Amp power and SNR measurements

[RayDunzl]

If you have an amp rated at 50W and one at 500W, measuring SNR at 5W will almost always favor the 50W amp, as the voltage/current gain applied to the noise signal coming out of the preamp stage will almost always be fixed, and the 500W amp is applying 10X more amplification than the 50W amp.

10 times power increase occurs with 3.162 times voltage increase if the load is the same.
.

 

[EngineerNate]

Actually if you take an amplifier, beef-up the MOSFETs & heatsinks, and increase the power supply voltage (without otherwise changing the gain or anything in the circuit) and the noise will be about the same if not exactly the same.

This is the kind of thing that's interesting here! It would seem that within a given topology family, more gain = more noise amplified as well, but what impact does more parallel devices or just device "size" in general have?

There are some Accuphase, Halcro and Boulder amps that are class AB, low noise and tons of power. But cost dozens of not hundreds of thousands of dollars.

Hypex and Purifi are class D and therefore smaller and sane prices.

It'd be interesting to study the topologies of those amps to see how they drive noise down.

 

[Ropeburn]

It would be interesting to start testing SNR at some fixed percentage of rated output capability rather than at fixed outout power levels. 500W full range would be instant deaf for most of us in practice, but academically it'd be interesting to see how quiet big amps can be on a "level" playing field.

Amir already routinely measures under full power, isn't that enough together with the 5W figure? You can definitely see how quiet it is under full load.

 

[EngineerNate]

10 times power increase occurs with 3.162 times voltage increase if the load is the same.
.

You talking watts or dB?

 

[EngineerNate]

Amir already routinely measures under full power, isn't that enough together with the 5W figure? You can definitely see how quiet it is under full load.

View attachment 486919

Yes thank you, I had somehow missed this in the review that triggered this thought.

 

[EngineerNate]

There are some Accuphase, Halcro and Boulder amps that are class AB, low noise and tons of power. But cost dozens of not hundreds of thousands of dollars.

Hypex and Purifi are class D and therefore smaller and sane prices.

Are the numbers measured or based on specs?

 

[RayDunzl]

You talking watts or dB?

You were talking power in Watts.

To which I replied a 3.162 multiplication of voltage yields a multiplication of 10 in power.

 

[babadono]

10 times power increase occurs with 3.162 times voltage increase if the load is the same.
.

Oh sheeeet math again

 

[EngineerNate]

You were talking power in Watts.

To which I replied a 3.162 multiplication of voltage yields a multiplication of 10 in power.

Ah right sorry, V=Sqrt(P*R) so for fixed ohms you get 3.162 = sqrt(10) for a 10X power increase.

And since V=I*R you get the same for current

 

[LSPhil]

My amplifier is 60 W, and that’s plenty for me. 60 W gives about +17 dB, added to the 80 dB driver sensitivity, for a total of roughly 97 dB, which is really sufficient. Noise in the bass range doesn’t matter anyway. Measuring at 5 W is perfectly fine, that’s about 1/8 of 40 W, so everything checks out.

Honestly, I wonder why anyone would need 1 kW for speakers with a sensitivity of 93 dB/1 W at home. That’s over +30 dB, or about 123 dB in the living room. I think there are more important things in life.

Besides, if your amplifier can double as a hair dryer, you’re probably overdoing it!

 

[kyuu]

Honestly, I wonder why anyone would need 1 kW for speakers with a sensitivity of 93 dB/1 W at home. That’s over +30 dB, or about 123 dB in the living room. I think there are more important things in life.

Don't forget that there's a widespread attitude that "power is cheap" so people get speakers with sensitivities more in the low 80s.

 

[Ropeburn]

My amplifier is 60 W, and that’s plenty for me. 60 W gives about +17 dB, added to the 80 dB driver sensitivity, for a total of roughly 97 dB, which is really sufficient. Noise in the bass range doesn’t matter anyway. Measuring at 5 W is perfectly fine, that’s about 1/8 of 40 W, so everything checks out.

Honestly, I wonder why anyone would need 1 kW for speakers with a sensitivity of 93 dB/1 W at home. That’s over +30 dB, or about 123 dB in the living room. I think there are more important things in life.

Besides, if your amplifier can double as a hair dryer, you’re probably overdoing it!

Oh, I definitely used 2x200W here occasionally, if only just for peaks (important distinction to average power). New years party and such, everyone drunk and suddenly the DAC volume straight to poweramp showed 99.

Speakers are 91dB efficiency, so that means if they were 85 or so instead, we'd have needed 2x1000W for the same volume.

Two things are generally true at the same time:

1) We overestimate the power needed for normal to moderately loud (fun) volume. It's much less than you'd intuitively expect.

2) We underestimate how quickly the power needed escalates when conditions change only so much. Double volume, double room size, -6dB speaker efficiency - suddenly you went from 50 to 500W. It's much more than you'd intuitively expect.

It's simply because our intuition fails completely when it comes to exponential relations. That's why it's always a good idea to do the math.

 

[Matias]

Don't forget that there's a widespread attitude that "power is cheap" so people get speakers with sensitivities more in the low 80s.

I chose this willingly to have more bass extension. Right call.

 

[wwenze]

Amir already shows both noise at a fixed power output (5W) and at maximum output. The noise at an arbitrary output can be obtained using the two. The amount of noise in uV or nV at 5W is the same as amount of noise at 50W. The "fair fight" you mention is already being done with SNR measurement at maximum output.

Why can't you get the information you want from plots like this one which are in (almost) all amplifier reviews?

This is SINAD vs power curve. It includes both noise and distortion components. It does not show only noise vs power.

When the graph continuously goes down from left to right, that is the region where noise bottlenecks performance. When it starts moving up, that's where distortion bottlenecks performance. Or as the author describes the red curve, distortion sets in at 0.1W.

But yea it's true, as long as the reader targets the region where noise dominates (which should be the case for all good amps), noise comparison can be made between two amps.

That's fair! I guess it just seems like there's some expectation that big amps will perform the same as smaller ones, when that's generally not possible from a physics standpoint, if you're starting from the same input sensitivity. Front loading the 5w measurement (and using in the ranking charts) and burying commentary on the full power performance would seem to reinforce this and give an unnecessarily negative impression of the amp under test.

It may very well be the amp sucks for doing 5W, but it might be a great amp for doing something else.

I want to say that it should be common knowledge (or sense) that bigger amps will be more noisy and draw more power for the reasons that you said. I mean we also know the difference between a class AB headphone amp and speaker amp. 150mW amp vs 15W amp is as big a difference as a 15W vs 1500W amp.

But at the same time, there is an undeservedly long thread where people complain that speaker amps with high max output consume more idle power than speaker amps with low max output...

But hey that's a good thing, it shows this website is having traffic from all walks of life

Ah right sorry, V=Sqrt(P*R) so for fixed ohms you get 3.162 = sqrt(10) for a 10X power increase.

And since V=I*R you get the same for current

The dB in audio is almost always unarguably referring to power. While dB is really just a ratio of two quantities which can be any unit of measurement, in the audio context it is always with respect to power. Even when we want to compare two voltages, for example 2V vs 1V, the difference is 6dB. But compare two powers, 2W vs 1W, the difference is 3dB. Yea audio is weird.

 

[JP]

Good to see you around, @EngineerNate.

 

[DVDdoug]

in the audio context it is always with respect to power.

No... It's a different way of calculating the same thing...

When you double the voltage (into the same load) you also double the current and that's 4 times the power. It's +6dB either way you calculate it.

If you cut the impedance in half (with the same voltage) you get twice the current so double the power (+3dB).

BTW - Digital dB levels (dBFS) correspond to amplitude which is converted to voltage by the DAC so doubling the digital level is also +6dB. Digital amplification or attenuation is done by multiplication with each sample multiplied by the same factor. (If the factor is less than 1 you get attenuation.)

 

[JP]

@DVDdoug can I beg of you to quote properly by highlighting and selecting "reply" so that the attribution link is there? This is the third post of yours that I've read in the past hour where I've had to search for what you're partially quoting to understand the context. The added bonus is then the person you're quoting will get an alert that you quoted them.

 

[amirm]

If you have an amp rated at 50W and one at 500W, measuring SNR at 5W will almost always favor the 50W amp, as the voltage/current gain applied to the noise signal coming out of the preamp stage will almost always be fixed, and the 500W amp is applying 10X more amplification than the 50W amp.

Most answers have been given. I will add that to the extent the gain is adjustable, I set it to same value to others, usually 25 dB. In those cases, there is no multiplication of input but rather, higher input voltage is required to reach max power.

At the end of the day with no standards in amplifiers as far as gain, it it is difficult to have a test scenario that is 100% proper. Fixed gain amps of say 12 to 15 dB, will certainly have an advantage in noise department.

 

[TurtlePaul]

You assume that a 500 watt per channel amp will be used in PA system. However, there are plenty of examples where people are using 500 wpc amps in 2-channel living room systems for the sake of 'overhead' in available power. Both the 5W and full power numbers are relevant. If you have the 5W number, given how SNR is calculated, it is pretty easy to deduce the SNR at any arbitrary power level and then flip to the power chart to see if that level is obtainable on a given amp.

 

[Chrispy]

Had a thought about amplifier measurements reading some of the PA amp reviews.

If you have an amp rated at 50W and one at 500W, measuring SNR at 5W will almost always favor the 50W amp, as the voltage/current gain applied to the noise signal coming out of the preamp stage will almost always be fixed, and the 500W amp is applying 10X more amplification than the 50W amp.

Attenuation happens before/at the input stage, so the self noise of the input stage is fixed. That self noise will get amplified along with the signal, so when we attenuate the signal to match the outputs of two drastically different power level amps, unless something is broken the less powerful amp should always win.

That's not taking into account the fact that to get high power levels often means paralleled output devices, which inherently increases noise as well.

It would be interesting to start testing SNR at some fixed percentage of rated output capability rather than at fixed outout power levels. 500W full range would be instant deaf for most of us in practice, but academically it'd be interesting to see how quiet big amps can be on a "level" playing field.

Cheers,
Nate

You're judging by wattage rating alone? Weird.