Yes - and do absolutely no harm at all.rhythmsandy said:
That is not a promising start to a post.MrMagic said:
It all depends on what you mean by "undistorted". You cannot mean 'no distortion at all' as that is not possible in this universe. If you mean 'quite low distortion' then it may be helpful to know what is the source of that distortion (e.g. limited slew rate, smooth nonlinearity, crossover glitches). Of course, any circuit parameter can be used in a misleading sense (and most of them often are treated in this way) but that does not mean that the parameter is redundant.
THD tells us something, but not much. Anyway, thank you for confirming to THD-haters that there are genuine THD fans about; they will use this evidence against those of us who have a sober view of THD, by confusing us with genuine THD fans like yourself.
Why, what will happen if you feed an amp characterized at 20-20Khz and ultra-low THD up to the 10nth harmonic, with a signal that extends to 30Khz?
Yet it's a honest one.
Of course I mean ultra-low distortion.
Since the amp has already ultra-low distortion according to the THD parameter, you do NOT care about ...the sources of distortion anymore.
Slew rate might be useful during development, NOT to characterize the amp for the customer/builder.
THD along with frequency response, output power, power bandwidth and load specs, is all a customer/builder needs, that makes slew rate redundant.
I don't have any insecurities with "THD-haters" and I'm a fan of perfection, not just THD, but I'm all ears to enlighten me about something I'm missing. "it doesn't tell us much" is too vague to be considered an argument.
Last edited:
THD (the clue is in the name) tells you the total harmonic distortion. It does not tell you what the distortion orders are, yet you need to know these. It also does not necessarily tell you everything about IMD, yet this can be more important than harmonic distortion. A "fan of perfection" will need much more than a THD figure to tell him whether an amp has acceptable distortion performance.
Amplifier slew rate is what I call a hygiene factor. That is, it needs to meet a minimum standard but is otherwise not relevant in and of itself.
IMO it is analogous to the top speed of a car. It may be a nice sales gimmick to boast a car that can do 150mph but this doesn't tell you anything about how nice it is to drive at normal speeds.
Most home listening is done with digital sources that produce virtually nothing above 20kHz. At 100W average 8-ohms that's 5V/us (and that is extremely loud). Some margin is sensible but where does 50V/us or 100V/us come from? Where is the explanation?
In my experience it is quite easy to exceed the minimum slew rate requirement in a circuit so it is not a sonic concern.
It may be that in some circuit designs, parts of the circuit become very non-linear as they approach the design's slew rate limit. In this case, for this design, it may be necessary to design for, say, 10x the slew rate needed to keep the circuit linear enough. But this is not a requirement of slew rate. This is a side-effect of other factors in the design.
IMO it is analogous to the top speed of a car. It may be a nice sales gimmick to boast a car that can do 150mph but this doesn't tell you anything about how nice it is to drive at normal speeds.
Most home listening is done with digital sources that produce virtually nothing above 20kHz. At 100W average 8-ohms that's 5V/us (and that is extremely loud). Some margin is sensible but where does 50V/us or 100V/us come from? Where is the explanation?
In my experience it is quite easy to exceed the minimum slew rate requirement in a circuit so it is not a sonic concern.
It may be that in some circuit designs, parts of the circuit become very non-linear as they approach the design's slew rate limit. In this case, for this design, it may be necessary to design for, say, 10x the slew rate needed to keep the circuit linear enough. But this is not a requirement of slew rate. This is a side-effect of other factors in the design.
Last edited:
Happy New Year, Matt.
My analogy likens the rate of change of voltage to the car's speed.
Your analogy likens the rate of change of voltage to the car's acceleration.
A car's acceleration is about power response. It is to do with the engine torque curve and mass of the car. In a power amplifier I'd liken the torque to the current delivery and the mass to the load capacitance. The resistance of the load is like friction.
The voltage slew rate of an amplifier is ultimately defined by a current limit somewhere. The music signal has a maximum dV/dt requirement. It may be that it is beneficial in a certain design to have excess current capability that also happens to allow the amp to far exceed the music signals dV/dt, but that is not actually a slew rate requirement.
The reason I am going on and on about this is because I have seen some designs where high slew rates have been achieved at the expense of far more important sonic parameters. It's a bit like THD - there are plenty of speakers with 1% THD that sound wonderful and plenty of amps with 0.001% THD that sound unimpressive.
Basically, I am saying don't get fixated with the pursuit of excessive slew rate...look for what actually affects the sound quality.
If THD is medium, then yes, you need to check the causes of the existing distortion, but if it is "ultra-low" as was the requirement in my first post eg around or less than -120db, that practically means that there is NO distortion for you to care, at all.
THD is a linearity measurement and so is IMD. For example, I'm developing a high voltage SS ES amp for some time now and i have never measured IMD. THD was always my guide, along with frequency response (and noise), but I was measuring THD at 20Khz, not just 1Khz, on an amp with a bandwidth that would exceed 1Mhz (if the mosfets could cope with the extremely high capacitive currents).
So I just did that now and IMD harmonics are even lower: -128db is the highest harmonic in FFT @20Khz fundamental and -132db is the highest harmonic in IMD measurement with 20Khz and 19Khz tones.
And power bandwidth replaces slew rate and is far more intuitive and meaningful. If your amp has the power bandwidth you desire along with ultra-low THD specs, it also has the proper slew rate.
So if you properly measure the other important parameters, you don't need slew-rate which is just a dependent subset meant for other purposes.
As for perfection, I mentioned 5 parameters on the previous post, not just THD and there are more, like stability, reliability, efficiency, transistor/mosfet protection, noise, etc.
I am glad to see this thread going back on topic.
Post #1 is:
What is acceptable ?
For instance 20 V / uS gives no distortion on a 100 Watt rms sine signal in 8 ohm at 1KHz....BUT, at 20 Khz, distortion comes in at 10 Watt and becomes bad over 10W.
This is theoretical results about pure sine signals. What about real audio ? What slew rate is needed for real audio quality ?
I am glad to see in the last answers remarks on the technical realities of slew rate, well apart from bogus arguments.
So I read as much I could stand in this long thread to gather informations and make my own opinion.
IMHO Slew rate is redundant, umoungous slew rate values are non sense.
An amplifier needs no more slew rate as necessary to not induce distortion.
Correct design provisions enough head room about voltage and current to avoid distortion from clipping.
As well, correct design must provision enough margin about slew rate ( voltage/uS and current/uS ) to avoid distortion
We do not design for 100V head room for 10V signals, designing for 50V/uS is a similar non sense when 5V/uS is enough.
Another interesting consideration is: How much slew rates do we have in real audio sources. All audio souces are band limited, there cannot be umongous slew rates in true audio.
Another interesting point, is the relevance of THD. Yes it is not perfect, but if we must point at the best measure ? do we have any better ?
What we ask from an ampifier is an output which is a best fit with the amplified input.
Mathematicians, to measure ( compare ) how two curves fit, came with the mean square root method. This is precisely what THD is.
THD 20Khz and IMD 20Khz / 19Khz are worthwhile measures. Slew rate is nothing more than a side issue.
High slew rates are met with large amplitude signals AND large frequencies ( SR = 2 Pi F Vpk ).
About a 100 Watt in 8 ohm amplifier ( Vpk is 40V ) There exist no signal at full power 100 Watt beyond, say 10 Khz, your tweeters would explode and your ear drums bleed.
I admit, we need a bandwith of up to 80 Khz, but not at full power from 10Khz to 80 Khz.
IMO, I think slew rate required to avoid distortion at 10 Khz full power is enough, and that's all I need for an amplifier.
Post #1 is:
What is acceptable ?
For instance 20 V / uS gives no distortion on a 100 Watt rms sine signal in 8 ohm at 1KHz....BUT, at 20 Khz, distortion comes in at 10 Watt and becomes bad over 10W.
This is theoretical results about pure sine signals. What about real audio ? What slew rate is needed for real audio quality ?
I am glad to see in the last answers remarks on the technical realities of slew rate, well apart from bogus arguments.
So I read as much I could stand in this long thread to gather informations and make my own opinion.
IMHO Slew rate is redundant, umoungous slew rate values are non sense.
An amplifier needs no more slew rate as necessary to not induce distortion.
Correct design provisions enough head room about voltage and current to avoid distortion from clipping.
As well, correct design must provision enough margin about slew rate ( voltage/uS and current/uS ) to avoid distortion
We do not design for 100V head room for 10V signals, designing for 50V/uS is a similar non sense when 5V/uS is enough.
Another interesting consideration is: How much slew rates do we have in real audio sources. All audio souces are band limited, there cannot be umongous slew rates in true audio.
Another interesting point, is the relevance of THD. Yes it is not perfect, but if we must point at the best measure ? do we have any better ?
What we ask from an ampifier is an output which is a best fit with the amplified input.
Mathematicians, to measure ( compare ) how two curves fit, came with the mean square root method. This is precisely what THD is.
THD 20Khz and IMD 20Khz / 19Khz are worthwhile measures. Slew rate is nothing more than a side issue.
High slew rates are met with large amplitude signals AND large frequencies ( SR = 2 Pi F Vpk ).
About a 100 Watt in 8 ohm amplifier ( Vpk is 40V ) There exist no signal at full power 100 Watt beyond, say 10 Khz, your tweeters would explode and your ear drums bleed.
I admit, we need a bandwith of up to 80 Khz, but not at full power from 10Khz to 80 Khz.
IMO, I think slew rate required to avoid distortion at 10 Khz full power is enough, and that's all I need for an amplifier.
In my view bandwidth and THD are more important than slew rate but distortion can increase due to capacitive feedback currents. If these are avoided then a wide bandwidth will give low distortion at 20kHz.
Why a high bandwidth - this is simply to avoid compounded effects of multiple stages on the overall bandwidth. A 100W 8 ohm amp working at 20kHz needs a functional slew rate of 5V/us. If you limit the amplifier to 5V/us the net effect is to degrade the bandwidth to 10kHz, so to please younger ears which might hear 20kHz I aim my designs for 200kHz, which needs a slew rate of 50V/us. It is doable with fast output transistors, but I suspect 100kHz would be sufficient in most cases.
Why a high bandwidth - this is simply to avoid compounded effects of multiple stages on the overall bandwidth. A 100W 8 ohm amp working at 20kHz needs a functional slew rate of 5V/us. If you limit the amplifier to 5V/us the net effect is to degrade the bandwidth to 10kHz, so to please younger ears which might hear 20kHz I aim my designs for 200kHz, which needs a slew rate of 50V/us. It is doable with fast output transistors, but I suspect 100kHz would be sufficient in most cases.
I agree:
5V/uS is needed for a 100W 20 Khz sine signal undistorted by slew rate.
50V/uS for 100W 200 KHz.
But...can we have in real audio such high powers 100W at 20 KHz, 100W at 200 KHz ?
They do not do that in bi ampling or tri ampling.
It is well admitted the amp for high frequencies is of a lower power than the amp for medium frequencies.
5V/uS is needed for a 100W 20 Khz sine signal undistorted by slew rate.
50V/uS for 100W 200 KHz.
But...can we have in real audio such high powers 100W at 20 KHz, 100W at 200 KHz ?
They do not do that in bi ampling or tri ampling.
It is well admitted the amp for high frequencies is of a lower power than the amp for medium frequencies.
I remind myself that an amplifier goes into slew rate limiting when 100.0% of the input stage current is steered into one leg of the LTP, and 0.0% of the input stage current is steered into the other leg of the LTP. At that point, the current charging the compensation capacitor cannot increase any further, and the slew rate dV/dt is at its maximim: Itail/Ccomp.
For a typical power amp with a JFET LTP biased at 5mA per leg, the input stage transfer characteristic will be similar or identical to Figure 1, below.
When the amplifier is slew rate limited, the IPS current is 100% steered into the "green" output leg and 0% steered into the "red" output leg, at points A and B respectively.
I myself don't want to go anywhere near points A and B when playing music, so in my designs, I prefer to add a margin of safety in the following way:
I insist that when reproducing a full-amplitude (onset of clipping) sine wave at 100 kilohertz, the input stage shall not exceed points G and H on the figure. This ensures the input stage remains comfortably within its linear amplification zone, including a large margin of safety to the left and to the right of points G and H.
The total IPS current is 10 mA (point A minus point B), but I refuse to operate beyond points G and H: 7mA - 3mA = 4mA. Notice that the amplifier's actual slew rate (10mA/Cc) is 2.5X greater than the slew rate achievable by obeying my requirement to remain within points G and H. Math: (10mA / 4mA) = 2.5X.
And, believe it or not, that's all we need to calculate the actual slew rate requirement of the amplifier: A 100 kHz sinewave has a slew rate of (2 * pi * 100K) = 0.63 volts per microsecond per volt of supply rail. And then my required G-H safety margin multiplies this by another factor of 2.5X. Voila: 1.6 volts per microsecond per volt of supply rail. It's a handy rule of thumb. 1.6V/us/V
Example: a 100 watts RMS per channel amplifier with ±50 volt supply rails, needs a slew rate of 80 volts per microsecond (50 * 1.6) to satisfy my requirement. Fortunately this is achievable with conventional discrete amplifier designs. For example, Bob Cordell's 50 watt amplifier from 1984, achieved 300 V/usec.
_
For a typical power amp with a JFET LTP biased at 5mA per leg, the input stage transfer characteristic will be similar or identical to Figure 1, below.
When the amplifier is slew rate limited, the IPS current is 100% steered into the "green" output leg and 0% steered into the "red" output leg, at points A and B respectively.
I myself don't want to go anywhere near points A and B when playing music, so in my designs, I prefer to add a margin of safety in the following way:
I insist that when reproducing a full-amplitude (onset of clipping) sine wave at 100 kilohertz, the input stage shall not exceed points G and H on the figure. This ensures the input stage remains comfortably within its linear amplification zone, including a large margin of safety to the left and to the right of points G and H.
The total IPS current is 10 mA (point A minus point B), but I refuse to operate beyond points G and H: 7mA - 3mA = 4mA. Notice that the amplifier's actual slew rate (10mA/Cc) is 2.5X greater than the slew rate achievable by obeying my requirement to remain within points G and H. Math: (10mA / 4mA) = 2.5X.
And, believe it or not, that's all we need to calculate the actual slew rate requirement of the amplifier: A 100 kHz sinewave has a slew rate of (2 * pi * 100K) = 0.63 volts per microsecond per volt of supply rail. And then my required G-H safety margin multiplies this by another factor of 2.5X. Voila: 1.6 volts per microsecond per volt of supply rail. It's a handy rule of thumb. 1.6V/us/V
Example: a 100 watts RMS per channel amplifier with ±50 volt supply rails, needs a slew rate of 80 volts per microsecond (50 * 1.6) to satisfy my requirement. Fortunately this is achievable with conventional discrete amplifier designs. For example, Bob Cordell's 50 watt amplifier from 1984, achieved 300 V/usec.
_
Attachments
I see you have decided your amplifier should be full power able at 100 KHz.
Then, you apply a safety margin x 2.5
On top of this another safety margin x 1.25 ( from 50/40 )
Remove the un necessary safety margins, then you need 25V/uS to handle 100KHz at full power or 2.5V/uS to handle 10KHz at full power.
Well, your tweeters and ear drums will explode at these frequencies at full power.
Then, you apply a safety margin x 2.5
On top of this another safety margin x 1.25 ( from 50/40 )
Remove the un necessary safety margins, then you need 25V/uS to handle 100KHz at full power or 2.5V/uS to handle 10KHz at full power.
Well, your tweeters and ear drums will explode at these frequencies at full power.
Of course other designers are welcome to add (or omit) a margin of safety, according to their own preferences.
@Mark Johnson
OR, to satisfy your criteria, you just need a power bandwidth of 100Khz x 2.5 your safety margin = 250Khz. No need for slew rate calculations and tests.
And an amp might be able to operate at say 1000V/μs at low-to-medium output but might be destroyed instantly at a full output from the excessive output currents, so slew-rate alone cannot express that, as you need magnitude too (voltage or current). In contrast, power bandwidth can: you just limit the power-bandwidth either natively, or by implementing a current limiter -as I'm going to do with the ES amp. So in that case, slew rate is not only redundant, it's also inadequate.
OR, to satisfy your criteria, you just need a power bandwidth of 100Khz x 2.5 your safety margin = 250Khz. No need for slew rate calculations and tests.
And an amp might be able to operate at say 1000V/μs at low-to-medium output but might be destroyed instantly at a full output from the excessive output currents, so slew-rate alone cannot express that, as you need magnitude too (voltage or current). In contrast, power bandwidth can: you just limit the power-bandwidth either natively, or by implementing a current limiter -as I'm going to do with the ES amp. So in that case, slew rate is not only redundant, it's also inadequate.
Last edited:
I am puzzled about the needed power versus frequency.
Sure we do need full power at bass and medium frequencies.
However, it looks wrong to me, to want full power, up to 99KHz then no power beyond 100 KHz.
I guess, it should go gradually, not sharply at some band limit. And why 100KHz, why not 80 or 150 or 300 ?
I guess there is some optimal decay in my way of thinking which is to put performance where it is needed and no more.
Sure my thinking is an opposite approach to the blameless way which is to put every where as much performance as one can design.
I do not claim my approach is better, I simply don't know when applied to audio.
If you think, my way, what would be a good decay, from some start frequency to some stop frequency with some decay law.
Sure we do need full power at bass and medium frequencies.
However, it looks wrong to me, to want full power, up to 99KHz then no power beyond 100 KHz.
I guess, it should go gradually, not sharply at some band limit. And why 100KHz, why not 80 or 150 or 300 ?
I guess there is some optimal decay in my way of thinking which is to put performance where it is needed and no more.
Sure my thinking is an opposite approach to the blameless way which is to put every where as much performance as one can design.
I do not claim my approach is better, I simply don't know when applied to audio.
If you think, my way, what would be a good decay, from some start frequency to some stop frequency with some decay law.
@mchambin
Mark Johnson meant 250Khz, not just 100khz.
About the need for 100Khz+, I'll give you a reason: High fidelity.
That is, to amplify the signal without changing it (as faithfully as feasible/practical).
It's almost impossible to achieve ultra-high-fidelity for the full audio spectrum at full power, with an amp that has a power bandwidth equal or just above the frequency response it needs to reproduce.
That's because the amp needs a really close and fast error correction (feedback) in order to be able to correct eg one millionth (-120db) of the signal, after it detects it, on frequencies up to 20khz. If the amp can just reproduce 20Khz, then until it corrects the detected error, 50μs or more would have passed (as is the period of a 20khz signal) -and the flaw to correct won't be there any more -it will be already traveling in the air towards the sensitive ears of the listener
Theoretically you could eliminate feedback loops and just use the immediate local feedback from emitter and collector resistors, along with linearizing techniques (introducing opposite distortion to cancel out distortion) but the parts' specs change a lot with temperature and achieving ultra-low distortion is extremely difficult with such techniques (if someone has achieved this, I'd like to know).
About the sharp decay, smooth or sharp won't make a difference, in my case it is just a disaster prevention for an unlikely event that can only happen beyond the amp's specs.
Mark Johnson meant 250Khz, not just 100khz.
About the need for 100Khz+, I'll give you a reason: High fidelity.
That is, to amplify the signal without changing it (as faithfully as feasible/practical).
It's almost impossible to achieve ultra-high-fidelity for the full audio spectrum at full power, with an amp that has a power bandwidth equal or just above the frequency response it needs to reproduce.
That's because the amp needs a really close and fast error correction (feedback) in order to be able to correct eg one millionth (-120db) of the signal, after it detects it, on frequencies up to 20khz. If the amp can just reproduce 20Khz, then until it corrects the detected error, 50μs or more would have passed (as is the period of a 20khz signal) -and the flaw to correct won't be there any more -it will be already traveling in the air towards the sensitive ears of the listener
Theoretically you could eliminate feedback loops and just use the immediate local feedback from emitter and collector resistors, along with linearizing techniques (introducing opposite distortion to cancel out distortion) but the parts' specs change a lot with temperature and achieving ultra-low distortion is extremely difficult with such techniques (if someone has achieved this, I'd like to know).
About the sharp decay, smooth or sharp won't make a difference, in my case it is just a disaster prevention for an unlikely event that can only happen beyond the amp's specs.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.