LED VU Meter 101

[bountyhunter]

Since a lot of people build VU meters around the LM3914 chips, I will put up some app circuits and information about their use. I also have included a peak detector circuit which is necessary to get a good reading.

In a stereo system, the signals called "line level" are the signals on the outputs of preamplifiers, tape deck inputs/outputs, Dolby device input/outputs, etc. By standard, "line level" signals voltage levels are:

0dB = 0.548VAC (sine wave RMS)

0dB = 0.775V (peak voltage level)

Since all the VU schematics attached include peak detectors, the dB levels are referenced to the 0.775V level for 0dB.

All circuits use a basic dual rail power supply derived from a 12V wall cube and a 3.3V Zener diode.

The circuit in VU Meter 1 is the basic LM3914 circuit with a positive going peak detector. U1 charges the 1uF cap to the peak input signal level.

Resistor RD affects the dot sweep rate: less resistance makes it sweep faster. This can be user adjusted based on preference. The LM3914 dot select pin is left open to select single LED (dot) mode, but in reality, about four or five LEDs are lit at any one time as the thing sweeps the display with one being brightest and the others tailing out. Using the dot mode significantly reduces power dissipation in the LM3914.

The LM3914 circuit in the Meter 1 schematic uses the internal 1.25V reference for maximum signal which corresponds to +4.2dB signal level, and the lowest LED is -15.9 dB. dB levels are calculated from:

Vs (dB) = 20 log ( Vs / 0.775 )

The first circuit is Ok for a VU meter, but when recording, better accuracy in the range right around zero dB makes it easier to get the recording levels where you want them. The second schematic of LED Meter 2 is the same as Figure 1 except the meter range is expanded around the 0dB point. I wanted the peak LED to be equal to +5 dB, so I used a divider to attenuate the signal applied to pin 5 of the LM3914. The lower LED is also shifted up using a divider to raise the bottom resistor of the LM3914 resistor string up to 0.25V above ground. The meter now displays from -6dB to +5dB, the most important range for recording.

The third figure in LED Meter 3 shows the basic circuit modified to monitor both channels by summing them together. This means you only need one meter. In most cases where you are using the meter to adjust volume levels, you can just leave the left and right channel levels the same and adjust their volume based on this combined reading. This type of circuit works very well for recording and makes things simpler.

Frequency Response: remember this is not an audio circuit you need to listen to, it just has to give decent representation of signal loudness. I intentionally use .01uF input caps to roll of signal below about 160 Hz because those low end signals "bury" the meter and don't represent perceived loudness because the ear rolls off at the low frequencies. Also, the LM358 would not be good for actual audio amplifying applications because of bandwidth limit, but has plenty of bandwidth for this type of circuit.

 

[EN0]

Thank you!

 

[audioguru]

The LM3914 is a linear meter for measuring voltage and current.
You should be using an LM3915 logarithmic IC for measuring audio and light levels. Each step is exactly 3dB.

I made a peak detector with an MC33172 dual opamp which is the same as the LM358 except its bandwidth is higher and it does not have crossover distortion.
The first opamp is a mic preamp that you do not need. You can use an MC33171 single opamp as the peak detector.
The second opamp is the peak detector and it is inverting without a negative supply and it rectifies the signal producing a positive output with a transistor instead of a diode. The transistor is inside the negative feedback loop of the opamp so its forward voltage is cancelled.

 

[bountyhunter]

The LM3914 is a linear meter for measuring voltage and current.
You should be using an LM3915 logarithmic IC for measuring audio and light levels. Each step is exactly 3dB.

I would not want a meter with steps every 3dB, as I explained in the text: that's useless for a recording instrument. I need better accuracy and resolution around zero dB to get any kind of control over recording levels on separate tracks from different sources.

To be honest, I am not overly fond of LED meters for recording and finally built an actual needle VU meter with a log amp and it works much better for recording level accuracy but nobody wants needle meters these days... except me.

The recording setup in front of me here that I use for audio processing has the LED meter similar to shown above but it also has the mechanical VU meter which is what I actually use for setting loudness levels. Colored LEDs are pretty, but the needle VU meter gives a much better measure of average loudness which is closer to perceived loudness. Now all the tracks I drop onto the CD's I burn are the same loudness when I play them back, which was my whole point in building the thing anyway.

 

[bountyhunter]

The second opamp is the peak detector and it is inverting without a negative supply and it rectifies the signal producing a positive output with a transistor instead of a diode. The transistor is inside the negative feedback loop of the opamp so its forward voltage is cancelled.

I used a similar peak detector design with transistor in the loop when I wanted more charging current to get faster rise time on the peak charge storage cap, it was used in a output "power meter" which measured the amplifier output so it had to be able to catch really quick transients (design is about 15 years old, circuit attached).

The power meter has a "four-quadrant" peak detector for both positive and negative going voltage peaks for both channels, all summed into a voltage on a capacitor and displayed on the LEDs (calibrated for Watts across either four or eight Ohms). That power meter uses both an LM3914 and LM3915 for two ranges of power.

That kind of speed isn't really needed for a VU meter because there isn't much peak content out past about 5kHz. The op-amp itself can kick out about 15 mA of current to charge the cap, so I didn't use a transistor. Don't see many impulse functions in pop music.

 

[gabeNC]

Thanks for posting these circuits, this is very much of interest to me. Have you made any consoles or mixers with this? Or is this just monitor circuits for your existing board?

 

[Nigel Goodwin]

I would not want a meter with steps every 3dB, as I explained in the text: that's useless for a recording instrument. I need better accuracy and resolution around zero dB to get any kind of control over recording levels on separate tracks from different sources.

To be honest, I am not overly fond of LED meters for recording and finally built an actual needle VU meter with a log amp and it works much better for recording level accuracy but nobody wants needle meters these days... except me.

Bit inconsistant here - you say you don't want a log display, and then build a log amplifier to give you a log display?. If you want more resolution, then put two 3915's in series.

As for no one wanting needle meters, it's more a question of no one been able to afford needle meters - it's a lot cheaper (and more accurate) to use log LED ones.

I see no problem with using a log meter for recording levels, and (as far as I know) all professional equipment does, the main criteria is the attack and decay performance. I seem to remember that the BBC set the standard for that decades ago?.

 

[audioguru]

The 3dB steps of an LM3915 are small audible changes in level and are perfect for recordings. Nobody can hear a level difference of only 1dB.

 

[bountyhunter]

Bit inconsistant here - you say you don't want a log display, and then build a log amplifier to give you a log display?. If you want more resolution, then put two 3915's in series.

Nothing inconsistent about it, since I was clearly referring to a dial VU meter I designed and to get any kind of log scale performance on it, you have to have a log amplifier to feed it. BTW, it is designed to go from -15dB to +5dB and about half the meter range is from -3dB to +5dB. It is very accurate for setting recording levels which is what I need it for. If I get time, I will put the schematic together and post it. I didn't post it because it is definitely not a simple build.

As for no one wanting needle meters, it's more a question of no one been able to afford needle meters - it's a lot cheaper (and more accurate) to use log LED ones.

It is true that good quality needle meters are not as dirt cheap as the IC's, but they work better. I got some at a surplus so they didn't cost that much. I certainly disagree that LED meters are more accurate. basically, it is like comparing a digital representation of an analog waveform to the actual waveform....

I see no problem with using a log meter for recording levels, and (as far as I know) all professional equipment does, the main criteria is the attack and decay performance. I seem to remember that the BBC set the standard for that decades ago?.

Then you did not read my post or understand what I said. Here it is again:

I would not want a meter with steps every 3dB, as I explained in the text: that's useless for a recording instrument.

A meter with 3dB resolution is worthless for recording. And, if you have seen modern (LED) VU meters for professional applications, they have many more LED steps than that (looks like maybe 50 or more across the meter) on the ones I saw on TV. I have no problem with a log amplifier per se, but I want a lot better accuracy in the -3dB to +5dB range with emphasis on the 0dB to +5dB area where the signal is peaking. The 3914 does it better than the 3915. Look at the meter #2 which is the one I use: the range is from -6dB to +5dB and six of the ten LEDs are around 0dB to +5dB range. It is more useful that way.

 

[bountyhunter]

Thanks for posting these circuits, this is very much of interest to me. Have you made any consoles or mixers with this? Or is this just monitor circuits for your existing board?

I use this to measure the sound level coming out of my PC's sound port (the one for headphones) so I can equalize all of my audio tracks to the same loudness level. You would not believe how much range I have seen (probably as far as 15 dB) of variation from the lowest to highest. I guess some people have brain lock when they create digital masters from old analog masters and get the volume level completely wrong.

 

[bountyhunter]

The 3dB steps of an LM3915 are small audible changes in level and are perfect for recordings. Nobody can hear a level difference of only 1dB.

Actually, nearly everyone can hear a 1 dB change. We were taught in school it is the smallest change a person with normal hearing can perceive, and everybody in our class could hear it when the demonstration was done. But, a meter with 3dB steps can certainly not tell a 1dB change and neither can the person looking at it. It's best case accuracy would be 3dB.


When you are trying to center up a volume level of a track and your meter resolution is only 3dB, you have a very poor instrument to work with. Also, the volume levels are reasonably close to the "top end" of the dynamic range for peak clip and you really want to get it right.

 

[bountyhunter]

ANALOG VU METER:

Here is the schematic for the analog VU meter I built using a generic decent meter I had in a drawer. It is built to take the headphone output signal from my PC, I have no idea what standard level that is. All the components were "dialed in" to get the scale I wanted on the meter for the sound levels it was getting fed. The meter face was made after the circuit was finalized and glued onto the meter. If you have a good graphics program like canvas 11 it's very easy to make meter faces. This meter works VERY well for equalizing audio tracks since it measures more of the average sound level which more accurately represents loudness than peak reading.

This analog meter (exactly as shown) and the LED meter in schematic #2 above are built on the same board sitting in front of me. I never look at the LED meter anymore. It is interesting to look at the graphics of the audio tracks using PYRO software and you see why peak reading for volume is useless. Some tracks have full dynamic range and MANY are clipped like somebody ran a hedge clipper across the top and bottom. Using an average meter, you get very accurate loudness measurement but the peak reader is way off comparing a full dynamic range track to a compressed one.

Anyway, I find this meter works best for equalizing recordings, your mileage may vary.

 

[audioguru]

Your hearing sound level range is about 140dB. 3dB difference in levels is barely noticeable and 1dB is almost no change.
Your analog meter has no damping so it has overshoot which ruins its accuracy.

Most modern recording studios have an old analog averaging VU meter and a fast LED peak reading meter to avoid clipping the peaks.

 

[gabeNC]

How do you add damping?

 

[audioguru]

How do you add damping?

The same as with speakers. Add negative feedback to the driver to reduce its source impedance.
An audio amplifier has an output impedance of 0.04 ohms or less for good damping. Then the damping factor of the amplifier is "only" 200 but many amplifiers have a damping factor of 500 (output impedance is 0.016 ohms).

The meter or speaker behave as a generator when they try to overshoot and the negative feedback detects it and halts the extra motion.

 

[ronsimpson]

**broken link removed**
This link (page11) has a section on audio meter standards.
I like to use different attack and decay times depending on the use of the meter.
Peak meter, average meter and RMS meters have different uses.

 

[bountyhunter]

Your hearing sound level range is about 140dB. 3dB difference in levels is barely noticeable and 1dB is almost no change.

I know what a 1dB level change is, have done the test: back in school, we set up a generator/amp and used a large JBL speaker to test the class to see if they could perceive a 1dB change, which is the step the HP calibrated attenuator boxes were set up for. He hid the switch and turned it and the class had to say "up or down" to indicate whether they thought he went 1dB up or 1dB down. The class got it right every time. Had to really listen, but they got it.

Your analog meter has no damping so it has overshoot which ruins its accuracy.

No actually, it is extremely accurate for what I am trying to measure which is the perceived loudness level which is how loud it will sound in my car. Fast peaks do shoot right by, but it does a very good job of displaying average level. It also has roll off below about 250 Hz similar to a dBA weighting which tracks the ear's response as opposed to dBC weighting which is flat. dBA weighting is closer to hearing so it is closer to perceived loudness. I have also seen that pop music is very heavy in bass content which masks true loudness if you don't roll it off. The ear's response doesn't hear low bass well, but if the meter's frequency response is flat, the heavy low bass dominates the signal and that's all you see displayed.

I have a fast peak reading LED meter right by and I can see them both. The VU meter is very accurate for equalizing volume levels across tracks. I have literally done thousands of tracks and no longer have loudness variations I always had before when I was using peak meters.

Most modern recording studios have an old analog averaging VU meter and a fast LED peak reading meter to avoid clipping the peaks.

That's exactly what I have, but I absolutely guarantee that the analog meter is far more accurate at displaying ear perceived loudness. I generally use the LED meter for entertainment.

 

[bountyhunter]

Peak meter, average meter and RMS meters have different uses.

Bullseye. A meter which will display how loud something will sound to the ear is closest to average response with dBA frequency weighting. At least, that's how it works for me.

I was just checking some tracks and it is very easy to see why this is true if you have audio graphic software that displays the sound patterns. The "full range" recordings have higher dynamic range, so the peaks are a lot higher than the average level. The tracks that were compressed (clipped) when they were mastered have a much lower peak to average ratio. When the tracks are set up for the same loudness (both to ear and the analog meter) the peak meter shows peaks about 4 - 5 dB lower on some of the clipped tracks. Ergo, if (as I did before) I use the peaks to set the levels, the average levels (loudness) will be about 4 - 5dB different between tracks and that is noticeable to me.

 

[bountyhunter]

DYNAMIC RANGE:

Attached are two scope shots from two different audio tracks. The right one is fuller dynamic range, the left one was clearly clipped when mastered. This is how they came straight off the disk, the clipped track was at peak levels way below maximum so there was no reason to use that much compression. using a peak reading meter to try to set equal loudness will not be accurate between these tracks, because loudness is basically average value not peak.. Seems like I see more tracks now like the clipped one than the full range one.

Notice that the average levels of the two tracks are about the same, they were adjusted using an average reading (analog) meter. Peak values are very different, hence the error you will have using peak meter for loudness levels.

 

[audioguru]

Your clipped signal must sound extremely distorted. Compressed signals do not look like that and do not sound like that.

Your hearing has a A-filter response (lows and highs attenuated) only at very low levels. It has a C-filter response (almost flat at low and high frequencies) at medium and loud levels. Stereos have a "loudness contour" that boosts the lows and sometimes also boosts the highs when the volume is turned down to compensate for the above.