 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: Re: MOSFETs work fine too... posted by Robert Karl Stonjek on December 12, 2001 at 23:31:19:
Sorry, busy day today and this thing's become something of a monster so I didn't have time to get a reply banged out. I'll set aside the time tomorrow.se
Follow Ups:
Understood. Half the disagreements on newsgroups and message boards start when one or both parties do not have enough time to outline their case clearly enough.Reiterating:- the debate began on the subject of feedback and whether or not it is really necessary. I pointed out that MOSFET amps need feedback to maintain a constant output with frequency when the load changes with frequency.
Someone pointed out that there are several varieties of MOSFETs and not all perform in the manner I outlined. I specified Hitachi MOSFETs in later discussions.
You said that feedback was not necessary for MOSFETs either but later said that it is necessary to correct for "inherent limitations" of some designs, and also maintained that there are other (non-feedback) methods of achieving the same thing.
I also mentioned that MOSFET amps can correct for the errant voltage output of a driver, such as might be generated by an overshooting speaker.
I did perform an additional experiment. I build a subwoofer that uses a compound woofer - basically the compound woofer is two woofers glued together face to face.
So I have some of these lying around. In the experiment I powered one of the woofers. The other was passively driven, simulating an overshoot condition.
The input voltage was 6VRMS at 20Hz. The voltage generated by the passively driven woofer was 2.3V with no load.
I then connected various resistors to act as loads/damping and checked the voltage output of the passively driven woofer:-
No Load.2.3V
10r.....1.2V
1r......220mV
0.033R...12mV
Amp.......9mV (passive woofer connected to power amp; amp input shorted)
Short.....9mVNote that the amp gave the same result as the short. The error bars are two big to say more than that. Accuracy of around ±2mV and 1.9mV measured noise.
Also, the leads I was using were just regular hookup cables with alligator clips on the ends. To short the woofer, I connected them together, but when a voltage still appeared, I tried shortening the lead and found the voltage dropped (as expected). But this does say something about cables and damping - the two leads were only 300mm in length.
Notably, the amplifier’s feedback circuit showed a 0.5V error voltage passed to the gates of the MOSFETs. This seems to show that the amp was ‘pushing against the voltage’ produced by the speaker.
If I had a couple days spare to set up the test properly I might find something interesting, but is it really worth it? My original point was that feedback is required to keep some (ie Hitatchi) MOSFET amps form varying output voltage with load. You say there is another approach to achieve the same thing.
Anyway, if a person was really serious about sound they would dispense with the passive crossover and drive each speaker with a separate power amplifier. I think there is something to be said for specialist subbass, bass-mid and treble amps, though that is one further step I don’t think I’ll bother taking!! :)
Kind Regards,
Robert Karl Stonjek.
Understood. Half the disagreements on newsgroups and message boards start when one or both parties do not have enough time to outline their case clearly enough.Yes. And becuase one or both parties do not have enough time to actually read what's actually being said by the other party.
Reiterating:- the debate began on the subject of feedback and whether or not it is really necessary.
It did? Granted, my memory's not what it used to be, but I could have sworn that it began on the subject of a loudspeaker's impedance dropping during transients. Oh well. Let's move on.
I pointed out that MOSFET amps need feedback to maintain a constant output with frequency when the load changes with frequency.
Yes. And as I beleive I pointed out, such behavior is the practical consequence of an amplifier's non-zero output impedance. And as I also believe I pointed out, an amplifier's output imepedance can be lowered verus what it would be otherwise by the application of feedback.
Someone pointed out that there are several varieties of MOSFETs and not all perform in the manner I outlined. I specified Hitachi MOSFETs in later discussions.
Any device with a non-zero output impedance will behave in that manner to one degree or another. And since every device that I'm aware of has a non-zero output impedance, it really boils down to a matter of degree and what degree the designer feels is acceptable.
You said that feedback was not necessary for MOSFETs either but later said that it is necessary to correct for "inherent limitations" of some designs, and also maintained that there are other (non-feedback) methods of achieving the same thing.
First, I didn't say it wasn't necessary. Whether it's necessary is ultimately a decision for the designer to make. What I did was point out that, in the context of controlling the motion of a loudspeaker cone, particularly at its fundamental resonant point, the amplifier's output impedance doesn't have quite the substantial effect that many people believe. That even a thousand fold difference in output impedances results in a very marginal difference in the amplifier's control over the loudspeaker.
Second, I didn't say that feedback was necessary to correct for inherent limitations. I simply said that that's what the feedback is doing in this particular context. If a designer feels it's necessary to apply feedback in order to reduce the amplifier's output impedance, then obviously the amplifier's inherent output impedance (i.e. without feedback) is a limitation. I made no mention of its being inherently necessary as again that's a call for the designer to make. Other designers might be willing to live with the inherent limitation.
Thirdly, I believe I said that you can design an amplifier with low output impedance without having to resort to feedback, but I didn't intend this to apply to your specific topology/parts compliment.
I also mentioned that MOSFET amps can correct for the errant voltage output of a driver, such as might be generated by an overshooting speaker.
And I mentioned that it does this by way of its output impedance. Doesn't matter if it's a MOSFET amp, a BJT amp, a tube amp, or what have you. With feedback or without feedback. Any "errant voltage" applied across the amplifier's output will see the amplifier's output impedance and behave accordingly.
I think now might be a good time to address this whole "overshooting speaker" thing as I think you may be looking at this issue from some erroneous assumptions.
For example your assumption that any time the speaker is moving in one direction, then before it can move in the opposite direction, its momentum must first be overcome, which you say cannot occur instantaneously.
While it's true that any mass which is moving (in this context in response to an applied force) has momentum seeing as momentum is mass times velocity, not every mass which is moving in response to an applied force GAINS momentum. And it's only momentum GAINED which would need to be overcome before you could change its direction.
Take a hockey puck for example. Set it down on a smooth sheet of ice and with your finger apply a little force in one direction for a certain distance and then stop applying the force. The hockey puck will continue to move a bit beyond the point at which you stopped applying the force. That distance represents the momentum that was GAINED by the hockey puck. And if you were to apply a force in the opposite direction at the same moment you stopped applying the initial force, you would first have to overcome this gained momentum before you could get it moving in the opposite direction.
Now do the same thing only this time with the hockey puck sitting on a sheet of rough sandpaper. This time the hockey puck doesn't move past the point at which you stopped applying the force. That's because the much higher frictional losses of the sandpaper versus the ice prevented the hockey puck from gaining momentum.
Loudspeakers of course have losses as well, in all three domains; electrical, mechanical and acoustic. With respect to raw drivers, the relationships between losses and reactances are described in the various Q parameters, Q ES , Q MS and Q TS . What they describe is the damping effect the losses have on the reactive elements.
Q ES describes the damping of the electrical portion of the driver, Q MS of the mechanical portion, and Q TS is the product/sum of the two others, i.e. (Q ES x Q MS ) / (Q ES + Q MS ).
Anyway, the point here is that these figures (well, Q TS really) are useful in determining a driver's propensity to overshoot. A Q TS of 0.707 will have a maximally flat response with no overshoot. Above this is considered "underdamped" and you will begin to get overshoot as the driver's internal losses are overcome by the reactive elements. Below this is considered "overdamped" and internal losses dominate.
Q TS for typical drivers ranges from around 0.2 to 0.6.
If it's not obvious yet, another way of putting this is that for any Q TS at or below 0.0707, the driver's internal losses are such that they prevent the cone from GAINING momentum. And if the cone doesn't gain any momentum, then when your transient comes along that wants to move the cone in the opposite direction, then as far as the transient is concerned, it sees nothing different than if the cone were otherwise at rest. It's like the hockey puck on the sandpaper.
Of course raw drivers are typically used in enclosures which in combination with the driver's Q TS will result in a different system Q (Q TC , which will always be at least as high or higher than the driver's Q TS ). And depending on the designer, one may decide on a Q TC greater than 0.707.
Anyway, this is just a side note to address any notion that simply because the cone is moving it must have gained momentum.
And now back to our regularly scheduled program...
I did perform an additional experiment. I build a subwoofer that uses a compound woofer - basically the compound woofer is two woofers glued together face to face.
So I have some of these lying around. In the experiment I powered one of the woofers. The other was passively driven, simulating an overshoot condition.
The input voltage was 6VRMS at 20Hz. The voltage generated by the passively driven woofer was 2.3V with no load.
I then connected various resistors to act as loads/damping and checked the voltage output of the passively driven woofer:-
No Load.2.3V
10r.....1.2V
1r......220mV
0.033R...12mV
Amp.......9mV (passive woofer connected to power amp; amp input shorted)
Short.....9mVNote that the amp gave the same result as the short. The error bars are two big to say more than that. Accuracy of around ±2mV and 1.9mV measured noise.
Ok. Basically this is a good illustration of the effects of an amplifier's output impedance. As I said previously, the reason your amplifiers don't have a constant voltage into a changing load is because of the amplifier's output impedance.
Let's say your amplifier, which isn't using any feedback at this time, has an inherent output impedance of 1 ohm. And into a 16 ohm load you're drawing 1 amp of current. Ohm's Law says that you've got 16 volts across the load.
And since the amplifier's output impedance is in series with the laod, all of the current flowing through the load is also flowing through the amplifier's output impedance. And Ohm's Law says that you must be dropping 1 volt across the amplifier's output impedance.
So basically you've got an amplifier that's ouputting 17 volts (since the total voltage is the sum of all the voltage drops) but because of its output impedance, it's only delivering 16 volts to the load.
Keeping all else equal, drop the load impedance down to 4 ohms. 17 volts into 5 ohms (4 ohms for the load, 1 ohm for the amplifier's output impedance) gives you 3.4 amps. Now you've got 13.6 volts across the load and 3.4 volts dropped across the amplifier's output impedance.
The effect of this is a 2.4 volt (1.4 dB) difference between the voltages applied across a 16 ohm and 4 ohm load.
Now compare the difference if the amplifier's output impedance is reduced two orders of magnitude, from 1 ohm to 10 milliohms. In the case of the 16 ohm load, you've got 16.99 ohms across the load and 0.012 ohms dropped across the amplifier's output impedance. In the case of the 4 ohm load, these numbers become 16.96 volts and 0.042 ohms respectively. So you've gone from a 2.4 volt (1.4 dB) difference to a 0.03 volt (0.015 dB) difference.
When you apply feedback, what you're doing is reducing voltage gain, which also has the effect of lowering output impedance. How much you lower output impedance (and reduce gain) depends on how much feedback you apply with respect to the amplifier's open loop voltage gain.
And this also goes toward your claim that the amplifier is "pushing harder" when you apply feedback. Since the feedback lowers voltage gain, the amplifier's output voltage will be lower for a given input voltage. So even though you've reduced the amplifier's output impedance, you've reduced its voltage gain by the same amount. Which means that the amplifier is pushing just as hard into the same load with feedback as it was without it.
Notably, the amplifier’s feedback circuit showed a 0.5V error voltage passed to the gates of the MOSFETs. This seems to show that the amp was ‘pushing against the voltage’ produced by the speaker.
If the amplifier was indeed "pushing against the voltage" produced by the speaker, then it would be producing a voltage of the same polarity as the voltage from the speaker in which case you'd be measuring something more like the 2.3 volts you got from the non-loaded speaker. And if that were the case, your feedback would be positive rather than negative.
If I had a couple days spare to set up the test properly I might find something interesting, but is it really worth it? My original point was that feedback is required to keep some (ie Hitatchi) MOSFET amps form varying output voltage with load. You say there is another approach to achieve the same thing.
What you're really saying though is that the MOSFET amps you're familiar with have inherently high output impedances and you feel that feedback is required to lower output impedance to an acceptable level.
If you want lower output impedance without feedback, then use devices in the output stage which have an inherently lower output impedance. I don't know that you'll be able to reach the level you'd like using MOSFETs and it appears that MOSFETs are the only devices you'd consider.
Anyway, if a person was really serious about sound they would dispense with the passive crossover and drive each speaker with a separate power amplifier. I think there is something to be said for specialist subbass, bass-mid and treble amps, though that is one further step I don’t think I’ll bother taking!! :)
No argument there. One of my friend's pet peeves are all the outrageously expensive, five figure loudspeakers out there which utilize passive crossovers. I'm inclined to agree. :)
se
Regarding ‘gained momentum’.There is no such a term mentioned in any of my undergrad physics text books. The momentum of a mass is given by its velocity and mass alone. There are no other variables to add. The force required to bring a mass that is travelling at some velocity to a halt is the same as the force required to accelerate that mass in the first place.
Introducing friction complicates things. In particular, the environment in which a speaker cone moves can be thought of as a mass, a force and a spring. The mass is made up of the speaker’s cone, the coil and any other moving parts, and the air mass that is moving with the cone.
The spring can be thought of as having two components - compression and extension. The compression spring can be thought of as all the forces that impact on the cone and prevent it moving. The compressibility of the air in front of the cone would be the main component (for a speaker moving outwardly).
Extension springs can be thought of as pulling against the speaker cone and so hindering its motion. The speakers spider and the surround at the top of the cone, electrical damping, and the damping that the box may provide can all be added together with the compression spring to give one figure for the spring (resistance to the cones motion).
This is not the same as friction. Friction prevents motion. Your hockey puck does not spring back. Speaker cones do. If a voltage were applied to a speaker from an infinite impedance source in one polarity only, say zero to 9 volts, then the driver returns to its rest position of its own accord. All piston-cone drivers will overshoot (decaying oscillations).
As it returns, the speaker produces a voltage.
If there is a low impedance source then the driver will return to its original position slower and will overshoot less.
Regardless of the tension of the spring, a driver will still have to slow down before it can move off in the other direction. That is physics. Whether or not the period in which this slowdown occurs is significant concerns Audio Engineers.
You claim that the ‘friction’ nulls the overshoot condition is erroneous. Leave aside the overshoot and just look at the conditions at the time the reverse voltage occurs (for the low frequency sine wave and drum beat with opposite phase to the sine wave at the instant the drum is struck). If it were a friction effect, like the hocky puck analogy, and ignoring any overshoot then you are 100% correct.
But if we replace friction with spring then at the moment the drum beat arrives there is a spring effect. This spring is going to pull the speaker in the same direction the amplifier was wanting to push it anyway. So the impedance will be higher that it would be if the speaker was stationary (for the overshoot condition, it would be momentarily lower then higher once it started moving in the right direction).
In an extreme case, the voltage produced by the speaker would be significant. A MOSFET amp (with feedback)can take the voltage produced by the speaker into account when calculating the correct voltage over the load. It does this by sensing the voltage at the output and comparing that to the input voltage.
It is because of the intrinsically higher impedance of the MOSFET amp that they are able to do this. An intrinsically lower impedance amp effectively blinds the feedback from any activity beyond the output devices.
And since the amplifier's output impedance is in series with the load, all of the current flowing through the load is also flowing through the amplifier's output impedance. And Ohm's Law says that you must be dropping 1 volt across the amplifier's output impedance.
No, the amplifiers output is in parallel with the load. The only series circuit is between the amplifier’s output and the supply rails via the output devices.
The amplifiers impedance is in parallel with the speaker. If the amplifier’s output impedance was zero ohms then this would be equivalent to a dead short across the speaker (assuming no loss on the leads).
The voltage drop across the amplifiers output is always the same as the voltage drop across the speaker (assuming no lead loss).
Your calculations are erroneous. When calculating the current drawn by the circuit you only need the voltage output of the amplifier and the impedance of the speaker at the test frequency. I’ve done this many times with an accurate current probe and it works out exactly as expected. The amplifier’s output impedance never enters the calculation. (I have done this with high impedance non-feedback MOSFET amps as well - no difference).
Output impedance disappears when the amplifier is switched off. A push-pull transistor amplifier has each pair of devices biased so that they push against each other (electrically). The output resists being changed from the value set by the bias (when the amplifier is idling).
The reason why amplifiers using Hitachi MOSFETs and a feedback circuit have effectively low output impedance is that any voltage appearing at the output, whether generated by the amplifier itself or the load, is compared with the input. An error voltage is generated to correct for any voltages that don’t match.
Apart from output impedance, the feedback corrects for non-linearities in the voltage amplifier stage. All the voltage amplification in most MOSFET amps is done by a transistor voltage amplifier stage.
The speed at which this can be done is roughly equivalent to a 3 megahertz frequency (depending on the stability of the amplifier and the placement and value of various capacitors in the voltage amplifier and feedback circuits).
Keeping all else equal, drop the load impedance down to 4 ohms. 17 volts into 5 ohms (4 ohms for the load, 1 ohm for the amplifier's output impedance) gives you 3.4 amps. Now you've got 13.6 volts across the load and 3.4 volts dropped across the amplifier's output impedance.
No, if the load is 4 ohms and the voltage is 17VRMS then the current drawn is 4.25ARMS. Note that this is not the case if you are measuring the amps voltage at the gates rather than at the output.
Just let me reiterate: the amplifier’s output impedance is measured between the amplifier’s two output terminals (usually load and Earth/zero volt). This can be done by using a constant current sine wave voltage generator to pass a voltage to the amplifier’s output via a resistor. By measuring the voltage across the resistor you can use Ohm’s law to calculate the amplifier’s output impedance. If the voltage is higher, then the impedance is lower etc.
If you want lower output impedance without feedback, then use devices in the output stage which have an inherently lower output impedance. I don't know that you'll be able to reach the level you'd like using MOSFETs and it appears that MOSFETs are the only devices you'd consider.
I took one of my amplifiers around to a certain advocate of Plinius amps and substituted one of his giant class A amps for mine. Compared to the transistor monster, my amp had a clearer midrange (the Plinius was ‘thicker’) and sweeter highs (driving 10’ high electrostatics).
He bought the amp. :)
Kind Regards,
Robert Karl Stonjek.
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2024, Audio Asylum, all rights reserved ]
