input sensitivity -- speaker efficiency
No, obviously not! People usually just copy schematics from the "days of old, when tubes glowed bold", and do not think about the input sensitivity of the amps and preamps used at that time... on the other hand, today's sources have quite a high output (like 2V) -- and most tube power amp schematics require only 300-500mV for full output power!
I usually do address that issue in my amps -- if possible. For instance, if the amp would normally be a three stage affair, I tend to avoid the first "input" stage... that way, the input sensitivity is lowered to unusual values, like 2.5V RMS for full output. But, using an adequate source (both CD/DVD and decent phono stage) and an active preamp (usually, gain at least 20dB or 10x), it's no problem to drive the amp!
By ommitting one stage of the amp, you have quite a lot more to gain, than you could loose... both noise and distortion.
Of course, input sensitivity can be lowered by adoption of negative feedback -- but to tell you the truth, that is not my cup of tea!
Regards,
Aleksandar
No, obviously not! People usually just copy schematics from the "days of old, when tubes glowed bold", and do not think about the input sensitivity of the amps and preamps used at that time... on the other hand, today's sources have quite a high output (like 2V) -- and most tube power amp schematics require only 300-500mV for full output power!
I usually do address that issue in my amps -- if possible. For instance, if the amp would normally be a three stage affair, I tend to avoid the first "input" stage... that way, the input sensitivity is lowered to unusual values, like 2.5V RMS for full output. But, using an adequate source (both CD/DVD and decent phono stage) and an active preamp (usually, gain at least 20dB or 10x), it's no problem to drive the amp!
By ommitting one stage of the amp, you have quite a lot more to gain, than you could loose... both noise and distortion.
Of course, input sensitivity can be lowered by adoption of negative feedback -- but to tell you the truth, that is not my cup of tea!
Regards,
Aleksandar
Oh, I think my question is a bit unclear, sorry for that. However, the information you have given is so valuable, thank you for that, Alex.
I was talking about component sensitivities -that is represented by S with superscript and subscript in circuit theory-. How a variation in component values affect the performance of the amplifier -frequency response, gain etc-.. I havent seen such calculations..
MB
I was talking about component sensitivities -that is represented by S with superscript and subscript in circuit theory-. How a variation in component values affect the performance of the amplifier -frequency response, gain etc-.. I havent seen such calculations..
MB
Oh, tolerances in parts? 10 to 20% is fine. Stable designs can tolerate even more (and to the extreme, SS can tolerate 500% or more variation because such tolerance is absolutely required!).
So yeah, it depends on the circuit. Show me one and I can tell you if it's good or sucks (electrically).
Tim
So yeah, it depends on the circuit. Show me one and I can tell you if it's good or sucks (electrically).
Tim
Monte Carlo Analysis
If it is about how does the variation of VALUES affect the circuit, there is something called Monte Carlo Analysis. While preparing for simulation, you should input the expected variation of components (i.e. 5% with caps, 1% with resistors)... etc.
But, the results of such simulation are not directly applicable to everything: they might provide information on where does a RIAA equalization go astray, and why... but generally, there is not much sense in simulating.
On one hand, you must not forget that you are a DIY-er and you are making ONE very finely tuning piece of equipment: therefore you can choose from a teoretically immesurable number of pieces the best ones... and have the values exactly as you wanted them.
On the other hand, several intelligent and experienced folks have already said that when building tube amps, it is not that important to be impossibly precise (i.e. 0% tollerance in values). And I tend to agree. Reasonable tollerances are OK.
On the other hand, inadequate components will inevitably give bad results, no matter what is written on them, and what tollerance is supposed to be there...
Did I stray somewhere, or is that what you wanted to know?!
If it is about how does the variation of VALUES affect the circuit, there is something called Monte Carlo Analysis. While preparing for simulation, you should input the expected variation of components (i.e. 5% with caps, 1% with resistors)... etc.
But, the results of such simulation are not directly applicable to everything: they might provide information on where does a RIAA equalization go astray, and why... but generally, there is not much sense in simulating.
On one hand, you must not forget that you are a DIY-er and you are making ONE very finely tuning piece of equipment: therefore you can choose from a teoretically immesurable number of pieces the best ones... and have the values exactly as you wanted them.
On the other hand, several intelligent and experienced folks have already said that when building tube amps, it is not that important to be impossibly precise (i.e. 0% tollerance in values). And I tend to agree. Reasonable tollerances are OK.
On the other hand, inadequate components will inevitably give bad results, no matter what is written on them, and what tollerance is supposed to be there...
Did I stray somewhere, or is that what you wanted to know?!
Tim;
> So yeah, it depends on the circuit.
> Show me one and I can tell you if it's good or sucks (electrically).
That is what I wonder. Is there any calculations of sensitivity of any passive component to the input-output transfer response in an amplifier ?
Alex;
I know about Monte Carlo Analysis. Moreover, I got your point, to sum up you say practically we dont need to consider such analysis because:
1) We already have components with very tight tolerances and other parameters (tcr etc.)
2) Passive sensitivities are not important in Amplifier topologies.
Again, I wish I could see an analysis like that, not a computer simulation, but a regular circuit analysis. I know such calculations are important for active filters, but it may be impossible/stupid to do it in an amplifier or do we miss some point because of complex calculations ?
I think, as you said, being a DIY-er, I may consider to eliminate some theoritical statistics and find the best value of a component, however, this mustn't disregard engineering perspective. By the way, I am not an experienced engineer, just a junior one
MB
> So yeah, it depends on the circuit.
> Show me one and I can tell you if it's good or sucks (electrically).
That is what I wonder. Is there any calculations of sensitivity of any passive component to the input-output transfer response in an amplifier ?
Alex;
I know about Monte Carlo Analysis. Moreover, I got your point, to sum up you say practically we dont need to consider such analysis because:
1) We already have components with very tight tolerances and other parameters (tcr etc.)
2) Passive sensitivities are not important in Amplifier topologies.
Again, I wish I could see an analysis like that, not a computer simulation, but a regular circuit analysis. I know such calculations are important for active filters, but it may be impossible/stupid to do it in an amplifier or do we miss some point because of complex calculations ?
I think, as you said, being a DIY-er, I may consider to eliminate some theoritical statistics and find the best value of a component, however, this mustn't disregard engineering perspective. By the way, I am not an experienced engineer, just a junior one
MB
Monte Carlo and repercussions
Yes, you got it right, basically.
Not that the tollerance of components does not mean anything -- but aside the fact that you can (if building one tuned specimen) go to various lengths to assure the matching of components, tollerances are generally strict enough, and by far surpass the tollerances involved in active components. And, when I said i.e. RIAA equalization, I did not mean just active, but passive filters as well.
Furthermore, when it comes to calculating... well, doing a Monte Carlo analysis of some sort without a simulator would take you so much precious time, that it would not be worth it. That is why they invented sims and analysis of the sort -- to save time and money.
Sincerely, I think that if something has to be engineered well, the engineer should think about other issues, which might not be directly obvious, like making it "idiot proof". By that, I do not mean that users are idiots, of course!!! I just mean that an inexperienced person using it will not suffer from using it inadequately (i.e. a child touching the B+ or the tubes, for that matter) -- and all the other purely operating factors like what will happen if a tube does not start for any reason (i.e. damaged heater) etc.
Regards,
Aleksandar
Yes, you got it right, basically.
Not that the tollerance of components does not mean anything -- but aside the fact that you can (if building one tuned specimen) go to various lengths to assure the matching of components, tollerances are generally strict enough, and by far surpass the tollerances involved in active components. And, when I said i.e. RIAA equalization, I did not mean just active, but passive filters as well.
Furthermore, when it comes to calculating... well, doing a Monte Carlo analysis of some sort without a simulator would take you so much precious time, that it would not be worth it. That is why they invented sims and analysis of the sort -- to save time and money.
Sincerely, I think that if something has to be engineered well, the engineer should think about other issues, which might not be directly obvious, like making it "idiot proof". By that, I do not mean that users are idiots, of course!!! I just mean that an inexperienced person using it will not suffer from using it inadequately (i.e. a child touching the B+ or the tubes, for that matter) -- and all the other purely operating factors like what will happen if a tube does not start for any reason (i.e. damaged heater) etc.
Regards,
Aleksandar
Typically, you don't really need a full-on Monte Carlo analysis of a one-off circuit. Just do "corners," i.e., the effects of extreme values.
For example, let's say I'm designing a power amp with a single-ended voltage amplifier direct-coupled to a split-load inverter. There are two critical resistors in the first stage, plate load and cathode load. So, I calculate gain, source Z, and DC plate voltage for (max Rp, max Rk), (min Rp, min Rk), (max Rp, min Rk), and (min Rp, max Rk). These are the four "corners."
The split-load will presumably have matched plate and cathode resistors, so the next calculation is the effect of the worst-case variations of the first stage on the operating points of the split-load.
For example, let's say I'm designing a power amp with a single-ended voltage amplifier direct-coupled to a split-load inverter. There are two critical resistors in the first stage, plate load and cathode load. So, I calculate gain, source Z, and DC plate voltage for (max Rp, max Rk), (min Rp, min Rk), (max Rp, min Rk), and (min Rp, max Rk). These are the four "corners."
The split-load will presumably have matched plate and cathode resistors, so the next calculation is the effect of the worst-case variations of the first stage on the operating points of the split-load.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.