Class D vs Class A or A/B

[agedhorse]

a couple of the guys I work with swear by the method of biasing op amp output stages into class a.

It's certainly something that's considered "uber-cool" in some circles. IMO, with most better op-amps driven under normal conditions and with reasonable loading, the difference is really small (maybe not even audible) though in one case I tried the results were clearly inferior to the point of something was very wrong. I gave up on this approach myself.

 

[rumblinbass]

It's certainly something that's considered "uber-cool" in some circles. IMO, with most better op-amps driven under normal conditions and with reasonable loading, the difference is really small (maybe not even audible) though in one case I tried the results were clearly inferior to the point of something was very wrong. I gave up on this approach myself.

I have never tried it myself. From what I understand it can not be done with any op amp. It depends on the output stage.

 

[Duke21]

It's certainly something that's considered "uber-cool" in some circles. IMO, with most better op-amps driven under normal conditions and with reasonable loading, the difference is really small (maybe not even audible) though in one case I tried the results were clearly inferior to the point of something was very wrong. I gave up on this approach myself.

I've a friend at work which build his own all tube head but then for HiFi use. He have a head which he can run on 30W as class A or 100W as class AB, well by adjusting the bias. Amazing sound and amazing open and clear. He also build his own speakers/ cabinets. He run his mostly as class A as he claims it has better sound as class A.

 

[Bob Lee (QSC)]

Part of the explanation for the "slower" sound of the line transformer/Class AB combination might be sag in the supply rails as power demands go up causing some dynamic changes in perceived transient response. Perhaps the Class D amp used for comparison has much stiffer supply rails, and thus stays more consistent (linear) up to the clipping point?

I don't think so. Most reports of "slow" response that I've seen don't mention that the amp is clipping.

 

[B-string]

The things I have seen cause slow response from tube amps have been bias too cold (output tubes starting below cutoff) and leaky decoupling caps. You can "hear too cold" bias in most SS as distortion on note decay (going in and out of cutoff) IME.

 

[R Baer]

I don't think so. Most reports of "slow" response that I've seen don't mention that the amp is clipping.

Could any of this fast/slow response feel of the amps be attributed to differences in damping factors and slew rates?

 

[Bob Lee (QSC)]

Could any of this fast/slow response feel of the amps be attributed to differences in damping factors and slew rates?

Those are both fairly irrelevent specs for bass amps.

 

[seamonkey]

Anybody have a "slow" amp?
Can you post a recording
Channel 1: Bass DI
Channel 2: Output from slow amp

My guess would be EQ
When you pluck a string there's the fundamental, and harmonics. These are traveling waves along the string. And I can clearly see for an open A the envelope of the second harmonic can be in some cases 100ms the fundamental. If you can reproduce the fundamental at all, then you may be hearing the envelope of the second harmonic.

 

[12bass]

Not sure how this relates, but my SVP-PRO all-tube preamp sounds subjectively "looser" than the "tighter" (modified) op amp based preamp in my Eden. The SVP also adds a considerable amount of harmonic distortion when the output is viewed with a spectrum analyzer.

 

[fdeck]

I have never tried it myself. From what I understand it can not be done with any op amp. It depends on the output stage.

Connect a current source from the output to the negative rail. This will effectively shut down the lower leg of the op amp output stage. You could use a JFET with the drain and source tied together.

If you're curious about strange output topologies, take a look at some of the CMOS and rail-to-rail op amps.

 

[fdeck]

We're at the same stage with amps, that we were at with speakers 15 years ago. I remember when the only rule of thumb was "tens are fast." What happened is that bassists have become more knowledgeable about the technology, and new designs have come out that violate the old heuristics.

 

[agedhorse]

From what I understand it can not be done with any op amp. It depends on the output stage.

Yes, I think I proved this one myself

 

[rumblinbass]

Connect a current source from the output to the negative rail. This will effectively shut down the lower leg of the op amp output stage. You could use a JFET with the drain and source tied together.

I am aware of how to and how it works. I have just never tried it.

 

[Calaverasgrande]

Those are both fairly irrelevent specs for bass amps.

 

[fdeck]

Not to speak for Bob, but both slew rate and damping factor are high enough in contemporary amps, that these are not differentiating factors. It's like saying "must not burn the house down." Sure, it's important, but it's not something that needs to be on the spec sheet.

 

[agedhorse]

I agree with Bob's comments for the following technical reasons...

SLEW RATE: Minimum required slew rate is related to both level and frequency. For definition, a slewrate of one volt per microsecond is the slope of the voltage waveform and it must be larger than the steepest part of the curve at the highest frequency and level that the amp must reproduce.

For a rough back of the napkin calculation, let's use 1 volt per microsecond as the amp's slew rate. This means that the transition from -Vpeak to +Vpeak must occur in less time than 1 uSec. This is 1/2 the period of a full sine wave, so at 1 volt peak to peak, 1 V/uSec slew rate will support 500kHz. (this is not completely accurate because we have looked at the average slew rate of the waveform and not the instantaneous rate of change in voltage or dV/dT, but it's close enough to describe the basic principle).

Now to determine the required slew rate of an amp, we take the required maximum voltage (peak to peak) and the maximum frequency that FULL power is required (for bass, as frequency increases power density decreases). Using full power of say 500 watts "RMS" at 4 ohms (based on RMS voltage) this is 45 Vrms, 63Vpeak and 126 Vp-p.

Let's use 1kHz as the maximum full power bandwidth, the 1/2-period time is 0.5mSec.

The amp must be able to slew at a rate of 126V/500uSec or 0.25V/uSec. Double the power and the required slew rate does not double because of the squared term in the power equation (P=V**2/R), it goes up by the square root of 2 or 1.414, so at 1000 watts into 4 ohms the required slew rate is ~0.35v/uSec. All amps I am aware of have a slew rate a minimum of 10 times this, and most are 100x higher.

When Bob (and myself and other engineers here) say slew rate does not matter, they mean IN CONTEXT of the application. There are some really talented and experienced engineers here that I respect very much, it turns out that all of these folks generally agree pretty closely, they are good resources to learn from IMO. They certainly make me think about topics they bring up. I am also seperating slew rate from gain bandwidth product, slew rate is also a function of GBW product, but slew rate is the easily measurable and visable (audible) end result.

Now, as the application changes, and we need to increase the maximum power bandwidth of an amp, say we increase the maximum full power frequency to 10kHz, the required slew rate will increase to 3.5V/uSec for our 1000 watt example. In practice, it will be a little higher due to the dV/dT considerations, but I am staying with the specific concept and not trying to get too detailed.

At 20kHz, this would require 7V/uSec.

I like to use a design margin of around 5x for slew rate (accounting for the zero crossing dV/dT) so for an amp with a maximum power bandwidth of 20kHz, 1000 watts/4 ohms, I would look to somewhere around 35V/uSec but not be terribly upset if 25-30V/uSec was the best I could do if the trade-off to higher slew rate was lower stability.

Just giving a simple " 'round the coffee table " argument to support Bob's comment, and the more you guys understand what this stuff means in general and WHY it's either important (or not important), the better prepared you are to think through these kinds of arguments on your own. I hope this is useful information to some of you anyway.

I will address damping factor in another post.

 

[chadds]

I agree with Bob's comments for the following technical reasons...

SLEW RATE: Minimum required slew rate is related to both level and frequency. For definition, a slewrate of one volt per microsecond is the slope of the voltage waveform and it must be larger than the steepest part of the curve at the highest frequency and level that the amp must reproduce.

For a rough back of the napkin calculation, let's use 1 volt per microsecond as the amp's slew rate. This means that the transition from -Vpeak to +Vpeak must occur in less time than 1 uSec. This is 1/2 the period of a full sine wave, so at 1 volt peak to peak, 1 V/uSec slew rate will support 500kHz. (this is not completely accurate because we have looked at the average slew rate of the waveform and not the instantaneous rate of change in voltage or dV/dT, but it's close enough to describe the basic principle).

Now to determine the required slew rate of an amp, we take the required maximum voltage (peak to peak) and the maximum frequency that FULL power is required (for bass, as frequency increases power density decreases). Using full power of say 500 watts "RMS" at 4 ohms (based on RMS voltage) this is 45 Vrms, 63Vpeak and 126 Vp-p.

Let's use 1kHz as the maximum full power bandwidth, the 1/2-period time is 0.5mSec.

The amp must be able to slew at a rate of 126V/500uSec or 0.25V/uSec. Double the power and the required slew rate does not double because of the squared term in the power equation (P=V**2/R), it goes up by the square root of 2 or 1.414, so at 1000 watts into 4 ohms the required slew rate is ~0.35v/uSec. All amps I am aware of have a slew rate a minimum of 10 times this, and most are 100x higher.

When Bob (and myself and other engineers here) say slew rate does not matter, they mean IN CONTEXT of the application. There are some really talented and experienced engineers here that I respect very much, it turns out that all of these folks generally agree pretty closely, they are good resources to learn from IMO. They certainly make me think about topics they bring up. I am also seperating slew rate from gain bandwidth product, slew rate is also a function of GBW product, but slew rate is the easily measurable and visable (audible) end result.

Now, as the application changes, and we need to increase the maximum power bandwidth of an amp, say we increase the maximum full power frequency to 10kHz, the required slew rate will increase to 3.5V/uSec for our 1000 watt example. In practice, it will be a little higher due to the dV/dT considerations, but I am staying with the specific concept and not trying to get too detailed.

At 20kHz, this would require 7V/uSec.

I like to use a design margin of around 5x for slew rate (accounting for the zero crossing dV/dT) so for an amp with a maximum power bandwidth of 20kHz, 1000 watts/4 ohms, I would look to somewhere around 35V/uSec but not be terribly upset if 25-30V/uSec was the best I could do if the trade-off to higher slew rate was lower stability.

Just giving a simple " 'round the coffee table " argument to support Bob's comment, and the more you guys understand what this stuff means in general and WHY it's either important (or not important), the better prepared you are to think through these kinds of arguments on your own. I hope this is useful information to some of you anyway.

I will address damping factor in another post.

Damp you!!

 

[B-string]

He pretty much slewed the subject Good work as always!

 

[agedhorse]

Now, slewing towards the damping factor question...

DAMPING FACTOR: The damping factor is simply the ratio of load impedance divided by the amplifier's output impedance. The damping factor will vary with load impedance.

For example, take an amp that has an output impedance of 0.01 ohms before the Zobel stabilization network (pretty common with a quality solid state amp) and a load impedance of 8 ohms and you will have a damping factor of (8 ohms/0.01 ohms) or 800 (note that the units cancel). If you want the damping factor at 4 ohms, and the output impedance remains 0.01 ohms (in practice it will probably increase very slightly) then the DF would be 400 and at 2 ohms the DF will be 200.

Now let's look at the real world, still remaining at one frequency for simplicity (DF varies with frequency), if the load remains at 8 ohms but we look at the source impedance part, we will see that the source impedance is the 0.01 ohms of the amplifier circuit, plus the impedance of the Zobel series inductor and internal wiring (0.01 ohms) plus the impedance of the wiring to the speaker (speaker wire pus internal wires) (0.02 ohms) plus the impedance of the low pass inductor on the LF section of the speaker if it has one (0.2 ohms), plus the DC resistance of the voice coil that is not in the gap generating force (1 ohm) and you have (0.01+0.01+0.02+0.20+1.0)=1.24 ohms. Now, calculating the REAL DF, (8ohms/1.24ohms)=6.45.

If you look at the terms in the source impedance equation, the most sigificant term is the voice coil overhang, the larger the Xmax the more wire outside of the gap and the worse the DF will be. This (plus the resulting loss of sensitivity) is the chief drawback to a driver with large Xmax. Generally (this is a rough generalization as there are other things that come into play) a driver with a large Xmax will typically have a lower practical damping factor because the wire outside the gap that's needed for large signal linearity brings with it the drawback of reduced DF.

So, it really doesn't matter all that much in practice if the amp has a DF of 2000, 1000, 500 or even 200 because that term is so far to the right of the decimal point to be relatively insignificant compare with the other terms.

So, we are back to the no free lunch discussion, eh?

 

[B-string]

That should do well to Damp.....en the question and slew the discussion of relevance?
Sorry should have stopped