Slew rate calculations - how much do I need?

Status
Not open for further replies.
Marc Vi. said:
Jan,

I almost agree with you. As I stated in earlier posts, a SR of about 30 V/usec is enough for a 100 W amplifier. But the bandwidth and with that the SR must be fixed by the input filter of the amplifier and therefor the amplifier (100 W) itself must have a bandwidth of 150 - 200 kHz and a slew rate of 40 - 50 V/usec.

Marc.

Mmm not sure. If you put an RC at the input with a turnover point of say 25kHz (and no, I don't believe you will hear the resultant phase shift) I would think the input slew rate assuming 1 or 2 V input for full power would be much less than 30V/uSec, meaning the amp can get by with much less than you state.

Jan Didden
 
SR Meter

Jan:
The remark about a slew rate meter was mine. Its pretty simple to build, all you need is a delay line (I used a 1 uS surplus one from DDD a specialist in them) and a differential amp looking at the difference voltage over the time. I connected it to an LED meter made from the National chip that has been discontinued for years.
It didn't show much. But in this arena that doesn't necessarily support the claim that a high slew rate isn't needed. Any more than the lack of rigorious proof that harmonic distortion below 1% is inaudible means that we don't need amps better than the 1% threshold.

For me its just part of thorough engineering to make an amp perform as well as possible. Getting flat frequency response is still at the top of the list because there is a lot of good evidence that its audible. As is signal to noise ratio.

It may be that the excess slew rate gets "used up" buy the task of driving a complex reactive load (e.g. tweeter w/ xover and eq components or transformer coupled electrostatic panel).

Actually drivers do have a slew rate limit- it comes from the magnetic circuit limits. A friend of mine, Keith Johnson (of Reference Recordings) had problems with a recording monitor where the midrange driver would slowly demagnitize during a recording session as the transients essentially overloaded the magnetics.

And another thing to keep in mind is that most recording mikes don't have response to 20 KHz. The most highly reveared ones with the 1" diaphragms barely make it to 15 KHz on axis.

But I still believe that the higher slew rate, lower distortion, faster settling time amp with instant overload recovery will sound better if it has a flat response and doesn't interact with the load.
-Demian
 
Mr Evil said:

There is nothing wrong with minimizing phase shift from an amp in itself, but doing so will force compromises in other areas. For instance, as has been pointed out earlier, extremely wide bandwidth means that more noise will be amplified. It also often means that open-loop gain must be reduced, which means increased distortion at lower frequencies.
This is not necessarily the case. For a 200kHz BW amplifier, I used a 2MHz filter on the input. The amp exhibits no RF problems, and audibly and visually on the scope is clean. My design does not reduce th o/l gain, only increases the bandwidth and slew rate. Therefore, all other things being equal, low frequency distortion remains constant, and if anything high frequency distortion will be reduced.
Mr Evil said:

Even if phase shift in the uppermost octaves were really that important (you can test this yourself by inserting an all-pass filter - you will find that it takes quite a lot of phase shift before the effects on music are appreciable, assuming it's equal on both channels), the fact that loudspeakers give a 180 degree phase shift at higher frequencies compared to lower frequencies clearly reduces the relative importance of the amplifier's contribution.
Amplifiers should be designed as to not introduce any phase shift. Or at least the amount should be minimal. Therefore, a PBW of at least 200kHz is required. From this figure, minimum Slew Rate is calculated. Loudspeakers exhibit a 180 degree phase shift--low end relative to high end frequencies, and the shift occurs in the low-mid frequencies, below 500Hz. Therefore, no amplifier will have an effect to detriment this speaker phenomenom, and as such, it is not relevant to this discussion. Sutleties in the upper frequencies are much more noticeable, and hence the desire to design an amplifier with both frequency and velocity headroom, within reason of course (i.e. 200kHz, 50V/us @100W).
Mr Evil said:

The uA741 does not meet the minimum requirements for audio, so it's a bad example. Anyway, as I said, increasing the performance in one area will necessarily lead to compromises in other areas. It is important to know how far one parameter can be improved before the point of diminishing returns is passed.
Insert any audio-suitable op-amp desired. How about the RC4558? Again then, will an LM318 sound worse than a RC4558? There are always tradeoffs, but little if any compromise is made in the design of an amp with 200kHz, 50V/us (@100W)specifications. Much higher is possible (I have obtained 600V/us)without too much trouble, but beyond the above specs., one is reaching overkill for audio.
 
janneman said:


...
OTOH, an amp the has slew rate limiting will produce gross distortions, that probably don't sound all that kosher. So, even with the speaker limit, it probably pays to keep the amp clean.
And with competent construction and up to 100W or so, 20V/uSec would be way high enough.

I found Pavel's (I think it was him) experience illuminating: the slew rate meter on his amp had to be sensitised to 0.5V/uSec full scale to get any deflection on normal use! That's a real world test, and should sober up all those 100V/uSec or more advocates.

Jan Didden
Quoting your earlier response: "Your formula is OK, yes. The rest is hearsay, speculation and completely separated from reality."

Your "keep the amp clean" is OK, yes. The rest, is as per the quote above.

20V/us slew rate results in a PBW of 80kHz. Assuming a nominal 6dB/oct gain/bandwidth rolloff, the phase shift at 80kHz is 45º. Translating back to 20kHz yields a phase shift of 14º, and at 10kHz, a phase shift of 7º. A decent loudspeaker will not deviate beyond 10º from some nominal level (say at 1kHz) all the way up to 20kHz.

In light of a typical, well-designed loudspeaker's phase response, it is apparent that driving such a loudspeaker with an amplifier limited to 20V/us slew rate and associated phase shift, compromises the audio reproduction by a significant amount.

We can do better...much better, to the point where the amplifier is no longer contributing to the phase delay, nor limiting the rate of change.
 
wimms said:
I figure that when amp hits SR limits, all hell brakes loose. But is it a sharp edge to cross? What is a difference between amp working at 1/2 of its SR limits and amp working at 1/10 of said limit? What amp parameters start to suffer when approaching the SR limit, even though being still far from it?

Depends overwhelmingly on the type of compensation used......
 
lumanauw said:
Continuing MikeB, Is there any corelation between Slewrate and Amp's OLBandwith/OLgain?

Is a fast slewrate amp necessarily have low OL bandwith with high OL gain?

I would say, there's a direct connection. But an amp with high slewrate
needs a high OLBW, not low. A higher OL gain can increase slewrate,
but also drive faster into slewratelimits.
When i design an amp, i check to keep OLBW over audioband, this
automatically reduces OL-gain to some value given by the topology
and the devices used.
You can reach a state, where the errorvoltage is big enough to push
single stages into saturation. This is some kind of catastrophy, as the
amp needs some extra time to recover from this saturation, giving
overshooting for example. A cap in the feedback can greatly reduce
this effect, as it reduces errorvoltage for high slews. This cap also
reduces slewrate, but can keep the amp from driving into saturation.

Also the design of the outputstage is important, for high "slewing"
the devices need more basecurrent. That's one reason why i prefer
tripledarlington, the highspeed smallsignal-bjts in the predriverstage
keep this effect away from the vas.
The funny thing to me is, that means that the amp can easily begin
to oscillate if not properly compensated...

Mike
 
mikeks said:
There is no direct connection whatsoever...
Oh, your posts are so very helpful!


Slew rate and open-loop bandwidth are related, but one does not imply the other.

Slew rate is (usually) limited by the current available to charge/discharge the compensation capacitor. Since this capacitor also determines bandwidth, you can see that it is hard to adjust one without also affecting the other, but it is possible to have high slew rate and low bandwidth or vice versa.
 
z_p_e said:

Assuming a nominal 6dB/oct gain/bandwidth rolloff, the phase shift at 80kHz is 45º. Translating back to 20kHz yields a phase shift of 14º, and at 10kHz, a phase shift of 7º. A decent loudspeaker will not deviate beyond 10º from some nominal level (say at 1kHz) all the way up to 20kHz.

In light of a typical, well-designed loudspeaker's phase response, it is apparent that driving such a loudspeaker with an amplifier limited to 20V/us slew rate and associated phase shift, compromises the audio reproduction by a significant amount.

We can do better...much better, to the point where the amplifier is no longer contributing to the phase delay, nor limiting the rate of change.

So far, I haven't seen any LS capable of rolling off at 80kHz. Typical, well-designed LS rather do so around 25kHz, usually steeper than 6dB/oct; hence the phase shift (transient response etc.) introduced by the speaker overwhelms any effect the fast amplifier could have on audio reproduction. This implies (such small) phase shifts being audible and full voltage swing required at such high frequencies. Also, the naked-number game (20V/us) can only be played at the given output power; build a weller-designed LS :clown: and use a 20W amp. At 20V/us, plenty of PBW there.

Cheers,
brummkreisel
 
Mr Evil said:

Oh, your posts are so very helpful!

Thank you.... :hug:

Mr Evil said:
Slew rate is (usually) limited by the current available to charge/discharge the compensation capacitor. Since this capacitor also determines bandwidth, you can see that it is hard to adjust one without also affecting the other, but it is possible to have high slew rate and low bandwidth or vice versa.

Actually, you can adjust 'small-signal' bandwidth by means other than the comp. cap., eg degeneration in input stage, resistor across comp. cap (bad idea!), additional comp. eg input lag comp....etc....while affecting slew rate not one jot....


MikeB said:


So, whatelse defines slewrate except inputfilter and outputdevices ???
Isn't speed of vas in direct correlation to OLBW ???


slew rate is only determined (to first order) by current through, and voltage swing across minor loop comp. cap...

input filter merely reduces the amplitude of HF stimuli to prevent an amp's slew rate limit from being infringed.....

output devices have, for practical purposes, no effect on slew....
 
Status
Not open for further replies.