There is a certain step voltage below which it is impossible to slew limit an amp. This can make it difficult to measure slew rate in dynamic conditions as when you want to measure slew rate at say clipping minus 2V, the output step voltage needs to be below 2V. And this usually means a normal square wave input is unable to reach slew limiting.
For example, if your amp is slew limited by the LTP, then it can't be driven to slew with a step voltage less than say 500mV because that is not enough voltage to overdrive the LTP.
This voltage is also raised by the degeneration voltage of the LTP resistors combined. So theoretically you could make an amp which has no chance of slew limiting regardless of frequency, as long as the input voltage range is not exceeded. Below this point there is not even any need to worry about what frequency range the source is able to put out.
At normal listening levels many amps probably cannot even be driven to slew limit regardless of the input frequency.
On the other hand amps such as the Cyrus One which use a nondegenerated CFP LTP could possibly be overdriven by just a few tens of mV (maybe less, I didn't do the math) provided the frequency is high enough. However this didn't stop it from being known for sounding good. It is easy enough to prevent slew limiting here by using an input LP filter.
As for a "soft slewing" metric, I think if we consider 6db to be a good number for loop gain margin, then it should also be a good number for the loop gain loss during slewing. So if the LTP is what is causing the slew limit, then we would determine the slew rate by looking at which rate reduces the LTP transconductance by 6db (which corresponds to 45mV across the bases if undegenerated). Above this point the gain margin of the feedback loop is exceeded, although perhaps not for long enough to develop an oscillation.
For example, if your amp is slew limited by the LTP, then it can't be driven to slew with a step voltage less than say 500mV because that is not enough voltage to overdrive the LTP.
This voltage is also raised by the degeneration voltage of the LTP resistors combined. So theoretically you could make an amp which has no chance of slew limiting regardless of frequency, as long as the input voltage range is not exceeded. Below this point there is not even any need to worry about what frequency range the source is able to put out.
At normal listening levels many amps probably cannot even be driven to slew limit regardless of the input frequency.
On the other hand amps such as the Cyrus One which use a nondegenerated CFP LTP could possibly be overdriven by just a few tens of mV (maybe less, I didn't do the math) provided the frequency is high enough. However this didn't stop it from being known for sounding good. It is easy enough to prevent slew limiting here by using an input LP filter.
As for a "soft slewing" metric, I think if we consider 6db to be a good number for loop gain margin, then it should also be a good number for the loop gain loss during slewing. So if the LTP is what is causing the slew limit, then we would determine the slew rate by looking at which rate reduces the LTP transconductance by 6db (which corresponds to 45mV across the bases if undegenerated). Above this point the gain margin of the feedback loop is exceeded, although perhaps not for long enough to develop an oscillation.
I don't think that is meaningful in any way. Slew rate is by definition a large signal parameter, about how fast the output can slew from min to max level at the highest frequency that must be reproduced and still following the waveform.
You want to know if it can go through a max level change at the highest frequency without distortion. Knowing that is can do that for a smaller level change isn't so useful, you still don't know if it can do that for a maximum level change.
Jan
Slew rate limit is set by the IS current and the VAS compensation cap value, whether you get limiting depends on signal dV/dt, and is only "large signal" if the input is band-limited(*). You can feed a small square wave into an amp and measure the slew rate without needing high signal levels, but its easy to calculate anyway.
(*) ie the max dV/dt scales with max V if band limited, rather than being independent of it.
(*) ie the max dV/dt scales with max V if band limited, rather than being independent of it.
1. Consider that high power amplifiers require lots of large transistors which are generally slow, making high slew rates difficult. A smaller, more accurate amplifier is usually a wiser choice for amateur use.
2. >50W at 20KHz is nothing more than a liability for most audio systems. Professionals deliberately use smaller amps for the high side of bi-amp and tri-amp systems because burning out horn drivers is expensive.
2. >50W at 20KHz is nothing more than a liability for most audio systems. Professionals deliberately use smaller amps for the high side of bi-amp and tri-amp systems because burning out horn drivers is expensive.
But it is not the output devices that determine the slew rate. It is the input stage standing current in combination with the comp cap.
There's no reason why a big amp would be less accurate, whatever that means here. Lots of examples.
Jan
Last edited:
It is the small amps that clip all the time that burn out tweeters. Clipping generates lots of high harmonics which heat up the tweeters.
High frequencies in music have very high crest factors. That means that although peaks may be high, average power in the high frequencies is very low.
But if you use an amp that is under-powered, it will clip and that generates much more power for that poor tweeter to absorb. So it's better to have a high power amp.
That higher power amp in itself will not generate anything that is not in the music. If the music is too loud for the tweeter to handle, you'll hear distortion and you can just turn it down. With an under-powered amp you may be too late ...
Jan
Last edited:
The slower the output, the more compensation is required to keep the feedback stable. The VAS is not compensated for it's own sake, but to cover all the other poles in the feedback loop, the slowest of which are the outputs.
Agreed, in principle, but 4 times as many output devices doesn't mean a 4 times as large comp cap.
Jan
Only for passive crossover systems and such are very rare in professional sound. And if an amateur needs hundreds of Watts of bass, then they should also be using a bi-amp or tri-amp system too, and soon enough a hearing aide.
It is true that high-power amplifiers require more output transistors, but those individual transistors need not be slower devices. A high power amplifier with, say, three times as many output transistors as a medium power amplifier, perhaps 6 pair rather than 2 pair, will require perhaps 3 times as much current to drive their capacitances and up to three times as much current to supply base current even at low frequencies at the much higher power level.
This means the driver must supply more current. But a high power amplifier implemented with an output Triple will have no problem in this regard. There is no fundamental reason that the output stage in a high power amplifier will be much slower than that of a medium power amplifier. Moreover, with an output Triple, a VAS operating at still-reasonable bias current will not slew-rate limit driving the pre-driver capacitance. There is also no fundamental reason that the higher power amplifier has to have a lower gain crossover frequency or heavier compensation.
That having been said, given the kinds of applications for high power amplifiers, e.g. pro audio, there may not be as great a need for as much slew rate margin as one might want for high-end consumer audio.
50W [continuous average] at 20 kHz might indeed be a liability for audio systems, but it is not going to happen. Instead, high frequency peaks that are brief may very well occur on well-recorded music with high dynamic range, and tweeters are generally quite able to handle such brief high-power peaks. In my Athena 3.5-way active speakers, there are four 125-watt MOSFET power amplifiers. One of them is dedicated to the single tweeter. There is nothing wrong with putting a 125-W amplifier behind a tweeter. That amplifier never clips, and the tweeter is perfectly fine. The tweeter would be less happy if that amplifier clipped. However, with the active crossover, that amplifier never sees low frequency clipping, so even if it did clip occasionally those instances would be very brief and the average power would still be low. It is indeed fortunate that we do not listen to sine waves
.Cheers,
Bob
Thank you everyone for that great discussion about slew rate very interesting, informative and helpful.
I was also wondering something for awhile now.
From reading Bob's book its clear that the more you degenerate the input differential the more IPS open loop gain you loose. Of course the degeneration is necessary for the reasons Bob talks about and one of the solution that Bob talks about to recover some of that lost gain is to add a current mirror to the differential pair and add another transistors to the input of the VAS to ultimately increase its input impedance and in doing so increase the IPS gain.
Would the KSC1845 be a better choice then the 2N5551 as it has alot more gain?
My second question is.
When we caculated the gain of the IPS using the extra darlington transistor arrangement why was the main VAS transistors gain not included in the calculation?
I was also wondering something for awhile now.
From reading Bob's book its clear that the more you degenerate the input differential the more IPS open loop gain you loose. Of course the degeneration is necessary for the reasons Bob talks about and one of the solution that Bob talks about to recover some of that lost gain is to add a current mirror to the differential pair and add another transistors to the input of the VAS to ultimately increase its input impedance and in doing so increase the IPS gain.
Would the KSC1845 be a better choice then the 2N5551 as it has alot more gain?
My second question is.
When we caculated the gain of the IPS using the extra darlington transistor arrangement why was the main VAS transistors gain not included in the calculation?
The KSC1845 is difficult to get in the actual high Hfe grades anymore. The BC550C is a better bet if what you need is raw gain.
In a LTP you want high Hfe so that your source impedance doesn't cut into linearity and you have less potential offset to worry about.
In a current mirror you want degeneration to reduce noise, and you want transistors that work well at low Vce. The BC550C/560C are generally a good choice.
Has anyone else noticed that "low noise" BJTs like the KSC1845 are specially designed to be as noisy as possible, with Rb of 600 ohms and up? All the BJTs actually suitable for low noise in circuits other than a hi-Z mic preamp are therefore given completely bogus "noise factor" specs so that no one will find the ones that actually might work.
In a LTP you want high Hfe so that your source impedance doesn't cut into linearity and you have less potential offset to worry about.
In a current mirror you want degeneration to reduce noise, and you want transistors that work well at low Vce. The BC550C/560C are generally a good choice.
Has anyone else noticed that "low noise" BJTs like the KSC1845 are specially designed to be as noisy as possible, with Rb of 600 ohms and up? All the BJTs actually suitable for low noise in circuits other than a hi-Z mic preamp are therefore given completely bogus "noise factor" specs so that no one will find the ones that actually might work.
Hi Stuart,
While it is true in a conventional amplifier you get more slew rate (and input stage linearity) with increased IPS degeneration, at the end of the day you do not lose loop gain where it counts - which usually in the frequency range from 1 KHz up, since with the decrease in IPS transconductance, you use a correspondingly small Miller compensation capacitor, keeping the ULGF the same. This reasoning applies if you are using a 2T VAS with a current mirror. It may be a little less true with a 1T VAS and/or without a current mirror load on the IPS.
Note also that with increased IPS degeneration, you will get a little bit more input-referred noise in the amplifier. I usually degenerate the input stage by a factor of about 10.
Cheers,
Bob
