Electronic Pages

Die Homepage der Familie Beis

VU measurements compared to PPM measurements

What effect does the difference between VU and PPM measurements have in practice?

( Deutsche Version: VU- im Vergleich zu PPM-Messungen)

Inhalt

WTF... (Preface)

So, what's the point now? I have always considered level measurement according to the VU standard to be rather unreliable, i.e. more of a cheap or amateurish solution. But that's not the reason for this article. There was an occasion that prompted me to do more intensive research on this topic and then to write this article:

For some time now, I have been offering my AMBMs (Analog Meter Bridge Modules), where you can set the level measurement according to both the VU and PPM standards. Nobody had made any comments about this. I recently developed a variant of the AMBM, the ALM77, which fits perfectly into the Revox B77 and PR99 tape recorders. But now there are questions and requests to make the VU display “better”, i.e. to behave in the same way as the original VU meters. I can certainly understand this, so I would like to do some persuading. So I want to be able to answer these questions in a qualified manner, not just with a personal assessment, but based on measurements and their statistical evaluation.

I must add that the VU display of the AMBM and ALM77 does not comply with the VU standard according to ANSI C 16.5 / IEC 268-17 and also DIN IEC 60268-17 in two points, but in my opinion there is a good reason for this:

1. I do not use a linear scale for the VU display, but the same logarithmic scale up to -50 dB as for the PPM display. In contrast to a linear scale, this also allows low levels and interference signals to be recognized.
2. The VU display is without the usual lead, which is usually approx. 6 dB. 6 dB corresponds quite precisely to a factor of 2, but even more precisely to a factor of 1.995.

The 2nd point in particular causes irritation. I understand that you miss the usual scaling, but I think it's better this way for two reasons:

1. If a sinusoidal signal with a level corresponding to full scale is present, I would like 0 dB to be displayed for both measurement methods. So not the lead of approx. +6 dB as with standardized VU meters, where full scale with a sine wave signal usually goes far beyond the display range.
2. In any case, I consider the VU display to be a superfluous relic from times when more reliable level indicators than VU meters were hardly technically feasible. So the VU meter may still be good for demonstrating the difference between VU and PPM measurements, but using it in practice would mean deliberately accepting a less reliable level meter - and that cannot be in the user's interest.

This unreliability of the VU measurement was also recognized by Revox. To alleviate the problem somewhat, an LED has been installed in the A700, B77 and PR99 to indicate overload in addition to the VU meters. It indicates when the maximum level has actually been reached or exceeded. At least.

In this context, a function called "peak hold" is also very important for PPM displays. In contrast to much slower VU displays, the peak values can only be seen for a short time with PPM, despite the relatively large fallback time constant. They disappear faster than they can be reliably recognized. With Peak Hold, a peak value is displayed for a short moment after it occurs, i.e. it is “held”. This makes the actual peak value measured according to the PPM standard clearly recognizable.

I had further concerns about the VU meters in both simple and high-quality tape recorders: I feared that the rise and fall time of 300 ms required by the VU standard would be far undercut, which should rather lead to a lower lead. But at least with the meters I checked, the time constant of approx. 200 - 250 ms is smaller, but not as much as I had feared.

Conclusion

In the end, I made at least a degree of peace with the VU measurement during my investigations. A little bit because the measurement errors were not quite as bad and frequent as I had assumed. Only in slightly less than half of all cases did the VU measured value deviate more than approx. +/-1 dB from the more reliable PPM measured value than I had estimated. At the same time, however, I also found in my investigations that in many cases the error can also be significantly higher, and in rare cases much higher, than I would have estimated. In my tests, it was -5 to +6 dB in extreme cases.

Nevertheless: Under no circumstances does the VU measurement deliver recognizably better results than the PPM measurement.

The Examination

The aim was to determine the differences between the maxima of the two measured values (VU and PPM) from many music recordings. Ideally, both should be the same. In other words, they should also lead to the same recorded level at the same level adjustment

The Problem of the Loudest Spot

You can immediately recognize a major, fundamental problem with the level of any signal: Only if you can look into the future can you set the correct level at the start of the recording. With live recordings you can say to the musicians: “Play something loud”, when recording a record you could theoretically run the whole record first and look for the maximum - but who does that? Even a PPM measurement cannot solve this problem.

The Approach

I have written a program for the evaluation that carries out these measurements over complete pieces of music and over complete lists of pieces of music according to the VU and PPM standards and displays them graphically. It looks like this, for example:

The red line shows the PPM measured value, the blue line the VU measured value, but without lead. Without lead, not least because it is much clearer. And also because +6 dB lead is only a common, but not a standardized value. Ideally, the maxima of the VU measured values without lead should be half of the maxima of the PPM measured values. In this image, this is largely ideal for the maxima 1.0005 and 0.5011 = +6.006 dB; a lead of +6 dB would be perfect here. However, the maxima do not occur simultaneously, but at approx. 35 and 58 seconds respectively.

During the measurement, the maximum values of the two measured values are also recorded for both VU and PPM. These maximum values (and more) are written to a log file for later statistical evaluation.

My Sources

My program can only process WAV files, i.e. uncompressed files. Fortunately, I had backed up my CD collection on the PC as WAV files not so long ago, so this material was an obvious source. Of course, this means that it is only the types of music that I like. Therefore, the test results cannot be generalized quite so easily. The statistics would probably look more or less different for other styles of music. If I'm not mistaken, there is even a recommendation for which type of music or sound source which VU lead should be used.

I also only analyzed the left of the stereo channels.

The Result

I processed a total of almost 2300 pieces of music. Very roughly: Assuming a +6 dB lead, the maxima of the VU and PPM displays differed by less than +/-1 dB for about half of the music tracks. So here are the statistics that were the aim of my investigations:

At least with my source files, such a VU meter tends to show a little too little on average. The lead should be a little greater than +6 dB, but rather less than half a dB.

However, there are too often unacceptable deviations from more correct measurements.

→ Therefore, I come to the conclusion that I had already anticipated in the foreword.

Technology and Further Investigations

Now it gets interesting for those who want to know more about the technology.

VU- and PPM Measurement Methods

I have explained the measurement methods in more detail in my AMBM article. The display and measurement methods in the ALM77, which was made for the Revox tape recorders, are identical to the AMBM. I would also like to provide three links with much more background information - including criticism - on the subject of VU meters: https://de.wikipedia.org/wiki/Vu-Meter as well as https://13dB.de/effekte/metering/vu-meter/ and from Rod Elliot: VU And PPM Audio Metering.

Here is at least a brief summary of the measurement methods:

1. at a constant level the output voltage only increases in the very short moments in which the peak values are only 0 to 24 mV higher than the momentary measured value, and not immediately, but with a time constant of approx. 1.65 ms, and
2. the output voltage decreases in the relatively long times between the peak values with a time constant of 1.5 s/20 dB (= 0.65 s) and a voltage difference of approx. 1 V, see figure on the right.

For sinusoidal signals, this means that over a wide frequency range, the output voltage rises on average to approx. 0.976 V instead of 1.000 V. So if 2 V_pp is to lead to a display of 100%, a correction factor of 1/0.976 = 1.0246 is required.

VU is a rectified value measurement. The average of a rectified sinusoidal signal is 2/π * peak voltage, i.e. approximately 0.6366 * peak voltage. So if 2 V_pp is to lead to a display of 100%, a correction factor of 1/0.6366 = 1.5071 is required.

(In case someone wants to simulate the system or calculate it numerically: A 2-pole low-pass filter with the transfer function U_OUT = (1+0.1585*s +0.0095*s²)*3.14159/2 follows the rectifier at VU. The digital coefficients for a numerical calculation can be determined from this as I have described with Converting Analog into Digital (IIR) Filters).

Here are two more graphics, first the analog simulation:

Here you can see the graph of the display when a 1 kHz sinusoidal signal with 2 V_pp is applied for one second:
Attack: The PPM value quickly reaches the value of 1 V, the VU value reaches 90% after 300 ms and later overshoots slightly in accordance with the standard.
Decay: It takes 1.5 s for the PPM value to reach 10%, the decay of the VU value corresponds to its attack behaviour of 300 ms.

The PPM value is shown here from the start of a 1 kHz burst of 10 ms duration with a higher time resolution:
It reaches 90% after 10 ms (blue). If the signal is 0 after the burst, the PPM value drops immediately (green) in contrast to the VU value, which still rises to approx. 45 mV after the first 10 ms (red) and only then starts to fall again, which is not visible here.

While PPM shows 90% (-1 dB) full scale, VU shows only 4.5% (-27 dB). A huge difference. Even with +6 dB lead, only -21 dB is displayed.

Now once again to check that the software is calculating correctly:

and

PPM and SPPM

Overmodulation can also occur with PPM measurements because the rise time is not 0, although a rise time of 0 would technically be no problem. Why was this done anyway? Again, it is a compromise. If you wanted to ensure that the full level was not exceeded for every shortest pulse, the overall level would have to be lowered in practice. This would reduce the overall volume and would be at the expense of the dynamic range. However, it would be largely unnecessary, as very short clipping is practically inaudible. They are therefore deliberately ignored in PPM measurements. Especially as some recording and transmission media react “graciously” to slight clipping, e.g. only with a slight increase in the distortion factor, but not with clipping. FM transmission, records and tape recorders are among them.

Now there are also cases in which even very short overloads should not occur. This is what the SPPM (Sample Peak Program Meter) is for. This can be particularly interesting with digital signals. Each sample is evaluated and the rise time is 0. I have also recorded the sample peaks with my program. Although this is not relevant for the VU/PPM comparison, but you get an indication of how the CDs or their music tracks are levelled:

Where do these - sometimes Very Large - PPM/VU Deviations Come From?

It is certainly also interesting to know how the extreme deviations shown in the bar chart above can occur. You need to know about the measurement methods and their standards for that.

Measured Values Too High

Sinusoidal signals are assumed in the standards. But what happens, for example, with square wave signals or at least signals that are more or less similar to square wave signals? Quite obviously:

• With PPM, the measured value would rise to 102.46% in extreme cases.
• With VU, the measured value would rise to 150.71% in extreme cases. This corresponds to 300% with a lead of 6 dB(!).

In the following recording, something like this happens for a short moment. Here, a signal is present for a sufficiently long time that is still far from the square wave extreme, but the rectification value is still higher than it could be with a pure sine wave. This is why the VU display, even without a lead, clearly exceeds 100%. And thus even higher than the PPM display is at the same moment. The VU measured value reaches up to 1.08 (this is in my log file). My program is not set up to display this:

A spread view shows the signal curve at the maximum:

The PPM value also exceeds 1.0 towards the end of this high level at 2:48:525, because the original is overdriven at this point. The PPM value is then not only driven higher in the very short peaks as with sine signals, but over a longer period of time, making the measured value slightly higher than it could be with sine.

Measured Values Too Low

Measured values are too low if the signal contains short peaks that are still well captured by the PPM time constant (90% in 10 ms) but are too short for the VU time constant (90% in 300 ms). And of course only if there are no other situations that are more favorable for a higher VU value. In the following example, the VU value only reaches a maximum of approx. 0.23, while the PPM value reaches a maximum of 0.825. This corresponds to approx. 11 dB.

Even with a lead of +6 dB, 5 dB less is displayed than with the more realistic PPM measurement.

And the signal curve, which contains these peaks that are easy to recognize for PPM but difficult for VU:

These two examples look rather disastrous for VU measurements. But as I said, these are only the two most extreme examples from almost 2300 pieces of music. Nevertheless, VU is absolutely out of the question for professional demands.