Dan Lavry
Member of the Trade: Lavry Engineering
- Joined
- Dec 30, 2008
- Posts
- 150
- Likes
- 16
Components and circuits alter sound, and much of it is due to their built in non linear transfer function (a bend in the curve).
Tubes come with significant non linearity, and triodes are very different then penthodes. Semiconductors also have a non linear transfer, and again, bipolar transistors and field effect transistors (there are a few types) also have different transfer function.
A transfer function is analogous to a mirror. A linear transfer means a “regular mirror”, a flat surface. You look in and you see yourself un altered. A non linear curve is analogous to a curved mirror. There are many different possible ways to curve a mirror.
Triode, penthode, bipolar FET, JFET, MOS FETS all has different types of curves. They also have different amount of gain (amplification) capabilities. As a rule, tube circuits are a lot less linear and have a lot lower gain then transistors circuits. Also, transistors circuits can be easily cascaded (put in series) to end up with huge gain. It is not common practice or practical to build an amp put of many tubes in series, so the tube gain capability is more limited.
Let’s call the gain capability of a circuit “open loop gain”. Let’s call the actual amount of amplification we need from the circuit “close loop gain”. As a rule, a circuit has more open loop gain (gain capability) then the needed amplification (closed loop gain). One may need to amplify by say X10, but have a gain capability of X100. So in this example the open loop gain is 100 but the closed loop gain is 10.
Why bother with more gain then one needs? Can one just “throw it away”? You do not want to just “throw it away”.
Both tubes and transistors have too much non linearity. The mirror is curved too much and what you see (or hear) is just too distorted. Some of the extra gain capability is utilized not to amplify, but to linearize the transfer function or in our analogy, to make the mirror flatter. Designers use a concept called “negative feedback” to do so. The more access gain one has, the better one can linearize the transfer curve. BTW, negative feedback also makes the frequency response flatter as well as fixes other factors.
Given that tube circuit have less open loop gain then transistor circuits, there is less access amplification thus less negative feedback, leaving some non linearity (curvature). A good designer can control to some degree the amount of curvature, thus control how warm the sound is. Some people like a certain degree of “warm sound” but if you overdo it, well, too much salt in the soup is not good.
Semiconductor circuits tend offer a huge amount of available amplification (open loop gain) so the capability to linearize the transfer function is readily available. One can design a semiconductor circuit that will emulate a tube pretty closely, but as a rule, semiconductors tend to yield a more linear transfer function (there are exceptions). Also, one can linearize a tube circuit and make it “less warm” (that may require additional tube or other hardware).
Clearly there are other factors at play. One should not ignore the sonic impact of transformers, and as a rule, components alone do nothing. I can not stress enough that it is the proper use of components in electronic circuits that make a product. Good parts in a bad circuits or poor parts in a great circuit yield poor results. A component that is good for one circuit may be terrible for another circuit!
The above information is just a starting point for understanding some of the fundamental differences between tubes and semiconductors for audio.
Regards
Dan Lavry
Tubes come with significant non linearity, and triodes are very different then penthodes. Semiconductors also have a non linear transfer, and again, bipolar transistors and field effect transistors (there are a few types) also have different transfer function.
A transfer function is analogous to a mirror. A linear transfer means a “regular mirror”, a flat surface. You look in and you see yourself un altered. A non linear curve is analogous to a curved mirror. There are many different possible ways to curve a mirror.
Triode, penthode, bipolar FET, JFET, MOS FETS all has different types of curves. They also have different amount of gain (amplification) capabilities. As a rule, tube circuits are a lot less linear and have a lot lower gain then transistors circuits. Also, transistors circuits can be easily cascaded (put in series) to end up with huge gain. It is not common practice or practical to build an amp put of many tubes in series, so the tube gain capability is more limited.
Let’s call the gain capability of a circuit “open loop gain”. Let’s call the actual amount of amplification we need from the circuit “close loop gain”. As a rule, a circuit has more open loop gain (gain capability) then the needed amplification (closed loop gain). One may need to amplify by say X10, but have a gain capability of X100. So in this example the open loop gain is 100 but the closed loop gain is 10.
Why bother with more gain then one needs? Can one just “throw it away”? You do not want to just “throw it away”.
Both tubes and transistors have too much non linearity. The mirror is curved too much and what you see (or hear) is just too distorted. Some of the extra gain capability is utilized not to amplify, but to linearize the transfer function or in our analogy, to make the mirror flatter. Designers use a concept called “negative feedback” to do so. The more access gain one has, the better one can linearize the transfer curve. BTW, negative feedback also makes the frequency response flatter as well as fixes other factors.
Given that tube circuit have less open loop gain then transistor circuits, there is less access amplification thus less negative feedback, leaving some non linearity (curvature). A good designer can control to some degree the amount of curvature, thus control how warm the sound is. Some people like a certain degree of “warm sound” but if you overdo it, well, too much salt in the soup is not good.
Semiconductor circuits tend offer a huge amount of available amplification (open loop gain) so the capability to linearize the transfer function is readily available. One can design a semiconductor circuit that will emulate a tube pretty closely, but as a rule, semiconductors tend to yield a more linear transfer function (there are exceptions). Also, one can linearize a tube circuit and make it “less warm” (that may require additional tube or other hardware).
Clearly there are other factors at play. One should not ignore the sonic impact of transformers, and as a rule, components alone do nothing. I can not stress enough that it is the proper use of components in electronic circuits that make a product. Good parts in a bad circuits or poor parts in a great circuit yield poor results. A component that is good for one circuit may be terrible for another circuit!
The above information is just a starting point for understanding some of the fundamental differences between tubes and semiconductors for audio.
Regards
Dan Lavry
