AUDIO METERING FOR BROADCAST
by Gordon S. Carter
The subject of audio levels in broadcasting always creates a hearty discussion. Audio levels are an important part of broadcasting, both radio and TV. They are extremely important in radio since radio does not have the picture to help conceal poor audio levels. Some people feel that adhering to proper audio levels stunts their creativity in producing a program. However, the simple fact is that there really is a right and a wrong way to set audio levels.
Since the advent of digital audio, it seems (at least to this writer) that the proper setting of audio levels has become a lost art. Many producers have no idea what proper audio levels are, let alone how to set and control them properly. The purpose of this paper is to attempt to clarify the procedures and improve the situation.
The topic of audio metering is essential to setting proper audio levels. Meters are the primary tool we have to see what the audio levels are, especially in relation to any sort of standard. However, meters can be confusing to the untrained eye. The multiplicity of types of audio meters does not help the situation either. A good understanding of what the various meters are telling you and what they are not telling you will help set levels properly.
Possibly the most common audio meter in broadcast facilities is the VU meter. The VU meter was jointly developed by Bell Labs, CBS, and NBC in 1939. The standards for a true VU meter specify many characteristics, including the ballistics of the meter and the scale. We won’t go into a detailed description here, but want to make some very important points.
A true VU meter has a specific impedance (7500 ohms) and is designed to work across a 600 ohm line. Remember that when the VU meter was developed the phone company (the primary supplier of program distribution lines) and the broadcast industry all used 600 ohm audio circuits. A true VU meter will not give proper indications when applied across lines of other impedances. It was designed to indicate “0" when a sine wave signal was applied 4 db above 1 mwatt across 600 ohms.
The VU meter specification calls for a full-wave rectification of the audio signal. Full-wave rectification insures that the meter will respond to both positive and negative peaks, which is important when dealing with real-world audio signals, which are rarely symmetrical.
Another important specification for the VU meter is the meter ballistics. A true VU meter should indicate within 99% of the true reading with a 300 ms tone burst applied, and then fall to within 5% of the true reading when removed. It should also have a frequency response of +/- .2 db from 35 to 10,000 Hz and within +/- .5db from 25 to 16,000 Hz. There is even a specification for durability of the meter under overload conditions.
A VU meter is essentially an average reading device. A good VU meter can be a very accurate indicator with tone applied, but requires some knowledge and interpretation to be truly useful under real-world audio conditions. We will discuss this in a bit more depth later in this paper.
Starting in the mid to late 1960's several manufacturers introduced VU meters that did not meet these standards. These meters, while often cheaper than a true VU meter, did not indicate properly and in some cases were actually worse than not having a meter at all. Some would actually load down the audio circuits, and then indicate improperly, further compounding the problem.
The advent of other types of indicating devices (other than a mechanical meter movement) further complicates the problem. Some of the ballistic characteristics of the VU meter were partially determined by the mechanical characteristics of the meter. Newer devices such as LED indicators require some electronics to indicate properly.
While true 600 ohm circuits are not as common as they once were, VU meters as defined in 1939 are not as useful as they once were. A more useful metering device on modern circuits is an electronic simulation of a VU meter. It should not load the circuit, have full-wave rectification, and have proper ballistics. However, most modern meters are actually indicating dbu rather than dbm. (dbm is referenced to 1 mw, dbu is referenced to the voltage generated by 1 mw across 600 ohms or .775 volts) This makes little difference in the real world, as long as it is properly calibrated.
At about the same time as the VU meter was developed, the BBC (British Broadcasting Company) was developing the PPM or “peak programme meter”. While the PPM has a scale roughly similar to the VU meter, its characteristics are much different. The PPM has a much faster attack time (within .5 db of the steady state value with a tone burst of 10 ms) and a slower release or fall time (takes 2.8 seconds for the pointer to drop 20 db). All versions of the PPM require some electronic circuitry to perform properly, regardless of the display type. While the VU meter is essentially an average reading device, the PPM, as its name indicates, is closer to a peak reading device. Since it has a finite attack time, a better terminology for it is a quasi-peak reading device.
For many years all broadcast operations were required to have a modulation monitor. Modulation monitors also have meters that indicate audio levels. These meters, while having a scale very similar to that of a VU meter, actually were faster than a VU meter but slower than a PPM. The idea is to show the peaks of modulation. Since over-modulation can cause distortion in receivers and even interference to other broadcasting stations, it is important to insure that a station controls and monitors the modulation peaks. These meters were specified so that peaks of between 40 and 90 ms would indicate within 90 % of full value, and the discharge rate would be 500 to 800 ms to return to within 10% of zero.
Remember that the standards for the VU meter, PPM, and modulation monitor were all defined in the days when the only indicating devices were meters. The advent of newer display technology and solid-state electronics makes possible faster response and different scales for the indicating devices.
With the advent of digital recording and transmission, a new type of indicator was required. Previous to digital, all audio metering could tolerate some “slop”. A small amount of inaccuracy or overshoot would not create a significant problem. However, digital reacts in a different manner. Until clipping is reached, there is no appreciable change in the distortion characteristic. At the instant clipping is reached (all bits on) severe distortion is generated. A meter that would accurately show this point under dynamic conditions was required. As a result, most digital devices have some sort of metering that indicates true peaks. These meters respond very quickly to even the slightest peaks of audio to warn the user of potential problems.
There are many other types of meters available in addition to the basic types mentioned above. Most of them are variations on the basics. Some have different ballistics. Some have multiple indicators in the same display, showing both peaks and average. An interesting variation on the peak meter is one that is commonly called a “peak-hold” display, where the indicator responds to the peaks, but then keeps the top segment lit while the rest of the display shows current peaks. Depending on the device, some hold the peak almost indefinitely, while others begin to go down after a period of time.
Regardless of the type of meter being used, it is important that the meter be calibrated properly and that the user be aware of what the meter is telling him. Each type of meter has its own strengths and weaknesses. No one type of meter is ideal for all situations. Some people will even argue that one type of meter is better than others, but that is only true for specific applications. There is no one meter that can be used for everything. If there were, there would only be one type of meter and this paper would never have been written.
Virtually all audio indicating devices (meters) have scales calibrated in db. It is important to remember that db is a ratio. To be meaningful, there must be a reference for the meter.
The db is a logarithmic ratio. For those of you not familiar with this area of math, a logarithm is a power of ten. For instance, 10 0 (read as “ten to the zero power”) is 1. 10 1 is 10, 10 2 is 100, and so forth. Translated into db, 10 db is 10 times the volume, while 20 db is 100 times the volume. It does not take a very large change in db to make a significant change in the loudness or volume of sound.
Earlier we mentioned that 0 dbm was equal to 1 mw. Here is a short list of common db references.
dbm – 1 mw
dbu – .775 volts
dbv – 1 volt
dbfs – full scale (all bits on in a digital device)
For the balance of this paper we will assume that any meter is properly calibrated to its appropriate reference unless otherwise noted.
While a modulation monitor is no longer required at a radio or TV station, the station is still required to maintain modulation within proper limits. One of the reasons that modulation monitors are no longer required is the increased use of modern audio processing. Today’s processing can control modulation peaks more tightly than a modulation monitor can detect. In fact, some modulation monitors actually had more overshoot than modern processors. While these meters are not very common today, a brief description as to how they are used can set the groundwork for other types of metering. A modulation meter had a dual scale. The upper scale was calibrated in modulation percentage, while the lower scale was calibrated in db. The scale had to extend to at least 133%, which is approximately 3 db above 0. Remember that the modulation monitor is the only meter we will discuss with an absolute calibration. 100% (0 db) as shown on the meter must equal 100% modulation of the transmitter. With this scale, 50% modulation was at -6db. Under normal operation the audio peaks were held to the 100% point. Since this was reading peaks, uncompressed audio would frequently be quite a bit below this point. Uncompressed audio, especially classical music, will sound very quiet when compared to compressed audio.
VU meters were frequently used on audio consoles and audio tape recorders. In a properly designed system, the audio level would average +4 dbm at the output of the console. The audio circuits were designed so that the overload point would be somewhere between +18 and +24 dbm. The difference between the calibration point and the overload point was called “headroom”. In normal operation, the audio levels were maintained so that the average was at 0 on the meter, with some peaks going above 0, but not so much that they would stay there or cause the meter to slam against the mechanical stops. The average response of the VU meter is closer to what people hear than a peak meter, but does not tell the whole story.
Audio can sound louder or softer, based on spectral content, density, and other factors, even though the average as indicated by a VU meter is the same. The first example shows two voices adjacent to each other. The electronic bar graph meters show peaks, while the mechanical meters are true VU meters. Notice that the male voice sounds quieter than the female voice, even though the peaks of both are the same and the VU actually shows the female voice lower than the male. The primary problem here is that the male voice is very bass-heavy, while the female voice has spectral content spread over the entire frequency range. If we were to compare the voice to music, the voices would sound much louder than the music if we kept the VU meters the same. Depending on the quality of the voice, we recommend that voices be somewhere between 7 and 10 db lower than music on the VU meters.
When VU meters are used on analog tape recorders, the calibration procedure is a bit different. While the electronic signals are the same as in a console, the true limiting factor in analog tape recording is the tape itself. Additional calibration pots between the electronics and the tape heads allow adjustment of the signal to the head independent of the meter. This adjustment would be made so that 0 on the VU meter was a specified number of db below overload of the tape. Since analog tape does not have a hard overload point (distortion increases gradually as the flux density on the tape increases) the overload point was usually defined as a given percentage of distortion on the tape. This would typically result in the tape reaching overload slightly before the electronics, but since it was a soft overload, it would not be a problem.
The PPM was not as common in the United States as in Europe, but it was not totally unknown here. The PPM has a different scale than the VU meter, even though it is calibrated in db. 0 is higher on the scale, and the scale is calibrated to be approximately linear. For instance, the distance from 0 to -4 db is about the same as from -4 to -8 db. The action of this meter is also different from the VU meter. In normal use the meter is calibrated so that 0 on the PPM is near (but not quite at) signal overload. The typical calibration point is +16 dbm for 0 on the PPM. The audio would be kept at or slightly below 0, but some small excursions above 0 are permissible. The PPM would appear somewhat hyper-active compared to a VU meter.
With the advent of digital recording and its associated abrupt catastrophic overload point, new metering techniques were needed. As a result the peak meter was developed. This meter is calibrated so 0 is at 0dbfs (the clipping point). These meters are usually electronic displays and have very fast response time, often holding the peak for a short while to allow the operator to see that the peak has been reached. The readings of the peak meter may have very little relationship to the loudness of the audio. A very short peak, perhaps so short that you can’t hear it, may cause the meter to show a peak up to overload. Voices can be a very difficult problem with peak meters, as they tend to be asymmetrical. With an asymmetrical waveform, the peak meter may show very high, but the average (what you hear) could be very low.
All of this has been further complicated by the tendency on the part of some producers to attempt to record on the digital media at as high a level as possible. While higher recording levels (as long as they are below overload) produce the lowest distortion with digital media, it also tends to produce very inconsistent average levels. Sloppy recording practices have even further complicated the problem.
Producing for broadcast
The producer of any audio recording should be aware of the ultimate use of his recording and produce it accordingly. Since the topic of this paper relates to broadcasting, we will focus on that.
First of all, let’s take a look a the limitations of the medium. Under the best conditions, most FM receivers provide no more than 60 db of quieting. That means that we can expect no more than 60 db from 100% modulation to the noise floor, otherwise known as signal to noise ratio. If the receiver is located in a moving car, office, or just about any place in the real world, the noise floor is further degraded by the ambient noise. A car in heavy traffic may have no more than 30 db of usable signal to noise ratio at some times. With HD radio, this improves somewhat, but there are other considerations to factor in.
Most radio stations employ some sort of audio processing. The processing typically reduces the dynamic range of the audio. The high end is restricted to prevent overmodulation. The low end is typically raised to help keep the audio above the noise of the transmission and receiving system. If the audio levels are too high or too low relative to what is expected by the processing, the processing will not be able to do its job effectively.
Also, many stations employ some sort of automation, at least during part of their broadcast day. Even when they have an operator on duty, he may not be the most skillful or may be distracted by other duties. All of these factors demand that audio produced for broadcast have fairly consistent levels. Of course, the producer has to also allow the dynamics of the music to come through, especially for classical music.
In order to produce a program properly, the producer must have a suitable place to produce the program. He needs good monitoring, with minimal distractions and noise. While a volume control is convenient, he needs to be aware of the setting of the volume control at all times. In addition, some sort of metering is helpful. The specific type of metering can be the producer’s choice, but he needs to be aware of its proper use and limitations. Also, the recording device must have proper metering to prevent overload.
A standard must be set for proper recording level and all recordings must adhere to it. In the days of analog tape, producers would often put a reference tone (0 VU) on the head end of the tape. If the program was produced correctly, this was useful to anyone who played the tape later. However, I have seen some cases where the tone was recorded and then promptly ignored. This makes the tone totally worthless. I even know of at least one case where a reel of tone was recorded and then cut and spliced at the beginning of the program, with absolutely no regard to the actual levels on the tape. Most producers have established an average level somewhere between -12 and -20 dbfs as their standard. NPR as established -12 dbfs for their uplink.
I recommend that you establish -15 dbfs as your average level, while trying to keep all peaks at or below -3 dbfs. Set your audio monitors (speakers) for a comfortable listening level with music at this level and mark the volume control. This is your reference listening level. While watching the meters to make sure nothing exceeds the overload point, produce your mix so that everything sounds consistent. This may take some trial and error at first, but will become a very natural thing to do with practice. You may change the volume control to hear more detail, but the important thing is to return it to the marked point to make your transitions.
When producing, the most difficult thing to do is get the transitions correct. When you go from one piece of music to another, from music to speech, from speech to music, and even from one voice to another are the places where it is difficult. It is at these places where you need to be especially careful of how the program sounds. For instance, if the music ends quietly and the voice following it comes booming in, it can be very jarring for the listener. After this goes through the station’s audio processing, the voice can sound even louder. The listener will immediately reach for the volume control, which will soon translate into reaching for the tuning knob to find a station that isn’t so jarring. On the other hand, a well-produced program that has properly dealt with this issue will be less jarring and make the listener think that you actually know what you are doing, instead of just “playing” at radio.
Proper control of audio levels for recording and broadcast seems to be a fading art, but it doesn’t need to be. Some simple procedures and lots of practice can turn most people into good producers. Remember, there is no substitute for experience!