Yep. We agree on that.
That defies the laws of physics. If the speaker kicks back EMF, the amplifier has to change its output current accordingly. Unless your amp has a superconductive output impedance at room temperature. That would be impressive.
So you get a different amplifier if you connect a different speaker. Connect a speaker with a few strong resonances (such as just about any ported speaker) and you have a stability nightmare.
Sure. The MyRef FE has ~10x the output impedance of the MOD86. Both are still low.
Why the wobble in the output impedance at ~4 kHz then?
Tom
Last edited:
George,
Is the output impedance plot for the My_Ref FE the Evo version?
BTW this discussion has been very enlightening.
It is good to see the lowly LM3886 get such SOTA results.
One trivial question for Tom, knowing the LM3886 is not unity gain stable, what is the openloop gain? If protected IP, no problem.
I know from hard experience the My_Ref is not stable openloop. Had a poor solder connection (my fault) in the feedback loop. Opened up after playing fine for months. Full rail voltage on output. My loudspeaker barked like a 12 ga shotgun before the relay opened.
My wife amp thought one of my mad scientist projects had exploded.
Is the output impedance plot for the My_Ref FE the Evo version?
BTW this discussion has been very enlightening.
It is good to see the lowly LM3886 get such SOTA results.
One trivial question for Tom, knowing the LM3886 is not unity gain stable, what is the openloop gain? If protected IP, no problem.
I know from hard experience the My_Ref is not stable openloop. Had a poor solder connection (my fault) in the feedback loop. Opened up after playing fine for months. Full rail voltage on output. My loudspeaker barked like a 12 ga shotgun before the relay opened.
My wife amp thought one of my mad scientist projects had exploded.
Tom,
Don't think so. First of all: the EMF signal is a voltage, potential developed in the magnetic field change (while crossing the moving coil)
No current, if there is no closure of the loop.
So again: a current source could not be less interested, if any voltage appears (forced) on it's output. That is the definition of high impedance output..
So: the high output impedance of the current source do not close the 'magnetic' circuit of the coil. No current is flowing.
This all before feedback applied.
This all as in contrast to an EmitterFollower output stage: it has low impedance by itself, soaks up current if a voltage is forced onto it.
Ciao, George
Don't think so. First of all: the EMF signal is a voltage, potential developed in the magnetic field change (while crossing the moving coil)
No current, if there is no closure of the loop.
So again: a current source could not be less interested, if any voltage appears (forced) on it's output. That is the definition of high impedance output..
So: the high output impedance of the current source do not close the 'magnetic' circuit of the coil. No current is flowing.
This all before feedback applied.
This all as in contrast to an EmitterFollower output stage: it has low impedance by itself, soaks up current if a voltage is forced onto it.
Ciao, George
What does it mean in practice? It means that ALL of the reaction from the speaker movement get's 'registered' by the feedback loop: it will be added to the error signal, in full quantity.
In an EF based output stage (Mod86) the error signal develops ~the following way: the EMF appears on the output. Depending on the EF stage output impedance, some current starts flowing. Some voltage drop is developing as well, on that not infinitely small impedance. This extra voltage (only a distorted part of the original speaker return EMF) is getting then added to the loop feedback error signal, too.
In an EF based output stage (Mod86) the error signal develops ~the following way: the EMF appears on the output. Depending on the EF stage output impedance, some current starts flowing. Some voltage drop is developing as well, on that not infinitely small impedance. This extra voltage (only a distorted part of the original speaker return EMF) is getting then added to the loop feedback error signal, too.
I have to correct George here...
Mauro gave me his permission to proceed with my My_Ref variation which I (and others) developped indipendently from the My_Evo on the premise to change the original name (RevO) to Fremen Edition so to clearly state that it was my variation and not an official revision.
The My_Ref FE improvements are inspired by My_Evo ones but have been indipendently developped or omitted where original (like the LM318 regulator or the DC servo) or borrowed when previous art (like the limiter) or already made public (like the Lm3886 grounding or the My_Evo Rev A compensation).
Mauro never released to the public the My_Evo complete schematic and the FE should not be considered officially released or approved by Mauro.
Last edited:
Absolutely.
This is valid for the My_Ref in general and also for the Fremen Edition which would have not been possibile without help from several people I greet on each group buy initial post:
Thanks again to all the great My_Ref lovers community
About the output impedance measurement.
I am quite puzzled. When I see something like that, I like to have a bit more access to the setup, and look for more hints..
It should not be like this.. At least not foreseen by simulation. There are no time constants in the design, which could go 'haywire' right in the middle of the audio range..
I have done a quick impedance scan with my hp3577A..
It's my FE modul, not with the LM318 but OPA828. The measurement is ~ at the limit of my setup - a bit above.
The reference line (0db) is 25ohm. The reading is
~ -55dB below reference = 44mohm.
I think it is close to Tom's result, but without any quirks?
Ciao, George
I am quite puzzled. When I see something like that, I like to have a bit more access to the setup, and look for more hints..
It should not be like this.. At least not foreseen by simulation. There are no time constants in the design, which could go 'haywire' right in the middle of the audio range..
I have done a quick impedance scan with my hp3577A..
It's my FE modul, not with the LM318 but OPA828. The measurement is ~ at the limit of my setup - a bit above.
The reference line (0db) is 25ohm. The reading is
~ -55dB below reference = 44mohm.
I think it is close to Tom's result, but without any quirks?
Ciao, George
Attachments
As long as the amp is connected to the speaker and is causing the speaker cone to move, there will be a loop. Unless you're suggesting that the loop is one-directional.
I know how a current source works. Thank you for the lecture, Professor.
So let me see if I can wrap my heard around this.
The MyRef FE has a reasonably low output impedance of ~40 mΩ as documented in the plots you linked to above. As result, it will act as a voltage source when it is driving the speaker. Do we agree this far?
It sounds like you're then saying that when the speaker acts as a voltage source and applies the back-EMF voltage to the output of the amp, the amplifier acts like a current source. Is this correct?
So basically if a voltage is applied to the output of the amp it should 'see' a near-infinite impedance and practically zero current flowing into the amp.
And the combination of these two features is what makes the MyRef and similar amps unique and special.
Am I understanding this correctly?
Tom
Hard to tell without Y-axis labels.
But... It's not the same amp. Swapping the LM318 for an OPA828 will change the amp. Right? So the output impedance will change. After all it's set by a combination of the output impedance of the output stage and the feedback applied to it.
I also don't understand why you bring up the OPA828. On one hand you say the MyRef FE is special because it has the LM318 in it, and that in order to enjoy the MyRef FE you have to leave it untouched. On the other hand you're saying that if you change things up you get better performance. Which is something we do agree on, but then you no longer have MyRef FE.
I have a good handful of amps that are better than the Modulus-86 and I could certainly claim that any of them is better than the MyRef FE based on any objective measure, but that would be besides the point of this thread.
Tom
Tom,
We agree in that the complete amp, with feedback loop closrd, is a low output impedance voltage source.
The applied feedback will take care of it.
Now: let us disconnect the global feedback loop and forget about it. Apply a 0V signal directly to the output stage section. Fixed 0V.
Drive the output with the impedance test signal.
What impedance will You register:
With the MyrefFE output stage?
With the Mod86 output stage?
We agree in that the complete amp, with feedback loop closrd, is a low output impedance voltage source.
The applied feedback will take care of it.
Now: let us disconnect the global feedback loop and forget about it. Apply a 0V signal directly to the output stage section. Fixed 0V.
Drive the output with the impedance test signal.
What impedance will You register:
With the MyrefFE output stage?
With the Mod86 output stage?
Following this thread, my impression is that the "why" of using a LM318 has been made clear.
To me Tom, it seems you're starting to play games
Do we then also agree that the completed amps with the feedback applied will act similarly?
For example, we approach the two amps, feedback loops closed, as black boxes. We conclude, based on measurements, that they act as voltage sources. If this is the case, would it then be logical to conclude that the black boxes would react in similar ways to whatever external signal we apply to the output of the box? This would also mean that both boxes would have the same response to back-EMF. Right?
If you argue that the boxes would behave differently when presented with an external stimulus, wouldn't that imply that the black boxes somehow had memory or knowledge of the external world? I'm curious how you would theorize how audio amplifiers would somehow acquire this knowledge and where their memory systems reside.
In that case, the emitter follower will have lower impedance. Probably a factor (1+beta) lower than the output impedance of a common emitter/source output stage. We agree on that too.
But, as Panelhead reported in Post #202, bad things happen when the global feedback loop opens. In fact, the MyRef FE (or Modulus-86 for that matter) cannot be operated with the feedback loop open. So how is that even relevant here?
It seems like you've found another rabbit hole of academic interest. Too bad the outside world doesn't exist within that one isolated rabbit hole.
Tom
No. All I asked for was labels. You could have added them in Photoshop (or whatever you use) pretty easily. But I guess that's too much to ask.
I do. I also recognized the screen shot. I use the same software.
Tom