Is true-RMS meter required for audio signals?

[Nigel Goodwin]

I suppose that average multimeters are just calibrated for sine wave, and not for a real average.

They are - historically all the AC meter ranges had was a single rectifier diode, and a resistor to calibrate the reading to be the same as the RMS value.

It's pretty much still the same today, although perhaps a little more 'hitech'

 

[KeepItSimpleStupid]

I know what true-rms is, but I wonder if it is required to measure a real audio signal.
With "real" I mean a mix of many frequencies, not just a test pure sine wave.
I suppose an average meter should work fine, but I've got no true-rms to compare.
Would an average and true-rms meters differ in readings?

Thanks!

My question is what do you want to measure?

p-p generally matters and so does distortion, so a scope there would be a better bet,

power measurements are usually done with a load and a sine wave test signal, so frequency response matters.

 

[Reloadron]

<Off Topic A Little>
Years ago as AC voltmeters advanced one of the top of the line meters was the HP (Hewlette Packard) 400H VTVM (Vaccum Tube Volt Meter) and another very popular meter was the HP 3400A, the latter 3400A actually a solid state true RMS meter. When meters like these were manufactured the manufacturers actually published detailed manuals which included not only complete schematics but detailed theory of operation on their meters. The HP 400H was touted as an average responding RMS indicating VTVM. The meter responded to the average value of a sine wave and was scale calibrated to indicate the RMS value and thus the term average responding RMS indication. The above links are to both meter manuals. Looking at the HP 400H in a nutshell the signal comes in through V1 a cathode follower and is amplified by V2, V3, V4 and V5 and arrives, as an amplified AC signal to the meter bridge circuit consisting of two diodes and two capacitors. This meter also used a degenerative or negative feedback circuit to maintain stability. During its time the HP 400H was actually quite the meter in the industry. For the curious I suggest reading the manual theory of operation and following the block diagram as well as the schematic.

The HP 3400A was a much different animal in that it was a true RMS meter meaning it was RMS responding, it responded to the RMS value and was RMS indicating. The RMS value of a sine wave or about any wave form is based on the DC heating effect. The term “RMS” stands for “Root-Mean-Squared”. Most books define this as the “amount of AC power that produces the same heating effect as an equivalent DC power”, or something similar along these lines, but an RMS value is more than just that. This is a good link on that subject matter. Now keeping in mind the 1960s technology used in the HP 3400A read the detailed theory of operation with a focus on the "chopper" circuit, thermocouples used for AC heating effect, and the demodulator circuits.

For their time both of these meters were considered high end intermediate level lab instruments. The theory of operation of each can be some interesting reading.

As to measuring audio signals there are dozens of small, inexpensive data acquisition systems which can be used to not only display audio data but record the data for later review.

</ Off Topic>

Ron

 

[MrAl]

Sorry MrAl, but you need to stop smoking whatever illegal substance you appear to be using

Normal multimeters read the 'average' calibrated solely for a sinewave - on the assumption that it's going to be used for mains measurements - as it's designed to be.

Obviously measuring a squarewave will be massively incorrect, as it's not calibrated for that.

Hi,

Not sure what you are talking about here because meters do not have to be calibrated to measure average for a sinewave they measure average for any wave naturally. It's only the RMS value they have to be calibrated for based on the average reading which they can do naturally either electronically or mechanically. That's except for the true RMS meters of course that take many samples and compute from that.

However, i wasnt talking about any particular meter anyway until the last paragraph. Before that i was talking theory only. Theory tells us what the actual average is and what the actual RMS value is of any wave we choose, no matter how complex. We dont have to calibrate anything. The reason for doing this is to compare average with RMS and show there there is always a difference.
The non true RMS meters try to make up for this by applying a calibration factor that allows the meter to read out in RMS rather than average, because that's the more important measurement.

The fact is though that this calibration method assumes that the wave is a sine wave, and so it takes the average and multiplies it by a factor and then displays the result, and that result is RMS as long as the input is a sine wave. Unfortunately, not all meters do this either. Some will take the peak reading and apply a different factor to that and display that new value, and assume that is the correct RMS value. This could be way off for various waves. One way of looking at this is that the crest factor for each wave shape is different.

In the end however my conclusion was that we should never assume that the average will be the same as the RMS value for any wave.

 

[MrAl]

any idea why the two images above disagree?

I suppose that average multimeters are just calibrated for sine wave, and not for a real average. Maybe that explains what MrAl explained.

Hi,

Not sure what images you are referring to.

Some meters read average and convert that into RMS, while other meters measure peak and convert that to RMS. Only the true RMS meters try to measure RMS more directly by taking a number of samples over the entire wave.

What i was talking about was mainly the theory behind all this. Once we know the theory, we can get a good idea why each meter reads what it does. The factor for average to RMS is different than the factor for peak to RMS conversion. Meters that read peak and convert that to RMS will be off more because the peak reads the same for many waveforms that have very different RMS values. Meters that read average will be closer, although not always the same.
You can also look up "crest factor".

 

[Elerion]

Images I was talking about are on Post #18

 

[MrAl]

Images I was talking about are on Post #18

Hi,

Yes, but can you describe ONE such contradiction?
We can look at that then.

 

[The Electrician]

I know what true-rms is, but I wonder if it is required to measure a real audio signal.
With "real" I mean a mix of many frequencies, not just a test pure sine wave.
I suppose an average meter should work fine, but I've got no true-rms to compare.
Would an average and true-rms meters differ in readings?

Thanks!

Have a look at this: https://en.wikipedia.org/wiki/VU_meter

 

[Elerion]

Yes, but can you describe ONE such contradiction?

Sure. Simple one:
Top image, the square wave is 10% higher on average than RMS, but on the other, they match.
Isn't it that a contradiction?

 

[ronsimpson]

Have a look at this: https://en.wikipedia.org/wiki/VU_meter

Are we using a "RMS" meter to measure power? or audio?

When I use RMS to measure power the signal is consistent for hours (maybe years). 50/60hz from the power line or from solar. Many of these meters fail above 1khz and below the 50/60 mark.

When I measure audio there are two things I look for. (peak and average)

Peak: I want to know if I am at the distortion level. No worry about how loud you ears think it is. Just want to know where the loudest (very short term) signal is. In complex audio (song/voice) you ear is forgiving for distortion for less than 1mS (one time). There for a "peak meter", in my point of view, does not have to catch the very fastest peaks. Then the meter needs to hold for long enough for you eye to see the reading. (too many years to remember) I think my meters, the "needle" moves up in under 1mS and holds for 1 to 2 seconds. (moves down slowly)

RMS: Because you ear hears "power" a "RMS" type meter will read better, if loudness is what you are measuring. You might want to look at ANSI C16.5-1942. I see no reason to measure any signal below 20hz or above 20khz. In FM broadcast no response above 15khz. If you read up, the broadcasters in 1942 were trying to get a meter that responded like a human ear. (not using the meter to average) They wanted a "attack time" and "decay time" that responded like the human voice. The meter should hit full reading in one syllable and not pull back within a word. It should pull back, part way, between words. (actual time in mS can be found in many sources)

A Volt Ohm Amp meter with "RMS" will not measure audio the way most broadcasters and audio engineers want.

A low cost audio meter will use the mechanical response time of the meter for attack and decay time. ( the decay time is way too fast)
-----edited-----
For audio repair I see no reason for a true RMS meter. I might want a audio "db" meter but not a requirement.

 

[dknguyen]

The Fluke meter has an AC frequency measurement limit, which may have been the source of your observed large measurement error.
A Fluke 114 true RMS meter, for example, has an upper frequency limit of 1kHz, which would seriously limit its accuracy when trying to measure high frequency signals, such as from a switching power supply.

A fundamental rule is, you have to know the limits of the instruments you are using if you want to make accurate measurements.
The Fluke is indeed a true RMS meter when used within it's specification limits, which you apparently didn't do.

I'm not so sure because these were not SMPS signals being measured. They were switched LCR signals running off the 60Hz mains. The Fluke 179 is listed as 100kHz and we were measuring 120Hz signals with only one fast switching edge per half cycle so it should not have been off by something as large as 0.7.

 

[The Electrician]

Not necessarily. Know that "true-RMS" doesn't always mean true-RMS. At work, we have a fluke meter labelled a "true-RMS", but it's not. It was giving us weird readings with some special switched LCR waveforms so we compared it to a bench meter power meter and an oscilloscope's calculated RMS values (both devices confirmed to numerically calculate RMS from samples). The bench meter and oscilloscope agreed and the Fluke meter was off by a factor of 0.7.

As asked the guy that owns it and he told us that it "true-RMS" only under certain conditions so it seems that if the waveform strays too far from sinusoid then it's no longer valid since it's working under some set of assumptions rather than numerically calculating the RMS.

What are the "certain conditions"?

Do you have a scope image of the special switched LCR waveform?

 

[dknguyen]

What are the "certain conditions"?

Do you have a scope image of the special switched LCR waveform?

He didnt say and we didnt ask. I dont have photos, let alone remember what circuit we were working on.

 

[KeepItSimpleStupid]

Take a look here

for Crest Factor limitations.

Bandwidth and high frequency content is your first wall.
Crest factor is your second wall.

 

[The Electrician]

Sure. Simple one:
Top image, the square wave is 10% higher on average than RMS, but on the other, they match.
Isn't it that a contradiction?

Reconsider the case of a single sine wave of peak amplitude 1 unit. The average of the absolute value (rectified, in other words) is .6366 units, and the RMS value is .707107 units. So if you have a meter that is "average responding", it will measure a value of .6366 for that sine wave, but the actual RMS value is .707107 units--the meter is measuring a value lower than the RMS value. To make it display the correct value of the RMS value of our sine wave we have to multiply the measured average by .707107/.6366 which is approximately 1.1107. Since it's for sine waves that we want the meter to correctly display the RMS value because the waveshape found on the grid is a sine wave, we build average responding meters to multiply the actual measured average (rectified) value by 1.1107. This means that the meter will display the RMS value IF the wave we're measuring is a sine wave.

If on the other hand we measure a wave shape that has the property that its average is the same as its RMS value, the meter will display too high because of the multiplication of the measured average (rectified) value by the value 1.1107 (which was done to make it display correctly on sine waves). Such a wave shape (average the same as RMS) is the square wave.

The orange colored table in post #18 is showing what the two kinds of meters will display given various waveshapes. It is not comparing the average and RMS values of the waveshapes, only what the meters will display.

The next table of values in black and white is showing the actual values, PK-PK, 0-PK, average and RMS, of various waveshapes. These average values shown in this table are not what an average responding meter will display. An average responding meter will display the value under the heading "avg" after it has been multiplied by 1.1107. Because a so-called average responding meter that you might buy has most likely been designed to be used with sine wave voltages derived from the grid, it will multiply the average (rectified) voltage it measures by 1.1107 and display that value. This will be the RMS value of the wave being measured IF that waveshape is a sine wave.

If you measure a square wave the meter will measure the average (rectified) value of that square wave and multiply the measured value by 1.1107 and display that value. But this will NOT be the RMS value of the square wave, because the average (rectified) value of a square wave is the same as its RMS value. In this case we would like the meter to simply display the measured average (rectified) value of the square wave, and not multiply its value by 1.1107. Unfortunately multiplication by 1.1107 is built-in to the operation of the meter, so the meter will display an RMS value for the square that is too high by a factor of 1.1107. This is why the orange table says that the meter response to a square wave will be 10% high. In theory it will be 11.07% too high, but 10% is about right.

 

[MrAl]

Sure. Simple one:
Top image, the square wave is 10% higher on average than RMS, but on the other, they match.
Isn't it that a contradiction?

Hello again,

Well, when considering data from different web sites or different charts on any web site we have to know what they are referring too, in exact terms. That's because there are many ways of interpreting the way we measure things and the results of those measurements.

For example, mathematically the average of a sine wave is zero. That's 0v for any sine wave regardless of the amplitude, 1v, 10v, 100v, 1000v, etc. So the mathematical average of a sine wave that has a peak of 1 million volts is zero, but does that do us any good? No, because we know that when we use a sine wave to power something it delivers energy to that thing, and so we'd like a more comprehensive definition of average that we can use to understand power. This means we end up using the average of the absolute value of the sine wave.

Enter the average reading meter. Early meters could read the average of the absolute value of a sine wave which is similar to a full wave rectified sine wave average, but they wanted the meter to read out in volts RMS not volts AVG, so they applied a 'factor' to the meter. This factor would take the reading of the meter and by way of the face scale multiply it by a factor of approximately 1.1107 and since the averate reading meter movement would see a value of about 0.9 for a sine wave with a peak of 1.4142, multiplying 0.9 times 1.11 would result in a reading of very close to 1.000 which is exactly the right value for the actual RMS value of the sine wave.
So we see that AVERAGE reading meter movements use a factor of 1.1107 in order to show the RMS value from the actual value detected which happened to be the AVG value.

Now enter the peak reading meter. The peak of that same sine wave is 1.4142, so to get a reading of 1.0000 we need to multiply that 1.4142 by 0.7071, and so peak reading meters use a different factor which is that 0.7071 which is 1/sqrt(2).

The question now is what happens when we try to measure a square wave with amplitude peak of 1.4142, what do those same meters read.
What happens in the average reading meter is the square wave has an average that is equal to the peak, so it is 1.4142. The RMS value is the same, 1.4142, but when we apply that same factor of 1.11 we get about 1.57 which is 11 percent high. Thus the meter will ready about 11 percent high.

What this means for the two charts is that one is posting the theoretical ACTUAL value, and the other is posting the MEASURED value measured on a meter that has a meter movement that detects the average value.

The results are more clear when we see both waves and their results. In the attachement, we see the two waves in question and various values we get from different operations with those two waves.

The dark blue values are the ACTUAL values, these are the theoretically accurate values that the waves posses. If we had a meter that could measure everything perfectly, these are the values we would measure.

The green values are the meter factors for using the actual values to get to the RMS values.

The red values are the results we get from the meter using the right factor for each type of meter.

The light blue values are the ratios of what the meter calculates (displays) vs what the actual real RMS value is. This represents an error calculation as a ratio. For a ratio of say 1.1/1 that would mean the meter reads 10 percent high.

It is clear that for the square wave using an average reading meter movement we get a reading that is 11 percent too high. Note also that a peak reading meter movement would give us a result that is about 29 percent too low.

 

[dknguyen]

Take a look here

for Crest Factor limitations.

Bandwidth and high frequency content is your first wall.
Crest factor is your second wall.

Must have been crest factor then since that certainly is something that is quite high in our work. I just never use the hand held DMM anymore and just always use the bench meter.