what is the difference between rms and peak

[1ManBand]

can someone explain it to me please in layman terms.

[Lupo]

Very quick and dirty: RMS is a sort of average and peak is the top level. RMS corresponds better with what we hear as humans. We don't actually hear the peak level, we hear a sort of average power.

RMS means to take the root of the mean and square it. (EDIT: this is wrong! see post number 18 in this thread) This gives a sligthly different result than to take an average or mean in the usual way.


Not so easy, but hopefully useful too:

With a sine wave seen as a circle, the RMS is exactly at 45 degrees, halfway to the top.



This pic is from my website and shows the relation between the circle and a sinewave. At 45 degrees, the sine value is .707. This is also the amplitude value of the resultant waveform. .707 may seem like an odd number, but it's actually -3dB. It's also half the square root of two, which makes utterly sense if you look at it geometrically.

In comparison, a square wave have an RMS power of -6dB compared to the peak, noise is around -12dB. Music is naturally somewhere between -14 and -20. Until loudness mastering took off, that is. In these days, music resembles square waves more than music..


Andreas

[Lagerfeldt]

That's a nice graph!

[nosebleedaudio]

You could also look at the RMS as being an average of the signals, for example one strong peak will change the RMS value very little, but could clip something down the line.
This is why speakers for example have a pretty big difference between RMS and peak wattage.

[MASSIVE Master]

And keep in mind that RMS levels in a mix can have little relation to the peak levels. The crest (the space between the peaks and the average level) on the average pop/rock mix can be 18, 20, maybe even 24 dB.

A mix that sits at -20dB(FS)RMS can (some would argue *should*) have peaks up around -2 or -1dBFS. Not uncommon at all.

[1ManBand]

damn well explained.. thank u. i dont beleive what u said massive! i bet i would be able to hear that crazy fluctuation.

[MASSIVE Master]

That's not a crazy fluctuation - That's normal. That's what happens. A peak is an instant reading - RMS is an "average" reading.

Even hyper-compressed recordings (the "too loud" stuff on the shelves) have a crest of maybe 10-12dB. It's not *enough* of a fluctuation.

[1ManBand]

well maybe for pop rock stuff.. for that tight little snare and kick transient. i dont think its the same for hip hop though.. do u master hip hop ever?

[Coyoteous]

- about 6 dB these days, if you're lucky...

[MASSIVE Master]

Originally Posted by 1ManBand

well maybe for pop rock stuff.. for that tight little snare and kick transient. i dont think its the same for hip hop though.. do u master hip hop ever?

Most of the HH I get in here...

Let me rephrase that -

Most of the good sounding HH I get in here comes in at around -20 to -18dBRMS and peaks at (give or take) -3,-2,-1dBFS. 15, 16dB crest (generally a little narrower crest than a lot of pop/rock, but as you'd expect, more samples = less chance for errant peaks).

I get plenty in that sounds like [SELF-CENSORED] that has a crest of 10 or 12dB... And of course, they want it louder.

And not surprisingly (well, I guess it's a surprise to some) the stuff that usually leaves here the loudest came in the quietest. More accurately, the stuff that can handle the narrowest crest came in with the widest.

[badhorsie777]

if I'm doing it all myself in the box with logic 8, is there an easy way to measure my rms, peak, and crest values? This is new for me, also - long time musician, just new to producing the whole recording.

thanks

[badhorsie777]

never mind - "level meter" shows me peak and rms... how do I figure out the crest value you guys spoke of?

[bob katz]

Originally Posted by badhorsie777

never mind - "level meter" shows me peak and rms... how do I figure out the crest value you guys spoke of?

Crest factor is the simple difference between the average and the peak in a given section of material.

BK

[MASSIVE Master]

Originally Posted by MASSIVE Master

And not surprisingly (well, I guess it's a surprise to some) the stuff that usually leaves here the loudest came in the quietest. More accurately, the stuff that can handle the narrowest crest came in with the widest.

Some of you are asking why I'm quoting myself... Let's call it a small flurry of interrogative statements that required clarification...

As Bob already explained, the crest is simply the "space between" the average and the peak. More dynamics = larger (wider) crest, less dynamics = smaller (narrower) crest. A mix that has an average level of -20dB(FS)RMS that peaks up around -0dBFS has a crest of 20dB.

What I meant by the above statement was that there are a lot of (usually "less seasoned" for lack of a better term) engineers out there who shoot themselves in the foot by tracking and/or mixing "for volume" -- Destroying the dynamics, using up all the available headroom at the first stage - along with pretty much every subsequent stage.

That's not the way to make recordings that can actually handle "loud" later.

Track with plenty of headroom. Mix with plenty of headroom. Don't throw limiters all over the place just to get the mix "loud" -- Do whatever it takes to make the mix sound *good* and don't be concerned so much with volume.

I'm not saying not to use limiters -- But as a "rule of thumb," if you find a mix actually sounds better - And I mean *BETTER* -- Not "better because it's louder" -- If it actually sounds better being rammed into a limiter, take the limiter off and find out why. Maybe one thing in that mix is truly "too dynamic" for the rest of the mix. Put the limiter on THAT and see how it sounds. Use a compressor when something has a dynamic range that's too wide for the mix -- Not because someone told you that everybody compresses everything so it can be louder.

"Punch" and "impact" comes from the difference between loud and quiet - Not the absence of quiet. And although I'm not a fan of the current "level insanity" going on, it's better to have decent sounding recordings that have the potential to be loud than loud sounding recordings that should be shut off. Almost invariably, it's those dynamic, wide-crest mixes that have that potential.

Mixing "hot" doesn't do anything to make your finished product louder. Tracking hot doesn't either (I don't even want to get into the nastiness that can happen from tracking too hot). Headroom is good room. Keep it, love it, cherish it - Your mastering guy (even if that's yourself, which I also won't get into) is almost undoubtedly going to use it all up... Give him some room to move.

[bob katz]

Dear John: What a nice summary. If you don't mind, I'm going to quote your post in a discussion on this topic that I have at our website.


Best wishes,



Bob

[hackenslash]

I have printed the whole thread, for use in teaching. this thread provides the clearest, most concise explanation I've come across. I have also posted links to it on every forum I'm a member of.

[MASSIVE Master]

Originally Posted by bob katz

Dear John: What a nice summary. If you don't mind, I'm going to quote your post in a discussion on this topic that I have at our website.

Best wishes,

Bob

Certainly -

[hackenslash]

Originally Posted by Lupo

Very quick and dirty: RMS is a sort of average and peak is the top level. RMS corresponds better with what we hear as humans. We don't actually hear the peak level, we hear a sort of average power.

RMS means to take the root of the mean and square it. This gives a sligthly different result than to take an average or mean in the usual way.


Not so easy, but hopefully useful too:

With a sine wave seen as a circle, the RMS is exactly at 45 degrees, halfway to the top.



This pic is from my website and shows the relation between the circle and a sinewave. At 45 degrees, the sine value is .707. This is also the amplitude value of the resultant waveform. .707 may seem like an odd number, but it's actually -3dB. It's also half the square root of two, which makes utterly sense if you look at it geometrically.

In comparison, a square wave have an RMS power of -6dB compared to the peak, noise is around -12dB. Music is naturally somewhere between -14 and -20. Until loudness mastering took off, that is. In these days, music resembles square waves more than music..


Andreas

This was pointed out to me by Dreamkeeper on KVR. I overlooked it. The original formula for RMS is incorrect. The correct formula is the square root of the mean of the squares of the values.

Still, great thread.

[Lupo]

Oops! Thanks for pointing that out. Now that I look at my own sentence, it's certainly looks wrong in more than one way. Am glad the net is here to catch my mistakes! :)


Regards,

Andreas

[Lupo]

The electrical way of thinking may be useful too; RMS value is the DC equivalent value to an AC stream.
The RMS value of an alternating voltage or current is the same as the level of direct voltage or current that would be needed to produce the same effect in an equal load.


As the RMS value is mostly used in audio, it's not a bona fide DC. The value changes with time unless all of the audio is analyzed at once. A fairly slow response with lots of input samples(or time, in analogue) is needed to get a decent "steady state" value from the musics alternating waveforms. VU meters are 300 millisec. I usually prefer faster. Somewhere around there is not too far off the response in the hearing.


Andreas Nordenstam


PS: peak level is not only different from RMS, it's different from peak level too! there's sample point peak level and there's the peak level of the signal in between the sample dots. this link can't be repeated too much, hope it's al right: http://www.cadenzarecording.com/pape...distortion.pdf

[hackenslash]

Bump. I'm referring peeps to this thread from all over.

[mikefloutier]

Hi all,

I'm new to this and like 1manband I'm also trying to understand Peak & RMS.

Firstly, can I clarify. Are you saying that the RMS figure is calculated by applying the corrected formula (sorry Andreas!) to a rolling window of the "Peak" (ie actual dbFS) signal.

Assuming that's so, I did a few very simple calculations on a bit of paper based on the said formula and, correct me if I'm wrong:-

1. A completely steady signal would yield a figure for RMS that was the same as Peak whereas,

2. A mixed level signal would give (unsurprisingly) a lower figure for RMS than that of Peak.

I've two reasons for wanting clarify this:-

1. I need to know (as my wife will tell you - lol) and I know it will help me, and

2. When I did the "steady signal" experiment in practice with my DAW there was around a 7db difference between RMS & Peak.

Thanks

Mike

[Lupo]

Hello Mike!

Welcome to the forum!

Originally Posted by mikefloutier

Firstly, can I clarify. Are you saying that the RMS figure is calculated by applying the corrected formula (sorry Andreas!) to a rolling window of the "Peak" (ie actual dbFS) signal.

That's the usual way, yes. It needs some time and you can get that either through a bunch of succesive samples in digital domain, or through capasistance or mechanical ballistics of the VU meter needle, in analogue.

And don't be sorry, I'm very happy that someone spotted my fault! (Also a bit dissappointed that no one in here did it..) Had not done these calculations prior to writing that message. Just had an intuitive understanding of it. Have now done some manual calculations and finally learned the math way too.

Originally Posted by mikefloutier

1. A completely steady signal would yield a figure for RMS that was the same as Peak whereas,

If you mean a single repeating number when you say steady state, that's correct. It's a "sort of average" and an average of a single number is that number. Another way to think of it is that a steady state signal is a direct current. RMS is the DC equivalent of an AC. Since the signal you're testing is already a DC, there's no difference.

Originally Posted by mikefloutier

2. When I did the "steady signal" experiment in practice with my DAW there was around a 7db difference between RMS & Peak.

Now that's odd! One possibility is that it's the DC nature of the steady state signal being measured that messes with the audio workstation. Many audio things remove or ignore DC by default. That could perhaps give rise to a short click at start/stop or similar that could give the measurement you got.


Andreas

[jamsmith]

You are going to get a different RMS readings for different waveforms also. The RMS of a sine wave is 70.7% but a square wave RMS is 100%. A triangle wave RMS is peak divide by the square root of 3.

[TSchlum]

Great to see some science around here for a change!

Another way to look at it is: RMS represents the AREA under the wave. That is, the POWER of the signal.

Voltage is only the poster child of the current. The current is doing the work.

[andy_simpson]

Does anybody remember the hit record crest factor study that Chris (Airwindows) did?

This was a study of the crest factor of the biggest hit records in the history of recorded music and the results were enlightening to say the least.

Thriller, Eagles, Beatles, Bee Gees, etc. BIG crest factors - often around 18-20 dB of crest.

As others have said, crest factor will vary according to many factors including instrumentation, arrangement, amount of reverberation (which brings average up), etc.

I have a track with ~27dB of crest on my site:

high crest.mp3

No compression/limiting or reverb - what ChrisJ would call 'high contrast'.

Andy

[jamsmith]

Originally Posted by TSchlum

Great to see some science around here for a change!

Another way to look at it is: RMS represents the AREA under the wave. That is, the POWER of the signal.

Voltage is only the poster child of the current. The current is doing the work.

Over a given impedance (ac) or resistance (dc), voltage and current are directly related. voltage = current * resistance. If you double the current, the voltage doubles, if you half the current, the the voltage halves.

I guess you could say the current is doing the work as its the amount of actual electrons passing a point in given (1 coulomb per second). Where as voltage is the force behind the electrons.

[TSchlum]

Originally Posted by jamsmith

Over a given impedance (ac) or resistance (dc), voltage and current are directly related. voltage = current * resistance. If you double the current, the voltage doubles, if you half the current, the the voltage halves.

I guess you could say the current is doing the work as its the amount of actual electrons passing a point in given (1 coulomb per second). Where as voltage is the force behind the electrons.

Michael, I’m in complete agreement with your observation that voltage and current are interconnected / inseparable.

The point I was trying to make is; as electronics people and more to the point, DAW screen viewers, it is easy to get focused on voltage. But signals are 3 dimensional. They contain voltage, current and time.

Ref the original question; RMS is a measure of two of these dimensions, voltage and time. The formula for RMS is the same as used to calculate the surface area of an irregular two dimensional shape.

There is more to sound then what you see on the screen.

[flatfinger]

OK , so what about rms VS AES-17 RMS?????

[allencollins]

Originally Posted by Lupo

Very quick and dirty: RMS is a sort of average and peak is the top level. RMS corresponds better with what we hear as humans. We don't actually hear the peak level, we hear a sort of average power.

RMS means to take the root of the mean and square it. (EDIT: this is wrong! see post number 18 in this thread) This gives a sligthly different result than to take an average or mean in the usual way.


Not so easy, but hopefully useful too:

With a sine wave seen as a circle, the RMS is exactly at 45 degrees, halfway to the top.



This pic is from my website and shows the relation between the circle and a sinewave. At 45 degrees, the sine value is .707. This is also the amplitude value of the resultant waveform. .707 may seem like an odd number, but it's actually -3dB. It's also half the square root of two, which makes utterly sense if you look at it geometrically.

In comparison, a square wave have an RMS power of -6dB compared to the peak, noise is around -12dB. Music is naturally somewhere between -14 and -20. Until loudness mastering took off, that is. In these days, music resembles square waves more than music..


Andreas

nice job