Well, as I alluded to in my first post, the exact time and power are important if you want to get a repeatable measurement. This is why I was after a standard test procedure to run through. The audio industry definitely needs a decent test standard.
The output transistors change temperature rapidly when a large amplitude sinewave is applied. Standard thermal compensation mechanisms certainly won't be able to keep up with this rate of change initially.
So it follows that the exact time you take the THD measurent after the test signal is applied is critically important, otherwise everyone will be quoting different figures for the same amplifier.
None of which will be representative for the distortion you get when listening to music. This is precisely why I built an amplifier that does not rely on thermal compensation schemes.
It was all to get rid of 'music power' as a measurement. An amp with say 100W 'music power' would probably never get to 33.3W continuous in the first place.
I wholeheartedly agree that distortion depends on temperature, though there are ways to correct for that. In a composite amp, for example, the controlling amp can be kept at a (reasonably) constant temperature, which means the overall amp wouldn't show a strong temperature dependence on the distortion.
If you design the amp such that the dominant source of distortion is the output stage you will get the strongest temperature dependency in the distortion, so I can see why you'd view the world through that lens.
I also agree that simulators can simulate parameters versus temperature faster than humans can measure them. But when you set the device temperature in the simulator, all device junctions are set to that temperature. That's not what happens in reality, so your simulation is inaccurate already. Also, I'm not aware of any simulator for discrete circuits that accounts for device changes in temperature as function of the applied signal, so you won't simulate the self-heating of the devices. That's pretty important in a Class AB output stage where the devices take turns dissipating power as the output swings about ground.
Don't get me wrong. I swear by simulation. I spent over a decade of my life running simulations 8-10-14 hours a day. It would be impossible to design modern semiconductor devices without simulators. The models we had at TI were remarkably accurate. But we also had an entire group of people whose only task was to ensure that we had good models. That's sadly not the case for most of the models released to the public (some of which are behavioural models rather than device models).
Simulators can be enormously helpful. But they can also be a fantastic way of generating a bunch of garbage data in a hurry.
Tom
Tom,
In the article I referenced above, I document some measurements about self-heating of power devices and the change in bias current.
To my own surprise, the impact on distortion is minimal and can often be ignored.
For device sim models that include self-heating you should talk to Ian Hegglun, he's developed such models.
Jan
In the article I referenced above, I document some measurements about self-heating of power devices and the change in bias current.
To my own surprise, the impact on distortion is minimal and can often be ignored.
For device sim models that include self-heating you should talk to Ian Hegglun, he's developed such models.
Jan
Note that FTC = Federal TRADE Commission. These folks are not engineers. Not all of them anyway. 🙂 They try to ensure some level of fairness in trade, including by limiting false advertising. They had to make some sort of standard for testing to prevent amplifier manufacturers from making outlandish claims. Of course, the marketing departments immediately went to work so manufacturers started specifying RMS Power (according to the FTC standard) and Music Power (what the amp could produce short-term on an ideal power supply).
The 1/3 of rated output power is a pretty good stress test. 1/2 would be better as that's where the dissipation is highest in a Class AB output stage. I'm guessing 1/3 came about because it corresponds roughly to running music through the amp at clipping levels with a 10 dB crest factor.
Amplifier manufacturers whined about the 1/3 operating point because it forced them to use sizeable heat sinks and power supplies, the most expensive components in the amp. This whining reached the levels of hysteria once surround sound became mainstream as amps would have to be able to run at 1/3 the rated power on all channels for an hour.
Frankly, I agree with the manufacturers on this one. There's really no reason a side, rear, atmosphere channel needs to ever deliver that much power. There's hardly any information in those channels. Anyway. So the requirement was then set at 1/8 the rated output power. I don't know how they came up with that number. It could have been anally extracted as far as I know.
I seem to recall that FTC rescinded the amplifier testing standard last year so we're back to the Wild West. That seems appropriate given how everything, including audio, is now sold with stories. "Our founder once had a glorious dream about an amplifier and went about to design one. The amplifier turned out even better than in the dream. It is so awesome that we can't even tell you what the gain of it is. Output power? Oh, it's plentiful. You should buy it. You will love it. Everybody (we've provided a free sample to) loves it." Yes. That was cynicism... 🙂
Tom
Many parameters change depending on conditions. You need to characterize over a range of conditions to really understand anything. One number, no matter how standardized, isn't terribly useful. THD measurements can give me a clue that a design is or isn't "broken", but the THD number in a spec sheet is often near to useless.
That is very true. This site has no shortage of circuits that could work only in simulation. 😉
Ed
To some extent, audio is and will always be a deck measuring contest... 😉
Interesting history behind how some of the 'standards' began and were removed/modified.
I found myself here b/c I have the same questions as the OP.
I suppose as long as I am consistent in my methodology and state any key parameters when posting / documenting results, then all's fair. If the procedure is known, then (hopefully) the results can be repeated.
Why not just enjoy the music. If the system sounds crap, bin it and get something you do enjoy listening to.
We're talking about measuring amplifiers in this thread, not systems. If the system sounds crap it might be something other than the amplifier. We should be in a situation where the amplifier is never the weak link really, as its not hard to ensure 0.01% or lower, better than any transducer.
My take on how THD should be measured is worst-case measure at 100% rated power for >= 1 hour in an ambient temperature of 25C. That'll stop unrealistically high ratings of output power by manufacturers and is very simple and proves a system's thermal performance. While most music loads an amp a lot less load than 100%, its an amplifier, not a music amplifier, and the specs shouldn't be making assumptions about what its used for, that way nonsense lies as is evidenced by all the "music power" ratings that break the laws of physics - that's simply fraud, and always has been.
My take on how THD should be measured is worst-case measure at 100% rated power for >= 1 hour in an ambient temperature of 25C. That'll stop unrealistically high ratings of output power by manufacturers and is very simple and proves a system's thermal performance. While most music loads an amp a lot less load than 100%, its an amplifier, not a music amplifier, and the specs shouldn't be making assumptions about what its used for, that way nonsense lies as is evidenced by all the "music power" ratings that break the laws of physics - that's simply fraud, and always has been.
If we're talking Class AB amps I would argue that the THD should be measured at half the rated output power as that's the worst case for power dissipation in the output stage. That'll be a better test of the thermal performance than running the amp at full output power.
Tom
You are still mixing two tests: checking for adequate cooling and for distortion. An amplifier that produces crossover distortion at low power levels and that has a large heat sink would pass the test, while one without crossover distortion with a smaller heat sink that is perfectly adequate for normal speech and music would not.
By the way, if you want to heat up a class B amplifier driving a resistive load, a square wave with a peak value of half the supply voltage (assuming symmetrical supplies) is worse than any sine wave.
By the way, if you want to heat up a class B amplifier driving a resistive load, a square wave with a peak value of half the supply voltage (assuming symmetrical supplies) is worse than any sine wave.
Two points:
1. Just where are all the system measurements anyway? I never see that. Ever. Since amplifiers and the rest don't operate in a pristine environment, free of ground loops and strange impedances, you'd think that system measurements would be important. OK - maybe only I think that.
2. Worst case is interesting, but perhaps not all that helpful. For most people, the power amplifiers are actually run at between 1 and 10 watts output, more or less. (With obvious crests behind that at times.) Shouldn't those be the actual power levels that are measured? I suspect that the focus on maximum power measurements are due to a legacy of maximum output power being used as a sales and marketing tool, and various companies fudging on this to get better published results. Unless you use a very low power amplifier, operating your power amplifier at maximum power or close to it is not a great thing to do to your hearing. And, shouldn't discrete tone measurements be made frequencies other than 50 Hz, like some places show? It's not hard to get great distortion performance at 50 Hz. 2234 Hz, maybe not so much.
Fretting about at which power level to measure THD is useless if you believe - and many do - that THD numbers are useless for judging audible performance anyway.
Nelson Pass generally measures at 1W because, as he says, 1W is where we live, mostly.
But he also uses efficient speakers, so there.
Don't forget that manufacturers always want numbers that look better than the competition, not necessarily numbers that are meaningfull.
Witness 'max peak music power' numbers.
Jan
Nelson Pass generally measures at 1W because, as he says, 1W is where we live, mostly.
But he also uses efficient speakers, so there.
Don't forget that manufacturers always want numbers that look better than the competition, not necessarily numbers that are meaningfull.
Witness 'max peak music power' numbers.
Jan
From a consumer point of view, those are the same test.
Great if an amp can get amazing results, but it's useless when the amplifier can only handle that for a short period of time.
You see this all the time, that amplifiers will overheat after a while, because they can't dissipate the heat enough, or the PSU is just not good enough to sustain the energy for longer periods of time.
So the stated performance is only good in a lab when there has been time enough between each measurement.
This is a real problem in serious amplifiers for like sound reinforcement and PA.
There are many brands that just fall totally flat on their face when music has been pumped out hours on end.
Obviously this is under extreme conditions, but within the consumer marker you see similar issues when there are budget constraints. (that doesn't mean cheap btw)
Last edited:
Right, we all know that SPL will go with a 10*log(dP) scale.
So 200W sounds a lot more than 100W, but it will only give you 3dB extra.
3dB difference is not something to be so super amazed about.
Even with 400W we only get 6dB more.
With 6dB it's getting there, but still not really so special to reach certain SPL goals.
When you have the space, using efficient speakers makes a lot more sense from a power amplifier perspective.
(given that they will perform the same obviously)