Need help with a Limiter/Meter problem!

[Brena]

I regularly use the Massey plug-in limiter in my Pro Tools HD system to brick wall limit audio for feature film and network TV shows. I have found the massey plug-in to be very accurate and easy to use with successful results every time... until now.

We use the Tektronix WVR 7020's at our post facility in our tape room and QC dept. to check phase, peak levels, etc. The problem we are having is the dbfs meter in the Tekronix shows peak levels to be 2 dbfs hotter than what is being output through our audio bay with the Massey plug-in brick wall limiter. This is a problem because it wont pass QC.

For example, I output a stereo AES track (from my HD rig) with peaks limited to -6 dbfs and layback the audio to Digi Beta. When the AES output of the Digi Beta hits the WVR meter it reads the -6 peaks at -5 to -4 dbfs. I also tried sending three 1k test tones from my Pro Tools bay strait to the WVR. One at -20 (no limiter), second limited to -10, and the third limited to -6. The WVR metered all three accurately to what was being sent. I then tried the same test with full scale pink noise and got different results. -20, -10, and -6 all read to be about 2 dbfs over what was being sent. I then sent my original stereo audio limited at -6 to a second audio bay and a Final Cut bay. This was to give it the old litmus test. Both bays showed peaks at -6. The same peaks that are being read by the WVR which is showing up to -4.

There have been many shows we have used the same hardware/software on before this with no issues and passing QC with flying colors. Why is this particular audio reacting differently?

Is the Tektronix meter more accurate than both Pro Tools HD and Final Cut, or is the Tektronix giving a false reading?

Any advice or suggestions would be great.

[Max Holland]

Well from what youre saying, there are three separate systems with different software showing the numbers as you are setting them, and the Tektronix is the odd man out... how many tektronix set ups do you have? if its just one than the odds are not in the texronix systems favor.


if you limit the tone or pink noise to 0 (Full scale) and then test that in the tektronix is it showing you are over (clipping)?

[Brena]

Originally Posted by diamondschwin

if you limit the tone or pink noise to 0 (Full scale) and then test that in the tektronix is it showing you are over (clipping)?

I haven't tried that yet. We have about 4 of these Tektronix systems in our facility and the tree that I tested all showed the same results. It seems that Tektronix is the only constant that shows peaks over -6.

Thanks for the suggestion!

[Andrew Mottl]

I don't know the devices mentioned, but maybe there is some kind of intersample peak measurement going on? Maybe, if hit by more than one clipped sample in a row, it will interpret it as a higher signal "chopped" off?

Just a quick thought.

[charles maynes]

what is your output setting on the L2007? you can reduce it, but it still requires being attentive to your levels- The thing you really CANT do, is expect to be able to slam the mix buss, and expect the L2007 to level loc (unless you are using extreme settings which will likely result in distortion.

I personally prefer running L2007's across the dialog buss and the the FX buss so the limiting is unique to each group. but you must still monitor both aurally and via meters your composite output to be safe.

If the QC pass fails doing this, you might want to simple start lowering levels feeding the limiters. And set the output gain of the program buss to a db or 2 lower than the delivery spec.

[dsteinwedel]

I output a stereo AES track (from my HD rig) with peaks limited to -6 dbfs and layback the audio to Digi Beta. When the AES output of the Digi Beta hits the WVR meter it reads the -6 peaks at -5 to -4 dbfs. I also tried sending three 1k test tones from my Pro Tools bay strait to the WVR. One at -20 (no limiter), second limited to -10, and the third limited to -6.

It sounds as if you've skipped this step in your debugging tests. Could the Digi Beta deck be out of calibration? (I've never done DigiBeta laybacks personally, so please excuse me if the machine does not require it.)

-Dave

[Brena]

I think we got to the bottom of things. After testing the audio in question on another very accurate meter (DK Audio), We found that it also read the peaks the same as the Tektronix. This eliminated the possibility that the Tektronix is in-accurately measuring peaks.

What was very interesting is when I put a LP filter rolled off at 16k, I was able to reduce the peaks from -4 to about -5 and -6. This seemed to catch most of the high transient peaks that the Pro Tools and even our Logitek AES hardware meters couldn't detect.

In response to a few of the posts... I was a bit unclear earlier on what elements I was actually limiting. In this particular situation I was conforming Spanish audio to the English vs of a feature film and laying back to ch 3/4 on the DBC English master. So the limiting was all post mix and only to meet network specks. Also, I believe that the Digi Beta Decks AES output requires no calibration, same as Pro Tools 192.

This seems to be a rare event, but just the same we are now coming up with a plan to efficiently deal with this anomaly.

Thanks for all of your suggestions!

[Branko]

There's a great and very helpful thread on GS where Paul Frindle explains the "reconstruction error" phenomena.
He says that, although ProTools meters do not show any problem (because they measure sample levels, not signal) the actual signal we're sending out to converters may contain illegal, higher values. So, it is perfectly possible that a simple signal (like sine wave) passes perfectly, while a complex one (pink noise) shows different values. A simple solution is, as Charles suggested, to turn down your volume a bit, to avoid high values at the limiter input.
Do some search, find PF's posts, they explain a lot of things about digital audio.

[Andrew Mottl]

You might want to try one thin: a rectangle signal. If your Tektronix happens to actually measure intersample peaks, it will definitely interpret the consecutive "peaked" samples as a higher value (due to reconstruction).

What surprises me - you are going digitally from the DigiBeta to the Tektronix via AES EBU? I would assume it to get the correct digital words.

[Max Holland]

something still sounds a little odd to me here.. I use DK Audio meters in my mix room and will see it go above -10 to about -9.5 if my limiter is REALLY slammed hard... your situation seems a bit drastic.

[quadraphonics]

I have seen brickwall limiters such as the L1 and L2 have issues with excessive low frequency information (20 Hz and below). In such cases, I have seen results that describe what the OP has been experiencing. Using a HP filter solved the problem.

Randall

Originally Posted by diamondschwin

something still sounds a little odd to me here.. I use DK Audio meters in my mix room and will see it go above -10 to about -9.5 if my limiter is REALLY slammed hard... your situation seems a bit drastic.

[Brena]

Originally Posted by Branko

There's a great and very helpful thread on GS where Paul Frindle explains the "reconstruction error" phenomena.
He says that, although ProTools meters do not show any problem (because they measure sample levels, not signal) the actual signal we're sending out to converters may contain illegal, higher values. So, it is perfectly possible that a simple signal (like sine wave) passes perfectly, while a complex one (pink noise) shows different values. A simple solution is, as Charles suggested, to turn down your volume a bit, to avoid high values at the limiter input.
Do some search, find PF's posts, they explain a lot of things about digital audio.

Thanks for the info on this "reconstruction error phenomena"... Ill definitely check it out.

Reducing the volume is the most logical choice but not fool proof. I also don't want to unnecessarily reduce the overall level of program for just one or more peaks that happen to slip past the limiter. From now on Im going to meter with the Tektronix during the layback. Then, manually fix any peaks by gently automating the volume during the moment it goes over and punch it back in.

[Andrew Mottl]

As posted earlier - intersample peaks are the "digital cause" of reconstruction errors.

You might want to check out this free plugin while mixing:
Solid State Logic | Music

Let us know what it shows you...
Hope it helps!

[spectrasonics]

Originally Posted by Brena

I regularly use the Massey plug-in limiter in my Pro Tools HD system to brick wall limit audio for feature film and network TV shows. I have found the massey plug-in to be very accurate and easy to use with successful results every time... until now.

We use the Tektronix WVR 7020's at our post facility in our tape room and QC dept. to check phase, peak levels, etc. The problem we are having is the dbfs meter in the Tekronix shows peak levels to be 2 dbfs hotter than what is being output through our audio bay with the Massey plug-in brick wall limiter. This is a problem because it wont pass QC.

For example, I output a stereo AES track (from my HD rig) with peaks limited to -6 dbfs and layback the audio to Digi Beta. When the AES output of the Digi Beta hits the WVR meter it reads the -6 peaks at -5 to -4 dbfs. I also tried sending three 1k test tones from my Pro Tools bay strait to the WVR. One at -20 (no limiter), second limited to -10, and the third limited to -6. The WVR metered all three accurately to what was being sent. I then tried the same test with full scale pink noise and got different results. -20, -10, and -6 all read to be about 2 dbfs over what was being sent. I then sent my original stereo audio limited at -6 to a second audio bay and a Final Cut bay. This was to give it the old litmus test. Both bays showed peaks at -6. The same peaks that are being read by the WVR which is showing up to -4.

There have been many shows we have used the same hardware/software on before this with no issues and passing QC with flying colors. Why is this particular audio reacting differently?

Is the Tektronix meter more accurate than both Pro Tools HD and Final Cut, or is the Tektronix giving a false reading?

Any advice or suggestions would be great.

A couple of observations. In the analog world, we consistently measure instantaneous peaks of 6-10db above audio program content. As a rule, peaks were measured with an O-scope operating in the megacycle range.
As a result, any measurement of audio level, including high excursion peak
amplitude requires similar bandwidth,(WVR). Metering devices operating within the range of 20khz cannot properly represent such voltage excursion due to ballistic characteristics and absence of a peak "hold" type circuit. I suspect that the peak amplitude measurement would have been even greater than indicated if prior stages/devices did not restrict transient material due to headroom limitations.

In short, metering devices operating under unknown variables of audio peak/average content is at best, a guess. Elimination of the peak overload component will allow for consistent audio program control.

[Etch-A-Sketch]

Originally Posted by Brena

Thanks for the info on this "reconstruction error phenomena"... Ill definitely check it out.

Here is the little test Paul Frindle suggests to do...

Create an aux in PT. insert a signal generator. Set it to Pink, -3dB. Now insert a Sony Oxford EQ plugin (with filters). if you don't have one, you can use the 7band EQ3 with protools, although it doesn't show the same results the OXF3 EQ does.

Anyway... in the EQ set the Low Pass filter to 20KHz, set it's curve to 24dB/Oct. Engage it and disengage it and see what happens to the meter in protools. If you can, listen while you are doing this. You'll notice even though the meter moves you aren't hearing any change in sound. That is because the sound isn't actually changing, only the display of the meter is changing based on the reconstruction of the waveform.

With my OXF3 EQ the meter jumps up around 3dB+, wiht the PT EQ it only jumps up about a 1dB. But the point is more to show you that what you are actually hearing isn't necessarily what the meter is showing you.
some meters don't actually look at what the reconstruction of the waveform would be and so there can be peaks up to 6dB hotter than what the digital peak meter is reading when the signal is converted to analog or reconstructed through some other source.

[charles maynes]

excellent illustration there....thumbsup

[Brena]

Originally Posted by Andrew Mottl

As posted earlier - intersample peaks are the "digital cause" of reconstruction errors.

You might want to check out this free plugin while mixing:
Solid State Logic | Music

Let us know what it shows you...
Hope it helps!

Looks like a helpful tool but they don't make A TDM or RTAS vs. I guess that's what you get when its free.

A few others that are pretty useful are the TL Master Meter and the Oxford Limiter. Both measure inter sample peaks and the Oxford is a pretty useful limiter. I just tried both of them in conjunction with our Tektronix meters and the results were interesting.

The TL Master Meter seemed to match the accuracy of our Tektronix meters used in our QC dept. but the event list it creates is pretty worthless. The Oxford limiter worked great, but like our other plug-in limiters, it also does not limit all inter sample peaks.

Ideally, what Im looking for is a TDM Plug-in meter that works like the Master meter but creates a usable event list based on time code.

Anyone know of anything like this?

[ggegan]

I'm not a Limiter plug-in aficionado, but I just had this conversation with someone who seems to understand the situation quite well, and this is how it was explained to me:

The Massey limiter (or any other Limiter as far as I know) is not instantaneous and when it is put on an Aux track, my understanding is that the Limiter cannot use its look-ahead function. This would suggest that high level, high velocity, high frequency signals could result in overshoot for a small number of samples thus giving the readings you are seeing on the meters.

If this is the case, one way you might deal with it is to create your own look-ahead circuit by keying the limiter from audio tracks that have been slightly advanced and then putting an equal amount of delay on the output that the limiter is affecting.

Personally, I feel that if you are getting that kind of overshoot, then you are hitting the Limiters too hard. I would revise your gain structure so that less limiting needs to be done at the final stage. If the lack of look-ahead capability is indeed the issue, then try to deal with the transient peaks at an earlier stage with Limiters installed on the audio source tracks in addition to the Aux.

[Paul Frindle]

Originally Posted by Brena

Looks like a helpful tool but they don't make A TDM or RTAS vs. I guess that's what you get when its free.

A few others that are pretty useful are the TL Master Meter and the Oxford Limiter. Both measure inter sample peaks and the Oxford is a pretty useful limiter. I just tried both of them in conjunction with our Tektronix meters and the results were interesting.

The TL Master Meter seemed to match the accuracy of our Tektronix meters used in our QC dept. but the event list it creates is pretty worthless. The Oxford limiter worked great, but like our other plug-in limiters, it also does not limit all inter sample peaks.

Ideally, what Im looking for is a TDM Plug-in meter that works like the Master meter but creates a usable event list based on time code.

Anyone know of anything like this?

The oxford limiter ought to suppress intersample peaks to within around 0.1dB when the 'auto comp' is on.

If there are slight differences between the the intersample detection and the tectronix measurement, reducing the output by 0.1 to 0.2dB should fix it.

[Brena]

Everyone's feedback has been very helpful! I have a much better understanding of how Digital audio levels are measured in relation to what is actually being translated back to analog.

Gain staging and NOT pushing the limiter too hard seems to be the best way to eliminate the possibility of inter sample peaks sneaking through. Its unfortunate but good to know that plug-in brick wall limiters have their limitations... no pun intended

Thanks again!

[Takeshi]

Originally Posted by Brena

Ideally, what Im looking for is a TDM Plug-in meter that works like the Master meter but creates a usable event list based on time code.

Anyone know of anything like this?

I think Dolby Media Meter would do this, though it's RTAS not TDM.

Best,

TK