Using a 12AV7

[SoundmasterG]

Hello,

I've been modding a Bogen CHB100 and was thinking of using a 12AV7 for the phase inverter. The amp has been completely gutted and the only thing I kept from the original are the transformers and the four 7868 power tubes. For the preamp I'm planning on using a bootstrapped 12AY7 with a gain stage and CF, then a FMV tone stack and volume control, then a gain stage set for moderate gain using a 12AX7 triode, another gain stage using the other half of the 12AX7 set up for slightly less gain and a different frequency response, then a cathodyne phase inverter using a 12AV7 going into the 7868 quad. I've read Merlin's book and site about setting up a CF to avoid the unpleasant aspects of when those overdrive, and was planning on using 47k grid stops on the power tubes instead of the 1k you commonly see with 7868/7591's.

I'm wondering what values and voltages I should shoot for on the 12AV7? I sketched something up on paper here using Merlin's Fig. 7.9 on page 168 in his book and adjusted some values a bit due to the different tube I'm using, but am not sure if I'm on the right track? So far I've got a 75k plate resistor on the driver stage, and a 2k cathode resistor with a 15uf cathode cap. In looking at Merlin's design, I'm not sure why the bottom resistor in the divider goes to the cathode of the splitter instead of ground, and I'm not sure what voltage I should shoot for on the grid of the splitter? I've got 75k plate and cathode resistors on the splitter so far also.

I can use a high voltage supply on the splitter and driver as the plates of the 7868's are at 450v on the schematic and 480v in reality. The screens are around 430v so pretty much anywhere down from there is available. I've read that using a higher supply voltage for the cathodyne will allow for more signal swing, and I do want to get full power plus overdrive out of this amp, but not huge gain. I'd like to set it up as a good blues amp for jams.

I was previously using a Vox AC100 seesaw version of the paraphase and didn't have enough swing for my liking the way I had it setup, and I think maybe the cathodyne will have more high end too? The amp sounded dull with the setup I had before which included a parallel 12AY7 on the input and a gain stage after the tone stack, then a feedback stage before the phase inverter. It didn't have enough gain and was really bassy and dull, so I've decided to change it up some.....

Does anyone have any advice or suggestions?

Thanks!

Greg

[HotBluePlates]

What is the bias on the 7868's? Did you mean to say that the output tube plates, screens, the phase inverter and the driver tube all have the same 480v supply? There's no dropping between all those points?

The only way to evaluate whether your plan is sound is to know how much output the inverter must make, and how much supply voltage is available. That's why I am asking about the 7868 bias and the supply voltage to the inverter.

[jojokeo]

The amp sounded dull with the setup I had before which included a parallel 12AY7 on the input and a gain stage after the tone stack, then a feedback stage before the phase inverter. It didn't have enough gain and was really bassy and dull, so I've decided to change it up some.....

Greg

Don't mean to but in here but there's some conflicting info I'm reading. You don't want to set up for too much gain earlier in your description and then you are not having enough. You are also using low gain tubes such as the AY7 which is easily changed out w/ others for more gain if wanted I'm sure you know. As for "a feedback stage" feedback by it's nature cancels out frequencies and lowers gain contributing to your complaints. Your tone stack values as well as your coupling & bypass caps will affect your "bassy & dull" feelings as well as not having and treble bleed caps on your dividers & vol pots. I thought I mention some of these things since you didn't. I hope this helps give you some ideas if they're even relevant currently?

[SoundmasterG]

What is the bias on the 7868's? Did you mean to say that the output tube plates, screens, the phase inverter and the driver tube all have the same 480v supply? There's no dropping between all those points?

The only way to evaluate whether your plan is sound is to know how much output the inverter must make, and how much supply voltage is available. That's why I am asking about the 7868 bias and the supply voltage to the inverter.

Right now the 7868 set I have is pretty mismatched, but they are biased to around 73% on a couple and 58% on the other two...thats of the 19w max dissipation. The schematic calls for 450v on the plates but its around 480v in reality. The screens are about 10 volts lower, and the phase inverter I had set to around 420v, but I can shift it around wherever it is needed. Every stage in the amp is seperately decoupled.

The amp sounded dull with the setup I had before which included a parallel 12AY7 on the input and a gain stage after the tone stack, then a feedback stage before the phase inverter. It didn't have enough gain and was really bassy and dull, so I've decided to change it up some.....

Greg

Don't mean to but in here but there's some conflicting info I'm reading. You don't want to set up for too much gain earlier in your description and then you are not having enough. You are also using low gain tubes such as the AY7 which is easily changed out w/ others for more gain if wanted I'm sure you know. As for "a feedback stage" feedback by it's nature cancels out frequencies and lowers gain contributing to your complaints. Your tone stack values as well as your coupling & bypass caps will affect your "bassy & dull" feelings as well as not having and treble bleed caps on your dividers & vol pots. I thought I mention some of these things since you didn't. I hope this helps give you some ideas if they're even relevant currently?

The way the amp is with the present setup, there isn't enough gain and the tone is bassy and dull. I realize I can change some parts in the tone stack and put in brite caps and what not, (it already has some) but I decided the overall sound of the amp wasn't what I was looking for, so I decided to change things up. I do want more gain than what I have right now, but I don't want the thing to turn into a gain monster like a Soldano or JCM 800 or something. I'm looking to use it for blues jams and would be happy with some good power amp overdrive. I'd like the phase inverter to drive the 7868's into typical Fender tweed or blackface territory, but I'd like to use the cathodyne inverter. You typically see those with a supply somewhere around 300v in those old Fenders, but I can go up to 450v if I need to for the supply. I just want to get enough drive to make the power tubes overdrive a healthy amount, but don't want a super gainy preamp. I've never worked with the 12AV7 and would like to use it for the phase inverter, but am not sure where to start at for resistor and cap values around it. I've got the 12AY7 on the input, but I am also bootstrapping it so it will have more gain than normal. So far I have the 12AY7, a CF, and two 12AX7 stages before the phase inverter, with a tone stack and volume control in between the CF and the 12AX7 stage. I plan to use dividers and/or tweak the gains of those sections to get where I would like to be, but would like to know more about setting up for a 12AV7 as the phase inverter if anyone can advise....
keeping in mind that the power tubes are 7868's and somewhat sensitive as compared to say a 6550....
Thanks!

Greg

[HotBluePlates]

No, no... I need the bias voltage. Either the negative voltage on G1 or the voltage across the cathode resistor.

[SoundmasterG]

Right now I have the bias voltage at about -25v....I have a quad of EH 7868's on order that are matched, so that will help to get a more accurate bias on every tube since some are mismatched by quite a bit right now. The spec on the schematic calls for -23v bias, but it was non-adjustable and I've set it to be adjustable now. I may add a cathode bias switching option to the amp later but for now it is fixed bias. Thanks for the help.

Greg

[jojokeo]

The 12AV7 has very similar plate resistance & transconductance of a 12aU but w/ the amplification factor of a 12aT that puts out more current the both. If it was me I'd go w/ the same values as if using a 12aU and use a large grid stopper for it. The lower the Ra & Ka resistors used the lower the output signal swing so start off w/ 56k's there intitally w/ maybe a biasing starting point of 1k or 1.2k - this hasn't been calculated on a load line but just my initial thoughts. Hope it helps?

[HotBluePlates]

I've started calculating. I'll post the book tomorrow some time. 

As a starting point to pencil in on your draft schematic, use a supply voltage of 350 for the inverter, a 12k load in the cathode and plate, and a 620 ohm cathode bias resistor.

In looking at Merlin's design, I'm not sure why the bottom resistor in the divider goes to the cathode of the splitter instead of ground, and I'm not sure what voltage I should shoot for on the grid of the splitter?

Merlin's diagram at the bottom of that page has a lot of things done by just a few components. The driver is d.c. coupled to the inverter grid, the larger of the voltage divider resistors is being used as the big grid stopper resistor and the smaller is being used as the grid reference resistor. Further, the divider is reducing the size of the signal input to the inverter's grid to avoid distortion in the inverter, and the whole divider is being used as a d.c. voltage divider between the plate of the driver and the cathode of the inverter to apply fixed bias to the inverter.

This is clever and tricky, and difficult to design. Bottom line, I don't believe it will work with our relatively low-gain 12AV7 the way it works with a high gain 12AX7. That's why I suggested the cathode bias resistor value above.

I still need to take some time to make sure I've got a good driver design. I'll post that along with an enormously long and detailed description of the process I went through to arrive at the values posted for the inverter.

[SoundmasterG]

I REALLY appreciate the help guys!

HBP, the diagram at the bottom of page 168 in Merlin's book is using a 12AU7, so I would think that same or similar circuit should work for a 12AV7 that has a gain of 41 as compared to the 12AU7 gain of 17? I initially picked 75k plate and cathode resistors because I have a bunch of the type that were used in 70's Marshalls that I wanted to use up, and it is a happy medium between a 12AX7's typical 100k and a 12AU7's typical 47k. I figured it would be a good starting point. I also would prefer that the power section distorts before the phase inverter or the preamp so I had started off with a 2k cathode resistor on the driver stage, but it will be interesting to see what you come up with.

I'm looking forward to your future post on this...specifically the reasons for the supply voltage chosen, and the choices of the 12k loads and the cathode bias resistor too. Thanks again!

Greg

[PRR]

Basic cathodyne driver:

KNOW:
peak grid voltage needed (often taken equal to standing grid-cathode bias)
grid resistor value Rg
proposed cathodyne tube and its parameters, especially Rp

CHECK:
Tube Rp should be much-less (4X-10X less) than grid resistors.
If this test fails, you can not use a simple approximate, you need the Sharp Pencil (or a more appropriate tube).


"bias voltage at about -25v"
7868 max grid resistance, fixed-bias: 300K per tube (150K legal for 2 tubes/side)


Pick supply voltage well in excess of 5 times the peak voltage needed. 10X is better.

The 7868 has a high Mu so this is not a problem. Same is true for most "audio power pentodes"; that's kinda mandatory for simple (affordable) audio amps; and why we don't normally use low-Mu triodes or sweep-tubes in audio.

Target the DC voltages at 1/4 and 3/4.... if the supply were 400V, aim the cathode at 100V and the plate at 300V.

However do not exceed the heater-cathode breakdown voltage. In many cases this will force you below 100V at cathode.

HBP's "350V" is wise. Rounding, take 90V across each resistor and 180V across the tube. That meets the 1/4-3/4 target and happens to hit the 12AV7's low 90V heater breakdown number.

Pick DC load resistors as the geometric-mean of Rp and Rg.

Square-Root(150K*6K)= 30K.

Check if Rp is valid at this current.

90V/30K= 3mA. This is much less than the 9mA and 18mA conditions where Rp is 6K or 5K. The G.E. sheet gives a curve showing that at 3mA Rp is near 13K. Round up and assume 15K. Repeat the DC computations:

Square-Root(150K*15K)= 47K.
90V/47K= 2mA.

At 2mA we find that Rp may be 17K, but close enough.

Altho if you are over-stocked on 75K resistors, that can work. 75K is not a lot less than the 150K grid resistors you should be using, but it is less. Also because we often cheat-up the grid resistors. Mostly because the 7868's high Mu does not require a best-effort driver.

Now, cathodyne bias. Take the 2mA 2*47K case, with 180V across the tube. A glance at the curves shows that 12AV7 can pass over 30mA at 180V! So we will bias more to cut-off than "on". Cut-off is ideally V/Mu, where Mu is nominally near 40.... but at this low current Mu is nearer 30. So 180V/30= 6V bias for cutoff, less to trickle 2mA. Drop a dart on "5V". Now the cathode bias resistor is 5V/2mA or 2.5K.

No further figuring is warranted. Build the dang thing! Use a 2.7K bias. Check for heater breakdown safety factor. Drop B+ if needed for heater safety. Check that it biases-up anywhere near the 1/4-3/4 points, diddle bias resistor 2.2K or 3.3K.

[HotBluePlates]

Bottom line: use PRR's suggested values. Also, for the driver stage ahead of the phase inverter, use a 100k plate load and the same 2.7k cathode bias resistor and cap-couple this stage to the inverter grid. If you need and want to add a grid stopper to the phase inverter, go for it. Feed this driver from the same 350v node as the phase inverter.

By my figuring, if you do this you'll need about 2.3v peak or so into the driver stage grid without feedback to push the output tubes to full power. If you use feedback, this number gets bigger by the amount of the feedback voltage. With the preamp setup you've outlined, you'll still probably need to throw away signal between the 12AX7 stages in "Gain" controls or fixed dividers to get a smooth overdriven sound to the driver.

The suggestion to use essentially the same operating point and load for the driver, and the same supply node, is due to the fact that the currents for each triode will be similar and opposite. That eases the filter requirements. This would have been a more prominent issue with the higher-current solution that I had come up with. PRR's relatively high load resistance lowers the tube current to be not much more than a typical preamp stage, while my solution would have been as much as 3 times the current and much lower internal resistance. That would have been a good solution if there were problems with grid current in the output tubes.

The posts below are shown just for reference on the design process (but you should also note PRR's rule of thumb approach, which is quicker) and shows how you can determine things like rp and mu at the operating point.

[HotBluePlates]

I'm wondering what values and voltages I should shoot for on the 12AV7? I sketched something up on paper here using Merlin's Fig. 7.9 on page 168 in his book and adjusted some values a bit due to the different tube I'm using, but am not sure if I'm on the right track? So far I've got a 75k plate resistor on the driver stage, and a 2k cathode resistor with a 15uf cathode cap. In looking at Merlin's design, I'm not sure why the bottom resistor in the divider goes to the cathode of the splitter instead of ground, and I'm not sure what voltage I should shoot for on the grid of the splitter? I've got 75k plate and cathode resistors on the splitter so far also.

Okay... First off, ya gotta follow the cardinal tube design rule: start at the output and work your way back towards the input. The way you have things described, it sounds like starting at the input and working to the output. The reason to do things this way is that with a later stage sorted out, you know what the next previous stage has to be able to do. This is why I asked for the bias voltage of the 7868's.

I’ll detail the entire design process used, so you’ll know where the numbers came from, and to allow you a template to use for future split-load design.

All tubes will start drawing grid current if the control grid gets to 0v; most output tubes will start drawing grid current even a little before that. Generally, once grid current starts flowing, the input impedance drops and the tube (and previous stage) start distorting heavily. This is termed by some as "input distortion" and is seperate from "output distortion" caused by the crowding of characteristic curves. We can take as a maximum needed input signal that signal which momentarily drives the output tube grids to 0v. With your -25v bias, this will be 25v peak or 50v peak-to-peak or 25v * 0.7071 = 17.7v RMS, whichever you prefer. When working with characteristic curves and loadlines, the peak voltage is the easiest to utilize.

I imagine you want to use the 12AV7 because you happen to have some lying around. Fair enough. We already have a handle on 1 requirement, which is an output of 25v peak, minimum. I also looked at some data sheets for the 7868, which stated that the maximum grid resistance for fixed-bias is 300k. You’re talking about switchable fixed and cathode bias, but we have to use the most restrictive case of 300k max. The bass roll-off formed by the coupling cap is determined by the cap’s value and the circuit resistance. A big resistance to ground allows the use of a smaller cap for the same bass roll-off, but this has to be traded off against the need for a small resistance to ground to minimize the effect of grid current and gas in the tube. A small grid resistance can also result in less gain from the stage ahead of it, if that stage’s output impedance is comparatively high. We’ll come back to this point later, but let’s pencil in the standard 220k resistor to ground and wait to figure our coupling cap value.

We can look at a 12AV7 data sheet and get a starting estimate for plate resistance and mu. For the typical operation listed, we see a plate resistance of 4800-6100 ohms and a mu between 37-41. Well, the plate currents listed are also big, much bigger than we’d probably like to operate a phase inverter or gain stage at. That means that when operated at lower, more reasonable current, the mu will probably be lower and the plate resistance higher. But we have a starting point to work from.

A basic rule of thumb when using a triode in a resistance-loaded voltage amplifier stage is to use a plate load of between 2x to 5x internal plate resistance. That allows for good gain from the stage. Using 5k as an estimated rp based on the data sheet, I chose to start outlining the split-load stage with a supply voltage of 350v (giving a good bit of drop from your available screen supply of 430v) and a load of 5 time rp, or 25k. 25k is not easily arrived at using a pair of standard-value resistors, so this becomes 24k or 12k + 12k. Since the split-load inverter has half the load in the cathode and half in the plate, we can use the standard value of 12k and draw a 24k loadline. From here on out, refer to the curves attached below.

The loadline runs from 350v and 0mA (tube drawing no current) to 0v and 350v/24k = 14.58mA (all supply voltage dropped across the load resistance). This is the d.c. loadline, and is a good starting point to determine the operating point. We may have to draw an a.c. loadline that takes into account the lowering of load resistance by the following grid resistor, but that’s unlikely since we should have a low output impedance given our tube choice and the effect of feedback in the split-load.

[HotBluePlates]

I arbitrarily chose an operating point where the -4v gridline crosses the loadline. Part of the reason was to reduce the amount of current drawn by the stage, which eases filter requirements (although the overall current is pretty small in a typical preamp). I also wanted a point along the line where input voltage resulted in an even output swing; that is, where the gridlines on either side of the operating point are equidistant. If you meet this criteria, distortion in the stage is small.

The -4v gridline crosses the 24k loadline at about 190v and 6.5mA. The plate current is quite a bit higher than anything you’re used to seeing with a common guitar amp stage. Also, it should be noted that because the load resistance is split between the plate and cathode that the 190v is not the voltage at the plate, but is the voltage between the plate and cathode. 350v-190v = 160v, so there is 80v across each load resistor; as a check 12k * 6.5mA = 78v, so there’s not a lot of error from my squinting and rounding. At this point, we’re looking at 350v – 80v = 270v on the plate and 80v on the cathode.

We should figure out what  rp and mu are at the operating point chosen. To figure mu, draw a horizontal line through the operating point to an equal amount of grid voltage on either side and note the plate voltage at the ends of the line. The definition of mu is the ratio of a change of plate voltage to a change of grid voltage, with plate current held constant. I drew the line giving 1v of grid voltage change on either side of the operating point, which showed plate voltages of 225v and 158v. (225v – 158v) / (-5v - -3v) = -33.5.

Rp is found by drawing a tangent to the gridline at the operating point, usually with the line as big as possible to minimize error. I squinted and estimated the plate voltage at the endpoints as 132v and 328v, and the currents as 0mA and 22.5mA. (328-132) / 22.5mA = 8.7k. That’s quite a bit higher than the data sheet number, but we’re at lower current. Our mu is lower than the data sheet numbers for the same reason. That might have thrown off our choice of operating point and load, but our load is still between 2-3 times rp, so we’re good.

If we move the grid by 2v in either direction, the voltage swings a little more than 50v during the positive grid input swing and a little less than 50v during the negative grid input swing. That means we’ll need a little more than 2v peak of effective input, and we’ll have a little distortion from the inverter (but not much). 50v peak is used because the output indicated on the loadline is really ½ in the plate and ½ in the cathode. Another way to look at it is to look at the current change caused by the grid input, and multiply that by each load resistor value to see how much output signal is present at each output.

[HotBluePlates]

To figure out how much actual grid voltage input we need, we have to know how much gain this stage really has after the 100% feedback caused by such a large cathode resistance. The formula is A = (mu*R1) / [rp + (mu+1)R2], where R1 is the plate load, R2 is the cathode load and rp is the plate resistance at the operating point.

A = (33.5*12k) / [8.7k + (33.5 + 1)12k] = 0.95

So to get 25v peak output from this stage we need 50v/0.95 = 52.6v of input. Round up to allow for a little extra leeway to insure the output tubes are fully driven; perhaps 53-54v of input.

The catch with Merlin’s diagram showing a driver stage d.c. coupled to the split-load is that the voltage divider using the large grid stopper and small bootstrapped grid resistor both fixed-biases the split-load inverter and also reduces the output signal of the driver to a level that won’t distort the inverter. We don’t have a tube with a lot of gain on tap if we use a similar operating point in our driver. The driver is already going to need more than 1v of input to drive the inverter, because the input needed at the inverter exceeds the tube’s mu. Further, to use fixed bias and d.c. coupling, we need the plate of the driver stage to sit a 4v below the cathode voltage of 80v. That would give us only 74v on the plate without a voltage divider to knock down the d.c. voltage along with the signal voltage. For our case, these things appear incompatible, and we should look to cathode bias the inverter and RC couple the driver to the inverter.

Our idle current is 6.5mA and we need a bias voltage of -4v, so we should use a cathode bias resistor of 4/6.5mA = 615 ohms. 620 ohms is a standard value in the 5% E24 series, so that is what we’ll use. We can use almost any resistor value we’d like for the grid reference resistor, so anything from 100k to 1M would be fine. Be sure to wire this Fender-style, with the 1M resistor running from the grid to the bottom of the 620 ohm bias resistor, which then has the 12k to ground.

Bootstrapping will ensure that the input impedance of the inverter is very high. Let’s say applied 1v to the grid. Because the stage gain is 0.95, we will get 0.95v at the cathode. That leaves 0.05v across the 1M grid resistor, which causes 0.05uA to flow through the grid resistor. To the preceding stage, it looks as though 1v caused a current of 0.05uA, which is equivalent to a resistor of 1v/0.05uA = 20M! Even a tiny coupling cap will produce the entire guitar range, and even a small grid resistor will present a high input impedance to the driver stage.

[HotBluePlates]

Another note, which applies to both my too-high-current solution and PRR's solution: use an elevated heater reference.

Yeah, I came in pretty close to being under the data sheet limit. However, you're already talking about another stage that's going to be a cathode follower. Exceeding the heater-to-cathode voltage rating may not result in sparks and destruction, but probably will result in noticeable hum.

I'd suggest elevating the heater center-tap to about 45-50v. That gives you plenty of margin on the split-load and cathode follower stages, and is also not too high for the stages with their cathode only a couple volts from ground.

Pick a convenient d.c. supply node, and make a voltage divider. Maybe bypass the lower leg of the divider with a large cap.

I'd also tend to suggest a standby switch in this amp, because while we probably won't kill old 12AV7's, it would be nice to let their heaters warm up and give the tube a chance to conduct and drop some voltage across the split-load resistors. That will ensure that the tube stays within its ratings. Of course, many amps abuse tubes and don't suffer destruction. Call it "best practice".

[SoundmasterG]

WOW, you guys are fantastic! 

Thanks a bunch PRR and HBP for going so in depth on this. This is exactly the kind of thing I was looking for....I'm going to have to go over the process you guys detailed with a fine tooth comb so I can figure out how to use this approach for designs in the future. For now this will get me started on the rebuild though, so thanks a bunch!

I do (of course) have some questions and comments however. I did forget to mention that the filaments are already elevated with 52v DC at the CT. I also already have a standby switch on the amp. I did choose to use the 12AV7 because I have a bunch of NOS ones lying around, and they seemed like they would make a good driver tube from my initial observations. From my initial component choices before I posted, it looks like I was heading in the right direction with the cathode resistor choice just from looking at the datasheet and trying to figure my way through it without knowing what I was doing, but the higher values you guys mentioned will work better so I'll use those. The plate resistor of 75k looks like it would work from what you've said, but I'll go with the 15k values you recommend for the splitter. I was comparing the use of a 6CG7 as a cathodyne in the Silvertone 1484 where they used 68k plate resistors, and comparing a 12AU7's typical 47k values and splitting the difference. I guess I was stuck on the mu of the tube rather than the other parameters. I can use my 75k's on something else in the future.

I did wonder about the possibility of grid current from the output stage and am still not clear on what I should use there for grid stoppers? The stock Bogen used 220k resistors to ground from each 7868 grid pair, and fed the bias through the junction of those. It also uses 1k grid stoppers on the power tubes. After reading Merlin's book in the cathodyne section, it sounds like adding more grid resistance is a good idea to avoid the nasty overdrive characteristics of the cathodyne phase inverter. So you say to use 150k PRR? I am assuming these would replace the 220k's that Bogen used, or are you meaning to use 150k in place of the 1k's that Bogen used?

I guess I still don't really get why Merlin's Figure 7.9 example wouldn't work if appropriate divider resistor choices were used to get the grid voltage where it would need to be to allow the splitter to bias correctly and still get enough signal? You would think that if Merlin's example is using a low mu (17) 12AU7 and my choice is using a higher mu (41) 12AV7 that there would be the possibility for more gain in my setup, thus allowing the circuit to work. What am I missing here?

Well with your guys help, I have a starting point to go build, so its off to the garage to take out the old and add the new.

Thanks again!

Greg

[PRR]

> maximum grid resistance for fixed-bias is 300k.

I believe he has four, two per side. Worst-case "legal" grid resistance is 150K. However maybe 99%, or maybe 90% or 50%, of a given tube-type will be stable with a higher resistor. What are the odds of two leaky tubes on the same side? And in HI-output fixed-bias we try to idle cool. A little leak and drift may not be a big problem.

> A basic rule of thumb when using a triode in a resistance-loaded voltage amplifier stage is to use a plate load of between 2x to 5x internal plate resistance.

Extend this. Make the plate load 2X to 5X smaller than the next-stage load. "Split the difference" between available driver (plate) resistance and required load resistance.

You don't have to make the two ratios equal. But if they are very un-equal you may not be best-possible. Your high-current plan is fine except 6.5mA is rather a lot to ripple-filter. In DIY the added cost of a larger B+ node-cap may be neglected, but commercial designers will tend to use low currents to cheapen the B+ filtering..... as you note "plate current is quite a bit higher than anything you’re used to seeing with a common guitar amp stage." (In fact you are coming to the reason why fat tuner triodes like 12AV7 are uncommon in audio.)

BTW: ripple can be less if cathodyne and its (triode) driver eat the same B+ node. But the reason is complicated and the improvement is small, probably insignificant. The real answer is that B+ filtering must be good enough that if "all" the ripple got into the audio, it would be negligible. For high power amps, maybe 87dB below full output. At this point, full output is 25V peak, say 18V RMS. 22,000 below 18V is about a milliVolt of ripple. Pretty stiff filtering from the several volts ripple at the first power cap. (And an urge to keep driver current down so we can use a large R in R-C filters rather than a large cap or a fine-wire choke.)

> What am I missing here?

Which specific Merlin plan are you looking at? There's one which almost-certainly biases the cathodyne grid +positive+, which can't work well. I'm not sure how that happened, it may have been drawn-up in haste.

[HotBluePlates]

The plate resistor of 75k looks like it would work from what you've said, but I'll go with the 15k values you recommend for the splitter.

No, use a 47k cathode load and 47k plate load, with the 2.7k bias resistor as PRR said. Both our solutions arrive at essentially the same gain and output from the inverter.

Why? Because PRR is right about the high current of the stage being harder to filter. What was going to take a little extra time for me on the driver stage was to recommend a driver with a plate load the same as the inverter's cathode and plate loads added, to cause essentially the same idle and signal currents to be drawn by both stages. The currents under signal conditions probably wouldn't exactly match, but it would reduce the amount of varying current and reduce voltage drop across the power supply dropping resistor.

But... this would still be 12-13mA total idle current for the driver and inverter!! I recognized this as a problem, but it also factored in to my choice to drop the supply voltage from 430v at the screens to 350 (or very slightly less) for the driver/inverter. The issue is harder filtering and a very *stout* filter cap needed for that node. The only real benefit is low internal resistance of the stages and because of this and smaller load resistors, the high-current solution delivers a low output impedance to drive the output tubes.  It also runs the 12AV7 more like it was intended, but doing that doesn't give us a marked advantage in a guitar amp.

But where I started "the book" I said go ahead and use PRR's suggested values because I recognized the high-current solution presents perhaps more problems than it solves.

I believe he has four, two per side. Worst-case "legal" grid resistance is 150K.

Yep. I had caught the error, but forgot to edit it out. 

Which specific Merlin plan are you looking at? There's one which almost-certainly biases the cathodyne grid +positive+, which can't work well. I'm not sure how that happened, it may have been drawn-up in haste.

I can't link the page from the computer I'm on, but it is a drawing where the driver stage is d.c. coupled to the split-load. But the d.c. coupling is accomplished through a voltage divider of 470k ll 47k to both divide the d.c. and the signal voltage. The big difference is the diagram uses a 12AX7.

I don't think I fully expressed the fact that this is not a workable solution for the 12AV7 (without significant re-thinking) because the 12AV7's gain is low enough to make dividing the signal down to an input to the inverter of 50v or more pretty difficult.

[SoundmasterG]

Ok I was confused there.....I was reading through both of your posts and I wasn't sure if 47k or 15k was the recommended plate and cathode load resistors for the splitter. So just to clarify then, the 2k7 is the cathode resistor for the driver stage? A 100k is recommended as the plate load for the driver stage? The splitter should get 47k for plate load resistor and cathode resistor? Using the Fender setup of the 1M resistor off the plate after the coupling cap of .02uf, the bottom of the 1M should go to the junction of the 47k and a 680 ohm resistor? Or is this where the 2k7 should be in place of the 680 ohm? Screw it...to make it easier I drew something up...does the file I attached look right?

HBP, the Figure 7.8 on page 168 of Merlin's book is the 12AX7 diagram...thats at the top of the page. If you notice, the one at the bottom of the page, Figure 7.9, is using a 12AU7. Since the 12AU7 has a gain of 17, I was thinking that the 12AV7 should also work with that same setup since it has a gain of 41? The divider is a 100 ohm off the plate of the driver stage, and a 47k through the diode, then to the cathode of the splitter. Given that the circuit you guys mentioned will work, I can use it, and am somewhat started on doing that, but I wonder if the 7.9 diagram on page 168 of Merlin's book would give any advantage? I wonder how it works with the 12AU7 but not the 12AV7 when the 12AV7 has higher gain?

Greg

P.S. I was planning on using a 47uf filter for the phase inverter stage as that is what I was using before when I was using the Vox AC100 seesaw paraphase inverter.

[HotBluePlates]

Everything is good, except use PRR's suggested 2.7k resistor for the cathode bias resistor of the inverter.

The idea behind the driver and inverter values are that the inverter is determined to work well with 47k+47k and a 2.7k cathode resistor. 47k+47k = 94k, or almost 100k, so we're reusing that load and cathode resistor in the driver. We're just moving all of the load into the plate circuit instead of splitting it between the plate and cathode.

HBP, the Figure 7.8 on page 168 of Merlin's book is the 12AX7 diagram...thats at the top of the page.

Well, I'm S.O.L. because I don't have a copy of his book. I saw the cathodyne page on his website. The very bottom of the page shows a fixed-bias split-load with a voltage divider between the driver and inverter. So that's the circuit I was describing.

Since the 12AU7 has a gain of 17, I was thinking that the 12AV7 should also work with that same setup since it has a gain of 41?

We're all used to saying all the 12A_7 tubes are "interchangeable" but with different gain because they all have the same pinout. And most circuit "will work" with any of those tubes in the socket. But really, they're all different tubes with different strengths and weaknesses, as well as slightly different ways to use them best.

The 12AV7 does have the low plate resistance of the 12AU7. But it also has roughly double the mu. There's a mathematical relationship between mu, rp and Gm which then also tells us that with double the mu, we have double the Gm.

If you had an existing amp, and simply plugged a 12AV7 into a 12AX7 socket, you'd have an amp that works, but probably wouldn't like the sound (I know I didn't when I did the same thing). Since we're working from scratch and want a "12AV7 circuit to use my existing tubes" then we're probably better off starting from scratch.

but I wonder if the 7.9 diagram on page 168 of Merlin's book would give any advantage? I wonder how it works with the 12AU7 but not the 12AV7 when the 12AV7 has higher gain?

Knowing that I don't have his book, I don't know if or how much the circuit may differ from that at the bottom of the webpage I linked. The one I linked shows a 12AX7. But it is not a plug-n-play circuit. The 470k resistor between the stages is an enormous grid stopper to prevent bad things from happening when/if the inverter distorts. The 47k is part of a voltage divider that knocks down the input signal and applies d.c. to the inverter's grid to fixed-bias it.

That divider (possibly with different values than the diagram you're looking at) makes the input to the inverter's grid 1/11th the size of the signal at the driver's plate. You need to have 26v peak or so at the grid of your 12AV7 inverter to drive the output tubes fully. If we changed nothing, that means you need 26 * 11 = 260v peak at the driver plate!!! You won't get that, not with the supply voltages we're talking and not with a 12AV7. 260v peak at the 12AV7 plate requires 260v / 32.5 = 8v peak input to the driver, which is more than 8 times what you normally feed to a 12AX7 driver or phase inverter. That means any typical preamp that is simply copied wouldn't come close to doing the job needed. All this assumes adequate supply voltage is available (which would need to be higher than 450v just for the 12AV7 by a quick guesstimate), and with a knowledge that even though the data sheet says "41" that this amount of mu really only happens at high current that you won't be running at. The mu of 32.5 that I used earlier is the actual mu measured at the operating point of the circuit PRR suggested, and the gain of any 12AV7 stage at similar current will be well below this unless the load is extremely high resistance (10's or 100's of Megohms).

We haven't even considered if the divider between the driver plate and inverter cathode will land at the voltage needed on the grid of a 12AV7 when the cathode is at the ~90v we estimated for the inverter.

This is not to say that the circuit I saw on Merlin's site won't work for a 12AX7. I imagine it would. But it is very cleverly put together and it is not trivial to refashion it for a tube of differing rp, mu or Gm. There's too many interlocking pieces to it for a simple swap. I suspect that the circuit you saw using a 12AU7 might have some clever non-obvious details built into it as well.

It's almost as though Merlin is fighting a reader's attemtp at cut-n-paste design. Which is great, because he's also giving the tools to do a proper design on his site and in his books.

[PRR]

> I wasn't sure if

You can practically steal the 5E3 plan, except because 12AV7 has half the Mu you will need a larger grid-cathode bias resistor (maybe 2.7K, may need to fiddle).

While I said that 300K/2 is the legal grid resistance, 220K will be OK.

While I figured 45K-47K range, if you like green-blue-orange resistors then use 56K.

[SoundmasterG]

Thank you PRR and HBP, it is much clearer now as far as what values I should use to get it up and running. I have about an hour's worth of work and I should be able to try it, since the amp was already up and running with my other design that I didn't like and I'm just changing parts mostly. I have to fix a Sunn Sentura 2 first though since its a paying job, but I think I should be able to post a progress report by the end of day tomorrow hopefully.

HBP, I highly suggest getting a copy of Merlin's book as there is a lot of cool stuff in there that isn't on his website. I think its the best of the amp books so far honestly. I'm waiting for his next one. :) I'll ask him if its ok to draw up and post a copy of the circuit I was talking about so you can see what was going on for discussion's sake.

I still don't have a good handle on what the tube ratings mean and how you use that and/or design a circuit around it. I'm determined to learn though, so I appreciate all the help. One of these days when I get some money, I hope to go back to school to get an EE and actually take my first electrical class, as so far I'm entirely self-taught with the help of books and forums and people willing to help. I know they don't teach tubes these days, but having the basics will certainly help!

Greg

[HotBluePlates]

Mu (grid to plate) = amplification factor = how much more effective a grid voltage change is than a voltage plate change in causing a change in plate current.

Mu is the least changing characteristic of the tube with operating point, but still changes. And it "lies to you" somewhat in that actual gain of a stage is dependant on how the load and the internal plate resistance of the tube form a voltage divider. Normally gain = Mu * Rl / (rp + Rl), for a common-cathode stage. We almost always get less actual gain than mu would suggest, unless the load is near-infinite.

rp = internal plate resistance = the ratio of a change of plate voltage to the resulting change of plate current.

This factor changes a lot, and rp is higher when tube current is low, and vice versa. A lower rp means you can use a lower plate load without as much loss in actual gain, due to the voltage divider effect. If the rp of a stage is low, the output impedance of the stage will be lower than if rp was high. That may be a factor in how well a stage can drive a tone circuit without serious losses, or how well a stage can supply a small amount of grid current drawn by a following stage that's seriously distorting. You might say that a circuit like a cathode follower will give you a low output impedance with a tube (like a 12AX7) that normally has a high rp, but that "low impedance" is due to feedback within the stage. You could get a lower output impedance with the same/similar circuit by using a tube with a naturally lower rp. Or by running the tube at higher current, if possible.

Gm = grid to plate transconductance = the ratio of a change in plate current to the change in grid voltage that caused it.

This is very relevant to output tubes, since it tells you how much grid voltage change is needed to get a plate current change. Or said another way, how easy it is to drive the output tube. It's also improtant in preamp pentodes because it is used in the easy way of predicting gain (aside from using a chart from the tube manual). Gm is higher when the tube current is higher. We usually don't care too much about how much Gm a preamp triode has, except that tubes that are worn out seem to exhibit less Gm (and at the same time higher rp) than when they were new. The end result is that their gain falls off, due to how Gm, rp and mu are inter-related. Gm can give an indication of how high the current can be run in a tube, as high Gm and high-current operation often leads to less tube noise. That's not a big factor in guitar amps, but could be in other tube circuits.

[Merlin]

Knowing that I don't have his book, I don't know if or how much the circuit may differ from that at the bottom of the webpage I linked. The one I linked shows a 12AX7. But it is not a plug-n-play circuit. The 470k resistor between the stages is an enormous grid stopper to prevent bad things from happening when/if the inverter distorts. The 47k is part of a voltage divider that knocks down the input signal and applies d.c. to the inverter's grid to fixed-bias it.

Actually that isn't a potential divider; the 47k is just part of the arc-protection circuit. There is no attenuation of the signal between stages in that diagram.

However, when it comes to a DC-coupled stage which does use a divider, you can just bypass the upper resistor with a cap to get the full gain back again. The divider would then just set the DC grid voltage, and could easily be made adjustable.

[SoundmasterG]

Thanks for the tutorial HBP, and thanks Merlin for elaborating. I have a better understanding of it now. I'll be working on the amp here in a bit so we'll see how it works out...

Greg

[HotBluePlates]

Actually that isn't a potential divider; the 47k is just part of the arc-protection circuit. There is no attenuation of the signal between stages in that diagram.

Bingo! That finally clicked with me.

My mistake was in seeing the diode as a short circuit, when it is really an open circuit when the tube tube operating. Because I saw it as a short circuit, it appeared to be a voltage divider from the driver plate to the inverter cathode.

But with no current flow through the reverse-biased diode, there's no division either.

Thanks for the correction.

[SoundmasterG]

I fired the amp up yesterday with the new circuit, but will have to add some shielded cable in places and reduce gain in others as so far it likes to squeal and oscillate a bit, which I expected. I'm still working on getting the stage voltages where I would like them and generally tweaking it, first to where it will function without oscillations, and then after that, for tone. But things are moving ahead slowly. More to come.....

Greg

[SoundmasterG]

Here is an update....

I've made some progress in reducing the oscillations...I'm going to try lowering voltages next as just little fixes like shielded cable and snubber caps and stuff didn't do the trick well enough. I also have a hum that I think is filament hum, but not sure why I am getting it yet. The amp sounds good already...MUCH better than it did before with my other circuit. The 12AV7 seems to be performing well as the phase inverter, though I haven't measured for swing yet or tried to balance the inverter. I think part of my hum is due to some badly balanced power tubes.....I'm waiting for a set of new 7868's to get here as the old ones I have are not balanced well at all....meanwhile I've been trying to get a handle on the squealing problem when the treble control is turned up and the volume is also up. More to come.....

Greg

[SoundmasterG]

Well I finally added a local feedback stage instead of a gain stage in the amp, and have adjusted it for gain. The amp sounds pretty good...mostly stable unless it is completely dimed and there is a little fizz on the note that I think is coming from the driver stage, but I know I can do better. I'm in the process of rethinking some things on the amp and will post updates as I get them.

Greg

[SoundmasterG]

I built a new preamp circuit into the Bogen today. I am using the switching parallel arrangement from Merlin's book with a couple twists as the input stage > Vox tone stack > volume control > triode half of 6BR8A > pentode half of 6BR8A > 12AV7 cathodyne PI > cut control > power amp. I have to tweak values and get the gain scheduling right, and may change the PI to a LTP, but so far it sounds pretty good...going from bluesy overdrive when the gain is lower down, to crunch when full up, which is what I was after. Another week hopefully and I'll have it ready to try out live.

Greg

[HotBluePlates]

Sounds great! So far, have you managed to control the oscillation issue?

And in the end, the phase inverter gave the output needed for the power tubes?

[SoundmasterG]

Yes, the phase inverter is giving 45v out on one side and 30 ish out on the other, and the bias is about -20 so its giving full power. Its a loud SOB! :) I haven't tweaked the phase inverter yet to make sure it is operating where it should be. I think the oscillations are coming from the driver stage, but I'm not sure yet. I've changed around the preamp a couple times to get to something I'm happy with and so far I'm liking this one, so more tweaking is necessary to tune the whole thing up. I can't see any oscillations on the scope no matter where I probe, but I can hear them when the volumes and trebles are maxed...a fizz on the end of notes and sometimes a squeal too. It sounds great at lower volumes, even with the treble up high. My layout is as good as it can be in this small chassis and I may just have too much gain for what it is, not sure yet. You would think I should be able to see an oscillation on the scope, but I've tried adjusting the input signal up really high and down low and probing everywhere and I can't see anything. It makes it hard to find the oscillation!

I'm about to go out and work on it some more. Theres a blues jam tonight and I'd like to try it out if I can. :) I've almost got the parallel input stage working well, and I just need to set the bias and voltages so I get the right overdrive characteristics going on in the preamp. I'll probably try to tweak this phase inverter a bit before I try a LTP. I like the overall sound of it except for the instability in the amp.

Thanks again for your help.

Greg