Advice on setting operating point?

justinis

Active Member
I am building a pure pentode class AB1 push-pull amplifier using 7591 tubes. I was planning to use following operating voltages found in the datasheet: plates 450V, screens 350V, grid -16.5V.

However, I noticed that in the datasheet these values are listed for a 6.6K load resistance. My transformer primaries (not ultralinear) are 7.7K. Do I need to adjust the operating voltages? If so, how do I do this?

I have been studying information about loadlines and from what I can tell, as long as I stay below the max plate dissipation, I think I'm safe to just stick with the numbers in the datasheet. I have a fixed bias similar to what's used in a ST-70, so I plan to adjust the cathode bias current to 25mA, which I think will be safe considering the max plate dissipation is 19W. Is this right?

Thank you!
 
I think you're going to want to do a real load line analysis. You may find that your voltages need to be tweaked. A key point is that you want your Class B load line to pass above the knee of the 0V grid curve

So, for the Class A portion, 450V into 3850 ohms is 117mA. For the Class B portion, 450V into 1925 ohms is 234mA. At those voltages, it looks extremely difficult to keep it below the max plate dissipation curve. It enters the Class B portion of the load conditions relatively quickly and then immediately exceeds max dissipation. Furthermore, at that operating point, the load line passes below the knee of the 0V grid curve, at which point the screen starts soaking up all the current.

Despite what the "Typical Operation" table in the data sheet (at least, the Sylvania one I'm looking at) says, I don't think I'd run that tube at that operating point.

To fix the 0V grid curve problem, you'd have to run a lower screen voltage to pull the grid curves downwards. That would also mean changing your grid bias voltage; how much depends on how far you pull down the curves. But you're still left with over-dissipation problem in the Class B portion of the load. (Seems to be this tube is very obviously meant to be run ultra-linear.)

I think you'd be better off with 350V + 350V, but then you're going to want an OT with a lower primary impedance to keep the Class B portion of the load above the knee.

Basically, I think those are the wrong output transformers for the job.
 

Attachments

  • Screen Shot 2016-05-25 at 1.48.45 PM.png
    Screen Shot 2016-05-25 at 1.48.45 PM.png
    351.4 KB · Views: 34
Oh, it occurs to me that perhaps the over-dissipation issue in Class B isn't quite that bad because the tube is fully cut-off for 1/2 of the cycle... But it's still something to be aware of... seems like you're still running it on the razor's edge there.
 
Thanks for your reply! I apologize for my ignorance, but could you please explain how you drew the blue curve in your graph? I understand the black one, but I don't understand how to get the blue one.

For what it's worth the tubes and the transformers came from a Heathkit AA-100, so they have been used together before. But I understand that doesn't mean it's a good design. The Heahtkit operating points were plates 445V, screens 365V, grid -16V.
 
An idea: what if I increased my HT to 480 or 490V? Keep the screens at 350V. Then my class B loadline would go straight through the knee. I could bias my plate current to ~20mA which I think would be -17VDC or so. Would this be happy?
 
The black line is the Class B line. The one with the blue glow around it is the Class A line. When you draw the Class A line, you have to shift it up the graph vertically while maintaining the gradient to the intersection of the plate voltage (vertical) and quiescent current (horizontal). I visually ball-parked it at -16.5V or so.
 
The black line is the Class B line. The one with the blue glow around it is the Class A line. When you draw the Class A line, you have to shift it up the graph vertically while maintaining the gradient to the intersection of the plate voltage (vertical) and quiescent current (horizontal). I visually ball-parked it at -16.5V or so.

Got it. I think I understand now. Thank you! So if I'm thinking about this right, increasing my plate voltage can only help me, as long as I don't exceed it's max of 550V. I think I can achieve 480 or maybe 490V with the power supply I have now. If I do this, and set my grid at -17 or so like I suggested before, this would be a better place, correct?
 
The problem with increasing your plate voltage is that you're sliding the cut-off part of the Class B line to the right and also making the Class B load line steeper (because you'll be pulling more current to drop the higher voltage over the same load), which puts you even further into the "over-dissipation" part of the graph. Increasing the plate voltage is going to do the same thing to the Class A load line, meaning you'll have to bias it further towards cut-off.

Increasing the voltages is the wrong way to go, I think.
 
Seems to be this tube is very obviously meant to be run ultra-linear

I'm honestly not sure I've ever seen a UL 7591 amplifier. All of the ones I know of are straight pentode, and most run the screens at nearly the same voltage as the plate. They do tend to run mostly about a 6-6.5k trafo, but later Fisher 400's ran a 10K, and my Sherwoods use a 5.3k transformer. They run fairly high voltages too for the most part, those tubes really had the snot worked out of them. Would be sort of interesting to plot the operational parameters in several of the common 7591 amps to see how far off ideal it is. Or do it and then have someone explain it to me. Load lines are spaghetti thrown on graph paper to me. I do not understand their purpose or use, other than its apparently important and tells you stuff.
 
I'm honestly not sure I've ever seen a UL 7591 amplifier. All of the ones I know of are straight pentode, and most run the screens at nearly the same voltage as the plate. They do tend to run mostly about a 6-6.5k trafo, but later Fisher 400's ran a 10K, and my Sherwoods use a 5.3k transformer. They run fairly high voltages too for the most part, those tubes really had the snot worked out of them. Would be sort of interesting to plot the operational parameters in several of the common 7591 amps to see how far off ideal it is. Or do it and then have someone explain it to me. Load lines are spaghetti thrown on graph paper to me. I do not understand their purpose or use, other than its apparently important and tells you stuff.

Huh, interesting. The Sylvania data sheet specifically calls out ultra-linear. It could be that the tubes could simply handle being roughed up like that.

I'd be happy to do a set of load lines (and explain them, to be best of my ability :) ) for your Sherwoods if you can tell me the parameters!
 
You won't find any 7591 commercial amps operated in ultralinear even though the tube manual says it can , The tubes won't last .

Operate it the same as 6.6K just expect slightly less power , and less distortion
 
The problem with increasing your plate voltage is that you're sliding the cut-off part of the Class B line to the right and also making the Class B load line steeper (because you'll be pulling more current to drop the higher voltage over the same load), which puts you even further into the "over-dissipation" part of the graph. Increasing the plate voltage is going to do the same thing to the Class A load line, meaning you'll have to bias it further towards cut-off.

Crap, I see what you mean. I started playing around with this, which was helpful: http://bmamps.com/Tech_tds.html. I guess the best thing to do is run at 350/350, or maybe something a little higher like 360/350.

Sure seems like a lot of Heathkit AA-100s must have been frying tubes running them at 445/365/-16. I plugged all the values in the datasheet's "typical operation" table into the link above and there are several that have the load line far into max dissipation. I don't get it.
 
The Fisher 400 clone output stage I built with the later model 400 output transformers did use 10K primaries. The 7868 plates (which is electrically a 7591) were running at about 430V, and the screens at about 310V, so I expect with 7.5K transformers, 350V screen would be about right, just putting the finger to the wind.

Ultimately it may not matter too much as long as you are in the ball park, which I think you are, given your opening post data, however you might be able to squeak out an additional few watts if you "tune" the screen voltage to match the output transformer primary impedance, which you can do via load line analysis as has been mentioned. But unless you have plate curves for the exact screen voltage you need, you are essentially extrapolating or interpolating graph data anyway, so it may not be worth the time to try to put too much of a fine point on it, given that I think your proposed voltages are within the ball park already.
 
But unless you have plate curves for the exact screen voltage you need, you are essentially extrapolating or interpolating graph data anyway, so it may not be worth the time to try to put too much of a fine point on it, given that I think your proposed voltages are within the ball park already.

Thank you. It looks like you're suggesting that I don't necessarily need to draw my load line so it stays away from the max plate dissipation curve. If that's true, what is the proper way to draw it? I would love to learn!
 
The load line is only an approximation anyway. Under a real speaker load, the instantaneous dissipation that the tube actually sees isn't linear, I'm tempted to say it looks more like a skewed football or oval shape. In the class B portion of the power cycle, the tube is only conducting for half of a full cycle anyway, so during the "off" times it has a chance to cool down. During the "on" times, it might have instantaneous dissipation of more than the design center max values on the data sheet, but that's okay for the most part as long as the designer isn't exceeding them by a huge amount. Outside of the longevity discussion, from the perspective of what's permissible, I think it's okay if you want to bias the tubes right at design center max dissipation ratings if that's what your design calls for. Although in that scenario you might get shorter tube life because of it.

This thread may inch you closer to understanding how to do it: finding the proper load for a pentode push-pull output stage. The relevant stuff is between posts 4 and 9 or thereabouts.
 
I do not understand their purpose or use, other than its apparently important and tells you stuff.

For me one of the main purposes of this whole load-line shebang is, given a pre-determined plate supply voltage, to determine how much current I need the power transformer to supply to support output stage max power output. A close corollary of that is to find a more or less optimal screen voltage that will deliver that max output power (and for me that is the screen voltage that will set the 0V grid curve to intersect the class B load line at the knee).
 
The load line is only an approximation anyway. Under a real speaker load, the instantaneous dissipation that the tube actually sees isn't linear, I'm tempted to say it looks more like a skewed football or oval shape. In the class B portion of the power cycle, the tube is only conducting for half of a full cycle anyway, so during the "off" times it has a chance to cool down. During the "on" times, it might have instantaneous dissipation of more than the design center max values on the data sheet, but that's okay for the most part as long as the designer isn't exceeding them by a huge amount. Outside of the longevity discussion, from the perspective of what's permissible, I think it's okay if you want to bias the tubes right at design center max dissipation ratings if that's what your design calls for. Although in that scenario you might get shorter tube life because of it.

This thread may inch you closer to understanding how to do it: finding the proper load for a pentode push-pull output stage. The relevant stuff is between posts 4 and 9 or thereabouts.

Thank you so much! That thread was very interesting. I think I see the purpose of loadlines now.

So in my situation, I have 7.7K transformers and 7591 tubes in pure pentode class AB1. That can't change. I can set my plate voltage to anything below 490VDC, and my screen voltage equal to the plate voltage, or anything less. How would I go about deciding what the best plate and screen voltages are? Are loadlines not the right tool?
 
Crap, I see what you mean. I started playing around with this, which was helpful: http://bmamps.com/Tech_tds.html. I guess the best thing to do is run at 350/350, or maybe something a little higher like 360/350.

Sure seems like a lot of Heathkit AA-100s must have been frying tubes running them at 445/365/-16. I plugged all the values in the datasheet's "typical operation" table into the link above and there are several that have the load line far into max dissipation. I don't get it.

With the AA100, I just accept a little less power- and back of the idle bias to about 35ma per tube. Usually, that means about -19v grid bias, in very general terms. I also drop the voltage as needed (with a series resistor after the rectifier tube and first cap) to get the plate voltage down to something like 435v or so.

That seems to give good tube life... and it's still got decent power.

Regards
Gordon,
 
With the Heathkit AA-100 output transformers, if the plate voltage is set to 430V, I think you'd want about 325V screen voltage. Conversely, with a plate voltage of 400V, I think about 300V screen voltage would be optimal. (I got those screen voltage estimates by eyeballing the EC2=350V and EC2=400V curve families to get a sense for how the grid curves adjust with a 50V EC2 change, and attempting to put the class B load line for each scenario right in the center of the knee of the 0V grid curve).

So either one of those look like they would work to me. I probably wouldn't go much higher than 430V on the plates.

Power-wise, probably about 23 to 25 watts output in either scenario, with the first two or three of those watts being delivered as class A.

You can play with the quiescent bias point in both scenarios, but somewhere between 30 mA and 35 mA seems about right to me. It depends a bit on the tube, but that's probably about -17V to -18V grid, but you should give maybe 15% adjust range on that, so adjustable between maybe -15V and -21V so you can dial it in.
 
Last edited:
Back
Top Bottom