The time a signal takes to pass through an amplifier really does relate to the amplifiers bandwidth. In the days of analogue colour TV, the luminance amplifiers (wide bandwidth of up to 6MHz) had to include a small delay to slow (or bring back into line) the luminance signal so that it aligned with the colour or chrominance part which was handled separately by much lower bandwidth circuitry (well under 1Mhz bandwidth). If that wasn't done, the effect was that the luminance and chroma didn't overlay correctly on screen and the chroma was displaced to the right by quite a margin as it arrived at the CRT later. If my memory is correct the delay needed was around 600ns (UK) and in the early days was coil wound on a long non metallic former... perhaps of 2500 turns of thin wire.
So extrapolating that... if you had a wide band audio amp and a low bandwidth sub woofer channel... the two 'aint going to be in step
heat propagation
Hi ilimzn,
You misinterpreted my question. It was about the relevance regarding heat propagation, thus not EM waves (which, by the way, travel around a wire and not inside).
I'm asking this question because I encountered two opposing views on this matter:
(BTW, I think it's just phase lag)
Cheers, E.
edit:
Remember, it's a FB loop, where a dead time might be devastating.
Hi ilimzn,
You misinterpreted my question. It was about the relevance regarding heat propagation, thus not EM waves (which, by the way, travel around a wire and not inside).
I'm asking this question because I encountered two opposing views on this matter:
So, what is it, just a phase lag (similar as in a low pass RC filter) or a real time delay (similar as in a coax delay line) ?
(BTW, I think it's just phase lag)
Cheers, E.
edit:
It's not that off-topic, as we were (also) talking about temperature compensation and associated bias generators.
Remember, it's a FB loop, where a dead time might be devastating.
Last edited:
thank you Bob!
i'm uncertain about the implications. If the Ut condition can't be met in reality, why should one stick to it at all? Choosing 26mV seems to be an arbitrary decision. Why not choose 30mV, since the junction temperature is most probably higher than room temperature or choose 15-20mV to lower costs?
Are there any rules of thumb maybe tolerance bands in which the choice of that Re voltgage drop has minimal impact on performance?
Regards
I think the true delay is tiny, it is a kind of low pass filter.
But it acts like delay, as in my example of a frypan handle that takes minutes to heat up.
This is not a contradiction, Bode, as usual, seems to have been the first to mention that the effect of a number of poles is a lot like a delay.
Bob makes the same point in his book.
I have often wondered about the physical interpretation of that.
Any number of poles will still be minimum phase, should be invertible.
A delay obviously can't be reversed, it's qualitatively different.
Funny they can look so similar.
Ok! And the reference to Bob's book was not even forced.
Best wishes
David
Hi Juergen,
First of all, the 26mV number is not perfect in reality, but that is certainly no reason to give up on it. When not terribly far off, the THD vs bias current is not a suddenly-changing thing.
There are also countering effects. While a warm junction might argue for a greater Vre, the ohmic component of the power transistor dynamic emitter resistance argues for a smaller Vre, since that ohmic component is effectively part of RE. As noted earlier, base stopper resistance divided by the output transistor current gain also contributes to the ohmic component seen looking into the emitter.
My general feeling is to err on the side of slight over-bias, since the consequences of under-bias can be much much worse.
BJT output stages are so imperfect in so many ways....
Cheers,
Bob
Electrons are fundamentally indistinguishable. It is meaningless to talk about this or that particular electron.
Again, signal travels as EM wave. Only the steady state current flow when all variation dies down, then EM wave disappeared and the slow electron float takes over.
Don't believe the "think of it like.....", these are people trying to make it easier to understand, it's not reality. Read Field and Wave Electromagnetics by David K Cheng 2nd edition p430. It explains how current and electric field and charge as the consequence of boundary condition of EM field traveling through a transmission line structure ( a wire in air is a transmission line).
I studied EM for years, I yet to find a good easy explanation on current and voltage traveling along a wire. There is a huge difference between electronic theorem and physical law. A lot of the so called equivalent circuits in electronics is not a law. AND a quite a few laws FAILED under scrutiny. Kirchhoff law is one that failed and even Ohm's law failed in certain condition as explained by professor Levin of MIT.
Far as I know, there is no explanation of Maxwell's equations that I know of that explains EM wave and current and voltage. It just is. He came out with the 4 equations. Under classical physics, it has not been over turned yet. People invented vector calculus to explain the Maxwell's equations, NOT the other way around.
As you get into advanced physics, you'll find more and more that there is just no explanation, some genus just dreamed up an equation, then they proof their inspiration until nobody can disprove it, then it became a law. Just like Newton law of gravitation held up for hundreds of years until Quantum mechanic came out. One day, even Einstein law of Relativity might be over turn!!! You never know.
As for now, I think the only thing that can explain circuit and signal travel is EM wave, not electron motion.
The only example that even kind of make sense is you put a cork in the ocean, as the wave travel with some velocity, the cork just bobble up and down instead of traveling with the wave. The electron is so slow it is actually bobble with the wave.
Don't believe the "think of it like.....", these are people trying to make it easier to understand, it's not reality. Read Field and Wave Electromagnetics by David K Cheng 2nd edition p430. It explains how current and electric field and charge as the consequence of boundary condition of EM field traveling through a transmission line structure ( a wire in air is a transmission line).
I studied EM for years, I yet to find a good easy explanation on current and voltage traveling along a wire. There is a huge difference between electronic theorem and physical law. A lot of the so called equivalent circuits in electronics is not a law. AND a quite a few laws FAILED under scrutiny. Kirchhoff law is one that failed and even Ohm's law failed in certain condition as explained by professor Levin of MIT.
Far as I know, there is no explanation of Maxwell's equations that I know of that explains EM wave and current and voltage. It just is. He came out with the 4 equations. Under classical physics, it has not been over turned yet. People invented vector calculus to explain the Maxwell's equations, NOT the other way around.
As you get into advanced physics, you'll find more and more that there is just no explanation, some genus just dreamed up an equation, then they proof their inspiration until nobody can disprove it, then it became a law. Just like Newton law of gravitation held up for hundreds of years until Quantum mechanic came out. One day, even Einstein law of Relativity might be over turn!!! You never know.
As for now, I think the only thing that can explain circuit and signal travel is EM wave, not electron motion.
The only example that even kind of make sense is you put a cork in the ocean, as the wave travel with some velocity, the cork just bobble up and down instead of traveling with the wave. The electron is so slow it is actually bobble with the wave.
These are all good points. In fact, if one staggers the poles of many first-order LPFs in series, one can achieve a nearly linear-phase low-pass filter. Sort of like a Bessel.
Also, modest delays used to be built from a cascade of all-pass filters, which of course were not minimum phase, since they gave delay without amplitude variation.
Cheers,
Bob
Yes, physicists are a very pragmatic lot, because they understand that all known physical theories are effective,
not fundamental. The assumptions that give the right results are the theory, at least until better ones are found.
Many incarnations ago I had a job where equalizing the propagation delays of RF signals had a significant effect on overall system performance. To help the technicians who did the final trimming of cable lengths, I referred to the "portable nanosecond". This was simply the span of the tech's hand - measured from the thumb tip to the tip of the long finger, with the fingers spread as far as comfortable. (The distance from thumb tip to the little finger tip is just slightly longer.) For most adults this is about 8 to 9 inches - and a good approximation of the distance traveled by an EM wave in 1 nanosecond, in a cable.
So when the test set told the tech that a signal was so-many nanoseconds too late, he could step off that many hand spans (plus a visual estimate of a fractional span, down to a quarter nanosecond or so) on the channel's cabling, cut the cable and attach a new connector. They found this easier than doing the arithmetic to convert nanoseconds to inches, then trying to manipulate a ruler or tape measure inside a cramped chassis. Since math and measurement errors happened too often, the portable nanosecond brought system performance into spec on the first cut at least as often as more sophisticated methods.
Dale
The speed through the cable depends on the permittivity of the inner insulation. Not all cable have the same speed. Speed U=1/sqrt{permittivity X permeability}. Calculation is very accurate. But if you use hand length and leave some spare to trim, then yes, this is a good way. We did something like that also. Just give some slack and read the delay, then trim back.
Yes, it's not a surprise that Non Minimum Phase (NMP) elements can simulate a delay.
My point was that it's a bit odd that Minimum Phase elements combine to simulate a delay.
Each MP element is reversible, the combination in cascade should also be reversible, yet the effect is very similar to an irreversible delay.
I subsequently realized that it's an example of a common effect in physics, where individual components are time reversible but the end result is not.
The classic example is fluid dynamics, where the equations of particle motion are time symmetric, reversible, yet gases only disperse and never spontaneously "unmix".
Physicists consider it a deep mystery so no surprise that I don't quite see how it works in electronics.
Best wishes
David
Agreed! But as Bob mentioned, the great majority of common coax cables are roughly 0.7 (or someplace between 2/3 and 3/4, if you think better in fractions). And, it varies with temperature and manufacturing lot.
I believe some foamed Teflon cables, produced for specialized instrumentation applications, get to almost 0.9. And some RG-types designed in the 1940's or 50's were around 0.5. (And a few cable types from that era, specifically intended for use as delay lines before we learned how to do it with clever analog circuitry, much less digitized signals, had even slower velocity factors.) But unless you identify a particular cable by model and manufacturer, the 0.7 value is probably as good an estimate as any.
So, what is it, just a phase lag (similar as in a low pass RC filter) or a real time delay (similar as in a coax delay line) ?
(BTW, I think it's just phase lag)
Cheers, E.
Edmond, I always though it is pure phase lag but I now doubt that. See attached. This a a graph of current into a thermal trak device (thus representing dissipation) and the threshold voltage of the tt diode. It looks very much like there is a delay between the dissipation and the reaction of the diode threshold.
What ye think?
Jan
Hi Jan,
Indeed, after the power has been turned off, the diode response clearly shows a delay. But I'm not sure if it's a result of a dead-time.
As David remarked:
"I think the true delay is tiny, it is a kind of low pass filter. But it acts like delay, as in my example of a frypan handle that takes minutes to heat up. This is not a contradiction, Bode, as usual, seems to have been the first to mention that the effect of a number of poles is a lot like a delay."
it also might be due to the concatenation of several poles. Furthermore, the graph suggests that the TT device is mounted on a heat-sink with considerable heat capacity, which adds another pole and complicates the interpretation of measuring results.
So, in order to make a final conclusion, we need to know more details about the test setup.
Cheers, E.
PS: I've googled several articles on heat conduction, but till now, no one mentioned a dead-time or a term like exp(-sT).
Indeed, after the power has been turned off, the diode response clearly shows a delay. But I'm not sure if it's a result of a dead-time.
As David remarked:
"I think the true delay is tiny, it is a kind of low pass filter. But it acts like delay, as in my example of a frypan handle that takes minutes to heat up. This is not a contradiction, Bode, as usual, seems to have been the first to mention that the effect of a number of poles is a lot like a delay."
it also might be due to the concatenation of several poles. Furthermore, the graph suggests that the TT device is mounted on a heat-sink with considerable heat capacity, which adds another pole and complicates the interpretation of measuring results.
So, in order to make a final conclusion, we need to know more details about the test setup.
Cheers, E.
PS: I've googled several articles on heat conduction, but till now, no one mentioned a dead-time or a term like exp(-sT).
Edmond, the mechanical setup was as follows. Two TT devices on a common heatsink, one receivng dissipation pulses as shown, the other not. The not-driven device TT diode threshold was subtracted from the threshold voltage of the driven device TT diode, to obtain delta-threshold as a function of dissipation.
Jan
Jan
Yes, and the more poles the closer the approximation to a true delay.
Any realistic thermal system is practically continuous, effectively has very many poles indeed.
I am not sure how to calculate this properly. See PS*
So, not infinite dimensional, like a true delay, but practically so.
Too small to measure. 10^6 m/second electron Fermi velocity, so sub microsecond for a reasonable heat-sink.
So no delay to mention, but it sure looks like one.
Best wishes
David
*The mean free path for electrons in Aluminium is about 15nm at room temperature.
Assume each little cell reaches a temperature and then heats the next cell.
So for a 15mm thickness probably in the order of 10^6 poles, perhaps more.
Last edited:
This kind of thing has been done in large computers as well, Cray for example.
Hi Mr. Cordell
I need to change out the pre-driver of my 3EF folded pre-driver OPS. I need to replace the KSC3503/KSA1381 with 2SC4793 and 2SA1930 to get lower Vbe. But the Cob increased from 2.6pF of the KS to about 26pF of the 2S transistors. But I don't think it would be a problem of loading the VAS.
I want to run it by you on my conclusion that it's ok to change the transistors. I am referencing to page 196 and page 437 of your book:
input capacitance C\pi = Cje+ diffusion capacitance where
difusion capacitance = gm/(2 X \pi X fT)
If I use 26mA collector current, gm=1, so diffusion capacitance =1/((2 X \pi X fT). If fT=100MHz, capacitance is 1600pF. So the diffusion capacitance is so much bigger than the Cob that if really does not matter whether it's 2.6pF or 26pF.....Hell, even 260pF!!!
For higher current and lower fT drivers and output transistors, the Cob is irrelevant as the diffusion capacitance is so so much higher than a few hundred pF of the big transistors.
I just want to confirm with you on this point. I don't need to choose transistor for EF base on Cob.
Thanks
I need to change out the pre-driver of my 3EF folded pre-driver OPS. I need to replace the KSC3503/KSA1381 with 2SC4793 and 2SA1930 to get lower Vbe. But the Cob increased from 2.6pF of the KS to about 26pF of the 2S transistors. But I don't think it would be a problem of loading the VAS.
I want to run it by you on my conclusion that it's ok to change the transistors. I am referencing to page 196 and page 437 of your book:
input capacitance C\pi = Cje+ diffusion capacitance where
difusion capacitance = gm/(2 X \pi X fT)
If I use 26mA collector current, gm=1, so diffusion capacitance =1/((2 X \pi X fT). If fT=100MHz, capacitance is 1600pF. So the diffusion capacitance is so much bigger than the Cob that if really does not matter whether it's 2.6pF or 26pF.....Hell, even 260pF!!!
For higher current and lower fT drivers and output transistors, the Cob is irrelevant as the diffusion capacitance is so so much higher than a few hundred pF of the big transistors.
I just want to confirm with you on this point. I don't need to choose transistor for EF base on Cob.
Thanks
I assume you are talking about Cbe?
Cbe is bootstrapped by the emitter. It only sees the voltage swing of Vbe which is very small, so only conducts a tiny current except at high frequencies. So even if Cbe is very large, Cob is still often more important because it experiences the full collector voltage swing, which can be 100s or 1000s of times larger then Vbe.
Cbe is bootstrapped by the emitter. It only sees the voltage swing of Vbe which is very small, so only conducts a tiny current except at high frequencies. So even if Cbe is very large, Cob is still often more important because it experiences the full collector voltage swing, which can be 100s or 1000s of times larger then Vbe.