Sound Quality Vs. Measurements

[DF96]

Originally Posted by magnoman

Slew rate at least in my view is number related to the overall loop or gain bandwidth of the amp. In order to correlate the two you have to bypass the input filter in order to stimulate the amp with frequencies just beyond its bandwidth.

My understanding is that the slew rate limit and the gain-bandwidth of the amp (either open or closed loop) are two independent parameters. In its simplest form, the slew rate limit is set by a capacitance and a current; the bandwidth is set by a capacitor (possibly, but not necessarily, the same capacitor) and a resistor (which may be an active device parameter or a separate component). Hence measuring one tells you precisely nothing about the other; they are not 'number related'. Of course, competent design will ensure that the slew rate limit is sufficiently high to handle the required bandwidth but that is a design goal not something guaranteed by the circuit theory.

[bcarso]

Originally Posted by DF96

My understanding is that the slew rate limit and the gain-bandwidth of the amp (either open or closed loop) are two independent parameters. In its simplest form, the slew rate limit is set by a capacitance and a current; the bandwidth is set by a capacitor (possibly, but not necessarily, the same capacitor) and a resistor (which may be an active device parameter or a separate component). Hence measuring one tells you precisely nothing about the other; they are not 'number related'. Of course, competent design will ensure that the slew rate limit is sufficiently high to handle the required bandwidth but that is a design goal not something guaranteed by the circuit theory.

Yes. They are independent. In principle an amplifier could have a 100MHz gain-bandwidth product (and even be unity-gain stable) and a maximum slew rate of 1V/usec. Note some of the bipolar opamps that have high GBW but modest slew rates compared to the BiFET opamps like the TL071.

A reference that Self cites is of relevance, Stochino's article on "non-slewing" amplifiers from Electronics World iirc. He presents a design that behaves for large signals just as it does for small. I saw some nuclear science signal processing amps years ago that managed something similar with ancillary circuitry alongside the main loop. I'll find the Stochino reference in a bit, or someone can chime in.

EDIT: EW March 1996, Non Slewing Audio Power, according to Doug's website list

[magnoman]

Yes, I agree they are in independent, sorry just used really sloppy wording.

Originally Posted by DF96

Of course, competent design will ensure that the slew rate limit is sufficiently high to handle the required bandwidth but that is a design goal not something guaranteed by the circuit theory.

Nice way of saying what I was sorta thinking about.

-Antonio

[bcarso]

A way to look at it: drive the device under test with a fast step but at very low amplitude, and observe the output. It will usually ring and take some time to settle to some specified error band, but overall will show some characteristic time constant behavior, ideally. Raise the input level and note the change in the shape of the output. The time to settle gets longer and longer. For most amps one reaches a region where there is a nearly linear slope for most of the output trajectory. The slope, usually measured in the middle, is the slew rate.

Demian Martin made the point in another thread that slew rate is only part of the story when it comes to settling time, and fast slewing is no guarantee of fast settling. I just was looking at the datasheet for the many-ways-excellent LME49710 opamp, and there is the rarely-seen specification, for an audio-oriented product, of settling time. It is a relatively mediocre 1.2us to 0.1%, for a 10V step, Av = 1, 100pF load. The settling time includes the slewing time; were it only slewing, the typical specification of 20V/us would entail 500ns. Nothing is said about settling to a tighter final value. And such performance is difficult to measure.

So this would be an amplifier to use with trepidation in a fast data acquisition system. The implications for audio are interesting.

[magnoman]

bcarso

Thanks for the insight.
Your settling time issue reminds me of the 70's articles discussing the "doublets" effect on settling time as well as the non-slowing input stages.

Have you not yet found your Janesick CCD book? I cant help but trip over it.
I find many of your posts nostalgic as I got started with radiation detectors, still have many fets with handwritten noise figures on them (we got a QuanTech analyzer just after our last build)


Thanks
-Antonio

[dvv]

Originally Posted by DF96

The issue arises with single-ended circuits too. The reason I raised it is that it is often omitted from open-loop analysis. You asked "Can we omit some components?"; the true answer is "No, you have to add some components". Fortunately, in many cases it only affects the results by a small amount.

Er, no, actually my point was SHOULD we omit any components.

Simply because I do not omit any, while I notice some manufacturers do. And, as far as I know, there is no standard procedure for measuring it, so everyone is pretty much left to their own methods and scrupules. Remove some compensation items and your open loop bandwidth often more than doubles and starts being impressive, however with the NFB loop back on, that impressive amp might not be impressive at all, and in fact, may not even be stable at all.

Regarding the above, well yes, of course it applies to SE inputs as well. I just happened to take a fully complementary circuit as an example.

[dvv]

Originally Posted by magnoman

dvv

My two cents;
For loop gain you have to include all compensation around the entire loop (including loading effects from those which may be considered as outside the loop, for example the input filter, though this should have a very small effect). Not having the compensation in place would be a starting point from which many different final closed loop performances can be achieved, thus it really tells one nothing of the final transfer function.

Slew rate at least in my view is number related to the overall loop or gain bandwidth of the amp. In order to correlate the two you have to bypass the input filter in order to stimulate the amp with frequencies just beyond its bandwidth. The same would be true for pulsed transient testing at least for purposes of showing stability. You would have to pulse the amplifier past the input filter to see the damping factor, similarly for pulsed load testing you would pulse the amplifier output before any RL but you would leave the RL and speaker load in place.

Thanks
-Antonio

Completely agreed.

[dvv]

Originally Posted by bcarso

A way to look at it: drive the device under test with a fast step but at very low amplitude, and observe the output. It will usually ring and take some time to settle to some specified error band, but overall will show some characteristic time constant behavior, ideally. Raise the input level and note the change in the shape of the output. The time to settle gets longer and longer. For most amps one reaches a region where there is a nearly linear slope for most of the output trajectory. The slope, usually measured in the middle, is the slew rate.

Demian Martin made the point in another thread that slew rate is only part of the story when it comes to settling time, and fast slewing is no guarantee of fast settling. I just was looking at the datasheet for the many-ways-excellent LME49710 opamp, and there is the rarely-seen specification, for an audio-oriented product, of settling time. It is a relatively mediocre 1.2us to 0.1%, for a 10V step, Av = 1, 100pF load. The settling time includes the slewing time; were it only slewing, the typical specification of 20V/us would entail 500ns. Nothing is said about settling to a tighter final value. And such performance is difficult to measure.

So this would be an amplifier to use with trepidation in a fast data acquisition system. The implications for audio are interesting.

If memory serves, we did discuss this, albeit briefly, a short while ago.

The problem of settling time is one with which I am familiar with. Just to remind you, I stated several times that I find AD op amps to sound better than most others and believe this to be, among other things, also a function of their stunningly short settling times. Typically, on their better op amps, times like 90 nS for 0.01% (not the usual 0.1%, for which it is often like 50 nS).

Right or wrong, I assume this is so. Therefore, I have paid particular attention in my current power amp project to settling times, as well as overshoots and ringing. With a wide bandwidth audio amp, one cannot escape any and all ringing, but one can make it very short. Of course, limiting the bandwidth would improve this, but then your wide bandwidth, low phase shift concept goes down the drain.

This is one of the reasons why I tend to push it out to 300 kHz at full nominal power and then install a 200 kHz low pass filter at the input. It tends to reduce ringing and shortens the effective settling time. Although I am not quite sure what exactly am I doing measuring that at 200 kHz.

Perhaps I should lay off H/K for a while.

[bcarso]

Originally Posted by magnoman

bcarso

Thanks for the insight.
Your settling time issue reminds me of the 70's articles discussing the "doublets" effect on settling time as well as the non-slowing input stages.

Have you not yet found your Janesick CCD book? I cant help but trip over it.
I find many of your posts nostalgic as I got started with radiation detectors, still have many fets with handwritten noise figures on them (we got a QuanTech analyzer just after our last build)


Thanks
-Antonio

Yes, that Analog Dialogue article on settling time by Demrow was linked recently, a classic

I probably have Janesick cataloged and located in a numbered banker's box, but not in the computer yet. I don't believe it made the cut for being one of the books on shelves in one of the storage spaces, and it's definitely not here in the apartment.

I miss my outside office. The high bay was full of bookshelves and my front office lined with them, and I almost knew where everything was, even with >10k volumes. However, except for the occasional reference, I rarely read them, since their presence was enough to be reassuring --- I could read them anytime, right? Now I read more, not having as much easy access. Ah, human psychology!

When I was communicating with Walt Jung about the various base-current-recovery circuits, he provided a scan of a page out of a book I absolutely knew I had but couldn't find. This triggered the book cataloging effort, and it turned out the book, actually two closely-related ones by Grebene, were just about as buried as they could be. If I'd found them right off the bat I probably wouldn't have gotten most of the cataloging done

[john curl]

I can't contribute until you guys get 'up to speed' on slew rate, how it can be increased, etc.