The think to keep in mind is that the time it takes the electrical signal to travel around the loop is near instantaneous, save for the actual delay it takes the electrical signal to travel from A-B on the circuit board. A 90 deg phase shift at 1MHz between input and output does not mean a 250nS delay somewhere in the amplifier - It is erroneous to think of th ephase shift being caused by a 'delay'. I would imagine the real delay is a few femto or pico seconds (I don't have a calculator with me right now).
Anyone remeber the famous article by Martin Colloms (spelling?) about feedback amplifiers chasing their tails because of the feedback 'delay'? Misguided.
Anyone remeber the famous article by Martin Colloms (spelling?) about feedback amplifiers chasing their tails because of the feedback 'delay'? Misguided.
That's right. It is erroneous to think the phase shift is caused by a delay. I see no need to use the word "delay" in this context so best avoided as confusion is so easy.
Hifi magazines are full of technical errors and procedural errors. Getting things right is not their business; selling magazines is.
For signal transport, a nanosecond is somewhat less than a foot. (Tip of the hat to Grace Hopper!)
if the order or the delay transmission was 1 picoS , then
our amps would have rising times in the same order, and
bandwiths of hundreds ghz...
As much as the "phase velocity" notion is accepted, so is "phase delay". Though, this has to be understood in the right sense: phase velocity is defined by the distance divided by the phase delay. The big problem (and source of many confusions) is that the definition is ambiguous within an additive constant of 2*PI*N.
In a linear phase system, both group and phase delays are equal to the constant delay of the system, and the phase shift of the system increases linearly with frequency. The other way around, deviation of the group delay from a constant value indicates the degree of nonlinearity in the phase.
Both group and phase delay concepts can be related to a "signal delay" invoked by the tail chasing NFB promoters, but not in a obvious way.
Wrong question. You cannot juxtapose frequency domain and time domain concepts. For the relationship between the group delay and the signal delay, see e.g. http://www.radiolab.com.au/DesignFile/DN004.pdf
Bonai,
You are completely correct. I never suggested or stated that a phase shift created by a simple pole, like 90 degrees from Miller compensation, was anything like a constant delay. I also agree that the time-of-flight delays on a circuit board are of relatively miniscule significance at frequencies anywhere near the gain crossover.
However, if you simulate three poles at different frequencies, separated by some factor like an octave (e.g., 5 MHz, 10 MHz, 20 MHz) and look at the phase lag through this network as a function of frequency with a linear frequency axis, you will see that, depending on the separation factor you choose, it is a reasonable approximation to a straight line at frequencies well below the first pole (e.g. at frequencies in the vicinity of 1 MHz give or take an octave or so. That is all I was trying to point out.
Those who understand the concept of making delay circuits out of all-pass filters would understand this.
Cheers,
Bob
Hi Bonsai,
Understood.
BTW, time-of-flight delay is often closer to 1.5 ns/ft in usual diectrics in cable and PWBs. the 1 ns/ft is more applicable to free space.
Cheers,
Bob
I am not swapping these terms.
I am asking if one or the other or both or neither happens when a signal passes through a semiconductor.
Hi Andrew,
I suspect that the transistor necessarily introduces a minisule amount of pure delay, on the order of the transit time of the device, maybe a couple of ns. At 1 MHz, this might be a degree or so. Passing through many transistors, it might add up.
I think the greater contribution from most transistor stages will be from poles that they introduce, rather than pure delay due to transit time. This will depend a lot on the kind of amplifier stage.
Lets not forget that this thread is about phase margin in a feedback loop. To that extent, it really doesn't matter a whole lot whether the excess phase comes from poles or delay.
Cheers,
Bob
if the topic is audio power output BJT then this is very wrong -10s of nS is more likely - base minority carrier diffusion time is a 1st order determinant of ft
the ST 2n3055 spice model has 99 nS Tf
this is due to capacitance effects, mainly..
bjts are intrinsically very fast devices,
the drawback is that they are plagued by parasistic
capacitors that make the devices slow, unless increasing the
currents, to help charge these caps faster, yet, it has its limit
because of the devices internal dynamic resistances thatwill limit these
currents di/dt...
I don't consider myself a device expert, but I think your assertion may be more applicable to JFET and MOSFET devices, which do not depend on minority carriers. They are intrinsically fast and stray capacitances do govern their speed. I think there is a reason why they call it "transit time" in the BJT model.
Cheers,
Bob
Yes, the circuit you have shown is lead compensation in the feedback path. It is used to advantage by some to help cancel the phase lag effects of extra poles in the forward path of the amplifier. It can also mitigate the effect of the pole formed at the summing node between the feedback netowrk resistance and the input capacitance of the LTP. However, caution is required in using lead compensation in the faeedback path because it can eat into gain margin and also provide a better path for EMI ingress from the output of the amplifier back to the input stage. The use of a resistor in series with the lead capacitor, as you have shown, helps reduce this concern.
I also regularly employ a small resistor in series with Cdom in the forward path to help counteract the effects of parasitic poles. Once again, this should be used with caustion, as it can eat into gain margin.
Cheers,
Bob
Good question. In the linear region time delays are insignificant. In switching applications there are non-linear charging effects that mimic delays and these can be very significant.
Brian, would these 'significant effects' occur in linear amps that switch like a class B amp?
jd