AP-2000 / FM-2000 FAQ
The AP1000 was our first high-end processor. As with most very complex products we quickly found ways to increase its performance and upgrade its feature set. Up to a point everything was achievable with a simple software upgrade. But we reached an impasse where some of our new DSP algorithms needed more breathing room and we had to redesign the hardware in order to provide these resources.
Every single AP1000 customer was contacted and invited to send their unit back to the factory so we could upgrade their unit to an AP2000 at no cost. To date, 96 of the first 100 AP1000's built have been upgraded and customers are quite happy with the improvements.
Of the remaining four units, two belong to the same customer, who is happy with his AP1000's as is.
The other two units belong to Wheatstone and reside in the racks of processors in our lab where we can compare them to not only our own line of processors, but also virtually any popular on-air processor that was manufactured in the past 30-plus years.
Actually, no. Maybe back in 1975 it would (and it did!), but not now - not with our exclusive band splitting technology.
We've invented a new method to parse and recombine the audio spectrum that doesn't create the recombination artifacts that have plagued traditional band splitting methods. Our frequency bands are loosely based on the Bark frequency scale (also known as the Critical Band Rate) and our parsing method is a highly refined technique perfectly suited to a very high-resolution audio processing application.
This technique is so revolutionary that we've written an application note about it, and you can find it on the Vorsis website. Look for application note AN2008-05 The Magic Behind Vorsis 31-Band Algorithm.
The simple answer is that the 31 band algorithm sounds good simply because it doesn't have a sound!
Because the number of frequency bands is very high each band is therefore quite narrow. When the frequency bands are this narrow we can take advantage of the human ear's masking behavior to hide the work that our limiters and clippers must do, much in the same general way that perceptual codecs such as MP3 and AAC hide theirs.
The peak control performed by our 31 band limiters is not perceived by the ear as distortion or dynamic level manipulation. Rather, it's perceived as an increase in depth of audio detail as well as overall loudness.
Our high band count technique absolutely works, and Vorsis processor customers report now being able to hear subtle details in musical program material that were simply missing when the same music is heard through other brands of processors. Those customers also report that Vorsis processors sound louder on the dial while being more natural, cleaner, and more open sounding at competitive loudness levels.
When there are fewer limiter bands, other program content within a band has to “go along for the ride” even though that material may not have caused limiting on its own. This means that even signals that are not high enough in level to cause limiting are reduced in level by whatever the limiting depth happens to be in that limiter band.
When a traditional processor does peak limiting in one of its half dozen or so bands, it's in response to a single frequency within that band that has exceeded that limiter band's threshold. While that band is in limiting and is reducing the offending frequency, any other program content that is also within that band's (rather wide, by the way) passband is ALSO reduced by the same degree of limiting.
In other words, when there are fewer limiter bands, other program content within a band has to “go along for the ride”. This means that even signals that do not cause limiting on their own are reduced in level by whatever the limiting depth happens to be in that limiter band.
Two important ways in which the behavior of a conventional limiter negatively impacts the on-air sound are:
- It reduces the overall loudness because a broader section of the audio spectrum is reduced in level than the energy that caused the limiter to take action, and;
- Subtle details in the program material that are within the bandwidth of the limiter band (but outside the masking envelope) while the band is in limiting are also reduced by whatever amount of limiting is taking place. These important subtleties in the program material are unnecessarily lost forever.
In summary, the benefits gained by going from five or six limiter bands to 31 are just as valid as going from one limiter band to five or six.
Dividing the spectrum into a lot of bands isn't that hard anymore given the perfection that can be realized in the DSP domain. What is hard is changing the relative levels of those bands with dynamics processing and then putting all those bands back together again and having everything still sound good.
That second part of the task is certainly not trivial and we have worked very hard to refine our method to ensure that the audio that is processed by it sums perfectly back together again and is always clean and natural sounding, regardless of how deep the limiting may need to be!
Please see our application note on the Wheatstone Processing website titled AN2008-05, “The Magic Behind the Vorsis 31 Band Limiter”. NOTE: This document is unpublished pending a rewrite which is in progress.
All of our on-air processors have less than 11 milliseconds of delay and will cause less of a disturbance to announcers who must listen to the off-air signal. Because our throughput delay is far shorter than other popular audio processors there is simply no need for “Low Latency” presets.
Wheatstone Audio Processor Latency by Model and Mode:
- AP2000 FM 9.40 milliseconds
- AP2000 HD 6.80 milliseconds
- FM2000 FM 9.40 milliseconds
- FM-10HD FM 8.90 milliseconds
- FM-10HD HD 9.60 milliseconds
- AM-10HD AM 8.80 milliseconds
- AM-10HD HD 10.4 milliseconds
- VP8 FM 7.40 milliseconds
- VP8 AM 7.20 milliseconds
- VP8 FM-HD 5.10 milliseconds
- VP8 AM-HD 9.40 milliseconds
- HDP3 4.60 milliseconds
- AP3 4.70 milliseconds
- M1 1.68 milliseconds
VBMS stands for Vorsis Bass Management System. It is a new set of algorithms we've invented that alleviates the bass challenges that can sometimes plague other audio processors.
Bass is a difficult part of the audio spectrum to process. Our VBMS algorithm was designed to overcome limitations that previously prevented a good, solid (and clean) bass sound on the air, especially when competitive loudness is desired.
The VBMS allows deep, natural bass without generating undesired intermodulation products. The algorithm is complex and expensive in terms of DSP resources but the improvement in audio quality makes the expenditure worthwhile.
Please see our application note AN2007-02 Vorsis Bass Management System (VBMS) for more in-depth information about Jeff Keith's VBMS algorithm.
The front panel USB port and its software drivers are designed to support standard USB input devices such as keyboards and simple two button mice.
At this time the port does not support more fancy USB devices such as:
- Wireless mice or other wireless devices;
- Mice equipped with “Track Wheels”;
- Apple MM mice;
- Other types of pointing devices.
Note that the front panel USB port does not support memory sticks and other data storage devices like hard drives.
The composite outputs are also available as balanced wideband composite stereo at the line-level analog output's XLR connectors whenever the analog output section is in its MPX mode.
The reason we provided the balanced output feature is that most modern FM exciters actually do have a balanced composite input - the BNC shell is floating over ground and the input circuit is a true wideband differential amplifier.
The balanced output of our FM processors can be used to interface with such exciters in a way that increases the signal to noise ratio over what the traditional unbalanced connection between the processor and exciter typically offers.
Our philosophy on audio processing is that the end user should be able to tailor the sound of the processor for his own application, whatever that may be. Also, our job as competent audio processor designers is to provide all the tools we can to enable the end user to achieve his special on-air sound, whatever that may be.
Leaving the sample rate argument alone for the moment, the 'sound' of a clipper primarily depends on its “hardness”, that is, how abruptly it clips the audio peaks.
We offer different styles of clippers with different 'hardness' factors so that a user can customize the sound of the final peak control to his liking and program format.
Our product Operating Manuals contain quite a bit of in depth information on this subject, including how to appropriately use and adjust each of the clipper styles, and why one might choose to use one clipper style over the other.
All three of the RJ-45 connectors on the exterior of the AP/FM-2000 can be utilized for connecting it to an external network (only one at a time, please!).
All recent Wheatstone processors contain their own five port, auto-sensing, 10-100BaseT Ethernet switch based on a powerful high speed Broadcom chipset.
The Ethernet switch is auto-sensing too, so it is not necessary to use a crossover cable when connecting the processor directly to a computer.
The two remaining ports of the five port switch are used internally by the processor.
Yes! In fact, all of our Wheatstone processors support up to four simultaneous connections from a remote GUI. Multiple instances of the GUI can be helpful when you want to see or control things from different locations at the same time for training purposes, or for remote management and troubleshooting, or for… use your imagination!
The external Ethernet ports can also be used as a convenient local switch if needed, making it easy to set up a 'mini-network' for special applications.
The only caveats are:
- You may never connect any of the external ports back to another one on the same Vorsis unit;
- Never connect more than one Wheatstone Ethernet port to the same external Ethernet switch.
Our processors use the TCP/IP protocol for adjusting the processor by remote control as well as for uploading and downloading preset information. We use TCP/IP because that protocol's robust handshaking guarantees that every single data packet will reach its destination error free. The default mode of our processor is to send metering data via UDP for best remote metering response. If no metering data can be seen it means that UDP packets are being blocked somewhere, usually by your ISP.
Click on the System button on the GUI and select “TCP/IP meter data” and you should see the meters become active in the GUI. Their responsiveness will depend on the bandwidth of your connection.
Please see our networking application note AN2007-01A Resolving Network Connectivity Issues on the Wheatstone Processing website for an in depth treatise on the how's and why's of networking for our audio processing products.
All of the most recent Wheatstone software has the option of retrieving metering via UDP or TCP/IP, which helps circumvent instances where UDP packets may be blocked by an Internet service provider.
The Backoff control adjusts the threshold of the AGC detectors relative to those of the compressors. When the Backoff control is at a setting of “0”, the AGC and Compressor detectors have the same audio detector threshold.
As the Backoff control is adjusted to negative dB numbers the AGC threshold is “Backed off”, moving its detector thresholds higher than the compressor's threshold by the amount set by the Backoff control. This causes more gain control to be done by the faster Compressors and less by the slower, long term AGC action resulting in an increase in audio density, loudness, and consistency.
The "Backoff" control has another interesting effect which could probably benefit from further explanation:
The multiband AGC's operate in the Sum and Difference mode which means that the audio is separated into its mono and stereo components (L+R and L-R, respectively) prior to being processed by the five band AGC. As the Backoff control is adjusted to more negative dB numbers both the AGC's and Compressor and the sum and difference channels become more 'uncoupled'.
The stereo image is kept under control because the Sum and Difference gain reduction mathematics are ganged (coupled together) by the slower AGC control signal. On the other hand the Compressor control signals are not coupled and always operate independently.
What this means is that when the AGC Backoff control is at “0” the sum (L+R) and difference (L-R) channels undergo virtually identical processing. In fact because the Sum (L+R) channel is almost always higher in amplitude it will usually have complete control over both signals forcing the L-R difference channel to perfectly track the L+R channel. This keeps the stereo image from wandering out of control.
When the Backoff control is adjusted to negative dB numbers the sum and difference channels become more 'uncoupled'. This allows the sum and difference channels to begin to take on their own amount of processing and with most program material the difference channel is allowed to have a bit more dynamic gain (and more aggressive processing) than it would normally have if the AGC Backoff control were set at “0”. This increases the stereo image width, but in a program-dependent and very well controlled way because the overall gain of the L+R and L-R are still coupled by the AGC control signal.
Please see our application note AN2008-03A Vorsis Stereo Enhancement on the Wheatstone Processing web site for a detailed explanation of our stereo enhancement scheme and how to operate it in our processors.
Actually both. Various FFT's (Fast Fourier Transforms) are used for certain processing functions within Wheatstone products including, but not limited to, our quite complex VBMS algorithm.
The FFT is a powerful tool for analyzing what is going on within the processor, particularly with regard to the kind of signals typically encountered in distortion cancellation schemes and how to achieve obviously lower processing distortion without giving up competitive loudness.
We have a lot of new ideas on how to better accomplish audio processing for broadcast! Experiments in our research lab and at secret customer sites has shown that we are able to achieve higher on the dial loudness than competing technologies offer today, but with drastically lower perceived distortion levels. As we continue to perfect this technology we will offer it in the form of DSP software upgrades to enable our Vorsis customers to stay current with the very latest in algorithms.
Actually, the stereo generator contains three different types of final Multiplex peak control that can be selected by the user:
- Hard Clipper
- There is the usual 'hard' composite clipper and ours runs at a sample rate of 1.536 MHz. It operates on the composite waveform prior to the stereo pilot and SCA injection. The output of the clipper is down-sampled to 768khz to perform the pilot and SCA protection filtering, and then down sampled again to 384kHz prior to the composite output DAC (digital to analog converter).
- Soft Clipper
- There is also a soft clipper available that can be used in place of the hard setting. This clipper operates at a lower sample rate (768 kHz) and has a softer transfer function than the hard setting. The clipper output is subjected to the same post filtering used when the hard setting is in use.
- Lookahead Limiter
- This is an oversampled feed-forward limiter with its control sidechain operating at 384 kHz (roughly 8x oversampling).
The lookahead time of this limiter is fixed at a half millisecond (500 microseconds). Unlike the other lookahead limiters in the signal chain the attack and release times of this limiter are not user-adjustable. Instead they are determined empirically and 'on the fly' as programming material changes in a way that produces the least audible processing artifacts.
The lookahead MPX limiter is useful for formats whose program material could not tolerate (or allow) using either soft or hard clipping but still require composite processing for competitive reasons.
The SCA inputs are digitized at a sample rate of 192kHz after being high pass filtered to reject low frequency noise and then low pass filtered at 96kHz for compliance with Harry Nyquist's well-known sampling theorem.
We are not aware of any SCA application that does not work with our digitized SCA inputs however the proof is always in the testing. If in doubt - and this is the case with virtually all SCA-related devices - please thoroughly test everything first to ensure compatibility. And if you discover a technology that doesn't work with our digitized SCA inputs please let us know!
There are several things to keep in mind regarding the AES Inputs and Outputs on all Vorsis audio processors:
- Their digital inputs accept 110 ohm balanced AES3 at 32 kHz to 96 kHz sample rates;
- The AES3 digital output sample rates are fixed at 48 kHz when there is no AES3 input applied to the processor;
- Whenever there is a valid AES3 signal preset at the AES3 Input the processor will lock its digital output rate(s) to that signal;
What this means is that when there is a digital AES3 input connected even if it is not selected as the audio source to be processed, the processor will use that signal to synchronize the AES3 digital output rate.
In the case of products with more than one AES output all digital outputs will be synchronized to the signal present at the AES3 input if one is present.
Again… if no AES3 digital input is present, the AES3 digital outputs operate at a fixed 48 kHz rate and are phase locked to each other and to a high accuracy, low jitter internal reference clock.
Each GUI can support a virtually unlimited number of Wheatstone processors. The main limitation is the maximum number of available IP addresses that are reachable by the computer that is running the GUI, and that depends on how the subnetting has been configured for that PC and the particular network that it is attached to.
On a network without fancy subnetting there are 256 available IP addresses (0 through 255). The PC that the GUI is running on would be using one of those addresses so the remaining 255 could be used as the addresses of 255 different Vorsis processors, but…
- IP addresses ending in “0” and “1” are rarely used for “everyday” networked devices because they are usually reserved for use with gateways and the like.
- Likewise an IP address ending in “255” is usually reserved as the broadcast address for the network and therefore no devices should be assigned to that IP address.
Yes. Each Wheatstone processor can support up to four simultaneous GUI connections.
By the way, a different software-based GUI is available (and required) for each of the Wheatstone products as each one is very specific to that product. For instance you can't use the GUI designed to control one product for controlling one of a different type.
Yes, there are always new presets being created for Wheatstone products! Presets are not just designed by us, but also created and contributed by our end users and we post them on the Wheatstone Processing website for downloading and use by any Wheatstone customer.
New Factory Presets - We're always working on new Factory presets for our products and as they become available we post them to the Vorsis website along with instructions on how to remove the old factory presets and install the new ones.
Celebrity Presets - This is a collection of presets that we've gathered from users in the field who feel they've found a great sound and would like their tweaking expertise to be known to others.
We like to give credit to those who create new presets for our products, so if you want to be on our Celebrity list simply create and send a cool preset or two our way. On the other hand if you happen to know one of the talented and generous folks who created one of the Celebrity presets, please give them a hearty thank you!
Please see the following application note on the Wheatstone Processing website for downloading and using preset packages: AN2008-02B Updating Vorsis Factory Presets.
Note that there are new preset packages also available for the AP3, and HDP3 products too that can be downloaded from the Vorsis website!
Go to the AP2000 front panel and move the mouse cursor to the upper area where the controls are and then 'right click' using the front panel's right hand mouse button.
Select the File option and a dropdown menu will appear.
One of the options will be “Center Window”….select that option and the screen will automagically return to its normal, centered position!
If the Wheatstone product is equipped with a headphone jack the headphone routing selector will be found in the System screen of the GUI.
Along with being able to monitor different places within the processing, we also let you hear the analog and digital inputs, even if one of them is not selected as the on-air program source! This will (for instance) allow you to check the availability and quality of your analog backup while the digital source is on the air. Other products require you to actually switch to the other input to see if there is audio.
Virtually every Wheatstone processor has stereo enhancement, although it's probably like nothing you've ever seen before.
We utilized what we understood about stereo enhancement techniques and the behavior of real-world broadcast signal chains to design a completely new and better way to accomplish it.
Our method might be a tad more complex to set up compared to the 'one knob wonder' stereo enhancement offered on other products. But we know that once you hear how good ours sounds on your station you'll be very pleased with the on-air results!
Please see our online application note AN2008-03A Vorsis Stereo Enhancement for a detailed explanation of how our method works and how to set it up.
There are also no hidden processing controls, nor is there a “secret passcode” to get in a back door that is reserved just for us Wheatstone factory folks.
Our security has been designed as a first line of defense to thwart genuine attempts at unauthorized access. If you lock yourself out of your Wheatstone processor and forget the password you are locked out - at least temporarily.
Although there is no “Master Password”, you can use the simple procedure that is outlined in the application note titled AN2007-04 Unlocking the AP-1000 GUI to gain access again.
It's an easy procedure to perform if you need to get back in but designed to be 'convoluted' enough of a process to keep the casual hackers out.
Our developers use the official Microsoft Windows Visual C shared runtime libraries to write the GUI applications. If the GUI generates errors at run time it's because an application installed before the Wheatstone GUI has replaced those libraries with its own versions.
When Windows is installed it puts all the right ones on your hard drive. But sometimes programs you install later put their own versions of the library files on your PC, or worse yet remove the original runtime libraries that were installed when your PC's operating system was first configured!
Our PC-based GUI requires the standard Microsoft libraries to be present on the host PC in order to run, so to fix the error message you must replace the original DLLs. But you probably weren't even aware that the originals were overwritten or deleted, so how do you know what to fix and where it is on your hard disk? Well, we've made that part of the task easy!
With each GUI installer we've included a copy of the Microsoft Visual C redistributable libraries. If you look in the folder where the GUI was installed you'll find a file called “vcredist_x86.exe”. This is the executable that you need to run to fix any missing or corrupted DLL errors. If you can't locate it, download it here: Microsoft Virtual C Redistributable PATCH
To apply this patch, navigate to the folder where the Vorsis GUI is installed and double click the vcredist_x86.exe file. When it has finished running, all the libraries that the GUI needs to run will be in all the right places. For 99.999% of the time installing this official Microsoft patch has no ill effects for other programs installed on the PC.
Actually, no.
Our SST operates by analyzing the behavior of the multiband AGC over a period of time to see how the operator has set its various controls to create the target on-air sound. It then watches the multiband section as program elements change and manipulates its internal operating parameters to maintain that target sound.
Those familiar with broadband AGC's know that those devices usually don't do anything to maintain the station's spectral consistency - even though their intended purpose is to create consistent operating levels into the multiband AGC which should result in a consistent output consistency.
Unfortunately those devices, because they are in front of the main processor (or even part of its preliminary processing functions) don't know a single thing about what the multiband AGC is even doing! That's why those configurations don't work as well as they might.
Because the SST is intimately involved in the operation of the multiband (in fact it's part of the multiband!) it knows exactly what is going on in that section of the processing at all times and can manage things to keep the output spectral balance, spectral consistency, and RMS energy (loudness) versus frequency levels exactly where the station wants it.
Because the SST is not a broadband AGC it has none of the limitations of that topology. Granted, it operates entirely differently from anything you've probably ever seen, and because it knows more about what's going on inside the multiband AGC at any particular instant than we do, you might see it do some unexpected things sometimes, but it does its job extremely well (see the next question…).
Please see our application note AN2008-01A Understanding Vorsis SST for more in-depth information about the SST algorithm. It can be found on our Wheatstone Processing website at http://wheatstone-processing.com.
The SST section will do whatever is required in order to maintain spectral energy output consistency of the multiband AGC where it is supposed to be - that is its purpose and is why we called it our Sweet Spot Technology.
Since the SST isn't a broadband AGC it won't act like one. Therefore it might be better to think of the SST gain meter as something that shows you how much “work” SST has to do to keep the multiband happy.
The SST meter should normally run between -10dB and -12dB with most program material and the more the meter departs from this the more 'work' it is doing in one direction or the other to keep the multiband AGC in its Sweet Spot.
When the SST meter moves towards the top of its scale it is modifying the multiband AGC's operating parameters in a way that makes the multiband section increase short term density. Conversely, when the SST meter is moving downscale it is changing the operation of the multiband in a way that makes it reduce short term density.
The multiband AGC section has a “Coupling” algorithm that helps manage spectral consistency by not allowing the lower and higher frequency bands from having more than a certain amount of gain difference from the mid, or reference, band. This is called Interband Coupling and we can explain better what it does by way of two examples:
Example one describes a Classical music station that doesn't want or need the automatic re-equalization that a multiband AGC contributes. In this case they would set the “Coupling” control to “0” which would prevent the two lower and two upper bands of the multiband from adding more gain than the mid band when they shouldn't.
Since the coupling control is effective only in the gain increase direction, each band can still compress, or reduce its gain when called upon to do so without dragging the other bands along with it. When the bands release however, their gains can only increase up to wherever the mid or reference band's gain is at that moment.
Example two describes an oldies station that plays material spanning several decades. As expected, the recording technology and spectral balance variations across this era are quite wide. The station needs some re-equalization to occur as different cuts are played, but it also needs to keep such re-equalization from sounding unnatural.
What the station would do in this case is set the Coupling control to a setting of perhaps “-6dB”, which would allow bands other than the mid, or reference band to have gains that are higher than the mid band by up to 6dB. Considering that 3dB is a doubling of acoustical power, 6dB would normally be sufficient to perform song-to-song dynamic re-equalization.
Obviously the Coupling control can be set to other settings, although at settings beyond about -10dB the bands are for all practical purposes 'uncoupled' and operate independently.
The Stereo Width Limiter operates by measuring the difference in content (both in level and spectrum) between the sum (L+R) and difference (L-R) channels and works to ensure that the L-R signal never exceeds a certain percentage of the L+R.
Why would this be wanted?
It is widely recognized that too much L-R can exacerbate multipath problems on FM. The Stereo Width Limiter can automatically manage this issue and this becomes especially important when a fair amount of stereo enhancement has been dialed in. Consider the following example:
You've set an aggressive amount of stereo enhancement that causes the majority of your music to sound awesome. But every now and then an older cut (early Beatles cuts were notorious!) just sounds awful on the air (and also weak on a mono receiver). This is because without a Stereo Width Limiter there is nothing to limit the maximum amount of L-R when something already has very wide stereo separation.
The Stereo Width Limiter serves as a stereo separation watchdog by never allowing something to be over-enhanced. It has an 'Off' position of course, but the most useful settings for most formats will be found between 60% and 80%. At these settings stereo width is very well controlled and it is virtually impossible to over-enhance the stereo sound field.
Now, about “…never hearing it doing anything”.
Assuming for a moment that it's actually on, the Stereo Width Limiter is unobtrusive in its behavior as far as 'hearing it work' is concerned. This is because even though electrical stereo separation on the order of 80dB or better is achievable in the rest of the processor (and the rest of the broadcast chain is probably in excess of 50-60dB) the human hearing system has a hard time 'hearing' the difference between 80dB separation and 20db separation when both left and right channels are present, i.e., normal stereo listening.
From the point of view of comparing the 'audible' and 'electrical' stereo separation, the electrical difference between 20dB and 80dB is huge. This is why the Stereo Width Limiter can work to reduce multipath - it can reduce the 'electrical' stereo separation down to where it would be 'lousy' in strictly electrical specification terms in order to minimize multipath, but from the standpoint of the human listening experience the signal sounds just fine. In fact, the listening experience is usually improved because with the reduction in electrical separation comes a reduction in multipath and multipath-associated effects and this results in a cleaner and more listenable signal.
The Stereo Width Limiter can have a dramatic effect in reducing excessive L-R induced multipath on a station while hardly impacting the listening experience of the common radio audience listener.
Note: Don't be afraid to use the Stereo Width Limiter just because it reduces stereo separation! In fact, our implementation is far better behaved, far more useful, and far more sophisticated than the static 'Stereo', '-3dB', '-6dB', and 'Mono' settings found in other processors. Those implementations impose a FIXED degradation in stereo separation regardless of the stereo separation present in the incoming program material.
Our implementation is automatic and does not reduce stereo separation unless it would be higher than what the setting of the Stereo Width Limiter control has been set by the end-user and only reduces it to that extent - never more.
It's important to recognize that voice and mono programming have a lot more in common from a 'processing' perspective than is usually recognized.
One of the things that helps hide processing artifacts in broadcast processors is the complexity of the stereo sound field which serves to mask a lot of what is going on to make the signal loud. We will call this effect 'Spatial Masking'.
Mono programming on the other hand does not benefit from 'Spatial Masking' so it is far more sensitive to the processing efforts taken to make a radio station loud. Voice is probably the most sensitive mono signal, but other mono material (such as oldies) isn't far behind.
By treating this material (from a processing perspective) in a way that can reduce artifacts rather than exaggerate them we can provide a competitive and clean on air signal regardless of the character of the incoming programming.
VoiceMaster does this - it reacts to and processes differently anything that cannot benefit from the 'Spatial Masking' that helps hide processing artifacts during stereo programming.
When the meter has been set to its Context position it follows the displayed screen. If the AGC screen were selected the Context meter would be showing the AGC and Compressor gain reductions.
When the FM Limiter screen is selected the Context meter will be showing the amount of clipping (or peak limiting if that function is on) occurring at any instant. Therefore this meter serves as a Clipping depth indicator showing the instantaneous clipping taking place in the pre-emphasized domain by the main clipper algorithm.
The AP2000, FM2000, FM-10HD, AM-10HD, HDP3 and AP3 all utilize 48dB/octave (or eighth order) crossovers. They are phase linear Linkwitz-Riley structures whose compensating delays are automatically optimized as the crossover frequencies are adjusted by the end user. This technology results in an uncannily accurate phase and amplitude behavior after the bands undergo recombination.
The reason that steep crossovers were selected is that we felt that in combination with the Interband Coupling feature and the SST the multiband should be able to do what it was designed to do - automatically re-equalize the spectral balance of program material on a source by source basis.
In order to do this job well, and without inadvertently changing the gain of adjacent frequencies, steep crossovers are required. Admittedly, without Interband Coupling and the benefits of the SST, our 48dB/octave crossovers might be undesirable, but because we have put a lot of thought into our implementation this is not the case.
Note that the VP8 utilizes 24dB/octave (4th order) crossovers instead of the 48dB/octave 8th order crossovers of the products higher up in the product line.
Our AGC and Compressor section is based on a 'lookahead' architecture. What this means is that we use the input signal instead of the output signal in order to determine how much gain (or compression) a particular band needs to have. By using the input signal we can avoid the inevitable overshoot that occurs with feed-back type compressors.
A feedback compressor's gain control signal is actually an error signal created because the compressor's output signal was higher than it should have been. By using the input signal instead to control compression a very accurate gain control signal can be generated that very faithfully follows the envelope of the program material without generating 'errors' in the output signal.
That explains why their action is so hard to hear. Now let's talk about how to undo that!
There are two things to consider: Short term and Medium term compression. Short term compression is mostly tied to the action of our five (or four) band compressors while medium term compression is tied more closely to the behavior of the five (or four) band AGCs.
Therefore one must decide which 'pumping' artifact one wants - that fast pumping stuff that follows every beat of the music, or the slower stuff that chases levels down as a song fades.
To make the compressors more audible use fast attack times and slower release times. Attack times under 10 milliseconds and release times of 200-300 milliseconds are good starting points.
To make the AGC's more audible leave its attack times fairly slow (several hundred milliseconds) and speed up its release times. Attack times around 300-500 milliseconds and release times around 1 second are good starting points. Then use the Backoff control (explained previously) to adjust how the two algorithms work together to create the “audibly obvious compression” sound you're after.
The AGC serves to control long-term gain levels and the Compressors work on short term stuff. A bit of explanation is probably necessary about how our AGC and Compressors interact because it follows a completely different philosophy than the designs of others….
The AGC and Compressors are actually separate but joined at the hip so to speak because the AGC derives its control signal from the Compressor, the faster operating algorithm of the two. The architecture is such that the Compressor gain can always be less than the AGC gain but never the other way around - in other words, the Compressor can reduce its gain below that of the AGC but can never return to a gain higher than that of the AGC.
What this does is twofold:
First it makes it very easy for an announcer to talk over a song intro without worrying about the song burying his voice because the faster acting compressor pushes the gain down for each syllable, exposing the announcer voice while the slower AGC's gain reduction values serves as a gain platform to return to for the compressor which avoids pumping.
Second it allows the five band section to process audio in a way that not only automatically re-equalizes the incoming audio spectrum to create a more balanced on-air sound, but also 'polishes' the audio levels to make them very consistent and without sounding busy, smashed, or fraught with unnatural gain recoveries as program material changes. In fact, the behavior of the five band algorithm makes transitions between program elements very consistent and professional sounding for the listener without increasing average program energy to the point where it causes listener fatigue.
The SST and five-band AGC do have separate Gate Thresholds in order to allow for separate customization of how each behaves during song fades or low program input levels.
Functionally the AGC's Gate Threshold is after the “output” (if there was one) of the SST which means that if the SST is allowed to increase its gain during a song fade while the five bands were gated, a point could be reached where the AGC Gate Threshold is reached and the bands would 'un-gate' and begin to increase gain to bring up the low audio levels.
In fact this is exactly how we've tuned the Factory presets - the SST 'stays awake' over a much wider range of input levels than does the five band AGC. This works extremely well at controlling levels because the SST's release time (Ramp Rate) is usually set slower than that of the AGC and also because the effective ratio of the SST (again, if it had one) is much lower and its control slope extremely nonlinear compared to that of the AGC.
Normally we set the AGC Gate Threshold in 'the 40's' on most presets - somewhere between about -42dB and -48dB (this is calibrated in dBFS - please see control calibration covered later on). The SST is usually set in 'the 50's' - somewhere between about -50dB and -55dB.
Our general advice is to:
Set the SST Gate Threshold so that the SST stops chasing falling levels when you'd like it to. The same goes for the AGC Gate Threshold;
Understand that the SST does a better job of raising levels when they fall because it affects all five bands simultaneously and this prevents unnatural or unbalanced spectrum coloration as low levels are brought up.
That is an excellent question!! We sense from our customers that it is sometimes confusing having processor controls calibrated to their real numbers instead of the arbitrary ones we've been used to for so long! All of our controls with a “dB” label are referenced to 0dBFS, or digital full scale.
What this means is that when the compressor threshold is set to -50dB, it is actually 50dB below digital full scale but that would be only 30dB below a nominal operating level assuming a 20dB allowance for headroom. This is an entirely reasonable threshold setting for a compressor that must deal with extremely wide range of input levels such as those commonly found in a real-world radio station.
What this means for the Gate controls is that they too are referenced to 0dBFS. Therefore when a Gate setting is -50dB it is actually -30dB below nominal program level - assuming again a 20dB allowance for headroom. This again is an entirely reasonable setting for a gate threshold control.
Another excellent question! The answer lies in the fact that our AGC/Compressor sections utilize a feed forward rather than typical feed back topology. This means that when the compressors are achieving 20dB of gain reduction (a very normal amount) their output levels have fallen by 20dB.
Because they are feed forward algorithms they don't know that the output levels have fallen - and they don't need to care! We then “Make Up” for this decrease in level by bringing the post-AGC gain, or Makeup Gain, up high enough to restore the desired level into the following multiband limiters. Because we allow an extra 10dB of headroom in the AGC for times when it might be operated with very low compression ratios the Makeup Gain control will typically be adjusted about 10dB higher than the amount of average gain reduction occurring in the AGC.
The Makeup Gain is never too much (if adjusted as we just advised) because when input audio levels fall the AGC's increase their gain exactly in accordance with that change and therefore the Makeup Gain can never be too high unless it has been set against our advice.
The High Pass Filter (Input Screen of the GUI) is there to remove subsonic signals from the incoming program material before processing is applied.
The reasons we do this are: (1) such signals are not normally audible on a listener's receiver, (2) they can have negative consequences as far as the quality of the station's air sound is concerned, and (3) there is no “useful” program material down where the Highpass Filter is normally set.
When setting the High Pass Filter think carefully about your listener's typical receivers - what do they listen to your station on? Then factor in that it is extremely unusual to see program-related frequencies below about 30Hz in typical material broadcast on the radio. Add to that how your air chain, and yes, all the way out through the transmitter, behaves with extremely low frequency energy. Does your exciter's AFC occasionally 'unlock' on heavy bass material?
With typical musical program material a good high pass filter setting is 30Hz and the High Pass Filter can be operated in Stereo mode. But…
Have an oldies format? Still playing vinyl on the air? You probably need either a slightly higher High Pass Filter setting -or- could perhaps benefit by operating it in Sum and Difference (M/S).
In the case of playing vinyl (or material that was dubbed from vinyl) there is likely non-program-related subsonic energy in the L-R that could and should be removed prior to air. In that case operate the High Pass Filter in Sum and Difference (M/S) mode and set the Sum (M) to 30Hz and the Difference (S) to perhaps 180Hz. This will remove the unwanted low frequency energy from the L-R without affecting the low bass which is typically recorded as a Sum (M) signal.
The Emphasis setting pertains to the high-frequency pre-emphasis that is used in FM broadcasting. In the US the correct setting is 75uS while in most places in Europe it should be set to 50uS. If no (pre) emphasis is used the station will sound very dull on typical receivers!
The Emphasis location selector allows the end user to place the FM pre-emphasis at the location in the processing structure that creates the sound that they desire from a sound dynamics standpoint. The three possible locations are:
- Pre-MbLim
- The pre-emphasis is before the multiband limiters creating a frequency-conscious multiband limiter that constrains energy as a function of frequency - high frequencies will be limited more aggressively than mid and low frequencies. The resulting sound will be smooth and easy to listen to albeit a bit dull compared to the other settings.
- Pre-Limiter
- The pre-emphasis is placed prior to the broadband lookahead limiter - if it is selected as active. This setting is recommended only for non-aggressive purist formats since broadband pre-emphasis limiting is quite audible when aggressive loudness is the goal. Note: when the lookahead limiter is not selected this setting is equivalent to the Post-Limiter setting!
- Post-Limiter
- The pre-emphasis is placed after the multiband limiters and therefore high frequency peak control is achieved solely by the distortion-masked main clipper. This setting has the most open sound as far as high frequencies are concerned but is accompanied by the touchiest behavior of the Lim/Clip Drive control which sets the drive into the main clipper.
The setting of the Emphasis Location is entirely a matter of preference and taste - each has its own sound, benefits and limitations.
The GUI image size is fixed at a screen resolution of 1024x768 pixels. Most PC's can display this resolution without difficulty however there are apparently still a few out there that are old enough to not support this resolution.
On the Wheatstone Processing web site, there is a skin called AP-2000 Little Skin. Inside each LittleSkin folder is a “xxxxxLittleSkin.skin” file for that product. Download this file and carefully copy it into the Skins folder where there AP-2000 or FM-2000 GUI resides on your PC.
Next, and in the following order…
- Fire up the GUI;
- Connect to the AP2000 (go “online”);
- Go to the System screen and click on the skins button.
- Select LittleSkin and click OK.
What happens next might be a bit surprising - the 640x480 pixel GUI that displays on the front panel of the AP2000 or FM2000 is now running on your desktop!
There are two caveats to this workaround. First, there are no I/O meters. Second, you must be in one of the normal AP2000 or FM2000 skins in order to connect and disconnect a session from the processor! The front panel skins do not have (because they do not need) facilities for switching between online and offline modes. This must be done through one of the normal skins.
To change skins while the front panel GUI is running on your PC you can right click in the 'controls' area of the GUI display and use the “File/Choose Skin” function to change skins. Selecting any of the skins other than LittleSkin will put you back into normal mode. When a normal sized skin is running it is normally not a problem to reach and activate the online/offline button on the top of the GUI.
Good question and we recently included this important topic in a paper presentation at the CCBE conference in Canada. Basically, some additional headroom must be allowed for what might happen in the phase rotator.
One of the first things the audio from the studio is applied to inside the processor is the phase scrambler. The phase scrambler's primary job is to make announcer voice and other asymmetrical waveforms more symmetrical making the processing of these signals easier for the symmetrical FM medium.
As long as the input audio has not been clipped (or mastered to death!) the processor's peak audio input levels could actually be set to just achieve 0dBFS precisely when the console output does. But this isn't a real world scenario - much of today's programming has been highly processed and some audio levels even run hot enough to achieve (or try to exceed) 0dBFS during the recording process.
The squared-off tops of those audio waveforms are radically altered in shape by the processor's phase scrambler. In fact in lab testing we've seen peak audio levels of clipped waveforms increase by 10dB or more after passing through commonly used phase scrambler architectures.
Because of this we recommend that some 'headroom' be left at the input of the processor and peak input levels achieving no more than about -12dBFS satisfy this requirement.
Because the processor itself achieves more than 140dB dynamic range internally, giving up a few decibels up front in order to avoid nasty distortion artifacts has no material effect on the signal to noise ratio.
The simple answer is that the exciter's modulation slope probably isn't symmetrical. This is quite commonly seen in analog exciters although the amount of asymmetry is usually quite small (1% to 2% at most).
The asymmetry can be caused by several things, all usually related to the exciter's design. The modulator can be nonlinear, meaning that the change in frequency resulting from the audio swinging between two defined positive and negative excursions isn't equal.
Another thing that can cause this symptom is asymmetrical behavior of the exciter's phase locked loop, noting that the first issue (above) can actually cause the second. Our advice would be to contact the exciter manufacturer to discuss this observation.
Note that in our experience 'most' recent technology digital exciters do not suffer from this nonlinearity and that most deliver virtually perfect modulation linearity.
Bass texture in a broadcast processor comes from the contribution by, and interaction of, a variety of things such as:
- The AGC and Limiter crossover orders and their frequencies;
- The phase vs. frequency behavior in the crossovers;
- AGC, compressor, and limiter time constants;
- Frequency domain distortion canceller behavior in the main clipper.
All of these things (and some others) affect how a processor's bass sounds on the air and are usually determined by the manufacturer's particular processing algorithms. Therefore a particular bass sound may not be adjustable in the field - in fact you may not be able to escape a certain processor's bass texture even if you wanted to. Certain 'signature bass sounds' of audio processors on the market are there no matter what - you cannot turn them off.
So… the real question is whether or not an audio processor can put deep, clean, natural bass on the air without the bass sounding 'hokey', synthesized, or there when it shouldn't be. All Vorsis processors can easily meet this challenge.
The flexibility of the VBMS algorithm, multiband limiters, and parametric equalization allows any reasonable bass texture to be created. In fact Vorsis processors with VBMS can put more bass on the air and do it cleaner and more naturally than any other brand.
Our best advice is to never use bit reduced audio if you can help it, especially for music. A station using bit-reduced program material will NEVER be able to sound as good as a station that uses uncompressed audio. Period.
If you can't escape the need for bit-reduced audio, always use the highest possible bitrates when recording to the playout system. Next in line for achieving a good on air sound are the high frequencies, and in this instance, less is really more. Audio processing, by raising the average levels and changing the spectral balance, challenges codecs in ways that unprocessed audio cannot. This can result in coding artifacts that were intended to be (and would have been) masked under normal listening conditions. The most noticeable and annoying artifacts usually are in the high frequencies, and bringing up the average high frequency levels via the on-air processing can make post-codec audio sound strident and unnatural.
Our advice is to go easy on the ‘high end’ when playing bit-reduced audio on the air. Using lower thresholds in the processor's upper frequency bands and/or reducing their output levels using the post-AGC mixer can go a long way towards making bit-reduced audio tolerable. Also take care not to using aggressive equalization settings in the parametric equalizer, especially in the frequency range from about 2kHz and up.
A secondary caution would be to be very careful with stereo enhancement, especially if you know that your coded audio was touched by a codec that utilized parametric stereo. The use of (aggressive) stereo enhancement can unmask very unnatural stereo sound field artifacts that were not part of the original program but are associated with this type of codec operation.
Most radio markets lean one way or another in their overall air sound and bass seems to be the Holy Grail these days. The AP2000 can create a wide variety of bass textures and can likely make your station sound just the way you want it.
There are a couple of caveats when going for that big bass sound. They are, and in no particular order:
- Bass takes lots of headroom. If you're already screamingly loud and want more bass, you'll probably have to give something up to get it.
- Keep in mind the typical receivers that your listeners use to listen to your station! Too many times we find ourselves in a station-to-station race without thinking about what our listeners hear and how they hear it. The listener has both volume and tone controls. You don't have to be wimpy on the air, but it's a good idea to not overdo it creating energy in the bottom end that only people with $15,000 speaker systems can hear. It's a good bet that those aren't the people listening to the radio or to your station…! :
Here is something that we sometimes have to tell customers who want a very bright high end, a heavy bottom end, crystal clear midrange, silky smooth voices, and ear bleeding loudness.
“Please choose something that you're willing to give up in order to get the thing that you want the most”. It's true!
The BS-412 control is provided for customers in European countries that have regulations on how loud a station is permitted to be.
In Europe the countries are arranged much like the states are in the United States. However, unlike the states of the U.S. each European region might have different broadcasting laws. Complicating it further is the short physical distances between stations and the narrow 100kHz spacing of the FM broadcast channels (the United States uses a 200kHz FM channel spacing) adjacent channel interference is a real concern. .