Processing and PPM Encoding

[LenoxAve]

I have absolutely NO background in audio processing whatsoever, so forgive my ignorance. I was hoping to find an answer to a question though.

I noticed that back during the implementation of the PPMs (I guess in Philly and NYC around '07/'08?), stations' sound quality seemed to deteriorate. Stations would no longer "pop" out of the radio as much as they used to, and they all sounded a bit more muddy. I wish I had a handle on the more technical terms to be able to explain what exactly I was hearing a bit better.

Does PPM encoding have ANY effect on a station's airchain at all? Or could it have been coincidence that stations were just starting loudness and awful processing wars with each other when PPM was introduced?

 

[LA_Guy]

The stations' audio does pass through the PPM encoder, I'll have to patch it out and see if things sound better that way.

 

[WNTIRadio]

I've never noticed a difference with the PPM encoders in or out of the audio path. The analog PPM encoders do have a A/D and D/A conversion process involved, but from very critical listening I haven't noticed any degradation. The AES PPM encoders are digital all the way though.

I think a lot of what you're hearing is the hyper compression on the latest releases getting re-smashed by the processors.

 

[LenoxAve]

Thanks for the responses and insight... it was something that had crossed my mind, and I figured some experts on here would have the answers!

It's funny how over a few years, audio on stations can deteriorate so badly. I always find it odd that the most highly rated, most listened to stations, at least in NYC, have some of the worst audio. WLTW and WCBS-FM certainly come to mind.

I've often found cleaner stations out in the 'burbs. They may not be EXACTLY as loud, but they're easier on the ears than their big city counterparts.

 

[davideduardo]

I noticed that back during the implementation of the PPMs (I guess in Philly and NYC around '07/'08?), stations' sound quality seemed to deteriorate.

Philly was initially encoded in 2002, not 2008, for the first round of PPM testing. Not all stations encoded at first, but almost total participation in the original tests was achieved. Because stations were already encoded, Philly was ahead of NY in being rolled out.

Since the Philly stations had a lot of time for evaluation, it's safe to say that none complained of any audio degradation or they would have pulled out of the test phase.

Remember also that the other PPM markets started encoding about a year before the actual pre-currency tests began; that was to enable panel creation based on actual listening.

 

[Jay Walker]

Most of the problem these days with CHR sonic quality in my not so humble opinion is the source material. One of the most abused tools made available on ProTools and other desktop editing systems is the "Finalizer". Too often the user cranks the controls "all the way to the right" and hammers the audio to the wall. While the master/CD sounds LOUD and DENSE on the studio monitors, the trade off is when it is aired through a digital processor on FM/HD radio.

Because of the misuse of extreme limiting on the master/CD, current hit music audio blows through the multi-band AGC/Limiter stages of a digital processor and heads straight for the final clipper where it is mauled on it's way to the exciter. It's a real world example of GI-GO, garbage in, garbage out.

When you look the majority of CHR currents on a waveform display, you'll basically see a "2x4 plank" instead of a waveform. With the absense of any peak to average ratio, the algorithm(s) in a digital processor has nothing to work with. There's an interesting white paper written by Bob Orban and Frank Fotti discussing this subject.

Link:http://www.orban.com/support/orban/techtopics/Appdx_Radio_Ready_The_Truth_1.3.pdf

One of these days I'm going to try an experiment that I've been thinking about. The basic idea is to run the audio file through a 3 to 1 or greater software Expander/Compander to see if this will add more/some dynamic range to the file so the muti-band AGC/Limiter stages in the processor has something to work with. Years ago DBX had a hardware device that "put back" dynamic range in compressed audio. I believe it was called the DBX 3BX. An interesting little box for consumer high end audio in the 1970's.

I can say that the Omnia 11 does react better to hammered hits, by my tastes and ears anyway. There seems to be a floating algorithm that keeps hammered audio from blowing past the multi-band AGC/Limiter and as a consequence delivers a pleasing sound while keeping PD pleasing LOUDNESS.

Jay Walker

 

[Jay Walker]

Most of the problem these days with CHR sonic quality in my not so humble opinion is the source material. One of the most "abused" tools made available on ProTools and other desktop editing systems is the "Finalizer". While the Finalizer plug-in is a good tool to have, it is too often the victim of the "turn the controls wide-open" school of thought.

Most of today's hit music releases are so heavily limited on the master/CD, the audio blows through the multi-band AGC/Limiter stages of a digital processor and heads straight for the final clipper where it is mauled on it's way to the exciter. It's a real world example of GI-GO, garbage in, garbage out.

When you look the majority of CHR currents on a waveform display, you'll basically see a "2x4 plank" instead of a waveform. With the absense of any peak to average ratio, the algorithm(s) in a digital processor has nothing to work with. There's an interesting white paper written by Bob Orban and Frank Fotti discussing this subject.

Link:http://www.orban.com/support/orban/techtopics/Appdx_Radio_Ready_The_Truth_1.3.pdf

One of these days I'm going to try an experiment that I've been thinking about. The basic idea is to run the audio file through a 3 to 1 or greater software Expander/Compander to see if this will add more/some dynamic range to the file so the muti-band AGC/Limiter stages in the processor has something to work with. Years ago DBX had a hardware device that "put back" dynamic range in compressed audio. I believe it was called the DBX 3BX. An interesting little box for consumer high end audio in the 1970's.

I can say that the Omnia 11 does react better to hammered hits, by my tastes and ears anyway. There seems to be a floating algorithm that keeps hammered audio from blowing past the multi-band AGC/Limiter and as a consequence delivers a pleasing sound while keeping PD pleasing LOUDNESS.

Jay Walker

 

[bilco]

Some of the perceived probelms with stations that are runnig PPM may be a result of a misunderstanding when the boxes were installed. The PPM data bursts only at the peaks of modulation, so some PD's or engineers may have thought if they had more peaks they would get more numbers. The only problem with that is, the PPM's only measure time listened, not number of bursts. Over compression to get more bursts gains nothing. Low level does hurt, but only if sustained for relatively longer periods. The PPM should work fine at normal not overly processed audio.

 

[FFoti1]

Some of the perceived probelms with stations that are runnig PPM may be a result of a misunderstanding when the boxes were installed. The PPM data bursts only at the peaks of modulation, so some PD's or engineers may have thought if they had more peaks they would get more numbers. The only problem with that is, the PPM's only measure time listened, not number of bursts. Over compression to get more bursts gains nothing. Low level does hurt, but only if sustained for relatively longer periods. The PPM should work fine at normal not overly processed audio.

Having studied the PPM tech in-depth, I can tell you the data is watermarked anytime the audio is present. The only time the watermark disappears is when there audio is silent. It has nothing to do with audio peaks, but more-so when audio is present. The watermarked signal does operate at a relative level below the peak level.

-Frank Foti

 

[jesseg]

One of these days I'm going to try an experiment that I've been thinking about. The basic idea is to run the audio file through a 3 to 1 or greater software Expander/Compander to see if this will add more/some dynamic range to the file so the muti-band AGC/Limiter stages in the processor has something to work with.

You could also demo an Omnia.9 to check out Undo™. It does that (using psychoacoustics for program dependency), and also ~perfectly declips the audio, hence removing that distortion too.

Years ago DBX had a hardware device that "put back" dynamic range in compressed audio. I believe it was called the DBX 3BX. An interesting little box for consumer high end audio in the 1970's.

Yet another rare DBX box that I need to get my hands on to check out. Thanks for the tip. I wonder if it was designed by Bob?

 

[Jay Walker]

One of these days I'm going to try an experiment that I've been thinking about. The basic idea is to run the audio file through a 3 to 1 or greater software Expander/Compander to see if this will add more/some dynamic range to the file so the muti-band AGC/Limiter stages in the processor has something to work with.

You could also demo an Omnia.9 to check out Undo™. It does that (using psychoacoustics for program dependency), and also ~perfectly declips the audio, hence removing that distortion too.

Years ago DBX had a hardware device that "put back" dynamic range in compressed audio. I believe it was called the DBX 3BX. An interesting little box for consumer high end audio in the 1970's.

Yet another rare DBX box that I need to get my hands on to check out. Thanks for the tip. I wonder if it was designed by Bob?

Thanks for the heads-up on the Omnia 9. I've not had the chance to use one. I am a big fan of Omnia and pretty much any of Mr Frank's gear since my first "Vigilante" which I still have proudly in my processor collection. A great piece of gear for the time and will still work in a pinch if needed...

 

[davideduardo]

Some of the perceived probelms with stations that are runnig PPM may be a result of a misunderstanding when the boxes were installed. The PPM data bursts only at the peaks of modulation, so some PD's or engineers may have thought if they had more peaks they would get more numbers. The only problem with that is, the PPM's only measure time listened, not number of bursts. Over compression to get more bursts gains nothing. Low level does hurt, but only if sustained for relatively longer periods. The PPM should work fine at normal not overly processed audio.

To add to Mr. Foti's explanation of the injection level...

The PPM "tag" which some call a "data burst" or simply "code" is approximately 4.5 seconds long. The encoder tries to insert it continuously, as if in a loop, so as many as 13 tags a minute could be sent.

However, if there is no audio to mask the tag, none is inserted. And if the audio "disappears" the tag is truncated. The encoder essentially finds an audio frequency among its several options where there is masking material and adds the tag to the existing audio in a process described in the PPM white paper on the Arbitron site.

The logic is "if there is audio at one of the tag frequencies, insert code." Peak level is not relevant.

Over the years that PPM has been in test and real operation, dating back now a full decade, the only concerns have been with the ability to insert enough tags often enough in spoken word programming to insure that a station gets credit for listening. Most commercials and music will allow for almost constant tagging.

The fact that spoken word stations like KFI, WSB, WBBM, WINS, WTOP, WXNT do extremely well demonstrates that talk stations do get adequate detection; I have not heard discussion on this subject for years now.

However, the "number of bursts" is very important. The PPM system gives credit for a quarter hour if tags were detected in at least 5 discreet (separate and different) minutes in any 15 minute period (there is a "missing minute" algorithm if the minute before and minute after detect the same station). Since there is a minimum number of detections required for credit, the total number of tags broadcast helps insure credit; if a person tunes in half-way through a discreet minute, any tag earlier in the minute is not detected... but one detected later in the minute will give credit towards the needed 5 separate minute detections.

The PPM meter itself stores to memory every single tag detected, and each tag has the unique station or stream code and a timestamp. Only when the data is uploaded to Arbitron is the processing done on the data where one or more tags in a single minute are "counted" towards quarter hour credits.

 

[Jay Walker]

It is my understanding that there is a "sweet" spot regarding program audio levels into the PPM encoder. I added an Aphex Compellor in front of the PPM encoder on the stations I work with to ensure the input level is correct. Seems I recall an ideal peak level is around +3? I was told that consistent levels would increase the validity of encoding.

Another item I was made aware of is the need to avoid long pauses in audio. Since I work with several talk formats, the need was stressed to avoid "pregnant pauses". I don't know if that point is true or if it is a "PPM urban legend"...

 

[rrsounds]

Years ago DBX had a hardware device that "put back" dynamic range in compressed audio. I believe it was called the DBX 3BX. An interesting little box for consumer high end audio in the 1970's.

Yet another rare DBX box that I need to get my hands on to check out. Thanks for the tip. I wonder if it was designed by Bob?

I have one if you want it. ;-) It predates Mr. Orban's time associated with dbx. There was a two rack-space version and then a single RU version (which I bought in the '80s, IIRC). It's pretty cool, and I've done a little 'play' emulation of how it works in DSP.

There was also a similar expander idea in an earlier pre-amp/processor for consumers, by Bob Carver at Phase Linear; he called the process the "peak unlimiter." Never got my hands on that one.

Kind Regards,
David

 

[WNTIRadio]

The best processor to use before a PPM encoder, if the budget allows, is the Ariane. Keeps it smooth and level without artifacts.

I have used compellors in front of them as well, but run them at 7 o'clock on the process balance, so they're mostly a full leveler with no compression. Just a "gentle hand on the fader" to correct wide level differences. Never had a PPM box complain.

I wouldn't run them barefoot, too low a level going in will cause it to not encode properly or often enough.

 

[davideduardo]

It is my understanding that there is a "sweet" spot regarding program audio levels into the PPM encoder. I added an Aphex Compellor in front of the PPM encoder on the stations I work with to ensure the input level is correct. Seems I recall an ideal peak level is around +3? I was told that consistent levels would increase the validity of encoding.

That's sensible. The PPM can not insert the tags if there is not enough audio... whether it be a pause or a low level piece of material. A leveling device like a Compellor would likely result in increased opportunities for the tag to be completed.

This thought reminds me a lot of the original Audimax with the 0 Ohm resistor modification...

Another item I was made aware of is the need to avoid long pauses in audio. Since I work with several talk formats, the need was stressed to avoid "pregnant pauses". I don't know if that point is true or if it is a "PPM urban legend"...

It's true. If the pauses between words and phrases are long, there may be long periods where the 4.5 second tag just can't fit. However, Arbitron does a "missing minutes" sort of thing... if a station was detected at least once in the 2:09 minute, not detected in the 2:10 on, detected in the 2:11 minute, not detected in the next minute and detected in the 2.13 minute, they will get 5 minute's credit as long as no other station was detected in the missing minutes. And with 5 minute's credit, the station gets full credit for the 2:00-2:15 quarter hour.

There is a full discussion on this in an Arbitron white paper at http://www.americanradiohistory.com/Archive-Arbitron/Arbitron_PPM_Encoding_White_Paper.pdf

This even has some discussions of the frequencies used for masking and other deeper details.