I see that many amp designs employ multiple gain stages. What is the inherent result of trying to obtain too much gain from a single stage?
How much SIGNAL input can say, a 12ax7 tube tolerate. I am sure there is a spec or calculation, I just haven't run across it. Are the OTHER advantages to multiple gain stages?
How much SIGNAL input can say, a 12ax7 tube tolerate. I am sure there is a spec or calculation, I just haven't run across it. Are the OTHER advantages to multiple gain stages?
Valves are limited in the voltage gain they can give. For a triode the upper limit is mu: a number typically in the range 15 to 100. A pentode can go higher bit it has a very high output impedance so that can limit gain due to loading from the next stage.
The grid signal input is limited by two bounds: valve cutoff (grid too negative) and grid current (grid insufficiently negative). Valve cutoff can be seen from the graphs in the datasheet. It could be a few volts negative for a 12AX7 or a few tens of voltage negative for a 12AU7 - in either case it depends on the anode/plate voltage. The start of grid current is usually not specified in the datasheet as it varies with age and from sample to sample, but it is often somewhere in the region of -1.5V to -0.5V.
Low gain valves generally give good results in simple circuits without too much thought. High gain valves can create distortion unless carefully used (this is why people often complain that the 12AX7 is non-linear - it isn't but it does need the right circuit).
If you need a voltage gain of 100 it may require less careful design to sling together two stages with gains of 10 each, than a single stage with gain 100.
The grid signal input is limited by two bounds: valve cutoff (grid too negative) and grid current (grid insufficiently negative). Valve cutoff can be seen from the graphs in the datasheet. It could be a few volts negative for a 12AX7 or a few tens of voltage negative for a 12AU7 - in either case it depends on the anode/plate voltage. The start of grid current is usually not specified in the datasheet as it varies with age and from sample to sample, but it is often somewhere in the region of -1.5V to -0.5V.
Low gain valves generally give good results in simple circuits without too much thought. High gain valves can create distortion unless carefully used (this is why people often complain that the 12AX7 is non-linear - it isn't but it does need the right circuit).
If you need a voltage gain of 100 it may require less careful design to sling together two stages with gains of 10 each, than a single stage with gain 100.
However there are exceptions (i.e. more linear high mu valves) like the 6SF5GT and the 6SQ7GT. The drawback is they are single triodes and expecially the latter is getting really expensive.....
If employed with plate CSS they are incredibly linear just like the best DHT's..😉..and one can get really high mu (with between 75 and 90 depending on the application) and plenty of output signal.
There are Russian and Chinese variants of the 6SQ7 (metal type) but I don't know how good they are in comparison to the original American.
Yes, as I said, you need careful design to get the best out of linear high mu valves like the 12AX7.45 said:
A rough approximation for the maximum peak-to-peak input headroom is:
Supply voltage / mu.
For a perfect triode this equation would be exact, but for practical triodes you may have to reduce this expectation somewhat if you want to avoid the very non-linear parts of the transfer characteristic. But it's a good rule of thumb to keep in your head, all the same.
Yes. You are unlikely to do better than this, but with careful design you might not do much worse. The snag is that grid current can be a problem with higher mu valves, as their grid is nearer the cathode space charge, so the forbidden region becomes a greater proportion of the 'supply/mu' estimate.
Re: Cascading stages
Thanks all. Merlinb's answer is very helpful as it gives me a rough 'baseline' to work from. I am not about designing the ultimate amp, but about learning general principles and guidelines.
A question to which the answer might be obvious.....but if I exceed the supply voltage/mu equation, does it lead to clipping?
DF 96 looks like good info in your answer as well. It will take a little longer to digest. Thanks for every answer guys. I couldn't get through this learning curve without ya.
Thanks all. Merlinb's answer is very helpful as it gives me a rough 'baseline' to work from. I am not about designing the ultimate amp, but about learning general principles and guidelines.
A question to which the answer might be obvious.....but if I exceed the supply voltage/mu equation, does it lead to clipping?
DF 96 looks like good info in your answer as well. It will take a little longer to digest. Thanks for every answer guys. I couldn't get through this learning curve without ya.
Last edited:
As a general rule unused electrodes should be connected to the cathode. If two cathodes either the relevant one, or the most negative one.
More or less, yes. It will certainly put you very close to clipping, although exactly how close depends on the chosen bias and how linear the valve is (an unbypassed cathode will also increase the input headroom, but that is yet another variable). If you bias dead-centre then you get the most headroom, although you still have to contend with grid current at one end and 'bunching' at the other. Because valves clip rather softly it can be tricky to define exactly where clipping begins. Easiest thing is to draw a load line and see what it tells you.
Last edited:
Good deal, thanks. I do realize that clipping, distortion and other termsd are relative and rather subjective............just asking in the general sense. I am understanding load lines much better, thanks.......
Actually, you can get substantially more gain out of anything; distortion starts to bite into performance, however.
By applying positive feedback, gain can be bootstrapped to infinity if you want. However, once it "passes" infinity, the transfer curve opens up into a hysteresis loop (it becomes bistable for part of the range).
Typically, a non-ideal amplifier's transfer curve has the highest gain in the middle (where it's biased), and lower gain to either side. This tends to compress and distort large signals, which is normal. The difference may not be much over the intended operating range (a typical stage might produce an output voltage only a few percent from ideal, when producing a signal level of half the maximum output), but just as negative feedback reduces gain, and errors in gain, positive feedback enhances gain and its error. A few percent error in the transfer curve might grow to a whopping 50% error (i.e., gain drops by half) when the gain has been increased by, say, a factor of 20 (perhaps from 100x to 2000x), and obviously, the difference is infinite when gain is pushed to infinity (which will only happen in the middle, where gain, and therefore positive feedback, are highest). If you require that your signal is still amplified with less than a couple percent error, its amplitude must be reduced by the same factor, so that instead of a 50% voltage swing, you might get only 2.5% out of the above example.
Is positive feedback therefore bad? Absolutely not. When small signal amplification is desired, a 200V swing is not at all required, or even desirable! It would actually be useful to apply positive feedback in a phono amp, where the signals are small, and therefore the amount of distortion will be small. (Distortion will be higher than the same stage without PFB, but when the difference is 0.1 versus 0.01%, no one will know; indeed, this is desirable as it produces more of Our Beloved Tube Sound.)
You also have to pay more attention to dynamic behavior, since gain and oscillation go hand in hand. Excessive positive feedback may yield a statically bistable circuit, or it may oscillate (guaranteed if the positive feedback is strong enough to cause hysteresis and is AC coupled).
Now, the gain-bandwidth of a given stage remains fairly constant, so that a triode that's capable of, say, 90x gain at 20kHz (dropping to, say, 70x at 40kHz) could be boosted to a gain of 900, but it will only do so up to a frequency of around 2-4kHz. (Again, the gain error is amplified, so that essentially nothing comes through at frequencies where, normally, only moderate attenuation would be seen.)
Bandwidth is the main motivation for cascading stages.
Cascaded stages can deliver unlimited gain-bandwidth, but the phase shift or delay through the chain may be intolerable for other reasons (e.g., an external error must be corrected, such as using the amplifier to control the voltage across a speaker), or the signal-to-noise ratio (SNR) becomes too terrible (if the noise factor of a given stage is greater than its gain, it's only making things worse!). For very large gains, shielding is also required (more an issue in radio receivers than audio amplifiers, but the exact same situation applies): if you have 100dB of gain through a chain of stages, but only 90dB of shielding (which isn't bad to begin with!), you will get an oscillator.
Tim
By applying positive feedback, gain can be bootstrapped to infinity if you want. However, once it "passes" infinity, the transfer curve opens up into a hysteresis loop (it becomes bistable for part of the range).
Typically, a non-ideal amplifier's transfer curve has the highest gain in the middle (where it's biased), and lower gain to either side. This tends to compress and distort large signals, which is normal. The difference may not be much over the intended operating range (a typical stage might produce an output voltage only a few percent from ideal, when producing a signal level of half the maximum output), but just as negative feedback reduces gain, and errors in gain, positive feedback enhances gain and its error. A few percent error in the transfer curve might grow to a whopping 50% error (i.e., gain drops by half) when the gain has been increased by, say, a factor of 20 (perhaps from 100x to 2000x), and obviously, the difference is infinite when gain is pushed to infinity (which will only happen in the middle, where gain, and therefore positive feedback, are highest). If you require that your signal is still amplified with less than a couple percent error, its amplitude must be reduced by the same factor, so that instead of a 50% voltage swing, you might get only 2.5% out of the above example.
Is positive feedback therefore bad? Absolutely not. When small signal amplification is desired, a 200V swing is not at all required, or even desirable! It would actually be useful to apply positive feedback in a phono amp, where the signals are small, and therefore the amount of distortion will be small. (Distortion will be higher than the same stage without PFB, but when the difference is 0.1 versus 0.01%, no one will know; indeed, this is desirable as it produces more of Our Beloved Tube Sound.)
You also have to pay more attention to dynamic behavior, since gain and oscillation go hand in hand. Excessive positive feedback may yield a statically bistable circuit, or it may oscillate (guaranteed if the positive feedback is strong enough to cause hysteresis and is AC coupled).
Now, the gain-bandwidth of a given stage remains fairly constant, so that a triode that's capable of, say, 90x gain at 20kHz (dropping to, say, 70x at 40kHz) could be boosted to a gain of 900, but it will only do so up to a frequency of around 2-4kHz. (Again, the gain error is amplified, so that essentially nothing comes through at frequencies where, normally, only moderate attenuation would be seen.)
Bandwidth is the main motivation for cascading stages.
Cascaded stages can deliver unlimited gain-bandwidth, but the phase shift or delay through the chain may be intolerable for other reasons (e.g., an external error must be corrected, such as using the amplifier to control the voltage across a speaker), or the signal-to-noise ratio (SNR) becomes too terrible (if the noise factor of a given stage is greater than its gain, it's only making things worse!). For very large gains, shielding is also required (more an issue in radio receivers than audio amplifiers, but the exact same situation applies): if you have 100dB of gain through a chain of stages, but only 90dB of shielding (which isn't bad to begin with!), you will get an oscillator.
Tim
re gain
I am beginning to realize there are few absolutes, but a host of trade offs in design. Very fascinating study. Your in depth answer will take some digestion, but on the surface it makes sense.
I am beginning to realize there are few absolutes, but a host of trade offs in design. Very fascinating study. Your in depth answer will take some digestion, but on the surface it makes sense.
I find distortion undesirable. I don't use valves because I like distortion, but because I like valves.Sch3mat1c said:
Positive feedback is rarely a good idea, unless you want to make an oscillator or are using careful bootstrapping.
one quick way is to look at the resistance coupled charts at the back of say an RCA tube manual, tables there will give you an idea of gains using predetermined values of resistors per tube type....http://www.tubebooks.org/tubedata/RC17.pdf
- Status
- Not open for further replies.