Mr kamwar
"The low impedance output of amplifier acts as a short circuit to the back EMF generated , hence the speaker ringing is eliminated and much of the effect of BACK EMF is reduced hence what you get is tight speaker cone movement.."
very good
but what is the significance of overkill damping
for example
emf effect=z (say 8) / damping (say low 150)
wouldnt be quite adequate to the ear
as ive said the returns are diminishing, however you can persue
high damping figures for product overkill
cheers
"The low impedance output of amplifier acts as a short circuit to the back EMF generated , hence the speaker ringing is eliminated and much of the effect of BACK EMF is reduced hence what you get is tight speaker cone movement.."
very good
but what is the significance of overkill damping
for example
emf effect=z (say 8) / damping (say low 150)
wouldnt be quite adequate to the ear
as ive said the returns are diminishing, however you can persue
high damping figures for product overkill
cheers
Food for though...
Here we have a basic circuit that could be part of a headphone amp or whatever.
A VAS stage feeding a complementay EF pair. There is no feedback (besides the degenaration on the VAS emitter).
Q1 is biased at 2mA, Q2 and Q3 at 5mA.
The output impedance should be (25/5 + .5)/2 = 2.75 ohms.
Here we have a basic circuit that could be part of a headphone amp or whatever.
A VAS stage feeding a complementay EF pair. There is no feedback (besides the degenaration on the VAS emitter).
Q1 is biased at 2mA, Q2 and Q3 at 5mA.
The output impedance should be (25/5 + .5)/2 = 2.75 ohms.
Attachments
Greg
It's related to the intrinsec emmiter resistence at 25 deg C, 1/gm or whatever (I really don't remember
).
One can simplify it using 25/Ic at room temperature (this I remember...).
So, for 5mA emmiter current, re is 5 ohms, plus 0.5 ohms of external resistence, divided by 2 because both transistors conduct at the same time.
It's related to the intrinsec emmiter resistence at 25 deg C, 1/gm or whatever (I really don't remember
).One can simplify it using 25/Ic at room temperature (this I remember...).
So, for 5mA emmiter current, re is 5 ohms, plus 0.5 ohms of external resistence, divided by 2 because both transistors conduct at the same time.
Greg
What the output impedance shows IS source impedance/HFE!
Which is the output impedance of a degenerated CE stage feed from a low Z source (such as an emitter follower)?
If this same output pair was feed by a voltage source, output impedance would be quite low (of the magnitude I've calculated).
Maybe that's one of the reason's for the VAS to ground resistors in the Leach amp...
The reason for the previous posts is to show that, depending on the output stage topology, all the damping is obtained through feedback, and so there's a very high likehood of back EMF been feed to the input.
Now, if the output stage was a double or triple Darlington, output impedance would be imuch lower, and very little back EMF would reach the input.
What the output impedance shows IS source impedance/HFE!
Which is the output impedance of a degenerated CE stage feed from a low Z source (such as an emitter follower)?
If this same output pair was feed by a voltage source, output impedance would be quite low (of the magnitude I've calculated).
Maybe that's one of the reason's for the VAS to ground resistors in the Leach amp...
The reason for the previous posts is to show that, depending on the output stage topology, all the damping is obtained through feedback, and so there's a very high likehood of back EMF been feed to the input.
Now, if the output stage was a double or triple Darlington, output impedance would be imuch lower, and very little back EMF would reach the input.
Graham Maynard said:
Graham:
A basic tennet we accept is system linearity, impying we can completely characterize system performance undergoing complex stimuly by superposition of system performance under each stimuly harmonic components.
Of course this is not completely true, and we are concerned with linearity departures as they give rise to harmonic distortion and intermodulation.
Nonetheless, your line of reasoning does not seem to be related to this issue but with the simultaneous occurence of different signals in an acceptedly linear system, and this is precisely something of no concern (unless of course nonlinear phenomena are involved).
Of course it may look counterintuitive, but you must acknowledge the robustness of the approach insofar it has allowed us to build current technology.
This is at least what I interpret from your postings. Pleas clarify if this is not the case.
Rodolfo
I'm not EE here. The feedback theory is NOT wrong at all. The problem is, does the feedback system in audio power amp works 100% as theory requires?
If the differential works 100% fine, we won't have different sound when feedback is taken at output or at VAS or using SPLIF design or using non-feedback design, they will sound the same (with comparable topology). The feedback will be able to compensate for whatever happened in output node, and compare it with input, make scaled magnificaiton, and make 100% amplified signal of input like in theory without any distortions.
But why in reality they sounds different? Because feedback system is not working 100% like theory.
Differential at 1st stage of power amps consist of transistors, bipolars, mosfets, Jfets, etc.
All of them do not have straight amplifying curve. Look at all transistors datasheet, the relation of Ib-Ic or Vgs-Id, or whatever magnification curve NONE is forming smooth straight line. They are curved, exponential, etc.
So how do you expect to built a feedback system that works 100% with devices that doesn't have linear magnification?
If we have a transistor that has perfect linear magnification from 0 then we can implement feedback theory better.
Some here is using high-efficiency speakers, like Fostex, Lowther. They sure make less residual energy, because more energy is converted to sound. Less back-EMF with these speakers or not?
So Jorge,
Your example gives an OL Zout =4K due to hfe=200 and 200K source impedance (Zoutvas) from the Vas. Alright.
Now lets put a chip at the front and apply 94dB of NFB at 50Hz. bingo. DF =100, Zout =0.08 ohm AS SEEN BY THE LOAD. or.
You can load down the Vas with leech like R's to GND, let's say 2K total. So open loop Zout=10 ohms and if we apply 42dB NFB we get DF=100. SO THE LOAD SEES THE SAME Zout=0.08 ohm. But at the same time we have degraded the amplifier by wasting 40dB of Vas gain exchanging it for flat OL gain. Also we have increased by 100 times the modulation of the Vas and signal level of preceding stage(s) so increasing distortion in the OL amplifier but with 40dB less NFB to correct it.
The load sees a similar low Zout and a more nonlinear amp.
Greg
Your example gives an OL Zout =4K due to hfe=200 and 200K source impedance (Zoutvas) from the Vas. Alright.
Now lets put a chip at the front and apply 94dB of NFB at 50Hz. bingo. DF =100, Zout =0.08 ohm AS SEEN BY THE LOAD. or.
You can load down the Vas with leech like R's to GND, let's say 2K total. So open loop Zout=10 ohms and if we apply 42dB NFB we get DF=100. SO THE LOAD SEES THE SAME Zout=0.08 ohm. But at the same time we have degraded the amplifier by wasting 40dB of Vas gain exchanging it for flat OL gain. Also we have increased by 100 times the modulation of the Vas and signal level of preceding stage(s) so increasing distortion in the OL amplifier but with 40dB less NFB to correct it.
The load sees a similar low Zout and a more nonlinear amp.
Greg
Hi Rodolpho,
As if I do not understand basic tennents you are asking me to clarify on your terms of accepting superposition based investigation of linear circuitry whilst including one significant element, which as Luminauw has pointed out, is not *linear*.
The effects of multiple amplifier non-linearities cannot be real-time minimised due to stabilisation and device capacitances slowing the NFB correction that actually generates one of the 'circuit elements' we are obliged to include.
Hi Jorge,
Please allow me to put some meat on the bones, as per this typical NFB amplifier - often accepted as being a *linear* performer.
As attached.
As if I do not understand basic tennents you are asking me to clarify on your terms of accepting superposition based investigation of linear circuitry whilst including one significant element, which as Luminauw has pointed out, is not *linear*.
The effects of multiple amplifier non-linearities cannot be real-time minimised due to stabilisation and device capacitances slowing the NFB correction that actually generates one of the 'circuit elements' we are obliged to include.
Hi Jorge,
Please allow me to put some meat on the bones, as per this typical NFB amplifier - often accepted as being a *linear* performer.
As attached.
Hi Greg,
That is an excellent explanation of design choice in relation to Jorge's contributions.
The amplifier with the loaded down VAS is an exact opposite of the circuit I have just posted, and as you say it will have 'flat' open loop gain.
This means that the loaded down VAS should introduce significantly less internal propagation delay and less back-EMF induced quadrature output error. This also means that it will much better prevent leading current induced fractional reverse bias voltage commutation as can be seen on the error trace a few microseconds prior to zero signal voltage crossover.
True these errors arise at inaudible frequencies, but that does not mean that their effects are inaudible.
Cheers ........ Graham.
That is an excellent explanation of design choice in relation to Jorge's contributions.
The amplifier with the loaded down VAS is an exact opposite of the circuit I have just posted, and as you say it will have 'flat' open loop gain.
This means that the loaded down VAS should introduce significantly less internal propagation delay and less back-EMF induced quadrature output error. This also means that it will much better prevent leading current induced fractional reverse bias voltage commutation as can be seen on the error trace a few microseconds prior to zero signal voltage crossover.
True these errors arise at inaudible frequencies, but that does not mean that their effects are inaudible.
Cheers ........ Graham.
I suspect we don't need to worry too much about damping factor; the Zout of any amp with global feedback is sufficiently low to damp a speaker and we really only need about 50:1 for good control. The big problem is keeping phase shift at bay with reactive loads so that oscillation does not threaten the output stage.....
Cheers,
Hugh
Cheers,
Hugh
Graham Maynard said:
The effect of feedback is precisely to attenuate whatever nonlinearities a real amplifier built off nonlinear devices may present.
This also goes to answer Lumanaw post in the sense NFB cannot account for a 100% correction.
Of course this is true, there is no 100% correction, but the correction factos is precisely the OL to CL gain margin.
Puting in other words, if an amplifier features an open loop gain of 100 dB and is connected in a 20 db closed loop gain configuration, the efect of feedback is to devote 80dB to correction.
What this means roughly is that if the amplifier (without feedback) featured say a 10% deviation from linearity, after closing the loop the deviation drops to 0.00001%.
This holds no matter whether the deviations from linearity are internal to the amplifier chain, or whether output errors are induced by loading challenges, it strives to hold the amplifier output as close a copy of its input.
As said earlier, for this to be true it must hold within the working band with negligible delay (phase shift) and sufficient output drive capability.
This implies among other things, a large gain-bandwith product, i.e. to keep both gain margin and phase shift within design objectives up to the maximum working frequency.
Modern audio intended OpAmps feature over 10 MHz and beyond gain-bandwidth products to account precisely for this.
Rodolfo
Besides non-ideal curve, like Mr.Maynard states, transistors also "still" have internal capacitance and "finite" speed. It will take time (no matter how small it is) to react.
If we have light speed transistor, 0 capacitance anywhere, straight line from 0 magnification curve, you can built amp with 20 stages with feedback, and feels good
Hi, Rodolfo,
But what "input" does the differential sees? I have strong suspicion that the "input" (differential) is not seeing the input signal ONLY, the differential sees inputsignal+backEMF as it's "input". So, it is processing it. I'm talking about non-inverting differential, where the input is on left base, the feedback is to right base.
If we have light speed transistor, 0 capacitance anywhere, straight line from 0 magnification curve, you can built amp with 20 stages with feedback, and feels good
Hi, Rodolfo,
The approach can be this way also. Let's assume the feedback is 100% perfect in feedback audio power amp.
But what "input" does the differential sees? I have strong suspicion that the "input" (differential) is not seeing the input signal ONLY, the differential sees inputsignal+backEMF as it's "input". So, it is processing it. I'm talking about non-inverting differential, where the input is on left base, the feedback is to right base.
Greg
I remember seen a similar viewpoint as yours many yrs ago - Audio (at the time it was still a decent magazine...) maybe John Curl?
And it's right.
Now let's increment the outut pair to a triple darlington with HFE of 100, 50, 30 transistors.
Overall current gain about 150,000. Open loop output impedance of 200/150 = 1.33 ohms.
If one increases VAS Ic the OL output impedance will be about halved.
And there will be a reduction in VAS fractional current excursion and no significant gain reduction.
The final result is that any back EMF will be attenuated by over 20dB before reaching the feedback loop (for an 8 ohms speaker).
On the other hand, 3 junctions in series wil generate more distortion than a single one...
The big question that remains is are these changes audible? Where is the point of diminishing return?
I remember seen a similar viewpoint as yours many yrs ago - Audio (at the time it was still a decent magazine...) maybe John Curl?
And it's right.
Now let's increment the outut pair to a triple darlington with HFE of 100, 50, 30 transistors.
Overall current gain about 150,000. Open loop output impedance of 200/150 = 1.33 ohms.
If one increases VAS Ic the OL output impedance will be about halved.
And there will be a reduction in VAS fractional current excursion and no significant gain reduction.
The final result is that any back EMF will be attenuated by over 20dB before reaching the feedback loop (for an 8 ohms speaker).
On the other hand, 3 junctions in series wil generate more distortion than a single one...
The big question that remains is are these changes audible? Where is the point of diminishing return?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.