How do the "big boys" do it? (Tetrode load lines)

JazzGuitarGimp · Post by **JazzGuitarGimp** » Mon Aug 20, 2012 5:04 am

I have asked this question of many people, and I am starting to think no one actually knows the answer. I've read Kuehnel's power amp book (and yes, I've even asked him via email) - and I am patiently awaiting Merlin's power amp book (which still appears to be a few years away).

My problem is with published grid curves for output tubes. I'd really like to be able to design a power stage from the ground up, using load lines. Problem is, manufacturer's published curves invariably are done at g2 voltages that are much lower than what guitar amps typically use. Yes, Kuehnel describes a way to "push the curves up" for higher screen grid values, but also admits that this is only useful for "small adjustments" before becoming too inaccurate to be of any practical use.

So, how do the big boys do this? You know, the Mesa Boogie, Dr.Z, Two-Rock, and Divided By 13 guys? Do they own curve tracers so they can plot their own curves to the anode and screen voltages of their choice? Do they know absolutely nothing about load lines, and just copy what's been done in the past? Do they use the rough math method to get into the ballpark, build a prototype, and see what happens?

I am not content to design an EL34 output stage with a g2 of 250V, just so I can have a set of curves onto which I can draw a useful load line!

There has got to be a way to get to my destination! Any ideas?

Many thanks,
JGG

katopan · Post by **katopan** » Mon Aug 20, 2012 8:28 am

The two most popular ways I've read of doing this are re-labelling the curves for new Vg values, and re-labelling the y-axis.

I've found re-labelling the curves to be quite accurate for more than just small adjustments. You can check the accuracy at least at the bias point back against curves, and for me it has lined up for multiple builds based on different output valves. Re-labelling the y-axis on the other hand gets really inaccurate very quickly.

Recently went through the theory on how to do this on another forum. I can post the example here if it's any help.

martin manning · Post by **martin manning** » Mon Aug 20, 2012 10:09 am

I'd appreciate it Katopan, please post. I've scaled anode curves using the method described in Kuehnel's book, and found it reasonably accurate based on scaling from one set of data sheet curves to another at a different Vg2.

JazzGuitarGimp · Post by **JazzGuitarGimp** » Mon Aug 20, 2012 11:47 am

Hi Katopan,

I too, would be very interested and appreciative to see this.

Many thanks!
JGG

tubeswell · Post by **tubeswell** » Mon Aug 20, 2012 4:45 pm

chief mushroom cloud · Mon Aug 20, 2012 7:14 pm

you need to contact Chris Merren
http://www.merrenaudio.com/
the guy knows this shit six ways to sunday

martin manning · Post by **martin manning** » Mon Aug 20, 2012 8:33 pm

Another way to go about this is to use a tube model in SPICE and have it trace out the plate curves for whatever Vg2 you want.

katopan · Post by **katopan** » Mon Aug 20, 2012 11:26 pm

I'm not familiar with Kuehnel's method in his book. This may be the same... But I was interested in this a couple of years ago, piecing together info from different sources, and as well Merlin offered some assistance in a thread about it on 18Watt.com. At the time he posted the 1950 article which I've reposted here as well.

We recently went through a SE 6V6 example on Wattkins. Here's what I wrote:
"Firstly the grid voltage curves need to be adjusted for the actual screen-cathode voltage. The standard graph is based on 250V screen-cathode. You've got 298V - 18.5V = 279.5V -> let's call it 280V.
Looking at the screen voltage curves (attachment 6V6_Screen_Curves) at 250V screen, 300V plate voltage gives 114mA for Vg = 0V. At 200V screen, 300V plate voltage gives 84mA for Vg = 0V. Then look at the grid voltage curves and you'll see the 250V screen point lines up with the Vg = 0V line (as you would expect). The 200V screen point lines up with the existing Vg = -5V line. This means at 200V screen the Vg = -5V line would become the new Vg = 0V line. So it's safe to assume going up to 300V screen the Vg = +5V line becomes the new Vg = 0V line. For 280V screen the new Vg = 0V line is 3/5ths of the space between the existing Vg = 0V and Vg = +5V lines. I've drawn this new Vg = 0V line, as well as the others down to Vg = -20V on the load line attachment. Re-drawing these lines also confirms the bias point you've got at -18.5V and 39mA."

This example isn't a huge change in screen voltage to what the chart is based on. But I have used the same method on bigger (and smaller) valves based on a much larger screen voltage change and it has been pretty accurate.

Do the same method with the EL34 curves and away you go. Keep in mind a single static load line is not going to represent the dynamic conditions of your amp. It might be going a bit far, but for one amp design I actually re-drew the curves twice - once for the bias conditions and again for the expected overdriven conditions. If you line up the load line through the knee of the Vg = 0 curve under no-signal conditions, it'll be different when the B+ and screen voltages sag under full signal. The plate and screens sag together so it tends to cancel each other out, but that also depends on your circuit design and selection of screen resistors, if you've got a valve recto or SS, etc. You can go too far with all this especially given the wide tolerance of actual valve performance, and a single load line at bias conditions is often good enough. But sometimes it's good to have the full picture if you're designing in new territory that hasn't been proven before.

The spice models are great tools, but you need to reality check the result just to make sure they're giving you the right answer in the context of how you are using them. I've used them extensively and the primary curves are good. Some of the secondary measurements can be out.

martin manning · Post by **martin manning** » Tue Aug 21, 2012 12:16 pm

Good stuff Katopan, Thanks!

JazzGuitarGimp · Post by **JazzGuitarGimp** » Tue Aug 21, 2012 3:18 pm

Excellent! Now I have a starting point. Thank you, Katopan!

jazbo8 · Post by **jazbo8** » Tue Jan 29, 2013 1:58 am

katopan wrote:I'm not familiar with Kuehnel's method in his book. This may be the same... But I was interested in this a couple of years ago, piecing together info from different sources, and as well Merlin offered some assistance in a thread about it on 18Watt.com. At the time he posted the 1950 article which I've reposted here as well.

The spice models are great tools, but you need to reality check the result just to make sure they're giving you the right answer in the context of how you are using them. I've used them extensively and the primary curves are good. Some of the secondary measurements can be out.

Having just joined the forum a few weeks ago, I am still going over the older threads - playing catch-up... Your solution to the screen voltage scaling is interesting, I will give it a try soon. As for the textbook approach, the subject of valve conversion factors (like many others), is well covered in the Radiotron Designer's Handbook (RDH), the conversion factors are relatively accurate upto 2.5:1.

If you guys are interested in the conversion chart, let me know, and I will dig it out and post it.

Post by **Colossal** » Tue Jan 29, 2013 5:44 am

I do pretty much everything in Excel, and plug and chug using methods by Kuehnel et al to scale published screen voltages to create a new set of curves. Most often, I just scale the Vg=0 curve to run the first approximation. Of late, I have been trying to find/build a suitable model to describe a pentode. A while ago, I made one which approximates a tetrode pretty well (see attached) but getting the taper to decay correctly as voltage goes up and grid current gets more positive has proven tough. The example is based on something Martin helped me out with and I was able to get my data to match his so I know it's working correctly. I've used a curve tracer program before but it was terribly tedious and I gave it up; I forget what it is called. I really just want to get an accurate equation for a tetrode and pentode and be able to use them for any situation. I'm slowly working up the iterative process in Excel to do composite load lines but right now just plotting the Class A and B lines.

Has anyone used TubeCad's software? http://www.glass-ware.com

JazzGuitarGimp · Post by **JazzGuitarGimp** » Tue Jan 29, 2013 6:35 am

Colossal wrote:I do pretty much everything in Excel, and plug and chug using methods by Kuehnel et al to scale published screen voltages to create a new set of curves. Most often, I just scale the Vg=0 curve to run the first approximation. Of late, I have been trying to find/build a suitable model to describe a pentode. A while ago, I made one which approximates a tetrode pretty well (see attached) but getting the taper to decay correctly as voltage goes up and grid current gets more positive has proven tough. The example is based on something Martin helped me out with and I was able to get my data to match his so I know it's working correctly. I've used a curve tracer program before but it was terribly tedious and I gave it up; I forget what it is called. I really just want to get an accurate equation for a tetrode and pentode and be able to use them for any situation. I'm slowly working up the iterative process in Excel to do composite load lines but right now just plotting the Class A and B lines.

Has anyone used TubeCad's software? http://www.glass-ware.com

Yes, I have most of their software. The app for plotting curves (I think they call it Live Curves) gives you a library of tubes to choose from, all of which are triodes. so, no ability to plot tetrodes and pentodes. Their emphasis is on HiFi tubes amps.

jazbo8 · Post by **jazbo8** » Tue Jan 29, 2013 8:50 am

Colossal wrote:I do pretty much everything in Excel, and plug and chug using methods by Kuehnel et al to scale published screen voltages to create a new set of curves. Most often, I just scale the Vg=0 curve to run the first approximation. Of late, I have been trying to find/build a suitable model to describe a pentode. A while ago, I made one which approximates a tetrode pretty well (see attached) but getting the taper to decay correctly as voltage goes up and grid current gets more positive has proven tough. The example is based on something Martin helped me out with and I was able to get my data to match his so I know it's working correctly. I've used a curve tracer program before but it was terribly tedious and I gave it up; I forget what it is called. I really just want to get an accurate equation for a tetrode and pentode and be able to use them for any situation. I'm slowly working up the iterative process in Excel to do composite load lines but right now just plotting the Class A and B lines.

Has anyone used TubeCad's software? http://www.glass-ware.com

"Of late, I have been trying to find/build a suitable model to describe a pentode."

So your tetrode curves were not based on manual data entry, that's impressive, how did you come up with the model? There are many SPICE models available, but I guess you already know that... I don't have the Kuehnel book, so what's his methodology for scaling the plate curves? I am guessing it's based on the conversion factor chart from the RDH...

Post by **Colossal** » Tue Jan 29, 2013 2:25 pm

JazzGuitarGimp wrote:Yes, I have most of their software. The app for plotting curves (I think they call it Live Curves) gives you a library of tubes to choose from, all of which are triodes. so, no ability to plot tetrodes and pentodes. Their emphasis is on HiFi tubes amps.

Thanks a lot Lou, I was wondering about that. Bummer about triodes only. You saved me a bunch of money!

jazbo8 wrote:So your tetrode curves were not based on manual data entry, that's impressive, how did you come up with the model? There are many SPICE models available, but I guess you already know that...

Jazbo, to find a starting point for modeling I've been scouring literature for equations (more miss than hit). I've used LT SPICE but have not really invested in the tube model side. I also want a tool that is intuitive, iterative, and doesn't leave you with more doubts than you started with. We go through this process every time we want to study an amp or design one from scratch, so I'm sick of always dealing with the same bottlenecks.

For models, I found two equations. One was from a tube modeling software package user manual which does the bulk of the heavy lifting to form the general pentode shape, however, they referenced tube variables which were totally obscure and had no explanation as to what they were or where they got the data from. The second equation creates a weighting factor that defines the value for a grid constant and this is then fit to the main model which plots the grid curve.

Iterating, I find that two variables just sort of weight the equation back and forth and ultimately you spend hours trying to pull the lines around, going nowhere. One of these sort of controls the taper when Va > the midpoint and the other is the shape of the knee but neither do exactly what I want.

It's still a work in progress and so I pick it up and slog along when I feel inspired and time permits. I think ultimately the equation might need to be broken into three distinct ones; one to describe the virtual cathode at Vg< say 100V, one to describe the hard diode effect in the tetrode and the soft knee of the pentode, and finally one for the linear region beyond the knee.

I don't have the Kuehnel book, so what's his methodology for scaling the plate curves? I am guessing it's based on the conversion factor chart from the RDH...

My books are packed (moving) but his method basically calls for curve tracing (you need points on a line for your known Vg=0 curve) and then this:

Ip2 = Ip1*(((Vg+(Vs2/Us))/(Vg+(Vs1/Us)))^1.5)

where...
Ip1 = the published screen voltage (x-axis from your traced curve)
Ip2 = the desired plate current
Vs1 = published screen voltage (y-axis from your traced curve)
Vs2 = the desired screen voltage
Vg = the grid voltage of desired curve
Us = ABS(Vs1/Vgc)

Tube Amp Archive

How do the "big boys" do it? (Tetrode load lines)

How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)

Re: How do the "big boys" do it? (Tetrode load lines)