QuoteThis is not an argue R.G. I stated my opinions accompanied with what I consider as pretty valid arguments. Why persist that there is something wrong with them?

I agree - this is not an argument. I have different opinions, and I consider the exploration of them to more than one ply of depth to be a discussion. No argument was intended.

Quote
Guitar power amplifiers do not have to have a dead flat frequency response but can be made to react to speaker impedance similarly as tube amplifiers do.

That's true in general; however, this forum was started with high consideration for the LM3886 amps, and that continues to be a major consideration. All of the "chip amp" style amps do have ruler-flat response unless you do something to change that, as they are all high-feedback, low output impedance amps unless you go to some lengths to make that not true.

QuoteYou can even make LM3886 exhibit the same character if you want to: It's all about the feedback configuration.

This being one way to make that not true.

QuoteSome amplifier models even have active tone control circuits in the feedback loop of the power amplifier.

And this being another. Note that this method still leaves a low impedance output with a tight "grip" on the speaker.

QuoteA reason why to make PA circuit's frequency response flat is as valid as the reason why not to. It's all about what the designer wishes to do. If the frequency response of PA is not flat then it's tough luck for limiters that detect only the preamplifier's signal level.

That's another version of what I said. We agree there.

QuoteAnyway, a limiter that has a detector hooked to PA output and that attenuates the signal fed to power amp: Is that a preamp-based limiter or power amp-based... Well, we have a semantics issue here and it's quite hard to determine where to draw a line in this. It is easier to make a division between limiters that control the power amplifier's gain or power amplifier's input signal amplitude.

I may not have been clear.

You can do the limiting entirely feed-forward by limiting at the preamp output or feed-back by watching the output of the power amp. There are multiple methods for each one.

Power amps are inherently high power opamps these days (neglecting tube amps, since this is a solid state guitar amp forum), and since they are, unless you are forcing them into some nonlinearity, the output follows the input as dictated by the feedback network. Whether you use a low power opamp or a high power opamp to do your limiting makes little difference if you make the assumption that you are not going to let the amp doing the limiting go into clipping - which is the goal you're after in the first place.

That being the case, it is a lower-risk technique to limit before you go into a high power stage, and to do the limiting in a way that whatever you feed the power amp will never force the power amp into clipping under any condition. The down side of doing your limiting in a low power stage ahead of the power amp itself is that (1) you have to use an additional US$0.15 opamp to do it instead of the power amp you would otherwise use (2) you have to do your power limiting at lower voltage, which is more of a design problem than it sounds like, and (3) you may give up a few watts of potential power out in the real power amp by designing not to drive it to the very edge of clipping.

It could be a problem to some people, but I don't consider 1 to be a problem, especially since it removes the risk of getting something wrong in the power amp stage and blowing it. Power amp chips are not as stable as general opamps, and high power oscillation will kill them. The LM3886, for instance, is not stable for gains under five. The TL072 is unity gain stable. You can also put in an RF decoupler and keep any mistakes from the limiting stage from getting into the power amp. 

2 is an issue when you are depending on using semiconductor junctions as references for where you do your limiting. The smaller voltages in an opamp limiter as opposed to a power amp limiter make this a bit trickier and require more ability on the part of the designer.

3 is something you've already signed up for by doing soft limiting. The logarithmic response of human hearing make this almost inaudible anyway.

QuoteVaporizing the power amp is a bit far-fetched but I guess it's valid to assume something like that would happen if someone doesn't quite know what he/she is doing.

Yep. I can promise you that it can happen. I've done it. I guess I'm one of those people who didn't know what they were doing at one time in the past. Not all of us can always know what they're doing.

QuoteIn my opinion, having the detector at the amplifier output is more accurate way of guessing when PA is approaching clipping levels than determing a certain preamp signal voltage level threshold where the limiter kicks in - at least in cases when power amplifier's gain is dictated by impedance (valid assumption) or when supply sag has to be considered (another valid assumption).

It is in my opinion too. But my opinion also is that there are other design reasons for making not having to make this determination accurately. If you don't need to make a razor-sharp determination of where something clips, so much the better.

QuoteIf power amp has dead flat frequency response and doesn't exhibit a notable sag then there is really no need to make things any harder than needed. In practice, many amplifiers work fine even without limiters. ...But they also would work better with them - especially with all kinds of them.

I believe that. Of course, I'm after soft-clipping for musical sound, not for protection of the power amp or speakers. IMHO, SS amps should never be let clip. Period. It's too nasty sounding. It's the sound I'm after. I'd like to control how the clipping happens so it's musical. I think this is easier to do in a low power stage than a high power stage. It's not the only way, just the one I like.

QuoteKeeping mind open for alternative methods is always beneficial.

I could not agree with you more. I think that keeping an open mind is essential, maybe crucial, to improving one's design abilities. 

QuoteWhat will happen if the voltage gain of the power amplifier is not consistent throughout the audio band - this is true with some specific feedback configurations. It is quite hard to build a limiter that has a compensation for the odd frequency response in the detector; it is much easier to build a limiter that has a detector monitoring the difference between output and rail voltage. For example, limiter kicks in when output has 1V difference to Vrail. This way the operation is not tied to constant signal amplitude threshold and can even tolerate supply sag etc.

As I noted, SS power amps are notorious for being accurate. With the LM3886 particularly, the voltage gain of the power amplifier is dead flat across the audio range unless you work at making it not flat. Non-flatness of the power amp voltage gain is simply not an issue with modern power amps. They're easier to make flat than not-flat.

As to limiters with odd frequency responses - well, don't do that. It is far easier to control limiter response at a line-level stage than by tinkering with a power amp. You have the same tools available in terms of circuit responses and components, and the advantage that a mistake will not vaporize your power stage. You have the same problems and solutions available, and lower downside risk of failure.

And for limiters kicking in with 1V to the rail - don't do that. Soft limiting doesn't happen right up at the power rail. That's hard clipping. Soft limiting starts kicking in maybe 20db down from max power and gradually increases as power approaches max. A good soft limiter starts clipping ever-so-gradually way down at lower powers. You may want a last-resort limiter to keep the power amp from ever clipping that hard limits at a fixed voltage down from the rail, but it is a last resort. It's not very musical.

Just curious - since a SS amp is great at just doing what it's told at anywhere less than clipping, why wouldn't you do an entire soft clipping circuit ahead of the power amp and then just feed that to the accurate power amp?

The soft clipper only has to soft clip the input at a voltage less than would drive the power amp to clipping. As long as that happens, the power amp is completely unaware what's happening to it.

As long as a SS amp never clips, it never sounds ugly, just accurate. It's possible to follow a soft clipped signal accurately.

Let's play transformer sizing.

The LM3886 datasheet says, as joecool85 notes, that 96V is the absolute maximum. That is, if you apply 96.01V, you must assume that the LM3886 will immediately transform itself into a Darkness Emitting Diode, or DED.

So let's see what that voltage is. DC and RMS AC are related by the square root of two, so for 96V, you have 96/1.414 = 67.89Vct on the transformer.

But wait - that's a maximum, which presumably happens at the very peak of your AC line power, which also varies. If you get 120Vac +10%, that's 132Vac. So we need to know what a 132Vac to 67.89Vac transformer is when it's at home at 120Vac. That's 61.72Vac.

But wait - that ignores loading. Transformers are rated with "regulation" which is really the no-load to full-load sag from losses. In transformers of the 100VA class, 5% "regulation" is pretty normal, so a transformer rated at 58.78Vac with 5% regulation is actually going to be 61.72Vac at no load.

We're finally down to the absolute maximum transformer voltage. I would not use anything over 58Vct on one of these and expect them to survive line surges.

What is the minimum?

I'd go with what National says if they said it clearly. But they don't. It's hidden. This stuff has to be gleaned from about three or four of their performance charts.

They say that the chip will do 80+ watts into 8 ohms with a +/-40V supply, and into 4 ohms with a +/-30V supply. Can you **guarantee** that you will never connect a 4 ohm load to it? If so, you could use +/-40V. If not, you might use +/-30V. There is more calculation behind this, but those are fair maximums because National suggests a 1 C/W heatsink to do those powers and voltage.  1 C/W heatsinks are HARD and expensive.

For longevity, it make a lot of sense to not dissipate all of the heat in one package. These things are CHEAP compared to heatsinks. So use a lower voltage, or use a cheaper heatsink and use two or three LM3886's in parallel.

So if you have to have the most, most power you can get, use a 56Vct transformer (that being a relatively standard value)  and two LM3886's in parallel for 8 ohms. If  you can stand the difference between 68W and 45-50W, back down to a 48Vct transformer which is easy to find. If you're going to use this for 4 ohms loads, back down to 44Vct transformers.

How much power? Music power varies. If you use a transformer rated at 1.8 times the amplifier rated RMS power, it can run a sine wave at that power forever. If you use a transformer rated at 1.0 times the amplifier rated RMS power, it can run music with a 50% crest to average power factor forever. Most music has between a 0.1 to 0.25 crest factor, so you're pretty safe with a transformer rating in VA equal to the amplifier rated power.

When I do LM3886's, I look for 44Vct or 48Vct transformers rated at 2A. If I run them into 4 ohms, I will use two or three LM3886's in parallel. Get over 56Vct and a power surge will kill your chips one day.

Mother Nature supplies us some hints if we look up the numbers.

Speakers come with ratings for how loud they are per watt. The number is "SPL/1W/1M" which translates to "sound pressure level with one watt drive at one meter distant from the speaker." That number varies - a lot.

If your speakers are rated at 86db 1W/1M then changing to a speaker rated at 89db same conditions is the exact equivalent of doubling your amplifier power. You get twice the delivered acoustic power. Some speakers are rated at high 90's, even low 100's. Electrovoice (EV) in particular is known for high efficiency speakers.

Quoting only your amplifier power and not the speaker efficiency is like saying "my car burns two gallons of gas per hour". It's not all that useful without knowing whether your car is a high efficiency hybrid or a dump truck. Both very useful, but for different things. Either way, you can't tell how much of what you consider good stuff you get without knowing more.

A favorite dodge of mine is the modularized amp/speaker. Got a 100W speaker? Put a 100W amp on it. Need 200W? replicate the amp AND speaker, rather than making a single 200W amp. a 100W amp is twice as easy to get running as a 200W amp and ten times as easy to get running as a 500W amp. And you're already trained on how to make a 100W speaker/amp, right?

Then build as many as you need. I'd rather make ten 100W chip amps as one 1kW amp.

Albert Einstein said that everything should be as simple as possible - but no simpler.

I don't think any designer makes amplifiers more complicated because it's fun. I suspect that the more complex designs are done because over time the complexities have proven their worth. If I had to summarize what you found, that would be it.

Diffamps produce more predictable inputs. Current source loaded voltage amps have higher gain, more predictable gain, and wider swing to the supplies than bootstrap loaded VAs. Dual diffamp plus complementary mirror-image VA designs produce symmetrical slew rates and predictable drive to output stages, as well as clean, predictable recovery from overload and clipping.

It is the essence of good engineering to produce predictable, reliable results from highly variable, unreliable parts. This was an important part of circuit design teaching at one time in EE curricula. It may not be any more. I guess I'm not surprised at all that a modern, modestly complicated circuit should produce measurably better results than a more rough-and-ready simpler circuit.

The game is different for solid state guitar amps than for solid state hifi amps. SS amps readily produce extremely accurate reproductions of what their inputs say to do unless and until they clip. With as much cheap, clean power available as SS amps provide, it's easy to make much larger clean output power amps than with tubes. The trick is then to feed them an input that makes them reproduce very accurately the sound of a tube amp. It's gotta be possible - PA systems do it all the time, yes?

Having once gone to the trouble of splitting off highs in a (presumably) well behaved crossover, why wouldn't you keep the highs separate from there all the way to the speaker? Bi-amping has a long and well documented history. The limiting on the highs could be done by a limiter or just by using a correspondingly smaller amplifier for the mid/tweeter drivers.

QuoteI've got to come up with a decent chassis which I may have to actually buy this time.

Get yourself a 2RU rack cabinet. Make the wooden box house a single 12 and the rack cabinet. Place electronics in the rack cabinet.

2U cabs can be had for $50 from Middle Atlantic as I remember.

Do yourself a favor - regulate your preamp voltages, and separage preamp and signal grounds from power amp and power grounds. That's more effective than the biggest capacitors you can buy for filtering the preamp supply.

ALL single ended circuits should be assumed to be intolerant of power supply ripple and ground hum until proven otherwise. It's the nature of the beast.

Yeah, I'll toss in a thing or two as appropriate. Rick knows a lot of my philosophy already.

I'll rephrase myself here, as the forum is new.

Power amps are best thought of as a power *supply* that lets some of the power out under special, carefully controlled circumstances. The power amp itself is really just a highly specialized wart on the power supply. In most cases, the power amp circuits proper are a trivial part of the cost of the unit.

The power supply needs to be the FIRST order of business in designing a power amp. Going to get 50W into 8 ohms? You need to know how to tell what power supply you need before you dig into what power amp or chip you'll use.

After the power supply conversion stages, voltages and currents, etc. you need to consider thermal matters. Again, this is substantially independent of the power amp circuit itself. If you're putting out 50W into 8 ohms with a class AB solid state amp, the power needed from the power supply is 50/.74 = 67.6W worth of DC coming out of the power supply at peak power. But that's not where the peak dissipation on the output devices happens. That peak comes at  forty (something) percent of the max power out. The heat has to be dissipated by heatsinks so the output devices are kept below the boiling point of doped silicon. Otherwise, you have just bought a new set of output transistors - or a new chip amp. The heat generated in the power supply itself also gets released inside the box. Where does that go?

After you have a good power supply design and a good thermal design, run wild with whatever power amp circuit you like.

I have some thoughts on the design of a preamp. Some of these have been stated here, some haven't.

1. Use a high input impedance. 1M is a minimum.
2. Protect your inputs. What happens if someone hooks up 120Vac from the wall socket to the input? The common reply to this is "No one would be stupid enough to do that." Second reply "Well, they deserve what they get if they do." The correct response is "Cool! How can we do that?" because someone can, and if they can, at some point they will. Or they'll hook up the output of their 1000W car audio amp to it; same thing. Protect your inputs.
3. Know what you're doing with feedback - and without feedback. Feedback is a great tool, but it's often misapplied. High open loop feedback gain stages clip sharp as a razor when they clip. So either use high feedback stages where you *want* razor sharp clipping or where it will (almost) never clip. Use no-feedback or low-feedback stages where you want to hear the device itself. In general, soft clipping is the characteristic of a single device, or a high feedback stage emulating a response curve and never clipping.
4. Know what your power amp wants to see in terms of inputs. Give it that. It is rare for a solid state power amp to clip gracefully. Remember all those quips about getting a pig to sing?