mikeks said:
Mikeks, unbelievable, you are even more stubborn than me... 😀
You suggested to check my "assumptions" in real world, rather than relying on sims. Okay, i did that. Gladly, some time ago i had replaced input devices by jfets, that way i had enough safety margin to skip the rloads. (2x47k to ground, of course i had to keep the caps to ground)
Measuring was done with 1w into 8ohms, 1khz. The whole setup was not moved, no wire changed, only one measuring with rloads, the other without.
Okay, attached first measuring with rloads.
Attachments
Second measuring with the rloads removed. (both pics are pixel aligned)
I must admit, i was surprised myself, the THD increased... This only shows that OL-distortions are even more reduced by rloads than the OL-gain is reduced. Chew on that ! (And this is only 1khz !)
What this does to sonics is another story, sure is, feedback has less work to do ! (maybe because of that lowered thd ?)
I'm glad that you drove me to prove that by real world measuring, now i am not convinced to be right, now i know it.
Mike
I must admit, i was surprised myself, the THD increased... This only shows that OL-distortions are even more reduced by rloads than the OL-gain is reduced. Chew on that ! (And this is only 1khz !)
What this does to sonics is another story, sure is, feedback has less work to do ! (maybe because of that lowered thd ?)
I'm glad that you drove me to prove that by real world measuring, now i am not convinced to be right, now i know it.
Mike
Attachments
Hi MikeB,
Nice! 🙂
Do you have measurements closed loop both with and without VAS Rload?
Cheers Michael
Nice! 🙂
Do you have measurements closed loop both with and without VAS Rload?
Cheers Michael
Hi Michael, these both measurings were closed loop, first with rloads, seconds without rloads. Or do you believe in distortion number below 0.002% for an poweramp operating openloop ? 😉
Mike
Mike
😎
I join in here.
To lower the gain of VAS ( or any stage ) by using RESISTOR across load
can be good.
I say can be good, because if the gain present isnt enough to get a good suitable feedback level
then such EXTRA LOADS are NOT GOOD.
We also call this arrangement LOCAL feedback.
And I think mikeks point is, that any local feedback
will lower the overall gain, so that we in the end do not have a good level of global FB.
But on the other hand:
If we have good local feedback, and so more linear stages inside amplifier
we may not need as much global feedback.
Why is good to use a RESISTOR
in parallel to a load, a transistor based output of a stage?
Answer:
Same reason as why we use resistors to assist in any way:
Resistors are superior to transistors, when comes to LINEARITY.
This is also why we, as MikeB mostly does 😀
should ALWAYS use totally linear ( zero inductance + zero capacitance )
Pure resistors 8 Ohm when simulating SPICE
or testing an amplifier we have built, in real world.
Because if we add:
- Loudspeaker Voice Coil, including crossover Coils + Capacitors
- Speaker cable capacitance + inductance, however little it may be
.... we will most certainly ruin a good perfect design
.... and nobody would want to come bear our creation
lineup
Lineup Audio Lab
.. currently
.. struggles with heavy oscillation in Multisim SPICE
.. while he is foolish as to put 2.2 uF across his 8 Ohm resistors
---------------------------------
* note:
According to Mr Leach, Professor in Electronics at a USA university:
.. like 2 uF across 8 Ohm LOAD RESISTOR is about the worst test
of an amplifier stability you can do
This is why we are doing it here at Lineup's.
We certainly do not do such tests because it is fun
to watch the result at the Multisim SPICE oscilloscope screen!
.. because it isnt ..
I join in here.
To lower the gain of VAS ( or any stage ) by using RESISTOR across load
can be good.
I say can be good, because if the gain present isnt enough to get a good suitable feedback level
then such EXTRA LOADS are NOT GOOD.
We also call this arrangement LOCAL feedback.
And I think mikeks point is, that any local feedback
will lower the overall gain, so that we in the end do not have a good level of global FB.
But on the other hand:
If we have good local feedback, and so more linear stages inside amplifier
we may not need as much global feedback.
Why is good to use a RESISTOR
in parallel to a load, a transistor based output of a stage?
Answer:
Same reason as why we use resistors to assist in any way:
Resistors are superior to transistors, when comes to LINEARITY.
This is also why we, as MikeB mostly does 😀
should ALWAYS use totally linear ( zero inductance + zero capacitance )
Pure resistors 8 Ohm when simulating SPICE
or testing an amplifier we have built, in real world.
Because if we add:
- Loudspeaker Voice Coil, including crossover Coils + Capacitors
- Speaker cable capacitance + inductance, however little it may be
.... we will most certainly ruin a good perfect design
.... and nobody would want to come bear our creation
lineup
Lineup Audio Lab
.. currently
.. struggles with heavy oscillation in Multisim SPICE
.. while he is foolish as to put 2.2 uF across his 8 Ohm resistors
---------------------------------
* note:
According to Mr Leach, Professor in Electronics at a USA university:
.. like 2 uF across 8 Ohm LOAD RESISTOR is about the worst test
of an amplifier stability you can do
This is why we are doing it here at Lineup's.
We certainly do not do such tests because it is fun to watch the result at the Multisim SPICE oscilloscope screen!
.. because it isnt ..
So there we have it - Rload on the VAS stage is a good thing provided overall loop gain is high enough to start with. Excellent thread. Now, can we have a simulation of Leolabs Lizi from someone? Thanks for enlightening us by the way MikeB - looks like the stuff the old guys (NP,JC et al) keep telling us true after all.
bonsai
bonsai
Bonsai said:
Hi,Bonsai!Thanks for interested in Lizi esp the VAS stage.
Well,before building Lizi,I was intended to run in her in simulation but I dont have the spice model of the JFETs and LATFETs,so i just simply build her.
Bonsai said:
I have spent the better part of this evening with hardware trying to replicate MikeB's findings and have consistently found the contrary position to be true.
Viz. If the second stage's output is provided with a sufficiently small shunt resistance to ground, so that the amplifier's foward-path half-power bandwidth is of the order of 20KHz, THD+N is repeatabily found to deteriorate.
This is consistent with the intuitively and scientifically sound notion that any stage in an amplifier cannot give of its best, in respect of linearity, if the load at its output is increased.
Other members (jcx perhaps?) might consider independently verifying this position.
Mike,
Have you gone the country mile and listened to the VAS loading paradigm?
All the theory, measurements and pontifications in the world won't tell you how it sounds......
If you find the sound is terrible, then all your points are vindicated and MikeB, myself, and some of the major luminaries in the field are all proven tossers.......
If you find the sound is somehow better, then you have quite a task ahead of you to rationalise this unfortunate development....
Here's to the worst part of your evening!!
Cheers,
Hugh
Have you gone the country mile and listened to the VAS loading paradigm?
All the theory, measurements and pontifications in the world won't tell you how it sounds......
If you find the sound is terrible, then all your points are vindicated and MikeB, myself, and some of the major luminaries in the field are all proven tossers.......
If you find the sound is somehow better, then you have quite a task ahead of you to rationalise this unfortunate development....
Here's to the worst part of your evening!!
Cheers,
Hugh
Hi Hugh,
I don't usually bother with uncontrolled listening tests; needless to say, i heard no difference that i couldn't attribute to experimenter expectancy; the fact that i was feeling somewhat famished at the time didn't help matters.
I don't usually bother with uncontrolled listening tests; needless to say, i heard no difference that i couldn't attribute to experimenter expectancy; the fact that i was feeling somewhat famished at the time didn't help matters.
Mike,
You raise a good point; the uncontrolled listening test is not terribly disciplined.....
BUT, it's the test done to death by the consumer, and taken over a period of a week or so, with the listener in a variety of moods and physical conditions, DOES give a good indication of whether you are on the right track or not.
I understand the engineering concern with rigour and objectivity, I really do, and I note the meter readings with as much gravitas as anyone else. However, there is also the subjective side, and the fact is that most consumers are neither willing nor able to go full house on objective measurements, preferring to read them in a test review. But as a manufacturer I am selling to these guys, so I'd better walk their territory now and then to find something of the background to all these ooh aah audiophile terms one can glean from the flowery excesses of certain high end magazines.
In truth I can now build a circuit, measure it, view the waveform on a CRO under a variety of tests, then listen to it - and know within a few minutes if I'm on the right track. But significantly the final listening test is always the final arbiter, no matter how good the distortion, FR, clip and stability tests look.
Incidentally, I find the triangle waveform highly predictive, particularly at 20KHz, as the linearity is readily apparent on a CRO. To me linearity is very important, where square wave response is of limited benefit, and THD below around 0.05% is almost meaningless. However, the spectrum of the distortion is very useful.
Cheers,
Hugh
You raise a good point; the uncontrolled listening test is not terribly disciplined.....
BUT, it's the test done to death by the consumer, and taken over a period of a week or so, with the listener in a variety of moods and physical conditions, DOES give a good indication of whether you are on the right track or not.
I understand the engineering concern with rigour and objectivity, I really do, and I note the meter readings with as much gravitas as anyone else. However, there is also the subjective side, and the fact is that most consumers are neither willing nor able to go full house on objective measurements, preferring to read them in a test review. But as a manufacturer I am selling to these guys, so I'd better walk their territory now and then to find something of the background to all these ooh aah audiophile terms one can glean from the flowery excesses of certain high end magazines.
In truth I can now build a circuit, measure it, view the waveform on a CRO under a variety of tests, then listen to it - and know within a few minutes if I'm on the right track. But significantly the final listening test is always the final arbiter, no matter how good the distortion, FR, clip and stability tests look.
Incidentally, I find the triangle waveform highly predictive, particularly at 20KHz, as the linearity is readily apparent on a CRO. To me linearity is very important, where square wave response is of limited benefit, and THD below around 0.05% is almost meaningless. However, the spectrum of the distortion is very useful.
Cheers,
Hugh
mikeks said:
Mikeks, it's okay that we have very contrary thoughts about this subject, but it's getting weird when we get different measurings. We live both on the same planet and have to follow the same physics.
Would you mind showing your measuring results and measuring setup ? That would say more than thousands words.
Again, these rloads cannot improve linearity of the vas itself, but reduce the problems introduced by the outputstage into the vas by changing vas output from nearly pure current source to nearly voltage source. These rloads can of course not really change the currents out of vas-collectors.
Mike
MikeB said:
Oh, it was a bit late last night, especially above made me think you made an open loop measuring without reflecting more over the figures..😀
But then, could you possibly make an open loop measurement with and without VAS Rload, shouldn't be impossible task?
Though I'm aware of your point measuring open loop, post #48.
Cheers Michael
MikeB, Last nite I tried loading the vas (simulation) on an experimental amplifier that I am working on. It is a high OL gain (130dB) design based on the standard Lin topology. Output is buffered mosfet. Clearly OL frequency response is limited, but there is still over 90 dB of gain at 20KHZ. Driving the circuit at 10KHz and 60% of max output, I find virtually no difference between the loaded and unloaded vas until the open loop response goes beyond 10KHz (so open loop gain drops at 10KHz). Then the lower order harmonics climb as expected.
Since I respect the work of both you and Mikeks, I am left with the following possible conclusions:
a.) Perhaps the sim cannot resolve the difference as all harmonics are below -120dB due to the large amount of feedback. Mikeb's results showed a fairly small but significant difference.
b.) Perhaps this is another of the situations that depend on other aspects of the circuit as well. That is it cannot be generalized.
C.) ???????
Finally I tried listening to the Simulation, but could not hear anything ..... with or without vas loading.
Bill
Since I respect the work of both you and Mikeks, I am left with the following possible conclusions:
a.) Perhaps the sim cannot resolve the difference as all harmonics are below -120dB due to the large amount of feedback. Mikeb's results showed a fairly small but significant difference.
b.) Perhaps this is another of the situations that depend on other aspects of the circuit as well. That is it cannot be generalized.
C.) ???????
Finally I tried listening to the Simulation, but could not hear anything ..... with or without vas loading.
Bill
wwood said:
Bill, did you find no difference closed loop ? If yes, you have the same phenomena that the rloads reduce the OL-distortions approx the same level as OL-gain is reduced...
mikeks said:
Hey, Mikeks, even andy_c thinks that these shunt resistors would help.
Yes, the input impedance of the outputstage drops, but becomes more linear.
Okay, teaching paralleling resistors.
Assume a dynamic input impedance of 10k to 20k. Now put a 10k in parallel. The result is a dynamic inputimpedance of 5k to 6.66k. Ratio dropped from 100% to 33.2%, factor 3. Input impedance dropped factor 2...3. Plain math.Depending on how to calculate that ratio, it drops from 50% to 25%.
Mike
Yes, the easiest way to explode openloop gain, add a positive feedback. 😉
This works good if you do not have already "too much" OLG.
Mike
This works good if you do not have already "too much" OLG.
Mike
- Status
- Not open for further replies.