Does the standard rule of thumb for plate load resistors (~2x the plate resistance) still hold true when using a CCS tail?
I ask the question as most of the reading I've done tends to gloss over on the exact details of designing a LTP with a CCS. Most writers say to use one for best balance between outputs, but then proceed to explain how to design one for a guitar amp using a smallish tail resistor.
I've got one on my bench that is working fine, but the values of my plate loads were arrived by looking at other designs and a little trial and error rather than a general understanding of the theory.
I ask the question as most of the reading I've done tends to gloss over on the exact details of designing a LTP with a CCS. Most writers say to use one for best balance between outputs, but then proceed to explain how to design one for a guitar amp using a smallish tail resistor.
I've got one on my bench that is working fine, but the values of my plate loads were arrived by looking at other designs and a little trial and error rather than a general understanding of the theory.
Yes the CCS does not change the plate calculations or the voltage swing you will get. The plate can get up to CCS and normally runs at half CCS.
If you use a resistor then when the voltage drops the current in the resistor drops causing the in balance you mention. If you use a CCS then the sum of the currents in the plate loads is always equal. If you do use a CCS then make sure you have enough voltage across it. This may mean running the grids above 0V or using a negative supply for the CCS. Of course you can tie the tail resistor to the negative supply if you have one to improve balance. I tend to use LTspice to experiment there's plenty of valve models on diyaudio.
If you use a resistor then when the voltage drops the current in the resistor drops causing the in balance you mention. If you use a CCS then the sum of the currents in the plate loads is always equal. If you do use a CCS then make sure you have enough voltage across it. This may mean running the grids above 0V or using a negative supply for the CCS. Of course you can tie the tail resistor to the negative supply if you have one to improve balance. I tend to use LTspice to experiment there's plenty of valve models on diyaudio.
Last edited:
Well, when you have set the current, draw a horizontal line across your valve curves and then decide what combination of bias and anode voltage you want for your operating point; this determines the load resistor.
A horizontal line is constant current. The tubes in a normal LTP do NOT operate at constant current. To operate at constant current you need one CCS in each anode and a resistor in the tail.
Last edited:
I'm talking about choosing the static operating point, in the absence of signal, just as you do with a SE stage.
You actually have to draw a line for the anode resistor in parallel with the next stage grid leak for AC analysis.
If you want to complicate things further, you also have to account for internal resistance. Changing as the slope of the valve in question transitions over its operating curves.
We spend the good part of three months learning circuit AC theory in class. Just taking an anode resistance and superimposing it on the plate curves produces an error.
You can optimize linearity by actually tweaking the next stage grid resistor, because this changes the slope of the load line. if you have curves with very even spacings at one point you can see the tube is quite linear. You want the slope of the load line to intersect at such an angle as to produce either good linearity. Or the right amount of distortion.
There is more to everything than meets the eye.
I dont want to come off as a wise-*** but this is what i was taught. We used field effect transistors for the analysis, however the same rules apply to thermionic valves. The region where a field effect transistor is not in saturation is actually called the triode region.
But in reality, you can only be so accurate as the accuracy of the parts, if you have a mu tracer you can get very accurate estimates of gain and distortion by graphical analysis for using the curves of the tube you are going to use.
But then you run into another problem, once you start amplifying with the tube, the anode heats up much more than the pulsed mode of that tracer. The heat from the anode is partly reflected back into the cathode and increases emission, which increases S and lowers internal resistance.
Furthermore most curves where made by taking the average of a production run of valves, no bogey tubes exist, only those that overlay quite nicely with the specs in the data sheet.
So in conclusion: You can try to chase accuracy in your methods, but all models are inherently inaccurate in nature. If you take data points you can increase the accuracy somewhat by increasing the amount of data points collected and running this through a average filter.
But in the end of the day, you can build a nice sounding amplifier with just a pencil and a ruler. If you remember to parallel Ra and RL you will get more accurate results. The formula for two parallel resistors is (R1*R2)/(R1+R2)
Cheers,
V4lve
If you want to complicate things further, you also have to account for internal resistance. Changing as the slope of the valve in question transitions over its operating curves.
We spend the good part of three months learning circuit AC theory in class. Just taking an anode resistance and superimposing it on the plate curves produces an error.
You can optimize linearity by actually tweaking the next stage grid resistor, because this changes the slope of the load line. if you have curves with very even spacings at one point you can see the tube is quite linear. You want the slope of the load line to intersect at such an angle as to produce either good linearity. Or the right amount of distortion.
There is more to everything than meets the eye.
I dont want to come off as a wise-*** but this is what i was taught. We used field effect transistors for the analysis, however the same rules apply to thermionic valves. The region where a field effect transistor is not in saturation is actually called the triode region.
But in reality, you can only be so accurate as the accuracy of the parts, if you have a mu tracer you can get very accurate estimates of gain and distortion by graphical analysis for using the curves of the tube you are going to use.
But then you run into another problem, once you start amplifying with the tube, the anode heats up much more than the pulsed mode of that tracer. The heat from the anode is partly reflected back into the cathode and increases emission, which increases S and lowers internal resistance.
Furthermore most curves where made by taking the average of a production run of valves, no bogey tubes exist, only those that overlay quite nicely with the specs in the data sheet.
So in conclusion: You can try to chase accuracy in your methods, but all models are inherently inaccurate in nature. If you take data points you can increase the accuracy somewhat by increasing the amount of data points collected and running this through a average filter.
But in the end of the day, you can build a nice sounding amplifier with just a pencil and a ruler. If you remember to parallel Ra and RL you will get more accurate results. The formula for two parallel resistors is (R1*R2)/(R1+R2)
Cheers,
V4lve
Last edited:
Ok decide on how much drive into you load you need, and the gain you require. This will determine whether you use a 12AX7 or a 12AT7 for example. At full CCS the grid must not go positive into conduction. So for a 12AX7 choice CCS for max 1ma say. Then you know .5ma goes down each plate, so pick plate resistor to give static voltage of around 65% of HT. So that would make the plate resistors say 180k.
Last edited:
If you really want to split hairs, you have to account for the output impedance seen at the cathode of tubes. Sometimes designers use resistors from each cathode to the CCS. To swamp the output impedance of the cathodes.
You can see this in transistor long tailed pairs, i beleive the output impedance of a emitter follower is IC/25mV but i could be mistaken.
65% of HT is a good rule of thumb that will give slightly higher overall gain before clipping due to the fact, that some tubes are limited by their internal resistance in how low they can pull the anode resistor.
You can see this in transistor long tailed pairs, i beleive the output impedance of a emitter follower is IC/25mV but i could be mistaken.
65% of HT is a good rule of thumb that will give slightly higher overall gain before clipping due to the fact, that some tubes are limited by their internal resistance in how low they can pull the anode resistor.
What about using a simulator that's a good way of experimenting. I don't tend to approach it from the valve curves but rather just play in the simulation and learn that way.
Morgan Jones is very good on LTPs; as a former BBC engineer, balanced operation seems to be his preferred method.
v4lve, do you really draw a AC loadline when you are choosing your DC operating point?
I only think about AC conditions after I've got my stage set up.
I only think about AC conditions after I've got my stage set up.
Last edited:
If you are interested in the theory of tube construction, there is "Grondslagen van de radiobuizentechniek" But I can also reccomend Rodenhuis Hifi circuits.
For current source analysis in tubes, there is little information online.
For AC circuit analysis, most internet resources will give you the basics. But CCS loading is far more complicated than above. Because the CCS impedance changes with frequency, however if you use low internal resistance tubes EG. 10K internal resistance, this will dominate the gain of the tube. And if you plot it on a logarithmic scale it will look ruler flat.
A resistor in the tail of a long tailed pair (differential amplifier) will have the effect of making a voltage divider, because the grid of one half of the tube is normally connected to ground. one tube gets the signal from the other tube through the cathode. In which case its attenuated by the impedance looking out of the cathode. And the impedance of the tail resistor.
One common misconception about long tailed pairs is that they increase the gain, which they don't really do that much. But they do increase the common mode rejection ratio of the two inputs.
For current source analysis in tubes, there is little information online.
For AC circuit analysis, most internet resources will give you the basics. But CCS loading is far more complicated than above. Because the CCS impedance changes with frequency, however if you use low internal resistance tubes EG. 10K internal resistance, this will dominate the gain of the tube. And if you plot it on a logarithmic scale it will look ruler flat.
A resistor in the tail of a long tailed pair (differential amplifier) will have the effect of making a voltage divider, because the grid of one half of the tube is normally connected to ground. one tube gets the signal from the other tube through the cathode. In which case its attenuated by the impedance looking out of the cathode. And the impedance of the tail resistor.
One common misconception about long tailed pairs is that they increase the gain, which they don't really do that much. But they do increase the common mode rejection ratio of the two inputs.
Last edited:
Oh that depends on the choice of valve. 12BH7 about 10x, 12ax7 maybe 50-100x. Cascade LPT 2sk170 + 12ax7 about 400x.
Not really, Im too lazy for spice, what i normally do is just current source the tube. and draw another line that is the loadline for the grid resistor. Once i am done with that. I use the MK1 eyeball.
The 65% percentage mark on HT for anode in quiescent is excellent as a rule of thumb.
If you really get OCD about doing everything by the book, you will end up very dissapointed because you lose track of the bigger picture. An amplifier is not only a low THD stage, but the power supply is just as important so are good bandwith good primary inductance output transformers. And while your at it, heat the heaters with DC..
The book is only useful if you have to put someone in their place for getting on a high horse. but truth be told. Ive had my hind handed to me on this forum on multiple occasions. By either not reading the whole thing carefully or by people that have discussed a particular topic to death already.
Sometimes this forum amazes me, some people are working with spice models that are based on a single sample of a tube that may have 30% spread in the parameters. Even if you take the curves from the book, those curves only apply to perhaps 3% of production.
And then they share this model, and these people make a game out of who can get 0.02% lower THD in a simulator. Just so they can ego trip behind a keyboard. So in essence they are running a simulation of a simulation that is a representation of reality.
😀
Take care,
V4lve
Yes its easy to design a perfect amp in spice. However you can do some tricks like make an LTP one half 12AU7 and one half 12BH7. You then start to see the effects of spread and mismatches in the system.
Correct. however if you use a single dual triode, you can be safe to assume that you can get the gain between a few percent. Because the mu of a tube is pretty stable unless you get to the end of the life of a tube.
Looking out of both halves of a LTP you have an output impedance that is roughly 1/S by adding a small resistor of say 22 ohms you can decrease the overall effects of the drop in S as the tube ages.
Some people use pentodes for drivers, but forget to realize that if you fix G2 with a zener string, there wont be any feedback mechanism to compensate for tube ageing. Much like people that use LED bias or SIC diode bias.
But these things are tried, because everything has been tried before and discussed to death. So people try to be original and invent new ways of doing the same, claiming sonic improvements. But truth of the matter is the drawbacks of most of these circuits are not included in the presentation. And A-technical people read this and think these people are the circuit gods.
A zener string is nice. if you want to have an absolute value for G2 voltage for calulation purposes. But when it comes to reliability and simplicity a resistor divider may be better in some circumstances. But a resistor divider requires that you know how to do an thevenin equivalent circuit and know how to calculate this.
I'd say only about 10% of people on this forum even know how to do that, let alone read data sheet with maximum specifications. These people are all talk sometimes.
Another thing: Most people that run pentode or tetrode amps measure cathode current. But say G2 is fed from a different suppy. Anyone that has tested tubes on a tube tester can tell you that two tubes that measure 100mA cathode current can have for example 3 and 7mA of screen grid current. Then these people try to balance an output stage with a multimeter. and are very happy when both tubes measure exactly 48mA cathode current.
If you have bifilar wound output transformers, you can have the same resistance from the center tap of a UGT to both ends, in that case it makes more sense to measure and compare the anode voltages of a PP stage.
Keep your feet on the ground at all times.
Looking out of both halves of a LTP you have an output impedance that is roughly 1/S by adding a small resistor of say 22 ohms you can decrease the overall effects of the drop in S as the tube ages.
Some people use pentodes for drivers, but forget to realize that if you fix G2 with a zener string, there wont be any feedback mechanism to compensate for tube ageing. Much like people that use LED bias or SIC diode bias.
But these things are tried, because everything has been tried before and discussed to death. So people try to be original and invent new ways of doing the same, claiming sonic improvements. But truth of the matter is the drawbacks of most of these circuits are not included in the presentation. And A-technical people read this and think these people are the circuit gods.
A zener string is nice. if you want to have an absolute value for G2 voltage for calulation purposes. But when it comes to reliability and simplicity a resistor divider may be better in some circumstances. But a resistor divider requires that you know how to do an thevenin equivalent circuit and know how to calculate this.
I'd say only about 10% of people on this forum even know how to do that, let alone read data sheet with maximum specifications. These people are all talk sometimes.
Another thing: Most people that run pentode or tetrode amps measure cathode current. But say G2 is fed from a different suppy. Anyone that has tested tubes on a tube tester can tell you that two tubes that measure 100mA cathode current can have for example 3 and 7mA of screen grid current. Then these people try to balance an output stage with a multimeter. and are very happy when both tubes measure exactly 48mA cathode current.
If you have bifilar wound output transformers, you can have the same resistance from the center tap of a UGT to both ends, in that case it makes more sense to measure and compare the anode voltages of a PP stage.
Keep your feet on the ground at all times.
Last edited:
Paul Joppa of Bottlehead fame wrote a really good paper on choosing operating points and loads in the 1998 Valve Magazine issues 3,4 & 5. that would be a very good foundation to build on.
Titled "Brainiac on load impedance and triode operating points." they are archived HERE free through the generosity of Doc. B. and Paul Joppa.
Awesome stuff! I’ve just started reading, but it looks very promising. I’m encouraged that he actually starts off with the “why” of choosing an operating point instead of “here’s what an operating point looks like on a load line...good luck!”