The No.1 Website for Pro Audio
All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
1 week ago
#1
Lives for gear

About measurement methods, scales and their comparability

I ran into an interesting discussion today. In a FB forum, John Eagan (are you around?) posted an illustration that provoked some irritation in me.

Before you get me wrong, this illustration has a point, and noble goals. Good intentions, and you can find this type of argumentation expressed all over the net. But...

It describes a fixed relation between:

B. a +4 ref dBu
C. a dBFS value

Technically, there is a logical problem with this perspective.

Core of the problem is the assumption that different measurement methods yield comparable results. This generally is not the case.

Different scales can be easily converted and compared. Examples for these would be: miles, kilometers, sea miles, light years. All are meant to cover different ranges, give insight into specific regions. But all are compatible and comparable, no matter their scales, because all base on the same measurement method.

Same is true for temperature. You have different scales, but all base on the same measurement method. This makes them comparable.

On the other hand, different measurement methods can sometimes use the same scales. Yet, they aren't compatible. A good example for this is the degree °. You can't compare a temperature of 3° Celsius to an angle of 7°. It's nuts! Even if they describe the same subject.

However, and it's subtle, the illustration above does exactly this. It compares totally different methods, which by coincidence, share the same scales.

A. A VU meter returns a very slow and simplistic energy or intensity measure. Primary feature is its laziness, which in turn gives a good idea of how much THD the signal can provoke in classic analogue gear. Contrary to popular opinion, VU is a weak representation for loudness.

B. The dBu +4 signal is also just the result of a VU meter, but shifted 4dB, for whatever reason. This value is perfectly comparable to A), because it uses the same measurement method. One only has to convert the scale, by attenuating the latter by 4dB.

C. dBFS is the absolute value of the raw PCM data. As such, it only roughly correlates with the true underlying signal, visually maybe up to 2kHz, not much more (just create a 19kHz sine, and zoom into the PCM). In fact, dBFS contains massive aliasing distortion, and can't be trusted for anything else than preventing clipping when entering the DA.
dBFS is an instant, non smoothed measure.

A and B are comparable. But C uses a completely different measurement method that's simply not comparable to a VU reading. Classic apples and oranges. Or better, temperature vs angle.

To be fair, there is some correlation to expect between some measurement methods, but it will be of confusing nature. I can see the noble intention, but technically, it's almost as crazy as the temperature vs angle above. To go back to the illustration above, in reality, this black line would wiggle around like crazy, because it has no reasonable relation to draw in the general case.

It is theoretically possible to link different measurement methods. But it involves pausing time and bending the universe around to make it fit. One of such examples would be the idea to specify a fixed reference for a specific signal. Say, a 1kHz sine that runs until infinity. But this isn't representative for music, it's maybe the worst "laboratory representation" for music. Once you mess with real music, it blows apart.

If you intend to calibrate stuff, make compare only values that originated from the same measurement methods. In case of recording with VU, it's fine. Buy a good software VU meter (one that produces the VU from an oversampled signal), and stick to VU throughout (just make sure nothing is clipping).

But don't try to compare different methods, it leads into trouble and depression!
Attached Thumbnails

1 week ago
#2

Good points, Fabien. It's true that 'degrees Celcius' and 'degrees of a circle' are not the same degrees, even though they use the same word, but 'degrees' is not an SI unit of measurement of a circle or angles, whereas 'degrees Celcius' _is_ (one of the International System of Units). 'Radians' is the official measurement unit of measuring angles, such as the segments of a circle describe.

Also, dBVU and dbu and dBFSD all measure decibels, but the two vertical scales shown, one for analog and the other for digital are undefined as meters. dBFSD and dBu can be used for average or peak metering, for example. Whereas the VU meter isn't a peak programme meter, but the VU meter by Weston was optimized to show perceived loudness because we don't hear peaks. It works well, according to the many CDs I've watched that have sustained fortes to 0 dBVU with fortissimos to +3+... We can measure RMS or VU or PPM in dBFSD and, if we measure dBVU in the digital domain, it should match the properly dampened ballistics of an analog VU meter. There's no requirement for 1 kHz at 0 dB VU to = + 4 dBm (i.e., 1.228 VRMS). I run one of my pairs of VU meters (by Sifam, with a passive parallel and series network of resistors across the terminals) set for 1 kHz at 0.775 VRMS, which is equivalent to +0 dBm, for example.
1 week ago
#3
Lives for gear

Verified Member
Guilty. It's an imperfect translation that only works for a steady state signal but it's what we have to work with. It actually works pretty well for its intended purpose, which is basically system calibration.

As a person who's been staring at those meters for decades it's easy to forget that young engineers don't understand exactly what each of them measure and they assume they are essentially telling us the same thing. All most engineers ever see today are peak (sample) meters with a poor relationship to perception. Many young engineers have never used perception based metering at all. I'd even say many engineers have little concept of what a logarithmic scale is or how hearing works. Maybe I'm drifting off topic but that's basically why VU meters exist in the first place.

What would you suggest we do differently?
1 week ago
#4
Lives for gear

Quote:
Originally Posted by Greg Reierson
What would you suggest we do differently?
In the specific case illustrated above, I'd suggest to get a good digital VU meter, and not try to replace it with digital peak meter (which really uses a much different measurement method).

Use a VU or true RMS meter, or a LU measure, throughout the process. All are sufficient, as long they remain comparable.
Maybe in pair with a clipping indicator watching the dBFS DA input.
1 week ago
#5
Lives for gear

Verified Member
There is no such thing as digital VU meter. The spec is clear and well defined. I’ve never seen software that has tried to mimic the VU meter ballistics be remotely successful. Probably for the reasons you describe. Cheap DC meters with a VU scale don’t count.

Metering has always been accurate for level but inaccurate for perceptual loudness. VU meters are remarkably good at approximating loudness for the human voice as was intended. For 1930’s technology it has remained very useful.
1 week ago
#6
Gear Nut
FabienTDR,
Which VU, LU, RMS meters are professional, extremely accurate (plugins)? What VU metering plugins are closest to analog counterparts? What you can recommend? Youlean, TBproAudio, psp triplemeter, klanghelm? Or?
Thanks
1 week ago
#7
Lives for gear

Verified Member
0VU = +4dbu = -18dbfs is entirely appropriate for calibrating a system with tones. That's all I've ever used it for. What's wrong with that?

All the various meter types tend to get used where they are most appropriate.

Digital peak meters for recording to digital are entirely appropriate to avoid clipping, which is the main concern since noise floor is a non-issue anymore.

VU meters are on analog tape machines where they belong.

RMS and LUFS meters are on brickwall limiters where they belong.

All is well in meterville.
1 week ago
#8
Lives for gear

Quote:
Originally Posted by Trakworx
0VU = +4dbu = -18dbfs is entirely appropriate for calibrating a system with tones. That's all I've ever used it for. What's wrong with that?
Because it sets a fixed relation between the results of two completely different measurement methods.

With real music signals, the relation between the reading of a VU meter and a dBFS peak meter is extremely wild. They are incompatible. To make it work, one has to construct a laboratory scenario, usually via a 1kHz sine, or pink noise. Once you feed in the true variety of music signals (drum solos, percussion, triangle hits, whatever), this relation blows apart.

A peak meter is a instant measure. A VU meter produces is a "lazy", an integrated measure. Comparing them produces a whole new class of measure (peak to average), that's useful. But also explains why they can't be used for calibration. The ratio between both is highly signal dependent.

Because of this, I'd recommend using one single method, throughout. If a good digital VU meter doesn't exist yet, I'm happy to take the challenge.

Last edited by FabienTDR; 1 week ago at 06:03 PM..
1 week ago
#9
Lives for gear

Verified Member
RTW, NTP and BBC style PPM meters use a 1K tone at reference level for calibration. Just like a VU meter.
1 week ago
#10
Motown legend

Verified Member
The PSP and Waves VU meters both pass the NAB ballistics test.

It's important to understand that VU meters were always intended to be used alongside peak indicators or meters of some kind. The compression effects of analog tape allowed manufacturers to get away with leaving out peak indicators and using low-cost meters that did not meet the ballistics test. People working in broadcast and analog disk mastering never had that luxury.
1 week ago
#11

SpectraFoo has a 300 ms integrating averaging VU meter in its digital audio meter, as well as the interesting K-System, which I still use.
The precise ballistics of an analog VU meter are probably only in the original Weston meters, based on the quote below.
They may be worth sourcing, if you want to use VU, 'according to Hoyle', because, fortunately, old ones can still be true to performance, or serviced to be so...

The maker of MRL Calibration Tapes wrote, on the Ampex list:
'Ballistics are determined by the mass of the pointer -- that won't change with
time--and the compliance of the restoring metal spring (and that would change the
meter sensitivity) -- don't think that would change the magnetic field
induction, which also would change the sensitivity.

So if the meter sensitivity is still correct, the ballistics are probably still
ok.'
1 week ago
#12
Lives for gear

Verified Member
Respectable observations, especially related to aliasing in a dBFS meter, which is interesting to know about. In practical use metering has only ever been a rough guide to the signal it represents. As such it comes as no surprise that one rough guide will never precisely parallel another rough guide (given a dynamic and complex sound source). However some guides are arguably better/preferable than others. I have mainly found metering useful to know roughly what the maximum signal a system can (or nominal level) take and not much more. They are not something to be leaned on as an absolute even in isolation, never mind comparison.

Most important is to know roughly how they approximately compare (if you are using a system with multiple meters and audio stages referenced to a known electrical level) and then always rest on the side of caution to avoid distortion or use your ears if distortion is your goal. This is a sensible practical use.

I always recall how a VU meter would not read hf transients well and could distort tape before it was in the red, we were warned about this
in audio engineering school.

Another good example is not to rest on a Gain reduction meter as the ballistics of the GR meter may not relate well to time constants of the compressor
really it just demonstrates something is happening.

Best viewed as rough guides really.
1 week ago
#13
Lives for gear

Verified Member
I’ve seen the Waves and Spectrafoo VU’s. They didn’t look right to me. One thing that may be responsible is latency of the system. If the integration time is correct but it’s late in time it looks terrible and is basically useless.
1 week ago
#14
Lives for gear

Regarding the fact that there's no good (accurate, precise enough?) digital VU meter I'd say that two really pops up and they can be treated as really good tools. One is Klanghelm's VU Meter and the other one is TBProAudio's VU Meter. Calibration both of level and responsiveness is very life-like. Very natural. Very, very, very good and almost life-like. Almost analogue.

Anyone thinks the same?

Krešo
1 week ago
#15
Lives for gear
I'm quite fond of Klanghelm's VU meter, have it on its own monitor. Honestly I don't pay a whole lot of attention to it, or any meters, because ears, but it's nice to look at and simple to recalibrate for mixing or mastering.
1 week ago
#16
Lives for gear

Quote:
Originally Posted by Paul Gold
I’ve seen the Waves and Spectrafoo VU’s. They didn’t look right to me. One thing that may be responsible is latency of the system. If the integration time is correct but it’s late in time it looks terrible and is basically useless.
I think that the impression or the feel of a VU will always be worse in digital, due to inherent latency. But the values they return can certainly be made sufficiently equal to their analogue counterpart. It's not exactly difficult to precisely predict the kinetics of a needle, but building time-machines, that's a whole different story
1 week ago
#17

Quote:
Originally Posted by FabienTDR
If a good digital VU meter doesn't exist yet, I'm happy to take the challenge.
I'd use it!
1 week ago
#18
Lives for gear

Verified Member
Quote:
Originally Posted by FabienTDR
I think that the impression or the feel of a VU will always be worse in digital, due to inherent latency.
It's even worse than that because the latency will change depending on if you are playing files from disc or using input monitor. A meter with a variable latency is barely a meter.

With a VU meter or any analog meter like an HP400 AC Voltmeter, it's the barely noticeable trembling of the needle that gives you a lot of valuable information. Screen resolution most of the time wouldn't even take that into account.
1 week ago
#19
Lives for gear

Verified Member
Do we really need another meter? I rarely look at any of them. I glance over to make sure things are where I except them to be based on what I'm hearing. I use them for system calibration. That's about it. With a calibrated monitoring system, meters are mostly there to confirm what I already know.

I think people need to spend less time looking at meters, not more.
1 week ago
#20
Lives for gear

Quote:
Originally Posted by Greg Reierson
Do we really need another meter?
Definitely not. But maybe a deeper understanding of the underlying (measurement) methods.

IMHO, the whole idea to "educate" the reader by obscuring the fact that the relation between a VU measure and a dBFS peak meter is dynamic, for any signal except sine, no matter the calibration, is not productive.

That black line doesn't exists in the real world. And maybe represents the main fallacy people face when learning to use these meters.
1 week ago
#21
Lives for gear

Verified Member
Quote:
Originally Posted by FabienTDR
Because it sets a fixed relation between the results of two completely different measurement methods.

With real music signals, the relation between the reading of a VU meter and a dBFS peak meter is extremely wild. They are incompatible. To make it work, one has to construct a laboratory scenario, usually via a 1kHz sine, or pink noise. Once you feed in the true variety of music signals (drum solos, percussion, triangle hits, whatever), this relation blows apart.

A peak meter is a instant measure. A VU meter produces is a "lazy", an integrated measure. Comparing them produces a whole new class of measure (peak to average), that's useful. But also explains why they can't be used for calibration. The ratio between both is highly signal dependent.

Because of this, I'd recommend using one single method, throughout. If a good digital VU meter doesn't exist yet, I'm happy to take the challenge.
I won't dispute any of that.

But.

If I understand you correctly, you are identifying a problem with the way everyone has been calibrating their systems basically forever.

1K tone. 0VU = +4 = -18dbfs.

So, ... If that's a problem, what is the negative consequence we are experiencing by calibrating this way?
1 week ago
#22
Lives for gear

Verified Member
Quote:
Originally Posted by Greg Reierson
Do we really need another meter? ...
this!

Quote:
Originally Posted by Trakworx
...All is well in meterville.
this!
1 week ago
#23
Lives for gear

These -18dBfs are deliberate. There is no fixed relation between a VU and dBfs, except for a 1kHz sine (a signal that never changes).

If you're recording a drum solo or a percussionist, you might very well get clippings and huge difference in loudness. If you're recording content with very strong HF, same can happen.

Nothing dramatic, but raises the question of why -18dBfs and not -19dBfs? I'm for -42dBfs!
What you really want is equal loudness between all contexts. A dBfs peak measure will never help to achieve that. It's simply impossible.

If you want to calibrate your system according to some integrated measure (VU reading, RMS, K-meter, LU), then do it consequently for any part of the system. That yields a reliable calibration for 99% of the signals. But comparing (i.e. calibrating) one part of the system via an integrated measure, and the other via a instant "peak" measure isn't very wise. Or necessary.

Quote:
If I understand you correctly, you are identifying a problem with the way everyone has been calibrating their systems basically forever.
Yes.

Point is, nobody calibrates anything statically, they simply move the input gain knob before recording. This -18dB is completely deliberate, because in reality it turns out to be super signal dependent (because again, dBfs is a peak measure, not a representative for energy or loudness).
1 week ago
#24
Lives for gear

Verified Member
Quote:
Originally Posted by FabienTDR
These -18dBfs are deliberate. There is no fixed relation between a VU and dBfs, except for a 1kHz sine (a signal that never changes).

If you're recording a drum solo or a percussionist, you might very well get clippings and huge difference in loudness. If you're recording content with very strong HF, same can happen.

Nothing dramatic, but raises the question of why -18dBfs and not -19dBfs? I'm for -42dBfs!
What you really want is equal loudness between all contexts. A dBfs peak measure will never help to achieve that. It's simply impossible.

If you want to calibrate your system according to some integrated measure (VU reading, RMS, K-meter, LU), then do it consequently for any part of the system. That yields a reliable calibration for 99% of the signals. But comparing (i.e. calibrating) one part of the system via an integrated measure, and the other via a instant "peak" measure isn't very wise. Or necessary.

Yes.

Point is, nobody calibrates anything statically, they simply move the input gain knob before recording. This -18dB is completely deliberate, because in reality it turns out to be super signal dependent. That's why it seems to make sense (because again, dBfs is a peak measure, not a representative for energy or loudness).
OK, but what's the actual practical negative consequence of calibrating 0VU to -18?

You said I might get clipping. But as you say, that's what the fader is for. And that's what the peak meter is for too. I'm not seeing a problem with this setup...

Yes, -18 is an arbitrary number, but no one is forced to use it. We can calibrate to whatever number we choose, but -18 seems to have become accepted as 'enough' headroom in practical use. An AE will always need to set the fader level no matter how the system is cal'd.

Peak meters show useful information when setting recording level to digital. The fact that peak meters measure something different than VU meters is a feature not a bug.

Last edited by Trakworx; 1 week ago at 05:41 PM..
1 week ago
#25
Lives for gear

Verified Member
I'm back to the 'what would you do differently" question. It's the late '70s and it's an analog world that is about to be invaded by digital recording. What do you do? You figure out a way to transfer a very well ingrained standard in the analog world into this new digital world. Pick a margin that is almost always going to be beyond what is needed to get 0VU off tape into digital without risking overs. 0VU = -18dBFS or -20 or whatever. It seems like a very logical solution.

And at this point, it's practically irrelevant (except for us dinosaurs who still work with tape and disc recording).

What are the negative consequences as a result of this? Are the engineers who look at peak reading meters and make judgements about the sound really going to understand this argument? Does it matter?

You know I'm not trolling you, Fabien. I'm just curious where this is coming from.

Last edited by Greg Reierson; 1 week ago at 06:59 PM..
1 week ago
#26
Lives for gear

Verified Member
Exactly. I'm trying to understand why this is a problem in real world studio operation.
1 week ago
#27
Lives for gear

It has no consequences for your established work-flows and habits. They are proven, and I'm not here to put them into question.

This is purely educative/academic (sorry I'm slightly pedantic with these things ). When teaching this stuff to newbies, we should stop "drawing these fixed black lines" (communicating the existence of a fixed relation), and instead clarify that a VU reading of course cannot be compared to a dBfs reading, for several reasons.

It's the idea to draw such a line, without further explanation, that irritates me. Because the newbie will interpret it as

"how nice, now I can convert VU to dBfs, and vice versa. I just add or subtract 18dB!"

which of course isn't possible.

In fact, it's the wild difference (i.e. ratio) between peak and average measures that explain how much headroom exactly you ideally need to calibrate for. As long the noise floor doesn't suffer too much. -18dBfs is maybe a good starting point, but in no way universal. Nowadays, -24dBfs (for a 1kHz tone) could be safer, and still have low noise. In reality, all that matters is whether AD/DA clip or not.

Last edited by FabienTDR; 1 week ago at 05:23 AM..
1 week ago
#28

Quote:
Originally Posted by FabienTDR

A. A VU meter returns a very slow and simplistic energy or intensity measure. Primary feature is its laziness, which in turn gives a good idea of how much THD the signal can provoke in classic analogue gear. Contrary to popular opinion, VU is a weak representation for loudness.
To understand what a VU (actually incorrect, it's a VI indicating Volume Units) meter is showing, you have to understand what it was developed for and why.

The purpose of the meter was to satisfy the needs of broadcast networks and phone companies for a standardized and application-specific Volume Indicator.

From the 1940 IRE (now IEEE) paper, "A New Standard Volume Indicator and Reference Level", by Chinn, Gannett and Morris, "This need may be better appreciated by considering the communication systems employed for broadcasting. These are very con~plicated networks spread over large geographical areas. A typical network may include 15,000 miles of wire line and hundreds of amplifiers situated along the line and in the 50 to 100 connected broadcast stations. Every 15 minutes during the day the component parts of such a system may be shifted and connected in different combinations in order to provide for new points of origin of the programs, and for the addition of new broadcast stations and the removal of others from the network. In whatever combination the parts of the system are put together, it is necessary that the magnitude of the transmitted program waves, at all times and at all parts of the system, remain within the limits which the system can handle without impairment from overloading or noise."

With the audibility of distortion as a primary factor, listening tests were made using RMS meters and Peak-responding meters to determine which correlated better with the audibility of distortion of a network line driving amplifier. RMS won.

From that basic need rose a series of criteria for a new passive volume indicator, including rise time, overshoot, impedance, circuit loading, reference level, scale, even the color of the face and shape of the pointer. The passive meter was meant to bridge program lines without undue loading, or adding undue distortion.
Quote:
Originally Posted by FabienTDR
B. The dBu +4 signal is also just the result of a VU meter, but shifted 4dB, for whatever reason. This value is perfectly comparable to A), because it uses the same measurement method. One only has to convert the scale, by attenuating the latter by 4dB.
Again, you have to look at the original application.

The standard line level for broadcast network "long lines" (some as long as 15,000 miles), was referenced to a power level into a characteristic line impedance. Because there was more equipment designed around 600 ohms, and a nice round number for the reference was desired, 1mW driven into 600 ohms became the new reference level. However, since the passive meter cannot be allowed to load a 600 ohm line, it's impedance must be 10X or higher than the line it's bridging. The meter movement impedance was 3K9 ohms internally, and a 3K6 ohm resistor is applied externally (so the meter "sees" it's own impedance), resulting in a total bridging impedance or 7500 ohms, and a resulting sensitivity for reading 0VU at +4dB above 1mW into 600 ohms at 1KHz.

Program monitoring panels used variable attenuators, so an operator would read the meter and add the attenuator setting for the total correct level reading. Readings were to be averaged over long periods of time, depending on content type.
Quote:
Originally Posted by FabienTDR
C. dBFS is the absolute value of the raw PCM data. As such, it only roughly correlates with the true underlying signal, visually maybe up to 2kHz, not much more (just create a 19kHz sine, and zoom into the PCM). In fact, dBFS contains massive aliasing distortion, and can't be trusted for anything else than preventing clipping when entering the DA.
dBFS is an instant, non smoothed measure.
That depends upon at what point the dBFS meter is applied. A simple analog dBFS meter calibrated to the equivalent input level required at the input to a A/D would have issues that digital meters that look at the resulting data stream would not, and modern metering algorithms that include inter-sample overs also would not.
Quote:
Originally Posted by FabienTDR
A and B are comparable. But C uses a completely different measurement method that's simply not comparable to a VU reading. Classic apples and oranges. Or better, temperature vs angle.

Tape had nothing to do with VU development, other than Ray Dolby didn't fully "get" what the "B" scale was about, so chose the "A" scale for the VU meters on Ampex tape recorders. Actually, the "B" scale, the one with 100% on the top, dB on the bottom, does make sense in terms of the percentage of "modulation", power, fluxivity, etc., but we got what we got. VU on analog tape is a reasonable fit because tape saturates gradually, thus the distortion audibility criteria of the VU works well.

Remember, one meter design criteria was that it had to be passive. Think about how they did that and got what they got, including 300ms rise, with no more than 1.5% overshoot, and 26dB range, and appreciate what was done here.

The original meter specified an arrow at the tip of the pointer and frosted bulbs for illumination. Got one like that...in software?
1 week ago
#29
Lives for gear

Verified Member
Meters are good at measuring level. They are less good at loudness. It’s engineering 101 to know the difference between level and loudness.

The VU meter was designed to measure level and loudness of a human voice. It works well for this.

PPM meters are strictly for level. There is very little attempt to measure loudness in the design.

The new crop of loudness meters say what they are supposed to do in the name.
1 week ago
#30
Lives for gear

Verified Member
Quote:
The original meter specified an arrow at the tip of the pointer and frosted bulbs for illumination. Got one like that..
Arrows: yes. The rest isn't quite stock.
Attached Thumbnails

Topic:

## Welcome to the Gearslutz Pro Audio Community!

###### Registration benefits include:
• The ability to reply to and create new discussions
• Interact with VIP industry experts in our guest Q&As
• Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD \$20/year
• Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.