Princeton Reverb Build using 6V6s and 6L6s

[tyru007]

Building a standard PR, except I would also like to be able to use either 6V6s or 6L6s.  I'm using a Hammond Vibrolux PT (180 mA, 325-0-325) and OT (35 Watt, 4000 : 4-8-16). 

Do I need to increase the load resistors (two 56K resistors) on the PI to provide more P-P swing for the 6L6s? 

I was going to move the supply voltage for the PI from Node D to Node C on the Fender diagram and provide around 300V, increasing the P-P swing, but was not going to change the values around the PI.  Any thoughts?

Besided adjusting the fixed bias voltage, anything else needed to allow 6V6 and 6L6 usage.

Thanks for your help.

[tubeswell]

I'd stick with the 56k first and then see how you go. Certainly upping the B+ to 300 would help get more strength into the output tubes whilst keeping it 'cleanish' methinks. Merlin's book says that for centre (cleanest) biasing, the anode should be around 1/2 to 2/3 of B+, and the cathode should be half of what the anode is (i.e.; 1/4 to 1/3 of B=+). He does say you get more signal swing from going to 100k for both resistors, but I wonder if you up the load resistance by increasing both resistors, on the one hand you'd get a bit more plate swing, but on the other hand you'd get a bit more cathode current feedback (generated from the larger cathode resistor) which will tend to cancel out the increased plate swing (methinks again).

Another option is moving to an LTP.

[phsyconoodler]

I've built quite a few Princeton's with 6L6's and I don't change anything.The amps are loud and sound just fine.I use a Hammond 272HX PT with 300-0-300v and with diode rect. it gives about 424v on the plates.Killer tones.

[Boots Deville]

phsyconoodler, what speaker(s) do you like with that build?  I've been thinking of doing the same sorta thing, and was planning on trying an Eminence Red White and Blues w/ it to start, but I'd be interested in your opinions.

[phsyconoodler]

I've used everything from Jensen C12K's to Celestion Vintyage 30's.
  The best sounds to date are from a Weber 12F150.

[HotBluePlates]

I was going to move the supply voltage for the PI from Node D to Node C on the Fender diagram and provide around 300V, increasing the P-P swing, but was not going to change the values around the PI.  Any thoughts?

I think this would move you in the wrong direction, if you want to use 6L6's. If you lower the supply voltage feeding the phase inverter, you also lower the maximum signal it can make. 6L6's will take more bias voltage than the 6V6's, and therefore require a bigger drive signal. I'd estimate you'll want somewhere near 400v of supply for the phase inverter (maybe a little less) and stock values. That should allow sufficient output swing and still leave enough voltage across the tube.

[mresistor]

  The best sounds to date are from a Weber 12F150.

I'll second that. That weber is a great speaker in the PR or the DR.

[bigdaddy]

Other speakers to use are a Celestion Blue(use a Gold if you really crank it or use pedals), an original Jensen C12N which are great!!!! If you can get one the old Naylor 12s(A mix of a Jensen and Celestion made by eminence at one point, I think he also used another manufacturer). Some like greenbacks, some like Vintage 30s. And the many different Eminence speakers seem to all sound pretty good.

But if you are really into a true Fender tone nothing beats the old C12N.

[Fresh_Start]

I was going to move the supply voltage for the PI from Node D to Node C on the Fender diagram and provide around 300V, increasing the P-P swing, but was not going to change the values around the PI.  Any thoughts?

I think this would move you in the wrong direction, if you want to use 6L6's. If you lower the supply voltage feeding the phase inverter, you also lower the maximum signal it can make. 6L6's will take more bias voltage than the 6V6's, and therefore require a bigger drive signal. I'd estimate you'll want somewhere near 400v of supply for the phase inverter (maybe a little less) and stock values. That should allow sufficient output swing and still leave enough voltage across the tube.

tyru was talking about the "Stokes mod" IIRC, raising the B+ for the PI from 240 to 320 (approximately).  The thought behind this mod is to increase the clean headroom from the PI so the power tubes distort sooner relative to the PI.  It doesn't get you all the way to 400 volts, but a 30% increase has to help some doesn't it? 

I've seen some guys talking about going further with this concept, lowering the 18watt resistor between B & C with a corresponding increase in the resistor between C & D.  That would increase the PI supply even futher, although you won't get all the way to 400 volts.  Just guessing, but a 5K resistor ought to provide adequate filtering and not drop the B+ anywhere near as much as the 18K.  Approximately 375 volts at the revised "C" point?

I've built two Princeton Reverbs and put a Weber 10F150-T in both.  Like 'em very much.

Cheers,

Chip

[HotBluePlates]

tyru was talking about the "Stokes mod" IIRC, raising the B+ for the PI from 240 to 320 (approximately).  The thought behind this mod is to increase the clean headroom from the PI so the power tubes distort sooner relative to the PI.  It doesn't get you all the way to 400 volts, but a 30% increase has to help some doesn't it?

Oops! I read that wrong as lowering supply voltage.

You'll still need more than 300v with a split-load inverter driving 6L6's. There's a reason you don't see it done in most old lean class AB amps. A split-load driving 6L6's with a low supply voltage might be found in a couple of 50's amps, but they're also running hotter with less bias voltage.

So I went back and rethought the circuit a little. The original amp uses a pair of 56k resistors, a 240v supply, a 12AX7 and a 1k cathode bias resistor. If you draw a loadline for a split-load inverter, the dc line is the sum of the individual loads; a pair of 56k resistors is drawn as a 112k line from 240v to 240v/112k = ~2.14mA. You can also draw a 1k cathode load line by assuming a bias voltage, and dividing by 1k; at 0.5v, the line should be at 0.5mA, 1mA at 1v, 2mA at 2v, etc. You draw that cathode resistor line by going to the grid voltage curve of your assumed bias and making a mark on that line at the amount of current that would produce that voltage through your cathode resistor.

And then I did a little cheating. It won't be exact, but if you start at where the cathode resistor line crosses you dc loadline, note the plate voltae, and then move along the loadline to the 0v gridline and note the plate voltage, you have output voltage swing in 1 direction. But, we have half our load in the plate and half in the cathode, and I took half of the noted voltage difference as the peak output from the split-load inverter. I call it cheating because the difference along the loadline in one direction may not be the same as in the other direction; also, we're not accounting for the loss of gain due to the following grid reference resistor (but this is small due to reduce output impedance from 100% feedback). Nonetheless, we'll consider this "best-case guessing".

If you look at the schematic of a Princeton Reverb, you'll see a bias voltage for the 6V6's of -34v. If you want to get all the clean output from these tubes, you have to at least get to within a volt of driving the grid to 0v. That means you need a peak voltage input of 34v to drive the 6V6 grid momentarily to 0v.

I drew my loadline and cathode line. I noticed that the schematic shows 1v across the cathode resistor, which agrees well with the lines I drew. At idle, the 2 lines cross at a plate voltage of 170v. This is not actual voltage on the plate (it is the actual plate to cathode voltage of the tube), but is a useful number to use as a starting point to observe changes. If the grid voltage is changed by 1v to 0v, then we get the point where the loadline intersects the 0v gridline, which is at 70v. 130v - 70v = 60v, which is the total voltage change across both the plate and cathode loads; that 60/2 = 30v peak to 1 output. So we're able to drive the 6V6's close to their maximum clean output power with the existing supply; maybe the phase inverter causes the stage before it to distort around the same time the output tubes start distorting.

I drew additional lines at the same load and cathode resistance, but with 300v and 400v for the supply. At 300v, the starting voltage is about 167v, and moving to the 0v gridline, the voltage is 90v, or a change of 38.5v peak output. For 400v, the starting voltage is 230v, and 117v at the 0v gridline; the change is 113v or 56.5v peak output.

A blackface Tremolux schematic shows 410v on the plates of 6L6GC's, with a bias of -45v. So we know we need at least that much output. Based on the numbers we have right now, I'm gonna guess we'll need a supply bigger than 300v by the same ratio as 45v peak is bigger than 38.5v peak; do the math, and you arrive at about 350v. We're looking for 45v peak output, and that's gonna require about 45v +1.3v (where I estimate the bias will land with a 1k resistor) = 46.3v of input from the previous stage. That is within the capability of the existing stage with a 240v supply, but you should always check because it isn't necessarily guaranteed.

Based on the schematic's voltage values for the power supply, the resistor leading to Node C could be lowered from 18k to 12k to arrive at about 350v, and if you like the following 18k resistor could be raised to 22k or left as-is.

[sluckey]

Do I need to increase the load resistors (two 56K resistors) on the PI to provide more P-P swing for the 6L6s?

I don't think you can change the PI output voltage swing by changing those resistors. The voltage gain of that split load phase splitter is slightly less than 1 (typically .8 to .9, just like a cathode follower). I like to think of the gain being equal to 1 just for the sake of keeping the math easy.

The signal voltage on the cathode will always be slightly less than the input signal voltage on the grid. And since the plate resistor is the same size as the cathode resistor with the same current flowing thru cathode and plate, the signal on the plate will be the same as the cathode signal.

I believe if you want to increase voltage swing at the output tube grids you'll have to increase the signal into the PI.

[HotBluePlates]

Do I need to increase the load resistors (two 56K resistors) on the PI to provide more P-P swing for the 6L6s?

I don't think you can change the PI output voltage swing by changing those resistors. The voltage gain of that split load phase splitter is slightly less than 1 (typically .8 to .9, just like a cathode follower).

...

I believe if you want to increase voltage swing at the output tube grids you'll have to increase the signal into the PI.

It could be both and neither. You have to draw the loadline to figure out if the stage can handle the added input. As it turns out, Fender designed the phase inverter to just barely have enough oomph to drive the stock 6V6's, about 30v peak output to each side. If you try to feed more input signal to the phase inverter, it draws grid current, distorts and still doesn't put out additional output. It just starts making the stage ahead of the split-load crap-out.

PRR writes the way he does for a reason; I think I'll need to adopt more of his style.

Bottom-line Up Front:
Keep the same parts values and a 12AX7 split-load, raise the supply voltage to 350v to the splitter. You need more input signal, and the driver stage ahead of the split-load will supply the extra needed signal with only a modest increase of grid-swing, so no changes need be made here. You want 410-420v of total supply, so using your Vibrolux PT, you may need to use a GZ34 as the Vibrolux does.

Explanation the way I normally do:
I got unlazy, and posted the curves I was drawing to write my last post. When I started my line-drawing expedition above, I double-checked myself using the book Amplifiers by York, which can be found on Pete Millett's website.

Yes, gain of the split-load itself is less than 1, so yes, the input to the split-load has to increase. But as I noted above, the bias shown on the Princeton Reverb schematic is 1v across the 1k cathode bias resistor. The bias is an indicator of how much signal input a stage can handle before grid current, and it looks like our split-load is 33v short! The thing to realize is the input signal causes a current change in the tube which results in an output signal. The normal method is to draw the loadline, and assume that our signal has changed by 1v.

As an example, say our input signal changes by 1 volt, and on the loadline we see that that 1v change in grid voltage results in 68v of "plate voltage change". That change is read on the plate curves, but it really indicates a change of tube current through the load resistance, and the line does the math for you to tell how much voltage is dropped across the load resistance. Since half of our load in this case is in the plate circuit and half in the cathode circuit, we only have half of that total voltage change found earlier present in each location; that's 34v.

But our cathode bias resistor is riding on top of the cathode load resistor. When the try to change the grid voltage by 1v, the cathode has an output which counteracts that (our 100% feedback). In this case, we have 34v of output, so it really takes 34v +1v = 35v of input to get the effect at the grid of a 1v change.

I noted above that we need around 45v peak output tube each split-load output. The stock values and supply voltage had to be investigated to know if there was any extra the circuit could give, and it was found that there was no extra. You could try to change the load resistors, but you can't reduce them because you're still dealing with a 12AX7, and triodes that are used as voltage amplifiers need a load of at least 2 times their internal resistance to get efficient amplification; that leads to the 56k + 56k = 112k that is used in the stock circuit. You could try to raise the resistor values, but that results in less voltage across the tube and less current. We all think "voltage, voltage, voltage" in a preamp, but if the current swing is not there, there is no voltage swing.

The easiest way of arriving at a circuit with increased output voltage (with minimum brain-strain) is to increase the supply voltage. My goal was to show how to get the needed output while keeping the same tube type and parts values, and only needing to increase the supply voltage. The thought is that it can be easily done by changing a power supply dropping resistor.

[Fresh_Start]

Note:  HBP's much meatier post came in when I pressed "Preview"

Do I need to increase the load resistors (two 56K resistors) on the PI to provide more P-P swing for the 6L6s?

I don't think you can change the PI output voltage swing by changing those resistors. The voltage gain of that split load phase splitter is slightly less than 1 (typically .8 to .9, just like a cathode follower). I like to think of the gain being equal to 1 just for the sake of keeping the math easy.

The signal voltage on the cathode will always be slightly less than the input signal voltage on the grid. And since the plate resistor is the same size as the cathode resistor with the same current flowing thru cathode and plate, the signal on the plate will be the same as the cathode signal.

I believe if you want to increase voltage swing at the output tube grids you'll have to increase the signal into the PI.

Can you accomplish that goal moving both the PI and the driver triode to node "C" and reducing the power rail dropping resistor to 12K +/- per HBP's calculations?  Maybe increase the plate resistor on the driver tube as well?  

The potential downside there might be loss of preamp distortion in the driver stage.  IOW increasing the plate voltage supply would increase "clean headroom" for that stage, and I don't know how important distortion at that point in the circuit is for the sound of a Princeton Reverb.

It's also worth remembering that physchonoodler doesn't change anything and likes the amp with 6L6s.  I'm interested to know how hot he biases his power tubes and if all the distortion comes from the preamp.  If your goal is louder clean Blackface tone, I'm not sure modding the AA1164 circuit is the way to go.

Not even 2CW

Chip

[sluckey]

Can you accomplish that goal moving both the PI and the driver triode to node "C" and reducing the power rail dropping resistor to 12K +/- per HBP's calculations?  Maybe increase the plate resistor on the driver tube as well?

Yes, all three of those ideas will increase the input signal to the PI. Or, you could just turn up the volume pot. The increased output from the PI will be due to the fact that you increased the input signal. The higher voltage on the PI just means that you'll have more headroom before PI distortion sets in. 

The potential downside there might be loss of preamp distortion in the driver stage.  IOW increasing the plate voltage supply would increase "clean headroom" for that stage, and I don't know how important distortion at that point in the circuit is for the sound of a Princeton Reverb.

That's the whole point of the Stokes mod. The Stokes mod increases the plate supply voltage to the PI only. This increased supply voltage doesn't increase the signal output of the PI. It just provides more clean headroom which means you can now increase the input signal into the PI more, causing a bigger output of the PI, without running into distortion. The Stokes mod was about getting more clean power out of the PR (kinda like it's bigger AB763 siblings.) But... There's a price to pay. The stock PR has a small OT which may croak after the Stokes mod because it can't really handle the extra power. The Stokes mod doesn't kill all those little OTs, but the risk is great enough that Stokes no longer recommends this mod without up-sizing the OT.

My opinion of the split load phase splitter operation is just that. And I don't have any experience with the PR and Stokes mod either. So bear in mind, I may be way off with my thinking!   

[HotBluePlates]

The potential downside there might be loss of preamp distortion in the driver stage.  IOW increasing the plate voltage supply would increase "clean headroom" for that stage, and I don't know how important distortion at that point in the circuit is for the sound of a Princeton Reverb.

It's also worth remembering that physchonoodler doesn't change anything and likes the amp with 6L6s.  I'm interested to know how hot he biases his power tubes and if all the distortion comes from the preamp.  If your goal is louder clean Blackface tone, I'm not sure modding the AA1164 circuit is the way to go.

What's the point of having the option of plugging in 6L6's? More output power, right? At least that seems to be Tyru's goal, because he is taking the extra steps of using a PT that is designed to feed 6L6's and an OT with the required primary impedance.

The final piece of the puzzle is making sure the phase inverter can actually drive the output stage to get the added power you paid for (twice) by buying the bigger PT and OT. It's like paying for a V-8 Mustang, but the gas pedal can't go far enough down to actually use the horsepower and gasoline you're paying for. If you don't feed enough drive voltage to the 6L6's when they're being used, you'll probably feel like the swap makes no difference.

I know I can sub 6L6's in my tweed Deluxe. The amp distort at a slightly higher volume control setting, but I don't get any output power increase. My 2x6L6 tweed Super copy is much louder by comparison.

The fastest route to more actual output power with 6L6's is raising the supply voltage to the phase inverter. If you're worried about the impact on the preamp, then raise the dropping resistor between the phase inverter and preamp to arrive at stock power supply voltages feeding those stages.

[HotBluePlates]

Yes, all three of those ideas will increase the input signal to the PI. Or, you could just turn up the volume pot. The increased output from the PI will be due to the fact that you increased the input signal. The higher voltage on the PI just means that you'll have more headroom before PI distortion sets in.

I still disagree, for the reasons outlined in my earlier posts. Namely, this would work, except the problem is you have to take the supply voltage and come up with 2 output voltage swings, plus leave voltage across the tube for it to operate. If you simply increase input voltage beyond what the split-load can really output, it just draws grid current, distorts, lowers the gain of the previous stage, and still doesn't put out any larger output signal. All the same things that might happen to a regular volt-amp stage, except we have 100% feedback trying to fight the distortion.

That's why I recommend increased supply voltage and increased input to the phase inverter. But the increased supply voltage is the part that can't be omitted.

That's the whole point of the Stokes mod. The Stokes mod increases the plate supply voltage to the PI only. This increased supply voltage doesn't increase the signal output of the PI. It just provides more clean headroom which means you can now increase the input signal into the PI more, causing a bigger output of the PI, without running into distortion.

You made my point for me. Stokes was only trying to drive 6V6's without the split-load crapping out. We need more than that to drive 6L6's.

Let's call this new iteration of the PR (if it works) the "Ingram Mod"! 

[sluckey]

That's why I recommend increased supply voltage and increased input to the phase inverter. But the increased supply voltage is the part that can't be omitted.

This may be one of those "chicken and egg" things.   

We're very close to saying the same thing, just slightly different viewpoint.

Let's call this new iteration of the PR (if it works) the "Ingram Mod"! 

OK, but we gotta call the PI a "split hair" PI too.

[HotBluePlates]

I'm vigorously defending my idea, because... wel, it's my idea. And I'm unduely attached to it.

But I suggest looking at the 5F4 tweed Super schematic, since I mentioned it puts out much more power than when I pop 6L6's into my tweed Deluxe copy. The Super has a supply voltage of 332v to the split-load phase inverter driving 6L6's; the tweed Deluxe has in the neighborhood of 280v (judged by the 5C3 Deluxe schematic; my amp is halfway around the world).

But I really am posting a lot about this because I want people to be able to design/alter a phase inverter intelligently instead of relying on pure guesswork.

I posted the 12AX7 curves with loadlines earlier. The line is drawn like a normal volt-amp stage, but the actual circuit is not a normal volt-amp stage. So you have to think and read between the lines.

I already explained how to interpret the voltage output swing based on the loadline earlier, but let's go at it from a completely different viewpoint: changing current as a response to changing input voltage. We'll start with the bottom loadline for a 240v supply, and then look at the top loadline for a 400v supply, to see what's changed.

The loadline for 112k and a 240v supply intersects the 1k cathode line near a grid voltage of 1v. We'll call it 1v for convenience. The current at this point is 0.95mA, which when passed through the 56k resistors results in a drop of 53v. With 2 load resistors dropping 53v, there is 240 - (2*53v) = 134v across the tube (we're ignoring the 1v across the cathode bias resistor). The loadline and the 0v gridline intersect at a voltage which is the minimum this tube must have across it; if there were less voltage, the current through the tube drops, reducing the voltage across the load resistors and restoring the voltage across the tube. That intersection happens at 70v, so let's keep that number in mind as we move forward. That number is only valid with this load and a 240v. The plate is sitting at 240v - 53v = 187v and the cathode is at 53v (add 1v if you don't want to ignor the cathode bias resistor).

An input signal occurs which results in the effective grid voltage moving positive by 1v to the 0v gridline. Current at this point is 1.5mA, and through the 56k resistors results in 56k * 1.5mA = 84v across each load resistor. 240v - (2*84v) = 72v across the tube, so we're golden as far as needed voltage on the tube. But look at the change that occurred: The plate moved downward to 240v - 84v = 156v, while the cathode moved up to 84v. The change in voltage is 84v - 53v = 31v. Since we looked at what happens when the grid is driven to 0v, we've reached the limit of what this stage can output without a low-impedance driving circuit to the input of the split-load which can supply the power needed for the grid current that will be drawn. Try it, but between grid current and 100% feedback fighting to keep the stage clean and the lowered imput impedance due to grid current wrecking the previous stage's gain... well, you just won't get any more.

The basic method is outlined, so let's look at the case of no changes except a 400v supply. Squinting, the 112k loadline crosses the 0v gridline at about 117v, so our needed voltage across the tube went up alone with the supply voltage. But the output has gone up too; let's see how much.

Looking at currents again, and figuring the resulting voltages, our idle point is at the intersection of the 112k line for 400v and the 1k cathode line, which is at 1.55mA and about 1.5v of bias (1.55v to be exact). 1.55mA through the 56k resistors is 86.8v (round to 87v), so there is 400v - 87v = 313v on the plate and 87v on the cathode. Assume a 1.5v effective grid input to drive the grid to 0v momentarily. The resulting current is 2.5mA, which is 56k * 2.5mA = 140v across each load resistor. That leaves 400 - (2*140v) = 120v across the tube, which meets the required 117v we noted earlier. Remember, the issue with triodes is that they must have a good deal of voltage across them to operate, more so than pentodes. The current change causes a voltage change of 140v - 87v = 53v at each output.

The voltage changes that we noted in each case are at the peak of the positive input voltage swing, so they are peak output voltages. The 53v noted in the 400v supply case is more than ample for driving 6L6's, but 400v for a phase inverter is problematic in an amp with a ~420v supply for the output tube plates! Which is why I interpolated in the earlier post to get 350v supply for the split-load, and that estimate compares favorably with the 332v shown in the tweed Super schematic.

[HotBluePlates]

I kept saying "effective grid voltage" because the actual needed driving voltage is higher. Using the 400v supply case, the cathode starts at 87v plus about 1.55v of bias, or 88.55v. A 1.55v effective grid voltage change results in a current that causes the cathode to change by 53v. So we really need 53v + 1.55v = 54.55v of input. The other 53v is the counteracting effect of feedback. But the stage ahead of the split-load is capable of doing this, so we really just needed to make sure the split-load was up to the task of driving the output tubes.

If you wonder why I've only calculated for the case of the input voltage moving positively to the 0v gridline, it's because that was the closest limiting factor. You could calculate the effect of the grid voltage changing the same amount the other way and check for distortion. But since there is not a limiting factor in our case closer than the 0v gridline, the tube is able to swing a bigger output voltage during the negative portion of the input signal swing.

Yes, all this is using a big frickin hammer to drive my point home. But hopefully it will also serve as a roadmap to others to know how to analyze a split-load inverter stage and check performance prior to building/modding.

[HotBluePlates]

OK, but we gotta call the PI a "split hair" PI too.

Guilty!! 

Really, though, people make mods and never consider the phase inverter because they don't understand it. Then they wonder why their amp doesn't perform like they thought it would.

Problem for me is, my hair keeps falling out, so I always have less of it to split!  But this is a case where it's not just niggling; the amp won't make more output power until you do both things we said: increase the supply voltage and increase the drive voltage.

[phsyconoodler]

All I can say is I really,really like the sound of the PR circuit with no PI changes for 6L6's.I have built many and they all sound great beyond words.
  The only 'change' is the slight increase in B+ I use from the stock 415v to 425v with a Hammond PT an diode rectifier.

The tone is to dies for and if the 6L6's are not working hard enough you wouldn't know it by the quality of the tone.

[HotBluePlates]

6L6's sound great in my tweed Deluxe too. But they don't put out more power than 6V6's.

Tyru is paying for more PT and OT. I'd be pissed off if I didn't also get the extra power I was paying for.

[phsyconoodler]

Well they sure do put out more power to my old ears.With 6V6's the amp is much quieter and can't be used in a medium sized club without a mic in front of it.With the 6L6's no mic is needed.And I'm not talking just cleans,it grinds plenty hard when cranked up.
 I use either a Vibrolux OT(4k) or a Hammond OT with a 6.6k primary.
   Both are very much louder with 6L6's.

[flora]

In the Princeton Reverbs that I've owned with 6L6's,( Yes.. It had a DR OT ) the Stokes mod made a huge difference in headroom .... I remember wondering why anyone would want a stock PR... too dirty , but that's just me.

[tyru007]

HBP is correct that I was looking for a way to build a princeton reverb that I could play using 6V6 (15-18 watts) or use 6L6s (25-30 watts) for increased power.  A 35Watt OT with a primary impedance of 4000 and secondary of 4, 8 and 16 taps would give me either 4000 or 8000, depending on the secondary tap used.  

The stock Princeton PI barely provides the voltage swing for 6V6s, so I know I need to increase the capability of the PI to provide more voltage swing for both the 6V6s and the 6L6s.  Most of the literature indicates that PI distortion is not desired.

Drawing the load line for 240V (stock) and 320V (Node C) looks like the P-P "capability" of the PI increase from 85V to 112V using the stock 56K resistors.

If I use 100K resistors, doesn't this results in a shallower load line?:  320V/200K = 1.6 mA at 0 Volts to 0 mA at 320V.  The max P-P increases to (320V-60v) = 260/2 = 130V P-P.  

It makes sence that although the PI may have the capability to provide more voltage to the power tubes, I still have a gain of about unity, and need to increase the signal to the PI.  I can get some increased signal by dropping the value of the 3.3M/10p resistor.  I could also remove the tremelo circuit and add a gain stage, but I would suspect this is farther from a Princeton than I want to be.




 

[HotBluePlates]

Drawing the load line for 240V (stock) and 320V (Node C) looks like the P-P "capability" of the PI increase from 85V to 112V using the stock 56K resistors.

If I use 100K resistors, doesn't this results in a shallower load line?:  320V/200K = 1.6 mA at 0 Volts to 0 mA at 320V.  The max P-P increases to (320V-60v) = 260/2 = 130V P-P.

The loadline cannot be a 56k line, because the total load is 56k + 56k = 112k. The fact that half the load is in the plate circuit and half in the cathode circuit makes no difference. You can't really go lower than 56k with good results, because the internal resistance of the 12AX7 is still around 5-60k at a minimum, and for good gain from a triode, you need a load of twice internal resistance or more; 2*60k = 120k, and so 112k made of a pair of 56k resistors is the closest standard value.

Look back at the curves that I posted. The lines are drawn for each supply voltage at a load of 112k, for the total of individual load resistors.

It makes sence that although the PI may have the capability to provide more voltage to the power tubes, I still have a gain of about unity, and need to increase the signal to the PI.  I can get some increased signal by dropping the value of the 3.3M/10p resistor.  I could also remove the tremelo circuit and add a gain stage, but I would suspect this is farther from a Princeton than I want to be.

You don't need to do any of that. That's the job of the stage ahead of the phase inverter. The stock circuit needed to make a peak output of about 32v, with a bias of 1v. That means the actual input to the phase inverter was 33v, and that came from the stage prior. We're talking about you getting more like 42v peak output or more, so something like 43-44v peak input. Even if the previous stage had a gain of 50, it could manage that with 44v/50 = 0.88v peak input.

The actual input needed to the stage ahead of the phase inverter will need to be a little bigger than this, due to the feedback injected at the bottom of the cathode resistor.

Speaking of feedback, 2.7k is the stock series feedback resistor value, but when you switch to 6L6's, the greater voltage at the speaker will result in more feedback. You may want to experiment with the effect of raising this resistor to 3.3k to 4.7k. You could arrange a switch for 2 different values, or you could optimize the feedback for one power level and accept something a little diffreent at the other setting. For example, if you get the stock amount of feedback for the high power setup by raising the series feedback resistor, if you make no change but switch to 6V6's, you will get less feedback. You should experiment with this to find out what you like.

[HotBluePlates]

So I went an extra step in checking on my assertion that more supply voltage would yield more drive to the output tubes and a better capacity for pushing 6L6's to maximum power. I looked at the supply and bias voltages for a number of Fender tweed, brown and blackface amps. There were both split-load and long-tail phase inverters in the group, and amps with 6L6's and 6V6's. Bias is an indication of the size of the driving signal needed, as a peak signal input that drives the grid to 0v is the maximum output power the output stage can deliver before gross distortion due to grid current. So a set of output tubes with a bias of -45v will need a driver stage that can produce 45v peak or 31.82v RMS.

6V6 Amps
-31v bias    260v phase inverter supply
-34v bias    240v phase inverter supply
-35v bias    325v phase inverter supply
-26v bias    325v phase inverter supply
-35v bias    280v phase inverter supply
-18v bias    280v phase inverter supply

6L6 Amps
-45v bias    340v phase inverter supply
-40v bias    332v phase inverter supply
-45v bias    340v phase inverter supply
-52v bias    450v phase inverter supply
-51v bias    430v phase inverter supply
-48v bias    385v phase inverter supply
-44v bias    410v phase inverter supply

Some interesting things can be seen here, even before we know the models. First, every 6L6 amp has more supply voltage feeding the phase inverter than any 6V6 amp, regardless of the bias level of the 6V6 amp. The simple take-away from this is that due to the large bias voltage of the 6L6, more drive is needed and more drive from the phase inverter is mostly a function of the supply voltage available. The highest phase inverter supply voltages belong to the amps with the highest amount of bias. The lowest supply voltage shown is 332v, and belongs to the 5F4 Super. This amp uses a split-load phase inverter, but the low voltage is not a matter of the inverter used, but the fact that it also has the lowest bias voltage of the group and therefore requires the least drive of these amps.

If you want to look at the schematics and/or layouts I pulled the info from, the amps are the tweed Super, the 5F6A Bassman, the AA864 Bassman, the AA165 Pro Reverb, the AB763 Twin Reverb, the AB763 Tremolux and the AB763 Vibrolux (not necessarily in that order).

The 6V6 group includes the 5C3 Deluxe, the 5F11 Vibrolux, the 6G2 Princeton, the 6G3 Deluxe, the AA1164 Princeton Reverb and the AB763 Deluxe Reverb.

[HotBluePlates]

If we look at the 6V6 group, the bias voltages and supply voltages seem to be more varied. It is also worth noting that the amp with the lowest phase inverter supply voltage also has the second highest bias voltage; we'll come back to that in a minute. The lowest bias of the group is the amp with 18v of bias, which is the 5E3 Deluxe. Given the small drive needed by this low bias level (pretty hot class AB), the phase inverter supply is still a healthy 280v. When Fender redesigned the Deluxe into the 6G3 model, bias moved to -26v (may be a typo and really -36v) and 325v for the phase inverter supply. The 6G2 Princeton had a bias of -35v but only 280v of supply to the phase inverter. Not a huge difference from the 6G3 Deluxe, but maybe enough to keep it from outperforming or equaling the next more expensive Fender model.

The big divergence is when the Princeton and Deluxe move into the blackface era. The Deluxe Reverb's bias is -35v, and the supply to the phase inverter remained at 325v. But the Princeton Reverb, with a bias of -34v has a supply voltage of only 240v feeding the phase inverter! I can't think of any logical design reason to evolve a new edition of an amp model and drop the supply voltage and therefore the drive to the output tubes... unless you've played both a Princeton Reverb and a Deluxe Reverb. If you have, you'll know that a Deluxe Reverb may get distorted and raucous, but can hang with a band in a small club setting, while a Princeton will be drowned out by anything more than a jazz drummer playing with brushes, or exercising a lot of control.

It occurred to me that the older Fender models had a bigger number of distinguishing characteristics, but by the blackface era everything had evolved to one basic preamp. The only real difference was power level by way of tube type and quantity, and the number and size of speakers. The Deluxe Reverb and the Princeton Reverb are essentially the same amp, with an overall power supply within 5v of each others, the same number and type of output tubes and a bias voltage within 1v of each other. Yeah, the Deluxe has a Normal channel too, and a 12" speaker. But neither account for the real volume difference in the amps. What does account for it is the way the Princeton Reverb is hamstrung by the low phase inverter supply voltage. I'm guessing the extra channel and slightly bigger speaker weren't enough to differentiate the amps and justify a higher price for the Deluxe Reverb; after all, that extra channel doesn't have any effects! I think Fender purposely made the Princeton a weak sister to the Deluxe in order to define the difference between the models.

And so I also still assert that a phase inverter supply of 350v would easily bring up the volume, make the amp able to drive 6L6's to 40w and even make the amp louder with 6V6's.

[tyru007]

HBP:

I do realize that there are two 56K resitors and drew my load line for 2X (112K), but thanks for the explanation.

You don't need to do any of that. That's the job of the stage ahead of the phase inverter. The stock circuit needed to make a peak output of about 32v, with a bias of 1v. That means the actual input to the phase inverter was 33v, and that came from the stage prior. We're talking about you getting more like 42v peak output or more, so something like 43-44v peak input. Even if the previous stage had a gain of 50, it could manage that with 44v/50 = 0.88v peak input.

I don't understand why you say that I should Not be trying to get more signal to the PI.  Since the PI provided basically unit gain increase, don't i need to find a way to increase the gain of the stages before the PI and thus an increase in voltage to the power tubes?

I have also noticed that a 1K cathode bias resistor for the PI is not centered on the load line.  The other amps that use a cathodyne into a pair of 6L6s tend to use 1.5K with two 56K load resistors.  A 1.5K seems like it would also be a better choice for either the 6V6s or 6L6s, since it tends to center the bias. Is there a reason to bias closer to the Vg=0 line?  compression? 

[phsyconoodler]

Quote:"I don't understand why you say that I should Not be trying to get more signal to the PI.  Since the PI provided basically unit gain increase, don't i need to find a way to increase the gain of the stages before the PI and thus an increase in voltage to the power tubes?"

Well increasing the signal does not increase the PI voltage.It gets it's voltage from a fixed source not the signal input.
  And I could not agree less with HBP about wanting more PI voltage.The tones are so good without increasing it that it may even be a detriment to the overall tonality to do that.I personally don't like a PR with the Paul C or Stokes mod at all.Of course it depend on what you want to hear,but overall TONE is what we want isn't it?
  I have built several now and I do not increase the PI voltage a whole lot and the clean headroom is more than sufficient and quite a bit louder overall than with 6V6's as well.If you want maximum headroom by all means do what HBP says.If the volume is not what you are after,leave it alone and enjoy some stellar tones.

  By the way,if you use an OT with a 6.6k primary,you can switch between 6v6's and 6L6's any time you want without a speaker load change.i.e. you can use 8 ohms or whatever for both tube types.

[HotBluePlates]

I don't understand why you say that I should Not be trying to get more signal to the PI.  Since the PI provided basically unit gain increase, don't i need to find a way to increase the gain of the stages before the PI and thus an increase in voltage to the power tubes?

What I'm saying is you do not need to do anything to the circuit to get the additional signal input. The stage before the phase inverter supplies the bigger input swing to the phase inverter. *If* the gain of this stage is 50 (it is really more than this), then to make a 32v peak output signal, this stage needs an input of 32/50 = 0.64v peak. If the same stage needs to make a 45v peak output signal, it needs 45/50 = 0.9v peak input. The extra 0.3v peak to drive the phase inverter easily comes from turning the volume control a little higher. You don’t need any circuit changes to produce the extra drive signal.

And I could not agree less with HBP about wanting more PI voltage.The tones are so good without increasing it that it may even be a detriment to the overall tonality to do that.I personally don't like a PR with the Paul C or Stokes mod at all.

Do you want tonal change or output power change? The question is what is the goal of modifying? Your goal? Tyru's goal?

If you want a tonal change but no appreciable increase in power, do nothing other than install 6L6 tubes and rebias.

You changed OT primary impedance and installed 6L6's, and it works for you. That's awesome; keep doing what you're doing.

IF the goal is to install 6V6's and have a 20w amp, then install 6L6's and have a 40w amp (and based on Tyru's initial post, I assumed this was *his* goal), then you have to install 6L6's, drop primary impedance to 4k, rebias, and give the phase inverter more supply voltage to enable it to produce a larger output swing to drive the output tubes.

I personally don't like a PR with the Paul C or Stokes mod at all.

No problem. Don't do it.

I haven't said anything about fixed-biasing the phase inverter. That's not a requirement, only more supply voltage.

My personal viewpoint is that a lot of value judgements about the "tone" of circuits have been and are being made on the basis of misapplied cut-n-paste of circuits. If the design is wrong, you can't help but wind up with a bad result, and the whole topology or approach gets condemned as a result.

The only outcome of changing the supply voltage as I've suggested is to have a Princeton Reverb at Pro Reverb/Vibrolux volume. It's the same idea as power scaling/VVR but a fixed increase at one point in the circuit to allow full power output from the bigger tubes.

[HotBluePlates]

By the way,if you use an OT with a 6.6k primary,you can switch between 6v6's and 6L6's any time you want without a speaker load change.i.e. you can use 8 ohms or whatever for both tube types.

You can switch tube type without a primary impedance change regardless of the primary impedance. There is 1 primary impedance that is "optimum" for producing maximum output, and that value is a 2:1 ratio for 6L6's and 6V6's running in class AB at this voltage range. You chose an intermediate compromise value which produces maximum power from neither tube type, and that's fine. You could run 8k and never change, getting maximum power from 6V6's and not from 6L6's. Or you could run 4k all the time and get maximum power from the 6L6's and less than maximum from 6V6's (supposedly why you're running 6V6's anyway).

I have also noticed that a 1K cathode bias resistor for the PI is not centered on the load line.  The other amps that use a cathodyne into a pair of 6L6s tend to use 1.5K with two 56K load resistors.  A 1.5K seems like it would also be a better choice for either the 6V6s or 6L6s, since it tends to center the bias. Is there a reason to bias closer to the Vg=0 line?

The initial goal was to get full output power from the 6L6's with a minimum of circuit changes. We might come up with something better if we threw the whole circuit design out the window and redesigned from scratch. But that's a lot of work, and the existing design works great for 6V6's.

When you select the operating point, there are a lot of factors to balance. If you centered the operating point evenly between the 0v gridline and the point where the tube passes 0mA and has the full supply voltage across it, you would appear to have equal output voltage capability during both positive and negative output voltage swings. But look carefully at the spacing between the grid lines. Say your operating point was around -1.25v, between the -1v and -1.5v gridlines), and the input signal was 1.5v peak-to-peak, or 0.75v peak. This signal drives the gird voltage to -0.5v and -2v on alternate swings. The distance travelled along the loadline from the idle point to 0.5v is different and longer than the distance travelled from the idle point to 2v. That is distortion of the output signal, but in extreme cases, when the operating point is at very low current and near the markedly curved parts of the gridlines, there is much more output swing capability in one direction than in the other. So your "push 6L6" may get all the signal it needs, but the "pull 6L6" won't. Not just distortion, but reduced power capability of the output stage is the result.

Fender chose a pretty good compromise between distortion and output voltage swing with the 1k resistor and the original operating point. So I did not want to change that up, and I noticed that with nothing changed and the supply voltage scaled up to 400v, the spacing of the gridlines on either side of the operating point were again fairly well matched.

In the end, there's no compelling reason to change the bias resistor for a different operating point. In this case, when the supply voltage is raised but no other changes are made, the phase inverter operates and behaves in the same manner it always did, but produces a larger output voltage swing. The price is a larger input voltage swing is needed. As I said before, the stage ahead of the phase inverter supplies this added input signal and only requires a very slight increase in its input signal to achieve it. The existing preamp doesn't need any modification to drive the stage ahead of the split-load inverter to enable the output stage to make full power.

I have said in the past and say again that, especially in Fender amps, the stage ahead of the split-load inverter should be thought of as part of the inverter. It is the injection point of the feedback from the speaker, and everything wrapped in that feedback is the output section of this amp.

[phsyconoodler]

HBP said:"If you want a tonal change but no appreciable increase in power, do nothing other than install 6L6 tubes and rebias."

On paper it's not supposed to increase output,but it does.The amp is much louder with 6L6's than 6V6's.I'm not using wimpy iron,so maybe that's the difference.Or maybe it's just the clean headroom and less compression?
  I'm not trying to argue with you,just share my experience after building quite a few PR's with bigger iron and 6L6's.

[HotBluePlates]

HBP said:"If you want a tonal change but no appreciable increase in power, do nothing other than install 6L6 tubes and rebias."

On paper it's not supposed to increase output,but it does.The amp is much louder with 6L6's than 6V6's.I'm not using wimpy iron,so maybe that's the difference.Or maybe it's just the clean headroom and less compression?
  I'm not trying to argue with you,just share my experience after building quite a few PR's with bigger iron and 6L6's.

It's okay; disagreement is not argument. I think we're both assuming things that each of us are not saying or meaning.

Where you quoted me, I said do nothing other than change the output tubes. I have done that in a tweed Deluxe, and the only change was the amp distorted on 7 instead of 6. No power increase.

Your amp is much louder with 6L6's. But at a minimum, you use a different OT with a different primary impedance. That is more than "no change other than output tubes." If you also have a PT capable of more current than a stock PR power transformer, you're again doing more than "nothing but..."

Our core difference lies in the assumptions we're making. I'm assuming a goal of a true 40w or more from the Princeton with 6L6's, so that it is as loud as any Super Reverb, Pro Reverb, Vibrolux, etc. I also assume that there is a goal of swapping back to 6V6's with the perhaps more output than a stock PR will make. I am also assuming that "tone" is shaped after you have met power goals, as tone can be had at high and low power output levels.

I'm also assuming the power goal should be met in a way that can be verified by measurement; that precludes a freer approach where the output is judged as simply "louder". How much louder? No one knows... But if we use a Vibrolux PT, essentially a Vibrolux OT, 6L6 output tubes and a phase inverter truly capable of driving them, then the resulting output power is measureable. And prior to measurement, we can be certain it is "as loud as a Vibrolux".

If these aren't the goals, then we need to define our goals. But the meager amount of engineering training I've had has made me see the value of goals that are "specific and measureable." How else can you tell when you've achieved your goal? In the case of tone, the specific and measureable can still be how wide I'm smiling when I'm done playing! 

[PRR]

> Do I need to increase the load resistors (two 56K resistors) on the PI to provide more P-P swing for the 6L6s?

If the amp is well-designed, you don't change the G2 voltage, you don't radically change the operating mode from near-A to near-B: no. Both 6L6 and 6V6 have Mu near 10. Therefore if G2 voltage is the same, G1 voltage will be very similar.

You get more power with 6L6 _IF_ you also lower the load impedance (and have enough power supply). You are leaving all the voltages the same, but increasing currents.

That said....

AA1164 is not "well designed" in an engineering sense.

Skipping to the real-world: I think HBP nailed it. The Princeton always had an identity problem. It was supposed to fill a gap between Champ and DeLuxe. First it was a one-6V6 Champ with a big box. Not too exciting, hardly worth the higher price. The next version was a two-6V6 amp; there's nothing in-between worth building. But two-6V6 is a DeLuxe, and a steady seller. How to keep the 2-6V6 Princeton from stealing DeLuxe sales?

Follow HBP's analysis and then look at the AA1164's B+ rail. 18K?? Where the heck did 18K dropping resistors come from?

I bet they mocked-up a cut-down Princeton with the handy DeLuxe PT/OT, and realized the dang thing was "too good", would steal DeLuxe sales. So they strangled it. Two 10K droppers didn't hurt it much, two 33K made it too lame, even two 22K was lame, but two 18K (36K total) was "just right".

(I still wonder why not 15K+22K.... standard values which may have already been in the bins. No matter.)

If you want the Princeton sound, where the driver can barely drive the power tubes, do nothing. It will barely-drive 6L6 about the same as it barely-drives 6V6.

If you really want Princeton simplicity with DeLuxe authority, change-down those B+ droppers, get more B+ at the driver. Two 10K, two 4K7.... I see AA1172 used 2K2 and 10K (12K total), and takes its driver voltage at the 2K2 tap (it is also a different driver).

I see that 1172 shows 16u caps and 1164 shows 20u caps... this is probably more about RETMA "rational values" fashion-change than engineering. 1172 also uses a choke, but the 1164's 1K resistor only drops 20V so that's not a big factor. The choke may give modestly lower hum, noticeably less cost, more weight/heft. It's an obvious place to save a buck, and I doubt the decision had anything to do with "tone" (except that a slight hum in Princeton gives a reason to charge more for DeLuxe).

All the modern (post 1936) audio power pentodes can make full output if the driver supply voltage is "near" the G2 voltage. That was a key design goal. (Triodes and older pentodes might need driver supply +higher+ than power-stage supply, which is awkward.)

"Near G2 voltage" is vague. If driver B+ is 80% of output stage G2 voltage, that's fine. If driver B+ is half of output stage G2 voltage, that's maybe not enough. AA1164 has driver at 60% of power G2 voltage.... that's dubious. You could probably get more "undistorted maximum power output" with a higher driver B+.

Of course a "barely good enough" driver may be a flaw or a feature.

And if it is a "feature", the exact amount of "barely good enough" may be open to musical taste and preference. Experiment!

If the goals is maximum driver signal voltage for a given load, B+, and tube, you can quickly compute an optimum plate resistor. Here, 230K (two 115K). But getting a good bias is tedious.

Don't bother, someone did this already. The amplifier table in a good tube datasheet will tell you what's "best". In this case, the split-load Cathodyne may be un-split to a plate loaded amplifier with 2*220K= 440K load. The GE tables have suggestions for 240K and 510K, 510K is pretty close to 440K. The maximum Eo for 510K load is for a 240K plate resistor (mighty close to the calculated 230K). Use a 2300 cathode bias resistor. With a 300V supply this will give 47Vrms or 66V peak.

Our 440K load is a little smaller than 510K, so we might pencil 207K and 2,200.

Re-split as a Cathodyne this is two 103K (use 100K) resistors with say 2K2 bias. Expect 33V peak output with 300V supply.

This is not like Fender's two 56K. That would be more like the table's 100K resistors. The table does not have data for 440K/510K load, but in 240K this does 40Vrms 56V peak. Less than the optimum but not a lot less. (And we've already guessed that this design was NOT supposed to be "best possible" but "just good enuff for the price.)

[HotBluePlates]

The maximum Eo for 510K load is for a 240K plate resistor (mighty close to the calculated 230K). Use a 2300 cathode bias resistor. With a 300V supply this will give 47Vrms or 66V peak.

But where the data sheet says that, it is for one load resistor and one output. When we split that load resistor in 2 parts and have 2 outputs, isn't the voltage at each output half what the table says? That would be 33v peak to each output, and still about the size of the original Princeton output.

[PRR]

I followed that with -

Re-split as a Cathodyne this is two 103K (use 100K) resistors with say 2K2 bias. Expect 33V peak output with 300V supply.

We agree on the "33". 103K is somewhat off from the 56K in the AA1164.

[HotBluePlates]

Oops! I didn't catch that they were part of the same example.

[phsyconoodler]

The Princeton's 'identity problem' is where it gets it's sweet tone.Too much more of anything changes it into something else.
  I absolutely love PR's with big iron and 6L6's and absolutely love them totally stock with 6v6's.
All the math in the world doesn't make the TONE better.
  With a 6.6k OT they shine like a diamond.With a 4k OT they sparkle like a Deluxe killer.I love the PI.That's where the magic is IMHO.
I'm going to try one with the improvements that you guys are suggesting and see if the tone is still there.
  I sure don't have a problem with the volume with big iron and 6L6's now.Quite a bit louder than a Deluxe Reverb.

[tyru007]

Like the MEGA Burritos at Bravos, it will take several attempts to digest all that you guys have offered.  Thnak you for all your insight.

It seems to me that Fender's selection of a cathodyne PI fo this amp would also tend to put it somewhere in between the Champ and Deluxe.

[HotBluePlates]

It is worth noting the Marshall Major uses a split-load phase inverter. The difference is that it also uses a driver stage between the inverter and power tubes to develop a bigger signal swing.

The McIntosh MC-30 (and likely other Mc amps) uses a split-load inverter. It also uses a driver stage between the splitter and output tubes, but for different reasons.

The Ampeg SVT uses a split-load inverter. Then it actually has a pair of driver stages ahead of the output tubes. One amplifies the diriving signal, the other lowers the driving impedance to the output tube grids, probably to counteract any grid current in the output tubes.

So split-load by itself does not have to mean "low output".

[sluckey]

So split-load by itself does not have to mean "low output".

How true. And it doesn't have to mean early distortion either. All the '60s Sunn amps used the split load PI. They were all very loud and very clean. And they didn't use a driver between the PI and PA.

[PRR]

The front-panel Ampeg VT-40 and kin use a plain cathodyne, and deliver enormous LOUDNESS per buck.

All the Dynacos were straight cathodyne no booster, and deliver every Watt possible from UL working at reasonable voltage. (Early Sunn is of course a hot-rod Dynaco.)

There's nothing wrong with properly powered cathodyne into a couple good audio power pentodes.

The Princeton's cathodyne is underpowered. Deliberately.

Aside from market model line-up considerations: the DeLuxe will ROAR, even make ugly sounds. The Princeton might not get so rude. If you only do mellow music in small lounges and weddings, the more polite model might even be preferable. But this is "artificial". Bring the Princeton's too-low driver B+ up to more like G2 voltage, and play hard, it should seriously overload the output bottles.

There's some difference in overdrive. The long-tail goes to symmetrical square-wave. The cathodyne loses symmetry and can do "odd stuff". To over-generalize: the longtail gets in-yer-face, the cathodyne may get less-sour. That's on paper. After years of abusing cathodynes I'm not sure I agree.

[Fresh_Start]

I think HBP is on to something with his musings about Fender deliberately hamstringing the Princeton Reverb so it fit better in the product line.  Can anyone name any other Fender amp out there with an unused filter node on the power rail?  Does it really do that much to clean up the power supply ripple for the preamp?  I doubt it.  

IOW is it possible that the prototypes included the "Stokes mod" but just sounded a bit too much like a Deluxe so Fender lowered the B+ to the PI without bothering to re-work the entire layout?  Maybe the prototypes were too hard on the low-budget output transformers chosen for the amp and shifting the PI power supply to the lower voltage node was the simplest solution to the "problem".

This discussion has changed my whole perspective on tweaking the AA1164 circuit, or any circuit for that matter.  "The hip bone's connect to the leg bone, the leg bone's connected to the ..." You get the idea.
 

At least now I have most of the right measurement devices.  I just have to use them scientifically and watch for interactions between components/stages/etc.

Cheers,

Chip

[tubeswell]

I think HBP is on to something with his musings about Fender deliberately hamstringing the Princeton Reverb so it fit better in the product line.  Can anyone name any other Fender amp out there with an unused filter node on the power rail? 

Chip

He did name one - the DR a1172.   

I wonder whether they were deliberately going after running all the pre-amp filter off one node, in order to eliminate sources of signal backwobble through the filter cap can  - i.e. it all synchs up if its taken off the same filter cap can node.  Plus the extra filter node then acts as another PII filter in the supply line - (would it not?) helping the sole pre-amp filter node cope a bit better. This is all just me supposising. (Or maybe I am just trying to find rationality where there is none?)

[Fresh_Start]

Not trying to be a pinhead, but each node on the A1172 circuit's power rail is attached to something:  http://www.schematicheaven.com/fenderamps/fender_deluxereverb_a1172.pdf

Plates
B - Screen grids, tremolo plus reverb driver
C - PI
D - preamp, including reverb recovery stage and the "driver" before the PI

I've come up with a simpler theory now:  Leo got a really good deal on a large batch of 4x20uf cap cans.  The Brown 6G2 Princeton used three 30uf filter caps.  The cap cans must've been cheaper...

Cheers,

Chip

[tubeswell]

Not trying to be a pinhead, but each node on the A1172 circuit's power rail is attached to something:  http://www.schematicheaven.com/fenderamps/fender_deluxereverb_a1172.pdf

Plates
B - Screen grids, tremolo plus reverb driver
C - PI
D - preamp, including reverb recovery stage and the "driver" before the PI

Cheers,

Chip

Duh!  - you're correct of course

[PRR]

> lowered the B+ to the PI without bothering to re-work the entire layout?

Change choke to resistor, increase two more resistors. Less grunt, less cost, no heavy thinking.

> Maybe the prototypes were too hard on the low-budget output transformers

Are Princeton OTs different from DeLuxe?

IAC, you won't hurt the OT, and the "laming" is not a real reduction in power, just takes the edge off. Like a good car with a dirty air cleaner.

> running all the pre-amp filter off one node, in order to eliminate sources of signal backwobble through the filter cap can  - i.e. it all synchs up if its taken off the same filter cap can node.

Wrong. In fact the AA1164 violates rule-of-thumb design: no more than two sequential gain-stages on one B+ node. Two is stable, three is inherently unstable. Wants to motor-boat.

The AA1164 gets away with it because there is loss in the tone-stack, huge loss in the reverb mixer, and the power amp's two driver stages have a little "Akido" effect. The reverb driver would tip the balance to unstable, but it is fed from "B" which is a long way from "D".

> Plus the extra filter node then acts as another PII filter in the supply line -

Signal at first plate is about 1 Volt, you need ripple to be much less than 0.001V. The raw DC at the rectifier has 10V ripple. You need ripple reduction better than 10,000:1.

When capacitance is expensive you want to get about 10:1 reduction per cap and use several caps. When capacitance is cheap you might go for 30:1 or 100:1 per stage and use fewer stages.

So from rectifier to first plate you want two to four R-C networks. Older amps will favor more stages. The AA1164 has three R-C filters (in addition to the bulk filter off the rectifier). They actually give ripple-reductions of 14:1, 250:1, 250:1. Total is 800,000:1, more than it really needs. If the 6V6 are not real matched, there may be more buzz from 6V6 unbalance than from the preamp and driver.

> Leo got a really good deal on a large batch of 4x20uf cap cans.

Yes, that's probably a factor. Hi-Fi and PA were using a lot of caps like that. 4*20u 450V was a very common part both for OEM and for repairs. On paper you could maybe use less, or another way, but when 4*20u is always in-stock at good price, use it.

I think the take-away is to fiddle the two "18K" resistors to balance driver saturation against power-bottle saturation. You don't want to drop the driver B+ way down: you paid good money for 6V6 power and sweetness. For "racing" you want the driver to have lots of excess oomph. However for "music" there may be a point where driver strain adds to the output stage strain and tone.

[HotBluePlates]

However for "music" there may be a point where driver strain adds to the output stage strain and tone.

I was thinking about this. If we optimize for driving 6L6, we have "too much" for driving 6V6. The solution is probably to add a post phase inverter master volume to knock down the signal slightly when running 6V6. The goal would be to "re-lame" the inverter stage a little so that it also has a chance to distort around the same time as the output tubes.

[tubeswell]

> running all the pre-amp filter off one node, in order to eliminate sources of signal backwobble through the filter cap can  - i.e. it all synchs up if its taken off the same filter cap can node.

Wrong. In fact the AA1164 violates rule-of-thumb design: no more than two sequential gain-stages on one B+ node. Two is stable, three is inherently unstable. Wants to motor-boat.

I knew about the rule of thumb for number of plates per filter cap already, but I was thinking more about keeping the circuits' various ground return paths going to the same ground return point as the decoupling cap that supplies that part of the circuit. I was wondering whether, since they were using a cap can, they may have thought it better to load it onto one cap, rather than splitting it to more than one cap and risking ground loop hum.  Hence my reference to 'backwobble', by which I mean AC micro-rises in ground return path potential, that get into other parts of the ground return path (where you don't want it) and cause hum.

The AA1164 gets away with it because there is loss in the tone-stack, huge loss in the reverb mixer, and the power amp's two driver stages have a little "Akido" effect. The reverb driver would tip the balance to unstable, but it is fed from "B" which is a long way from "D".

Thanks for that explanation. That makes sense. And as it makes the loading of everything onto a single cap more viable in this amp, it would therefore likely be 'better' to have everything going to the single cap for the reason I was trying to get at above, would it not?.

> Plus the extra filter node then acts as another PII filter in the supply line -

Signal at first plate is about 1 Volt, you need ripple to be much less than 0.001V. The raw DC at the rectifier has 10V ripple. You need ripple reduction better than 10,000:1.

When capacitance is expensive you want to get about 10:1 reduction per cap and use several caps. When capacitance is cheap you might go for 30:1 or 100:1 per stage and use fewer stages.

So from rectifier to first plate you want two to four R-C networks. Older amps will favor more stages. The AA1164 has three R-C filters (in addition to the bulk filter off the rectifier). They actually give ripple-reductions of 14:1, 250:1, 250:1. Total is 800,000:1, more than it really needs. If the 6V6 are not real matched, there may be more buzz from 6V6 unbalance than from the preamp and driver.

I was thinking that the unused stage would help to stabilise the supply to the end filter cap supply point. But I see now that the extra filtering is OTT in any event. Cheers