MikeB said:
Resistively loading down the second stage output does increase foward path 3dB bandwidth by decreasing gain at lower frequencies.
The mistake you're clearly making is to think that this increase in bandwidth is equivalent to an increase in foward path (open-loop) gain at HF.
This is completely untrue, as the the overall open-loop gain at HF remains virtually unchanged, while that at LF is dramatically reduced to give the increase in 3dB bandwidth.
Smokes and mirrors springs to mind: there is no such thing as a free lunch.
lumanauw said:
It's a class AB as I am running at more than 100mA.
Currently I am partnering her with KEF iQ1.
Music-Very detail and well balance sound.
Movie-Still very detail and able to play loud but not irritating.You can distinguish between the sound effect and background music easily.
bogdan_borko said:
Believe it or not.I never run her through simulation,I simply build her base on some simple calculation.Regarding the spice model of the LATFETs,you can search on this forum.
Mikeks, yes the gain bandwidth product remains nearly unchanged. But the -3db point for the openloop bandwidth gets much increased. In many cases it gets extended from far below 1khz to above 20khz.
About the decreasing openloop distortions, it works.
I guess we are agreed that a vas itself is a nearly pure current source ? With perfect vas devices, output impedance is infinite. (by cascoding them you get very close to that, reaching gain >120db)
The outputstage only currentamplifies the voltage created in vas.
But how is the voltage defined from a current source with infinite output impedance ? By its load...
Now imagine the following (much simplified, exaggerated) scenario:
The load represented from the outputstage giving a variable input impedance of let's say 50k to 100k. The vas supplies +/- 1ma currentswing.
Now the voltage output from the vas changes between 50v and 100v, giving a distortion level of 100%.
Now apply a parallel resistive load of 50k. This time the load to the vas changes between 25k and 33k, giving outputvoltage 25v to 33v, distortion level = 32%.
Of course the load from the outputstage is not really resistive and is in fact a variable current load. Because of this, increasing vas current has a similar effect.
About the "terrible dynamic load from the outputstage"...
Even with perfect transistors, the load to the vas would be dynamic, because of the load from the speaker. (not much for mosfets)
Now, all transistors have an unlinear changing B-C capacitance and varying hfe with Vce. Especially at higher freqs, the varying B-C cap creates a lot of problems to the vas.
Play with the attached circuit, it represents a perfect vas with an outputstage.
Rloading decreases gain factor ~10, at the same time distortions drop factor ~8 (1% to 0.13%), phasehift drops from 70° below 10°, bandwidth ~doubled. Remember, this is a perfectly linear dummy load...
Total THD might slightly increase, but at the benefit of increased openloop bandwidth, decreased phasehift, reduced high order harmonics with feedbackloop closed, and less sensible to dynamic speaker load. (output level was ~10v peak, 7v rms)
With a real vas, the effect of outputcapacitance of the vas devices will also be reduced.
In this circuit, as current input of 0.1ma gives 10v output (at 10khz), the virtual input impedance of the outputstage is ~100k. Loading it with 11k, reduces gain ~1:10.
Mike
Leo, sorry for the threadjack...
About the decreasing openloop distortions, it works.
I guess we are agreed that a vas itself is a nearly pure current source ? With perfect vas devices, output impedance is infinite. (by cascoding them you get very close to that, reaching gain >120db)
The outputstage only currentamplifies the voltage created in vas.
But how is the voltage defined from a current source with infinite output impedance ? By its load...
Now imagine the following (much simplified, exaggerated) scenario:
The load represented from the outputstage giving a variable input impedance of let's say 50k to 100k. The vas supplies +/- 1ma currentswing.
Now the voltage output from the vas changes between 50v and 100v, giving a distortion level of 100%.
Now apply a parallel resistive load of 50k. This time the load to the vas changes between 25k and 33k, giving outputvoltage 25v to 33v, distortion level = 32%.
Of course the load from the outputstage is not really resistive and is in fact a variable current load. Because of this, increasing vas current has a similar effect.
About the "terrible dynamic load from the outputstage"...
Even with perfect transistors, the load to the vas would be dynamic, because of the load from the speaker. (not much for mosfets)
Now, all transistors have an unlinear changing B-C capacitance and varying hfe with Vce. Especially at higher freqs, the varying B-C cap creates a lot of problems to the vas.
Play with the attached circuit, it represents a perfect vas with an outputstage.
Rloading decreases gain factor ~10, at the same time distortions drop factor ~8 (1% to 0.13%), phasehift drops from 70° below 10°, bandwidth ~doubled. Remember, this is a perfectly linear dummy load...
Total THD might slightly increase, but at the benefit of increased openloop bandwidth, decreased phasehift, reduced high order harmonics with feedbackloop closed, and less sensible to dynamic speaker load. (output level was ~10v peak, 7v rms)
With a real vas, the effect of outputcapacitance of the vas devices will also be reduced.
In this circuit, as current input of 0.1ma gives 10v output (at 10khz), the virtual input impedance of the outputstage is ~100k. Loading it with 11k, reduces gain ~1:10.
Mike
Leo, sorry for the threadjack...
Attachments
And the most debated R8 and R10 issue,why dont you guys build one and see the results??Keep in mind that the debated issue here are just base on theory and simulation.
Leolabs said:
I was able to actually listen the difference, it made an amp from sounding like broken glass to sweet sounding.
Of course there is always the possibility that my explanation is not the real cause, and that lowering the loop gain simply increased stability. But in this case these rloads become even less "nonsense". Degeneration resistors will never reduce distortions at these levels.
Leo, stability should not be any problem in your circuit...
Mike
Michael,
I am guessing you are swamping the highly variable load of the output stage with a stiff resistive load?
With the very high Zout of a VAS collector, this should result in far less feedback correction of a wildly varying output caused by changing loading as the waveform moves through the 'dead' zone.....
Hugh
I am guessing you are swamping the highly variable load of the output stage with a stiff resistive load?
With the very high Zout of a VAS collector, this should result in far less feedback correction of a wildly varying output caused by changing loading as the waveform moves through the 'dead' zone.....
Hugh
AKSA said:
Hugh, you are much better with words than me !
(English is not my native language)You perfectly got my point. And as FB has less to do, it creates less high order harmonics. (steeper decaying with frequency)
In my amps these resistors have extra large effect as i tend to use low current vas.
Mike
AKSA said:
Hi AKSA,
I am sorry you cannot swamp the effect of a variable impedance by
shunting it with a fixed resistance.
This would only be true if the the later were were at least 10 times larger than, and in series with the varying impedance, and not in parallel with it.
Whether the the output impedance of your output stage is a valid design consideration is, of course, dependent on the type of stage adopted.
AKSA said:
Incidentally, this would only be true if the second stage were an ideal current source (hardly: see D. Self) and the output stage an ideal transimpedance amplifier (not even close).
mikeks said:
Think relative, not absolute.
Yes, you can't stop the current variations flowing into the outputstage. But by adding a linear value to a unlinear one you reduce the ratio between unlinear <-> linear. That simple.
Example: (The H1 coming from vas, the H2 from output stage)
H1: 0.1ma H2: 0.05ma -> Ratio 2:1
H1: 1.0ma H2: 0.05ma -> Ratio 20:1
Try my circuit in sims. You will see how openloop distortion drops with the rloads. If you don't like the virtual vas, replace it with bjts. It will keep to be a current source, just a less perfect one.
Anyway, if you just want to reproduce a 1khz sinewave into a resistor with perfectly low thd number, skip these rloads.
But typically an amp is supposed to deliver music into a speaker.
Mike
MikeB said:
Mike,
You are not swamping the non-linear impedance by adding a shunt resistance: to do so would require that the resistor be in series with and not in parallel with the output stage's input impedance.
Remember the voltage across a parallel combination remains unchanged, while the current draw is increased?
This is elementary circuit theory.
MikeB said:
Mike, i have tried this with real circuits, and, invariably, THD at LF, including PSRR, is severely compromised, while that at HF is virtually unchanged.
You need to look at that reversed ! That's the trick here. Here, the voltage gets defined by the current draw. As the resistance drops, the voltage created by the current draw from the outputstage drops, at the same time the current output from the vas is increased to compensate for the lower gain.
I think we don't need to teach us here ohms law and calculating paralelling resistors.
Mike
I think we don't need to teach us here ohms law and calculating paralelling resistors.
Mike
mikeks said:
I think that the issue here ,is that Mikeks take the output of the VAS as a low impedance point so the premise in the quote apply , and MikeB see the same node as a high impedance node , so loading this point with a resistor to ground will help to swamp the output stage input impedance unlinearitys , reflected by the speaker load...
A different point of view... 🙂
mikeks said:
Yes, of course, that's right. severely compromised ? Factor 2 ?
You must not look at THD as a single number, you need to look at the spectrum. Also, these numbers are typically done with a resistive dummy load.
Don't forget, i was speaking about HF all the time, and don't you think the fact is interesting that the HF-thd barely changed ? And, its about openloop distortion. Closed loop distortion does increase with these rloads. (as a single thd number into resistive load)
If life would be that easy, all low thd amps should sound great.
I am convinced that low thd at low freqs is useless if amp starts to generate plenty of harmonics at higher freqs. For a subwoofer amp i would not use these rloads.
Sadly, i can't test this with my actual amp, taking out the rloads would make it oscillate. (heavily increasing openloop gain)
Tube_Dude, yes i see the vas as a current source (very high impedance), why else should adding these resistors heavily drop gain ?
Mike
MikeB said:
Given a constant output voltage, the current drawn by the input impedance of the output stage will never never change merely because you've shunted said impedance with an increasingly small resistor.
Indeed, the current demanded of your second stage for the same voltage MUST increase to accomodate the demands of the shunt resistor.
This MUST increase the second stage's non-linearity.
MikeB said:
It would appear, sadly, that this is exactly what is required.
- Status
- Not open for further replies.