minutes to install. The end result, it works as intended and the customer is happy.

[mresistor]

would require getting into the head of the engineers

spending time with them was an adventure, getting into their minds is a bridge to far 

lmao

[DuaneOh]

Can it? The 5F8A greatly exceeds the "maximum" 0.1Meg Rgk for 6L6 or 5881 in fixed bias. Maybe these amps were killing tubes, or just too fussy about their tubes. When Leo (an accountant) asked why the tubes were dying, RCA/GE tole him to respect the limits and everybody could be happier.

Two at 0.1Meg is 50k per side. The 12AX7 is not happy pulling 50k, see the amplifier table. Both gain and max swing are way down. Later there was a similar Mu tube built double-wide for more current, but even so the usual "grunty 12AX7" is the 12AT7.

Thanks for pointing that out, makes sense as to why they went to the 47k and 68k in the silver amps. It seems that many classic amps ignore or disregard Rgk rating. Does having a grid leak resistor that is out of spec effect tube reliability? Does it effect its performance / sound?

The only way to know for sure "why" would require getting into the head of the engineers. So, that leaves us to speculate and that may leave you questioning the speculation.

I have been involved in design most of my life, and when reverse engineering a competitor's product I can usually follow what they had in mind. My discipline was mechanical, so I don't have the expertise to understand where they were going with some of these circuits.   

But consider this, changing from tweeds to browns and then blacks were major redesigns, not just a simple tube change or resistor change.

Tweeds, browns and blacks, in my view all use the same building blocks, arraigned in basically the same order. I don't see major redesigns in the circuits, they each have their own "sound" but I just see evolution.  Maybe it's apples and oranges and is useless to compare.

I never hear anyone wondering why those major changes happened. Usually money is a big factor.

That is just my thought process. If I can understand why a successful product was designed a certain way, or what they were trying to achieve with a change then I can better incorporate their successes in my design.

In a blackface deluxe it looks like you could substitute a many different triodes in the PI position, why did they choose a 12at7 when they had been using a 12ax7? Maybe it's not for a technical reason, it could be for circuit consistency across the product line, or liked the tone, or ?

would require getting into the head of the engineers

spending time with them was an adventure, getting into their minds is a bridge to far 

I enjoy discussing the minutia with other engineers... 

[dwinstonwood]

As far as I know, and correct me if I'm wrong, Fender did not sell amps. Music stores (dealers) sold Fender amps. Fender received feedback from their dealers (and, from musicians). I imagine if the same issue kept getting reported back to Fender from dealers/repair depts across the country, they made an adjustment to the next production batch, or to the next iteration of the amp model. Look at the Tremolux 6G9. The original variant apparently had EL84's (6BQ5's) running at close to 400V. I'd bet a lot of those were brought back to the store for fixing. The 6G9-A switched to 6L6GC's.

https://el34world.com/charts/Schematics/files/Fender/Fender_tremolux_6g9.pdf

[shooter]

I enjoy discussing the minutia with other engineers...

most of the places I worked, the position was salaried employee, the bean counter's knew we'd spend 10hrs debating the relevance of changing something that was currently working 

[HotBluePlates]

... It seems that many classic amps ignore or disregard Rgk rating. Does having a grid leak resistor that is out of spec effect tube reliability? Does it effect its performance / sound? ...

I, too, made the mistake of learning the details of amplifier electronics before learning the details of electricity/electronics & how tubes work.  Put away "performance / sound" for a while.

Do you have a death-grip on Ohm's Law yet?  Volts = Current x Resistance.  An implication is for any given current, more-resistance ---> more-volts.

When we bias a tube, we would like our chosen bias voltage to stay where it is.  Otherwise, the tube is no longer biased where we wanted.  So we want the only changing voltage at the output tube grid to be the AC Volts from the phase inverter.

You asked, "bias excursion and grid-blocking (same thing?)."

     If bias changes to become more & more negative, the output tube is turned more-Off & results in "grid blocking" and "blocking distortion" (which is often half-wave rectification).

     If bias changes to become more & more positive, we don't have grid-blocking but the tube is "less-Off."  This means the tube overheats.  Along the way, if the bias voltage drops to 0v the positive-going half of the waveform is lopped off and we again have a half-wave rectified output.  But you'll mainly focus on the redplating tube.

Overdriving the output tube will result in grid current on the positive peaks of the input signal.  But even when not overdriven, tubes may exhibit grid current, and the direction of that current may be into or out of the grid (positive or negative current), which means grid voltage could tend to move positive or negative.

Grid Current is usually small, in the microampere range & less.  But larger grid resistance means the voltage is displaced more for whatever current does flow.  Moreover, in overdriven cases the grid current moves from micro amperes to milliamperes, and every little bit of grid resistance matters much more.

Tube manufacturers assumed "no (serious) overdrive."  They recommended limits on grid circuit resistance based on observed/expected amounts of grid current in normal (clean) use, and how far bias voltage could be shifted by Current x Rated-Resistance.

[dwinstonwood]

HotBluePlates, do you mind if I gather your some of your posts and save them to my hard drive? I would never post them anywhere. It's just that you take so much time to thoroughly explain things, and it's an amazing resource of tube amp info. I already have some of your posts bookmarked, like this old one from 2011 on screen voltage, for example:
https://el34world.com/Forum/index.php?topic=12281.0
Thanks for all that you share!

[HotBluePlates]

HotBluePlates, do you mind if I gather your some of your posts and save them to my hard drive? ...

Knock yourself out!  But you will do better to catalog PRR's posts, as he knows very much more than me.

He just stays concise, and assumes the reader has enough background knowledge to understand the implications.  If you do not catch the implications you would do well to ask, "Why is that the case?"  Either in the thread, or via your own research.

I err in the opposite direction:  I over-explain, and under-estimate the reader's knowledge.  While that might help some understand better, it also allows others to correct me right at the point my reasoning broke down.  Most of what I know is thanks to others pointing out how I got it wrong.   

[tubeswell]

IMO "current" gets overstated when talking about phase inverters.

Yes well, even the Silverface amps weren't optimum scientific designs. A 47k plate load with an effective HT of 285V, with each 12AT7 triode pulling 2.4mA through the tail, is way below Pmax. But the load line looks better (in terms of output swing) than a 12AX7 under similar circumstances.

[dwinstonwood]

But you will doo better to catalog PRR's posts, as he knows very much more than me.
He just stays concise, and assumes the reader has enough background knowledge to understand the implications.  If you do not catch the implications you would do well to ask, "Why is that the case?"  Either in the thread, or via your own research.

Thanks HBP. And, yes, I always pay close attention to PRR's posts. And, yes, he usually has me scratching my head. Because, I know there's an important piece of wisdom and knowledge there. But, trying to see the "implications," as you say, usually sends me running to dig deeper in order to connect the dots. Ultimately, that's probably a better way to learn this amp stuff. But, it's also useful to have it "over-explained," and I doubt you could really under-estimate my knowledge of tube amps.  Thanks!

[HotBluePlates]

IMO "current" gets overstated when talking about phase inverters.

Yes well, even the Silverface amps weren't optimum scientific designs. A 47k plate load ... with each 12AT7 triode ... the load line looks better (in terms of output swing) than a 12AX7 under similar circumstances.

Certainly!

On one hand, same-supply-voltage with smaller-plate-resistance & using a tube-with-lower-internal-resistance all amounts to "more current."  That's Ohm's Law at work.

But novices read "more current" and think the phase inverter is delivering current into the output tube grids, which I showed earlier it is entirely incapable of doing.  Other errors build on the logic of assuming that's the case, and eventually they have a bit to un-learn if they wish to advance their knowledge.

_____________________________

A more helpful way of thinking of the issue is focusing on Resistance, not Current.

As I said before I think Fender chose 68kΩ grid-leak resistors, then made follow-on decisions to support that choice.  In picking a plate load resistor for the phase inverter, the common rule of thumb is the grid-leak should be 2-5x the plate load resistance (RDH4, Page 482).

     For onlookers, the grid-leak is in parallel with the plate load for AC signals.
     For grid-leak 2x the plate load resistance, the effective load is 67% of the plate load resistor value.
     For grid-leak 5x the plate load resistance, the effective load is 83% of the plate load resistor value.
     For grid-leak 10x the plate load resistance, the effective load is 91% of the plate load resistor value.

So for 68kΩ we would be thinking 34kΩ (or the standard-value of 33kΩ), though Fender chose a bit higher at 47kΩ.

In the same way, we would like the plate load resistor to be 2-5x the tube's internal plate resistance.  That's because this internal resistance forms a voltage-divider with the plate load resistor, so lower internal resistance means the tube's in-circuit gain is a bigger % of Amplification Factor.

     We have a plate load of 47kΩ, so we would really like the tube's internal plate resistance to be down to 23kΩ (1/2), or even as low as 10kΩ (1/5).

     Using the graph on Page 4 of the 12AT7 data sheet, we can estimate the internal plate resistance.
     Schematic figures imply ~2.7mA per triode, so I drew a Red vertical line at 2.7mA.
     Schematic says 245v plate & 120v cathode, so I follow the vertical up to the intersection with the Rp curve for "100v plate"
     Reading the value at the left, the 12AT7 has about 17.5kΩ of internal plate resistance.


Turns out, we're reasonably close to the plate load resistor being 3x the internal plate resistance (a good pairing).  With a Mu of 60, we anticipate a basic gain of around 60 x (47kΩ/(47kΩ + 17.5kΩ)) = 43.7 (actual gain in a long-tail is much lower due to feedback across the tail resistance).
__________________________________

So we see that thinking in terms of "bridging resistance" (where load resistance is much bigger than source resistance) is an easier rule to apply.  It also betters support logical intuition, and makes for easier back-of-envelope calculations when we consider that amps are typically designed from the output backwards towards the input.


(I had to "waste" a lot of time un-learning bad amp electronics info, and unhelpful concepts.  That's why I keep pushing what I believe to be the more-useful ways of conceptualizing amp circuits)

[DuaneOh]

I was away for a day and am trying to catch up. There are many different ways to look at a problem, and the post above has given me a new perspective. And you wrote that in a way that made perfect sense, thanks!

Here are a few answers and questions from previous posts:

Do you have a death-grip on Ohm's Law yet?  Volts = Current x Resistance.  An implication is for any given current, more-resistance ---> more-volts.

I do have a good understanding of ohm's law but only a basic understanding of how tubes work.

You asked, "bias excursion and grid-blocking (same thing?)."

     If bias changes to become more & more negative, the output tube is turned more-Off & results in "grid blocking" and "blocking distortion" (which is often half-wave rectification).

     If bias changes to become more & more positive, we don't have grid-blocking but the tube is "less-Off."  This means the tube overheats.  Along the way, if the bias voltage drops to 0v the positive-going half of the waveform is lopped off and we again have a half-wave rectified output.  But you'll mainly focus on the redplating tube.

I have a basic understanding of cut off, grid current, bias excursion and blocking distortion, but I was unfamiliar with the term "grid-blocking" and wanted a clarification, thanks.

But even when not overdriven, tubes may exhibit grid current, and the direction of that current may be into or out of the grid (positive or negative current), which means grid voltage could tend to move positive or negative.

Grid Current is usually small, in the microampere range & less.  But larger grid resistance means the voltage is displaced more for whatever current does flow.  Moreover, in overdriven cases the grid current moves from micro amperes to milliamperes, and every little bit of grid resistance matters much more.

Tube manufacturers assumed "no (serious) overdrive."  They recommended limits on grid circuit resistance based on observed/expected amounts of grid current in normal (clean) use, and how far bias voltage could be shifted by Current x Rated-Resistance.

When the valve gets very hot, grid current can increase, which across the grid leak will tend to create a voltage in opposition to the bias. So the valve gets even hotter, grid current increases, and the valve spirals into self destruction.

Thanks, this is very helpful.

Using the Marshall 2001 as an example, it has 8 x 6550 with each wanting a 50k g1 circuit resistance. With a bank of 4 x 6550’s, would it need a 12.5k grid leak to the bias capacitor? Their bias circuit shows 150k + ½ of the 22k bias adjust pot to the bias capacitor, so 161k. Is there something I’m missing or is this really bad design?

If I was to use a different way to tame grid current and grid blocking (smaller coupling cap, big grid stopper, Paul Ruby, etc.) would that relax the tubes grid leak resistor spec? A big grid stopper would be added to the grid leak resistor making the g1 circuit resistance even higher, would this make things worse?

[HotBluePlates]

I have a basic understanding of cut off, grid current, bias excursion and blocking distortion, but I was unfamiliar with the term "grid-blocking" and wanted a clarification, thanks.

A more-complete writeup is on Aiken's website.

Using the Marshall 2001 as an example, it has 8 x 6550 with each wanting a 50k g1 circuit resistance. With a bank of 4 x 6550’s, would it need a 12.5k grid leak to the bias capacitor? Their bias circuit shows 150k + ½ of the 22k bias adjust pot to the bias capacitor, so 161k. Is there something I’m missing or is this really bad design?

"Really" bad?  Ehh... more like ill-advised.

Guitar amp companies routinely exceeded the recommended grid circuit resistance limits of the output tubes they used.  Mostly, they get away with it.  But the limits are created/recommended to allow safety-margin for worst-case.

As I showed in my last post, there is tension between what's good for a later stage, and what is good for the preceding stage.  The later-stage wants lower grid resistance, but this is "heavier load/less gain" for the earlier-stage.  What happens is then a trade-off between competing priorities.

Meanwhile, Ampeg's SVT used 47kΩ per 6550.  The trade-off they made for pursuing "the right way to do it" was needing a cathode-follower to drive the output tubes using a triode with more grunt than even a 12AU7.  And overall, the driver/phase-inverter of the SVT involves all 3 double-triodes shown in that power amp schematic.

Marshall didn't invest the time/money in an elaborate power amp design.  The pretty much lifted the same phase inverter they'd use for any guitar amp, then cheated the grid circuit resistance up & hoped it all holds together.