A novel approach to automating NPR’s All Things Considered and other satellite-fed network programs without cue tones or other break signals.

By Chris Scott C/E, WKYU-FM

This white paper, illustrations, schematics, and examples are Copyright © 1997 by Chris Scott. EUonline has converted his material into this format for publication with his permission. Chris is making his complete work freely available to all NPR member stations free of charge, with the request that stations implementing his work consider a small donation to WKYU to defray the costs of research and development.

Abstract Public radio stations have had mixed success using conventional systems to automate the National Public Radio network's All Things Considered. NPR is unique among radio networks in that, for a variety of reasons, only a minority of its programs are produced with cue tones to provide member stations with a signal to begin a local insertion.

Instead, NPR programs use scheduled "timeposts" for breaks. Theoretically, these one second windows would be simple to automate, given the available time synchronization accuracy. In practice however, several factors conspire to cause cutaway transitions to occur during network announcer speech, resulting in "upcut" or truncated audio. A system is in use at The Public Radio Service of Western Kentucky University that effectively solves this problem, using a time-constrained "pause detector" to trigger the break and rejoin the network, as well as to trigger recording embedded promos in the background.

 
Concept Block Diagram
Conceptual Block Diagram of the system


The complete system is shown in the Concept Block Diagram (left). The integration of the components was a result of empirical study of the Timing Environment and automation Time Synchronization with the network. We will explain in detail what these limitations are, and how we overcame them to make a more seamless automated operation.

This paper is divided into the following sections:

Explanation of the Timing Environment. A discussion of where the pause for stations to leave and join should occur, and where is usually does occur.

Time Synchronization. Keeping your automation system's clock synchronized to the network is important when trying to join and leave the network cleanly.

Sync Time Relocation. Things get busy at the top of the hour---other events may draw your automation computer's CPU away from the important task of keeping the automation clock synchronized with the network. We present two circuits that address this problem.

Timed Breaks. A brief discussion why relocating the coordinating sync pulse isn't enough.

The Novel Solution. We present a schematic for a straightforward, but sophisticated circuit that has the smarts to say "go" when it senses a pause in programming. In addition, a through narrative explains how the circuit works, and how to make adjustments if necessary.

Automation System Integration. We discuss how WKYU put the above pieces together to create a robust, reliable automation system that works with "the real world" timeposts, including a sample Cartworks automation script to illustrate the concepts.

Conclusions

 


 

The Timing Environment
NPR 1-second break timing illustration

At the time of this writing, an NPR cutaway pause that occurs at the top of the minute actually begins at :59 seconds plus-or-minus variance, with program resuming at :00. The :59 out-post is variable due to the difficulty of fitting copy and prerecorded material to reach the second precisely, with the resume post being more consistent.

Therefore, both the cutaway time and the theoretical one-second pause duration are variable.

Empirical data suggests that the pause may  be as brief as one second, or during weekend network programs, even less. Figure 1, left, depicts these relationships.

 

 

 

 

Time Synchronization
At the member station end, periodic correction of the automation PC clock is required. Global Positioning System-based frequency standards are now available at reasonable cost that produce rubidium-class performance. This approach merits consideration.

For a lower cost system, the OS/2-based Satellite Operating Support System (SOSS) PC hardware provides a convenient momentary signal each hour at 00:00 allowing precise automatic sync. Unfortunately, this is often a busy time for a station's automation system since other demands on the CPU timeslice may intermittently cause a reaction time variance, affecting sync accuracy.

Further complications may develop when universal program compatibility must be maintained, as there seems to be no standard among program syndicators regarding pre-correction of the quarter second satellite delivery system latency (i.e. the delay created when the digital signal must make a round trip to the satellite and back, and be converted back into analog audio).

Assuming that member-network synchronization is achieved, a more serious difficulty is that many automation systems allow events to be scheduled only in integer, whole-second increments, which seriously limits break timing adjustment unless the basic automation clock sync is skewed. This skewing must be avoided if universal program compatibility is needed.

 

Sync Time Relocation  

Delaying the time sync until the automation computer's CPU sheds some top-of-the-hour tasks will help prevent the aforementioned potential variance. Some stations add an SOSS event each hour for this purpose. At WKYU-FM, this operational overhead and schedule clutter is avoided through the use of a hardware solution.

The SOSS "time sync closure" is set to 15 seconds duration, with the SOSS sync time delayed until 00:15 using an interface circuit which transfers the trailing edge of the pulse to the sync input of the automation system, Figure 2 (below).

Delayed SOSS Sync pulse timings

The SOSS simple Time Sync Relocator illustration (thumbnail size below left) shows a simple interface circuit, and the SOSS precise Time Sync Relocator schematic (thumbnail-size below right) shows a more precise quartz-controlled, firmware-based version that can be used for greater accuracy in longer delayed relocations.

The SOSS output pulse duration is selectable in 100 millisecond increments, allowing fine adjustment of the automation clock. If synchronized every hour, it's estimated that this agreement between NPR and the automation clock can be maintained to less than plus or minus 100 milliseconds.

Depending upon the PC clock drift, less frequent sync may be tolerable. With this accurate agreement, it would seem reasonable to initiate a centered cutaway at :59, and expect the near half-second margin of error to prevent audio clashes.

(Click on a thumbnail image to see it  full size in a new window)

Click to view the simple Time Sync Relocator schematicClick to view the Precision Time Sync Relocator schematic

 

 

Timed Breaks Even after experiments in clock skewing, the experience at WKYU was that on average, only 80 to 90 percent of the breaks were handled well. The most common break corruption was truncation of NPR announcer speech. The minor timing synchronization error contributed only slightly to break corruption. Human factors at NPR introduce timepost variances; it’s very burdensome for announcers and producers to end each audio segment at :59 seconds and resume at :00 perfectly, every time, without compromising program flow. WKYU management decided that these timed break failures compromised the air product too much to allow automation to be substituted for a human "board operator." Clearly, further experimentation was in order.

 

The Novel Solution An approach was pursued which augments timed break triggering with additional "smarts" to track break timing variances. A pause detector was developed that senses absence of audible audio, and signals the automation system when the duration exceeds seven-tenths of a second.

The pause detector had to provide a reliable output signal to the automation system, cue-tone-detector style, and although simple silence sensors exist for detecting automation system failure, they had not been adapted for this purpose. The circuit design goal was to prevent transition during detectable audible program material, and we took several steps to design-in high reliability to prevent false triggering.

First, the threshold of audibility detection was carefully controlled. The sensitivity of the unit is augmented by shaping the frequency response to emulate human hearing. Using the well-known Fletcher and Munson response curves (see our quick sidebar explanation on Fletcher-Munson curves), we modeled the weighting factor after the mid-level intensity curve, realizing that most listeners would have their receivers somewhere at this level or lower.

Second, since it’s also known that quick pulses of frequencies normally audible will not generally be heard if they are brief enough, the attack time of the level detector was tailored to approximate the DIN standard 45406 for peak program meters which describes this effect. Thus, "blurts" of sound will not falsely inhibit the pause trigger.

Under normal conditions with a perfectly clean one (or even two) second break, the above frequency response tailoring and attack time control are not necessary. But to make the circuit maximally robust, and able to cope with background noises, such as paper shuffling, these refinements had to be built in.

A Closer Look at the relative timings around the breaks.

The "inaudible silence" or pause must last at least 700 milliseconds (0.7 Seconds) to be considered a "valid" break, and if any sounds pass through the filters during this time, it is reasonable to believe the audience would hear the sounds too, and the timing circuit is reset to wait for another "inaudible silence". In the case of too short a break (600 ms), if the attack (timer reset) time were 100 ms, the break would still be triggered with clipped audio.The Pause Detector Timings & Breaks illustration (above) depicts graphically the timing relationships. Click to view the full size Pause Detector Schematic

Figure 3 is a full-size (84.9K jpeg) schematic of the circuit. (click on the thumbnail graphic, left, to see the entire schematic in a new window. The circuit description follows below).

 

 

A Description of the Pause Detector Circuit
(See the full-size schematic, if you haven't already done so.)

A four-section LM324 Operational Amplifier is the heart of the circuit, and performs these tasks:

1. The Input Differential Amplifier

2. The Filter-Detector

3. Timer

4. Peak Comparator

Also see (below): Inhibiting the Pause Detector output | Power supply requirements.

Black Squiggle.gif (270 bytes)The input differential amplifier is biased to half supply (+9 volts) by R4 and R5, and converts a balanced line level input (0 dBm to +8 dBm) to a single ended usable level. Slight high pass filtering is provided by C3 and C4, which also block dc offset.
Below unity-gain operation is achieved to allow the upper 30 dB of program dynamic range to be compared. The output is AC coupled into a bandpass network that rolls off above 5KHZ and slowly rolls off frequencies below 500Hz.
The right PC trimpot VRl, controls detector input gain, with the center position approximately correct for +8 dBm line level.

Black Squiggle.gif (270 bytes)The detector is biased at silicon junction voltage by D1. The output is rectified by D2, which also removes most quiescent dc. The combination of the reactive input components and the feedback network in the first two stages (C3, C4, C5, C6, C8, C9, and  C10) forms the bandpass filtering approximating middle-intensity (60dB to 70dB) Fletcher & Munson response.
Seven-tenths of a volt output is the threshold for both the green (level present) LED and the timer reset, achieved simply with NPN transistor junctions.

Black Squiggle.gif (270 bytes)The timer section compares the charge on C12 against a resistor-network-derived reference voltage, feeding the open-collector style "Pause out" port through a red LED.The left PC board trimpot (VR2) sets the charging rate of C12. When centered, the time constant should be approximately 700 milliseconds.

Black Squiggle.gif (270 bytes)A second Comparator forms the peak level indicator (yellow LED) which should blink on normal-level program peaks.

Should it be necessary to inhibit the output, a relay or transistor closure applied to the mute input prevents the LED and transistor from coming on. This outboard closure should be able to sink 20 milliamps.

Power supplied to the circuit should be well filtered +12 to +16 volts dc, with a maximum current of 60 milliamps. In high RF environments, a shielded enclosure and bypassed port connections are recommended.

The completed Pause Detector

 

 

Automation System Integration

WKYU uses the Cartworks PC-based digital audio cart replacement and automation system. This system was chosen due to its user-friendly operation and cost-effectiveness. A scripting language is provided that permits great flexibility in handling automation tasks and simultaneous background recording with parallel "script processors" allowing various combinations of simultaneous task execution. Cartworks accepts external trigger signals using a predefined "watch window" of time that can be set as briefly as 12 seconds, up to a maximum of 24 hours.

The pause detector is permanently connected to the network downlink demodulator. This is essential if program re-entry is to be pause controlled. Its output feeds one of Cartworks' twelve external trigger inputs. The conceptual diagram is shown in Figure 4. (below).Concept Block Diagram (Figure 4)

This trigger is used not only to sense cutaway breaks, but also to simultaneously record embedded promos fed during these breaks.

By time-constraining the pause detector trigger---that is, looking at the output of the Pause Detector only around the time a break is expected--- audio switch operation during the body of a program segment is prevented.

The automation script event allows pause detector triggering at :57 and ends 12 seconds later. A narrower window of perhaps five seconds would be optimal, but the Cartworks minimum of 12 seconds has proved quite satisfactory. Breaks are typically triggered at :59, plus or minus network variances. Incorporated into the automation script is a "safety event" for the network rejoin, or in the case of background recording, a recorder stop command.

We implemented the additional "safety event" in the script, although examination of the automation logs shows it has not yet been needed, because it was believed to be essential that a rejoin to the network at re-entry time must happen, even if a major network variance resulted in no pause at re-entry time.
This safety feature consists simply of a timed rejoin at the last possible moment.

Figure 5 (below) shows a sample script depicting event structures to handle cutaways and simultaneous background recording.

Figure 5    SAMPLE CARTWORKS SCRIPT LISTING
NOTES:
[comments are in brackets]
Event 2 is a time sync.
Events 3 thru 6 are the :06 past-the-hour ATC 30 second break
Events 11 thru 15 are the :29 past-the-hour ATC break and promo record

If the network does not give a sufficient pause to trigger the detector, the event is ignored.
Note that "Output:" refers to audio switching.

[We're now looking for the trigger from our SOSS simple Time Sync Relocator circuit]
002 - 16:00:15 - Input Closure
Input: (7)-NPR Time sync    Watch Window: 16:00:09 - 16:00:21
Sync Clock

[load our 30 second local insert material]
003 - 16:05:00 - At Event Time
Load Carts 030 sec

[If we have the trigger from the Pause Detector, turn off the Demod and play our local insert material]
004 - 16:06:03 - Input Closure
Input: (ll)-DMD 6 Pause     Watch Window: 16:05:57 - 16:06:09
Play Carts
Output: (4)-Cartworks ON
Output: (8)-DMD 6 OFF

[safety event, just in case the network doesn't give us a pause, we'll rejoin anyway]
005 - 16:06:31 - At Event Time
Output: (4)-Cartworks OFF
Output: (8)-DMD 6 ON

[absolute event refers to the "parallel" task processor, and we're confident that the net WILL give us a pause]
006 - 16:06:33 - Absolute Event
Input: (ll)-DMD 6 Pause     Watch Window: 16:06:27 - 16:06:39
Output: (8)-DMD 6 ON 

[load our 60 second local insert material]
011 - 16:28:00 - At Event Time
Load Carts 060 sec

[multiple tasks done by same "processor" in same event.]
[when we have the trigger from the Pause Detector, we do the following things:]
[1. Switch the audio; 2. Play the local material; 3. Record the forward promo]

012 - 16:29:03 - Input Closure
Input: (ll)-DMD 6 Pause     Watch Window: 16:28:57 - 16:29:09
Play Carts
[set up to capture the MORNING EDITION/WEEKEND EDITION forward promo]
Start Recorder Cart Name: PRME-WE (Record in Stereo)
[note that "rec" indicates record bus assignment]
Output: (4)-Cartworks ON
Output: (8)-DMD 6 OFF
Output: (9)-Rec Console Audition OFF
Output: (10)-Rec DMD 1 OFF
Output: (12)-Rec DMD 6 ON

[we have captured the promo, and we wait for the Pause Detector to tell]
[us when to rejoin the network and stop the promo recording]

013 - 16:29:33 - Absolute Event
Input: (ll)-DMD 6 Pause     Watch Window: 16:29:27 - 16:29:39
Stop Recorder
Output: (9)-Rec Console Audition ON
Output: (10)-Rec DMD 1 OFF
Output: (12)-Rec DMD 6 OFF

[safety event, just in case there's no pause from the network]
014 - 16:30:01 - At Event Time
Stop Recorder
Output: (4)-Cartworks OFF
Output: (8)-DMD 6 ON

[There IS a pause from the network, and we rejoin the net]
015 - 16:30:03 - Absolute Event
Input: (ll)-DMD 6 Pause     Watch Window: 16:29:57 - 16:30:09
Output: (8)-DMD 6 ON

 

Conclusions This system appears to be a robust interim automation technique to be used in conjunction with, or even in place of, cue tones or other external signaling. Automation clock sync accuracy is increased while stability requirements are considerably reduced.

Due to the considerable technical differences between the various makes of radio station automation, the method may not be transportable to every system. It is believed, however, that the majority of automation systems will permit use of the technique, since most allow for external triggering from cue tone detectors using some form of "watch window."

WKYU is making this comprehensive information package including the detailed timing diagrams, system interconnections, schematic diagrams, and sample Cartworks scripts free to every NPR member station. Should your station adopt this method, please consider a small donation to the WKYU to help defray its development costs.

In addition, a commercial vendor has the completed pause detector module available.

 


Chris Scott is Chief Engineer for the Public Radio Service of Western Kentucky University, You can contact him at WKYU-FM, Academic Complex room 248, Western Kentucky University, Bowling Green, KY 42101, (502) 745-3834, email: chris@scott-inc.com 

Chris Scott & Associates, [P.O. Box 52, Bowling Green, KY 42101; (502) 781-5301] manufactures the Pause Detector module. Please see http://www.scott-inc.com/ for pricing and availability.
Western Kentucky University is in no way connected with the sale or warranty of this product.

Return to Features

Return to the EUonline Home Page