Signal to Noise Measurement Details

[Brickbat]

I want to better understand the signal-to-noise spec for an integrated amplifier. I think I can guess the 'noise' part - it'd just be a weighted measurement of the amplifier's output noise level with no input signal.

But what about the 'signal' part? What reference is used?

For example, I am looking at the spec sheet for my Yamaha A-560. Apparently there was a change made by IHF, as it shows two numbers: 82 dB and 100 dB. What's the difference in the way the measurement is made?

 

[Brickbat]

Hmm. No response. Wrong forum?

Or...Maybe... :scratch2:

Does anyone have a picture of a cat sitting on the IHF Amplifier Signal-to-Noise Test Method?

 

[JBL GUY]

"But what about the 'signal' part? What reference is used?"

The signal is a test signal.

It's level is usually the maximum signal available at the rated distortion of the device under test.

The signal to noise ratio is actually two level measurements, followed by a simple calculation.

First, we measure the output level of the device under test with no input signal.

Then we apply a signal to the device and take another level measurement. Then we divide.

The signal to noise ratio is the ratio of these measurements, expressed in Decibels.

Bandpass filtering and weighting may used with noise measurements, to ensure that the measurements are both appropriate and comparable.

Bandpass filtering limits both the low and high frequency range being measured, defining a measurement bandpass. A typical measurement bandpass for audio devices is 20 hertz to 20 kilohertz.

A weighting filter may also be used, either alone or in conjunction with a bandpass filter. Weighting filters apply one of a number of standard weighting equalization curves to the measurement, typically representing the response of human hearing.

Any filtering used in a noise measurement must be stated as a component of the measurement conditions.

 

[ferninando]

I have a copy of the IHF test methods. I have to see what it says.

edit: well its a 1966 paper.
no sig/noise measurement.
hum and noise only.
lotsa engineering speek relating to how to do it.

 

[Jon_Logan]

Your question about, "what reference level?", is a good one.

It is important too. Consult the calibration procedure of the particular equipment.

In general, you input a certain frequency tone signal at a certain amplitude. You raise the volume control to a certain point until a particular output level is reached. This level becomes your 0dBr level. "r" for Reference level.
You kill the input signal and measure the leftover noise. You can either filter the noise to comply with particular standards or not.

Different manufacturers will have different procedures their own equipment, so it's be to look at their manual and see how they achieved their spec.

 

[JBL GUY]

With regards to signal to noise ratio, the actual level of the noise floor may also be an issue.

We normally do not listen to our gear at it's maximum level, so even though the gear might have a respectable S/N ratio based on it's maximum output, if the noise floor is high enough, noise may be heard during quiet passages of the music.

This may be more noticeable when using very sensitive speakers or headphones.

You can see the noise floor in my example. Also note that the amplitude of the noise is not the same at all frequencies.

It is important that the measurement results include all test parameters.

 

[Brickbat]

It's level is usually the maximum signal available at the rated distortion of the device under test.

Thanks - that was one of my guesses. I was just looking for some historical background on the reference level that IHF specified.

WRT the Yamaha spec I attached, maybe the 100 dB measurement was A-weighted, and the 82 dB number (New IHF) not weighted?

It is important that the measurement results include all test parameters

Ha. Tell me about it.

 

[JBL GUY]

Thanks - that was one of my guesses. I was just looking for some historical background on the reference level that IHF specified.

WRT the Yamaha spec I attached, maybe the 100 dB measurement was A-weighted, and the 82 dB number (New IHF) not weighted?



Ha. Tell me about it.

IIRC Hmmm...


There is A-weighting as used in testing, set out in the very old IHF standard: THE INSTITUTE OF HIGH FIDELITY, INC. IHF Standard Methods of Measurement for Audio Amplifiers, IHF-A-202 (1966 then 1978).

It went through various revisions before being superseded by EIA-RS-490 (1981), which in turn has been superseded by CEA-490-A.

Take a look at the the blue curve (the A-Weighting curve, possibly used for testing your amplifier). Note that is down by almost 10 dB at 15 Khz and almost 20 dB at 120 Hz (power supply ripple frequency).

The use of A-Weighing could result in improved numbers for signal to noise ratio measurement, especially if an amplifier had appreciable levels of noise at either or both ends of the audio spectrum. However manufactures claimed that if more closely followed human hearing characteristics.

Back in the day, manufactures played fast and loose with specifications, this is what lead to the much more stringent so called FTC (Federal Trade Commission) testing standards.

Note only could various weightings be applied, the maximum level of the test signal could be chosen to drive the amplifier to higher output levels, which in turn resulted in better signal to noise ratio measurements, even if at these higher signal levels the distortion increased to 5% or even 10% or more.

Note that in your first set of specifications, the input signal levels are referenced along with specifying that the input be shorted for the noise measurement.

 

[ConradH]

I've no idea what standards are used today but A-weighting tries to put the most weight in the frequencies most annoying to the ear. It makes the numbers look far better by reducing the noise components you don't hear as well.

The reference (0 dB) level might be full output, or it might be some standard level like 1 Vrms or 1 Watt into 8 ohms. The higher the reference level the better the numbers look, but consider a 50,000 watt amplifier with a -80 dB signal to noise level. It won't be very good listening at low levels! A 5 watt amplifier with -80 dB s/n ratio will have a lower noise level. Thus a fixed level makes some sense, since you don't automatically listen to high power amps all that much louder.

+1 about filters. Any talk of noise has to include the bandwidth to mean anything.

 

[Freezer7Pro]

If you look at the manual for the Yamaha A-560, you'll also find a "Residual noise" spec (I love Yamaha for including these) of 210 µV A-weighted. If you crunch the numbers and calculate the ratio of it to the full output power (60 W/ch, or 21,91 VRMS), you get 100,368 dB. That's clearly how they've derived the 100 dB SNR.

Since they also specify the input level (0,5 V), you can calculate what voltage gain they used (32,83 dB) and from that figure out where the volume knob was turned when they did the measurement, and replicate it if you have the gear!

I love Yamaha's specs.

 

[Brickbat]

If you look at the manual for the Yamaha A-560, you'll also find a "Residual noise" spec (I love Yamaha for including these) of 210 µV A-weighted....

Yeah. I missed that - Thanks for pointing that out.

It all makes sense now...

 

[Brickbat]

...The reference (0 dB) level might be full output, or it might be some standard level like 1 Vrms or 1 Watt into 8 ohms....

Apparently the ref level was full rated power.

And given that, yes, it's good to understand that a higher power amp will have a 'louder' background noise level than a lower power amp with identical s/n ratio.

 

[pdm4606]

signal to noise

More of a crapshoot than anything. The hi-fi institute attempted to control specs. voluntarily but the manufacturers wouldn't hear of it. They wanted everything their way. Then the trade commission stepped in and in some instances confused things even more.
The situation now is somewhat livable but still according to some manufacturers whims.
The specs are not related to anything but what is best for print.
How can you get a wallymart item with the same specs as high end equipment?

Paul

 

[jaguar06]

Apparently the ref level was full rated power.

And given that, yes, it's good to understand that a higher power amp will have a 'louder' background noise level than a lower power amp with identical s/n ratio.

Valid point about "identical SNR". But thankfully we don't necessarily get identical SNRs even in the same family. It intrigues me that the A-560's residual noise is exactly 10 times the MX-1000, 210µV for the 560 versus 21µV for the 1000, at the same time SNR increases from the 560's 100 to 128 (same weighting), and power from 60W for the 560 to 260W for the 1000, both at @ 8 Ohms.

My point is that a casual observer might read your post missing that "with identical s/n ratio" and think "more power=more noise", so I want to underscore that your case of more noise would be *only* if the SNR's were the same, for clarity.

Fortunately there's ways to lower that residual noise, increase SNR so that scaling up power doesn't mean we'll necessarily increase noise.

 

[Brickbat]

...It intrigues me that the A-560's residual noise is exactly 10 times the MX-1000, 210µV for the 560 versus 21µV for the 1000, at the same time SNR increases from the 560's 100 to 128 (same weighting), and power from 60W for the 560 to 260W for the 1000, both at @ 8 Ohms....

So, 28 dB better SNR on the MX-1000.

And, let's see here... 20*log($900/$21.50) = 32 dB more expensive...

So, just about what you'd expect, eh?

 

[zenith2134]

I remain wary of s-n-r measurements as they pertain to turntables. Often times, an ad campaign may be misinformed or intentionally mask the truth. To combine rumble instead of listing it on its own against a reference bothers me.

Also, it has been my experience that loudspeaker input-sensivity can confuse things more, since a higher noise floor may not be as audible on a low-efficiency speaker design.

 

[Brickbat]

...Also, it has been my experience that loudspeaker input-sensivity can confuse things more, since a higher noise floor may not be as audible on a low-efficiency speaker design.

Low speaker efficiency decreases the audible noise floor but, as we've discussed it here, doesn't affect the SNR. Seems simple, am I missing something?

 

[callab_gte_net]

I have a copy of the IHF test methods. I have to see what it says.

edit: well its a 1966 paper.
no sig/noise measurement.
hum and noise only.
lotsa engineering speek relating to how to do it.

Do you or does anyone else have a copy of IHF-A-201 (1966) or IHF-A-202 (1978) "Standard Methods of Measurement for Audio Amplifiers" that they could provide me with?
(I am in the process of writing a calibration procedure for 1970's-1980's era power amplifiers and would like to use these as a reference.)

I am not interested in any of the standards that later replaced these earlier versions.

Any help would be greatly appreciated!

My current e-mail address is callab @ frontier dot com

Ed