I do not agree that TMC is 2 pole.
TMC is 1 pole system. It is more like that one pole is nested in another pole. There are 2 poles, but they are not 2 pole system. The pole around output stage is nested in the Miller compensation, so it is 1 pole system.
You can nest 1 2 3.. even more, but it is still 1pole system.
It does cause some stable issue, but it is different from 2 pole system
Doesn't matter if you agree or not, it's still a second order system. Use the search function and read the posts about in the last year or so.
Including the output stage in the Miller loop does not lead to a second order system. Break the loop using the Miller theorem and you'll find out why. The output capacitive loading is very small (almost equal with the compensation cap) and combined with the low output impedance of the emitter/source follower creates a pole that is to far away to have any effect.
I think I understand what you are saying, undedamped second order system overshooting may drive the amp into the rails, so we need to LP filter the closed loop response at the input, to avoid that. But if driving the amp into nonlinearities is a source of instability (and unfortunately to often it is) then the design still has to be revisited. Chances are good it's going to be, sooner or later, a "tweeter eater".
Yes, I know you will see 2 pole like frequency response if you put a probe correctly.
The problem is what cause it?
I believe even you remove Miller Capacitor. You will still see that behavior. 180 degree phase dip, and rise up to 90 degree at high frequency. (Thanks to all software for letting us able to see frequency response without Miller Cap)
That's is caused by the R & C in series connected to input&output of EF stage.
The CURRENT gain of EF stage is like a 1-pole rolling down system(it is related to Ft of the transistor you use). The input impedance of EF stage will decrease as frequency climbing up. Let's assume it is 1 pole like rolling down on input impedance. That means, as the same amount current driven from VAS, the open loop voltage gain will 20dB/Oct rolling down.(V = RI)
If you put a capacitor between input/output of EF stage. The capacitor will suck more current from VAS as frequency climbing up, it will cause additional 20dB/Oct rolling down. So far, you get 40dB/Oct rolling down. What if you put a resistor in series with that capacitor. It creates a ZERO at high frequency. That's how you get that 2 pole like Frequency Response!!
The 2 pole like response is caused by RC network around output stage, if it is driven by current source. (low gain VAS with small Miller Cap is good example for current source.)
I must add, TMC, at low frequency, the miller feedback is mostly taken from output stage, so the output of VAS is pretty much a current source.
Last edited:
Slew, linearity, minimum phase etc
One definition of linearity is that big signals should behave exactly like small signals except bigger. If in your rise time is 100ns@20mV and also 100ns@zillionV, than your system is linear... at least for rise time and slew. If at 40Vp, the rise time is slower than 100ns, then slew limiting has occurred.
All real linear systems have a bandwidth and this directly gives rise to a 'rise time'. If you feed a square wave to this system, it will show an exponential rise/fall from each transition to the flat part.
The slew of this transition is maximum at the start of the transition and drops exponentially as it approaches the flat part. This maximum slew is of course bigger for bigger square waves. The signal with the highest possible slew that an amplifier may be asked to deliver is a rail to rail square wave.
An amplifier's slew is ALWAYS limited (see Self, Cordell et al) but if it is bigger than the slew for this rail to rail square wave, it is sufficient for the amplifier to remain linear. For this linear system to show greater slew than the maximum slew demanded by a rail to rail square wave, the input signal MUST overload the amplifier.
You can do this simply by turning the volume control fully up. Then you'll see the amplifier exhibit its maximum slew from one overload point to the other.
I'm assuming you don't have any source demanding rail to rail Sines at zillion MHz ... and you are happy with nice clean undistorted rail to rail 40kHz Sines.
The bandwidth of an amplifier MUST be limited to guarantee no slew limiting. The bandwidth & slew must be matched. If you have 300V/us on your 100W amp, you are allowed 600kHz. If your amp is less able, you can still avoid trouble by accepting somewhat less bandwidth. It makes sense to have a bandwidth less than this criteria so you have some slew in hand.
Of course if you are happy for your amp to slew limit, you can have whatever bandwidth your Marketing VP wants.
______________
Amps have loadsa opportunity for non-minimum phase behaviour cos they have one necessary (but not sufficient) feature. This is that there should be more than one route that the signal takes. Alas, this non-minimum phase behaviour always works against us.
Baxandall shows one example in the Baxandall Papers on Douglas Self's website.
I used 'small signal slew' in quotes cos it's really linear slew or linear rise time rather than something that only happens at low levels.
One definition of linearity is that big signals should behave exactly like small signals except bigger. If in your rise time is 100ns@20mV and also 100ns@zillionV, than your system is linear... at least for rise time and slew. If at 40Vp, the rise time is slower than 100ns, then slew limiting has occurred.
All real linear systems have a bandwidth and this directly gives rise to a 'rise time'. If you feed a square wave to this system, it will show an exponential rise/fall from each transition to the flat part.
The slew of this transition is maximum at the start of the transition and drops exponentially as it approaches the flat part. This maximum slew is of course bigger for bigger square waves. The signal with the highest possible slew that an amplifier may be asked to deliver is a rail to rail square wave.
An amplifier's slew is ALWAYS limited (see Self, Cordell et al) but if it is bigger than the slew for this rail to rail square wave, it is sufficient for the amplifier to remain linear. For this linear system to show greater slew than the maximum slew demanded by a rail to rail square wave, the input signal MUST overload the amplifier.
You can do this simply by turning the volume control fully up. Then you'll see the amplifier exhibit its maximum slew from one overload point to the other.
I'm assuming you don't have any source demanding rail to rail Sines at zillion MHz
... and you are happy with nice clean undistorted rail to rail 40kHz Sines.I'm only limiting Bob's 300V/us monster to 600kHz. I hope he doesn't mind.
The bandwidth of an amplifier MUST be limited to guarantee no slew limiting. The bandwidth & slew must be matched. If you have 300V/us on your 100W amp, you are allowed 600kHz. If your amp is less able, you can still avoid trouble by accepting somewhat less bandwidth. It makes sense to have a bandwidth less than this criteria so you have some slew in hand.
Of course if you are happy for your amp to slew limit, you can have whatever bandwidth your Marketing VP wants.
______________
Amps have loadsa opportunity for non-minimum phase behaviour cos they have one necessary (but not sufficient) feature. This is that there should be more than one route that the signal takes. Alas, this non-minimum phase behaviour always works against us.
Baxandall shows one example in the Baxandall Papers on Douglas Self's website.
Last edited:
I used 'small signal slew' in quotes cos it's really linear slew or linear rise time rather than something that only happens at low levels.
One definition of linearity is that big signals should behave exactly like small signals except bigger. If in your rise time is 100ns@20mV and also 100ns@zillionV, than your system is linear... at least for rise time and slew. If at 40Vp, the rise time is slower than 100ns, then slew limiting has occurred.
All real linear systems have a bandwidth and this directly gives rise to a 'rise time'. If you feed a square wave to this system, it will show an exponential rise/fall from each transition to the flat part.
The slew of this transition is maximum at the start of the transition and drops exponentially as it approaches the flat part. This maximum slew is of course bigger for bigger square waves. The signal with the highest possible slew that an amplifier may be asked to deliver is a rail to rail square wave.
An amplifier's slew is ALWAYS limited (see Self, Cordell et al) but if it is bigger than the slew for this rail to rail square wave, it is sufficient for the amplifier to remain linear. For this linear system to show greater slew than the maximum slew demanded by a rail to rail square wave, the input signal MUST overload the amplifier.
You can do this simply by turning the volume control fully up. Then you'll see the amplifier exhibit its maximum slew from one overload point to the other.
I'm assuming you don't have any source demanding rail to rail Sines at zillion MHz
I'm only limiting Bob's 300V/us monster to 600kHz. I hope he doesn't mind.
The bandwidth of an amplifier MUST be limited to guarantee no slew limiting. The bandwidth & slew must be matched. If you have 300V/us on your 100W amp, you are allowed 600kHz. If your amp is less able, you can still avoid trouble by accepting somewhat less bandwidth. It makes sense to have a bandwidth less than this criteria so you have some slew in hand.
Of course if you are happy for your amp to slew limit, you can have whatever bandwidth your Marketing VP wants.
______________
Amps have loadsa opportunity for non-minimum phase behaviour cos they have one necessary (but not sufficient) feature. This is that there should be more than one route that the signal takes. Alas, this non-minimum phase behaviour always works against us.
Baxandall shows one example in the Baxandall Papers on Douglas Self's website.
One of the things that sometimes slips us up is our somewhat sloppy use of the term "slew rate" - I often do it as a matter of convenience. Slew rate is not really a property of a signal. It is technically the maximum rate of change that a device (say an amplifier) can deliver at its output. This is often its maximum rate of voltage change, but it can also be maximum rate of current change (also important for amplifiers).
I seem to recall the term "slew rate" being associated with the maximum rate at which a gun turret could move, in degrees per second.
It is easier to say that a 1V peak sinewave at 20kHz has a slew rate of 0.125 v/us, but we should really be saying that it has a voltage rate of change of 0.125 V/us. I plead guilty to doing this at times.
One might also refer to the demanded slew rate of a signal, I suppose. For example, a 40V peak sinusoid at 20kHz, corresponding to 100W at 8 ohms, demands a slew rate of at least 5 V/us. By the same token, a 20kHz square wave first-order bandlimited to 20kHz will demand a slew rate of at least 10 V/us.
It is also important not to mix or confuse slew rate with risetime. Slew rate is always a large-signal term, while risetime is often (but not always) a small-signal term.
The input filter of a power amplifier is there at least to suppress EMI, but it has often been prescribed for preventing slew rate limiting or hard TIM. In the latter case the pole frequency of the filter (if it is first-order) might be lower than that needed for EMI purposes, and might be determined by the slew rate of the amplifier.
Another way in which I sometimes trespass on the semantics is to refer to the maximum slew rate of an amplifier. I'm not sure that makes sense. The amplifier's slew rate is the maximum rate of change of the output signal.
I hope I have not muddied the waters further
.Cheers,
Bob
Here you go again. First order or one pole is 20dB/decade or 6dB/oct, and second order or two pole is 40dB/decade or 12dB/oct.
MIC was attributed by J.L. Hood in his book, "Valve and transistor audio amplifiers", to Dr A.R. Bailey: Wireless World, May 1968, pgs 94-98. Perhaps Douglas Self can scan you a copy.
Oh dear - I am going blind
I for one would like to know those reasons.
Last edited:
I've already done the simulation as indicated here:
http://www.diyaudio.com/forums/soli...-self-wants-your-opinions-73.html#post3463332
See below:
I think "small signal slew rate" really doesn't exist despite the fact that you've previously shown references to the contrary. I think it is a misleading conflation of rise time and slew rate.
100% correct me old mate!
As I have pointed out numerous times "TMC" is merely the application of TPC about the second stage and the output stage while exposing the input stage to a single pole loop gain characteristic.to correctly design a "TMC" network, one needs to first design a stable TPC network using TPC algebraic expressions for the location of the coincedent poles and the zero, and then merely connect the resistor in the TPC network to the output.
This approach ensures that the minor loop inclosing the output stage with "TMC" is as stable as the global loop using TPC.
Last edited:
"TMC" is a single pole system from the perspective of the input stage, but it is a double pole single zero system fro the perspective of the second stage and the ouput stage. viz. it's simply double pole compensation applied around the second and output stages.
Page 263 of D.Self's APADH 5th edition will lead you to the answer, eventually. Read it four or five times and meditate upon the ideas. You will holler Eureka!I for one would like to know those reasons.
The voltage difference between VAS and output stage is negligible. It is unable to form another pole.
Here is what I am talking about at http://www.diyaudio.com/forums/solid-state/230492-audio-power-amplifier-design-book-douglas-self-wants-your-opinions-77.html#post3465453
2 pole is there, but it is not because of TMC network.
I simulated TMC without miller capacitor connected. Here we got 2 pole.
Attachments
Actually as was refering to the JL Hood design that has emitter degeneration for the LTP and then pervesely bypasses it with a large capacitor. Also mirror degeneration resistors are used and then bypassed by capacitors. The whole design is a mess.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.