Arthur Skudra
Sy added that Mr. Richardson was at AES Toronto's 40th anniversary day long event in 2007. He mentioned Jack's interview on the commemorative DVD produced of the event.
Sy introduced Arthur Skudra providing his credentials. Mr. Skudra has presented an earlier version of Smaart at an AES section several years ago.
Arthur thanked everyone for coming. While he spoke passionately, enthusiastically, and engagingly about the software, he stressed his philosophy of making corrections at the source first before attempting electronic corrections.
He provided background for Smaart & Rational Acoustics who bought the rights to Smaart from EAW/Loud technologies in November 2009. He then recounted a brief history of updates; and then of Rational Acoustics.
Mr. Skudra's title for his presentation was "Measuring Something That Doesn't Exist". A summary of the main points are:
Systems interact most where they are equal level ie: crossovers and overlap;
Solve the problem at the source;
Use the right tools;
and
something called the 20/80% rule.
Mr. Skudra explained the first 20% of the time spent optimizing a system, typically if done right, allows one to achieve 80% of the desired goals. The remaining 80% is getting the extra 20% performance out of the system.
To illustrate the concept of solve the problem at the source, Arthur recalled a situation setting up for a concert event where speakers were constantly being blown out in a line array and the staff couldn't figure out why. Mr. Skudra spoke of checking through the racks and discovering a Y cable where one side was wired pin 3 hot, and the other pin 2 hot, resulting the in the line array canceling itself out. The staff was compensating by cranking up the bottom end more and more.
Concerning the right tools for optimizing a system he provided the correct order of use:
Acoustic Design/Treatment;
Equipment choice/maintenance;
System design.
Once that is complete his next step is level balance and having enough subs to support the type of event. Then he checks for correct polarity; delay and timing. His final step is EQ. He noted the first two steps typically take a whole day, while the EQ tends to go pretty quickly.
He discussed what EQ can do including: achieve intended response; improve intelligibility; improve gain before feedback.
He also discussed what EQ can't do: make sound perfect everywhere; compensate for poor speaker system design or placement; compensate for poor room acoustics - you can't EQ the room (unless you have a couple of sticks of dynamite); you can't change the direct to reverb ratio. Finally, EQ can't change the loudspeakers' coverage pattern.
He then went over the caveats of EQ:
You can't make system response be flat everywhere, though you can make the system flat at one seat — but which seat (the person paying the money?!).
Two or more seats can't have identical responses.
Corrections at one seat may have serious problems at another seat.
The final point - the ultimate goal is to find a compromise that gives the greatest consistency for the majority of the audience.
When dealing with an array, his goal is to find an average frequency response within the coverage area to possibly find one mic position (out of the several) to represent the typical response of the system. Positions to avoid are edge of the coverage area (good though to study off-axis response) and locations outside the core coverage area.
For dual-channel measurements, Smaart deals with propagation time issues by lining up the two channels in time so they coincide when they go into the mathematical engine, you can get the proper transfer function between the two channels to produce a clean frequency response of the system.
Since version 3, Smaart has had a "fixed point transfer function" which takes 5 separate FFT analyses and creates one composite curve with 24 points per octave resolution which closely resembles human hearing capabilities. Due to increased processing power in computers, Smaart 7, taking advantage of multi-core processing, now can deal with 7 FFT's in the background and provide much higher resolution curve with 48th octave resolution from 60 Hz up and 1 Hz resolution from 20 to 140 Hz. This one improvement allows one to see more in terms of cancellations in system response, for instance.
When dealing with transfer functions in Smaart, there is almost an anechoic response in the high frequencies, but as you go lower in frequencies you're including more of the room in your measurements. While the effect of the room can be windowed out, the tradeoff is reduced resolution in the low frequencies. You can't remove the room from measurements. Mr. Skudra feels the human ear is trained to deal with these anomalies; and for practical purposes, the goal for measurements is to emulate the human ear - which Smaart does.
Giving a practical demonstration, Mr. Skudra illustrated and highlighted the main points of his discussion, through a series of test responses taken earlier with the Smaart system. After displaying the widely varying response curves he suggested "the key is to be a forester and not an arborist"! The point is to look at the overall shape of the responses and intelligently figure out what EQ is going to be effective at all the measurement positions in order to get a representative response out of a system that is pleasing to everyone in the audience.
Taking an average of the individual responses can provide one with an envelope of system response. Mr. Skudra stresses striving for simplicity in EQ'ing. Find the major problems in the system response and take care of those areas; instead of applying surgical EQ which he feels results in an absolutely lifeless sound, and is a typical fault he sees in other people's system optimization.
Smaart is capable of providing smoothing to find the trend. Mr. Skudra cautions on using this function cavalierly as applying octave or ? octave resolution can hide peaks and valleys that would warn one of corrective action that should be taken.
A practical demonstration of the software was provided next for the audience. Mr. Skudra uses the Roland Octa-Capture interface which is an 8-channel/mic-pre USB-based interface with an individual 8-channel line out mixer which allows him to choose any combination of inputs and outputs.
Seven microphones were connected: five were placed at various points in the room, another one - a 'dummy mic' plugged into the back of his interface, and the final one was his standard reference Earthworks mic. "Six cheap mics and one really good mic". The mic inputs were not calibrated for this demo due to time constraints.
One of the benefits of the Smaart software is one can configure as many channels as one has on their interface and run them simultaneously as individual 'measurement engines'. Arthur mentioned that, for instance, if he had a MADI card plugged into a DiGiCo mixing console, he could technically run 56 channels on his laptop(!) The only limit imposed by Smaart is one's CPU power. As he quickly qualified that point, it could become "challenging" to see trends in the measurement results with so many inputs.
He then played back some pink noise to obtain a reference channel. The Smaart display even has a live average of all the mics combined together. The interface is completely configurable, and mic inputs can even be assigned meaningful names.
In Smaart's configurable display, the RTA can be in any resolution one wishes. A spectrograph is also available which can be 'windowed' to highlight desired upper and lower limits even while running measurements live. Smaart has a histogram (history meter) which is limited only by the amount of RAM one wishes to assign to the history buffer.
Mr. Skudra often uses the spectrograph to plot coverage of a loudspeaker. He demonstrated this by playing back pink noise and moving one of the mics around to allow the audience to see, not only the response coverage of the loudspeaker, but also the cancellation patterns happening, as well as effects of reflections, in the histograms.
A new feature of Smaart 7.2 allows one to choose compensation curves for any of the inputs, which would permit the use of 'cheap' microphones. According to Mr. Skudra he could send such a microphone to a test lab to obtain a frequency response of the mic and apply a curve to compensate for the anomalies. In his case, however, his reference mic is the M-30 which he knows is flat to 30kHz, so he created quick transfer curve functions between it and his lesser mics, allowing him to 'normalize' his mics to his reference standard, the M-30.
Responding to an audience question whether the measurements are just frequency based or include phase response, Arthur replied that while Smaart has the ability to include phase response, it hasn't been implemented, though it's very possible for a future update.
Mr. Skudra has found most mics are pretty flat to around 5kHz, which is the region where most room anomalies occur. In live situations with most venues he's lucky to have a bandwidth of 10kHz to work with, and the region above 5 kHz is often corrected by ear. Utilizing transfer compensations, he'd rather spend $600 on six mics than on one. He then has a 'powerful tool' to let him optimize a system quickly. In fact, he noted he spends more time setting up his Smaart rig and stringing mic cables than doing measurements!
He then gave a practical demonstration showing how time delay is compensated in real time by Smaart by the use of tracking filters which is helpful in situations where one needs to work quickly.
Another audience member asked "how does it know?" It's a dual channel FFT essentially doing an impulse response continuously in the background, taking in the pink noise and the measurement signal, locks in the cursor on the peak of the impulse response.
He continued his demonstration to show how this feature can display and identify reflections in the room. It's useful also to identify multiple delayed speakers, actually allowing one to see the peak of the main loudspeaker and seeing how much time delay to apply by seeing the difference between the two speakers. Mr. Skudra uses this live IR to do subwoofer alignments.
When Arthur was asked when he uses the manual find vs. the tracking filter feature, he stated he generally he uses the manual find if his distance propagation is greater than 50 ms. He'll also use it when he first starts up Smaart to make sure he has a really good measurement signal, stating a mic could be placed incorrectly and actually pick up reflections louder than the direct field of the loudspeaker.
While continuing his demonstration highlighting the averaging feature, he was asked if Smaart can do different types of averaging between engines. He replied the possibility is there in future updates. Currently Smaart has the ability to factor in coherence into averages. This coherence curve shows one how reliable one's data is for each measurement. Mr. Skudra referred to it as a frequency specific signal-to-noise meter. He pointed out that Smaart has the ability to select coherence weighted averaging, so that more 'strength' is applied to coherent data points. The problem he noted is that this method tends to favour and exclude dips in the response, not allowing one to see those in the resulting frequency response curve.
There are plans to do more sophisticated averaging in Smaart.
A question about tracking and windowing came up. Mr. Skudra replied the previous version of Smaart had this and will most likely be included in an update.
At this point everyone took a break.
Mr. Skudra reviewed the response curves of cheap measurement mics and discussed the trends. He qualified his point regarding using cheap mics by stating he has less worry about setting up an inexpensive mic (typically at a very high level to avoid floor bounce) out in an audience area where it's beyond his control and have someone trip over it than with an expensive mic. He would not use such a mic to calibrate studio monitors!
Referring to the high mic stand, this brought up the question from Sy Potma, who asked if Arthur would want that 'loading to the floor' as part of the measurement component so that he could compensate for an audience sitting close to a boundary.
He replied with a 'yes-and-no' answer: no since mics too close to the floor produce lots of ripple in the low end which is why spatial averaging is so important because it averages out those anomalies. So, if one wants to take measurements at the audience ear level, do so with averaging a number of mic measurements. He went on to state his preferred method was to straddle plywood across the chairs to create a ground plane measurement to eliminate as much of the floor bounce as possible, basically creating a PZM mic and pushing the comb filter in 15k - 20kHz region. He added finally that with boundary loading you get boundary effects.
An audience member asked Arthur if RTA-ing an empty room with one mic at centre of room just above audience height was a good compromise or if there were any alternatives?
He responded by suggesting to move the mic to several different places, and getting an average of the snapshots for a more representative response of the system. The more mic position the better. Generally with spatial averaging, 3-4 mics is good enough to get an average response within the coverage area.
Responding to a question how Arthur finds correspondence between what he models in EASE vs. what he measure in a real live situation, he finds a lot of correlation, as much as 90%. Where he falls short is in extreme coverage areas, citing himself as the main fault of that by not spending as much time to tweak the front fill coverage zone.
Another question concerned if there was any correlation between Smaart and other modeling programs: not at the moment because of impending updates.
Another member asked a question regarding wireless implementation (to avoid stringing long cables). Arthur singled out the Electrosonics Wireless TM 400 Test Set that has no companding. He noted that the problem with companding makes measurements non-linear. Another caution he warned of is some wireless systems have a low cut at 100 Hz.
A further member queried if there was a Lab version of Smaart? Mr. Skudra replied there is only one version currently but noted that wizards may be incorporated in future versions to cater to specific types of tests.
Next, he demonstrated the resolution capabilities: one can smooth phase separate from magnitude response. This is important as it allows him to do alignment to fractions of a milliseconds between loudspeakers, or aligning the crossover between the subs and the main speakers in a system. It also allows him to spot possibility of phase reversal. He actually demonstrated this by forcing a phase reversal while playing back a pink noise test.
Next he displayed the new interface in iPad, a program called Audio-tools from Studio 6 Digital which is an iPhone/iPad app that includes a suite of modules for audio measurement, including SPL meters, RTA, Impulse Response, Speaker tests, STI-PA, and is under license from Rational Acoustics, making the Smaart algorithms available for the the iPad. Version 2 will include a dual channel FFT analyzer.
He discussed the Impulse Response module which actually lets one perform acoustical measurements. Loading a file, and recalling the response, he displayed an ETC curve of an impulse response recorded in the application. Schroeder integration and reverb time can be determined, as well as T30, early reflection field, clarity index, among some of the other data. He followed by giving practical demonstration in the meeting room.
Responding to a question, the data can be exported.
He demonstrated a beta program that can measure STI. Arthur mentioned that STI has some serious limitations in the way it computes things, one of them being is the way it calculates intelligibility. One can be completely off axis of the loudspeaker and yet still get good results. However solutions are being sought.
Mr. Skudra expressed amazement that with the cost for all the modules for the iPad application coming to a couple hundred dollars, he could do more with this than he has been able to do with his hardware analyzer and its components costing around $5000!
An audience member wanted if the app was limited to iPad 2 and iPhone 4, Arthur replied it works on his 3G iPhone but slowly. He stated the iPad 2 was a 'perfect balance of speed and capability'. He's hanging on to his 3G iPhone since the microphone is not high passed, allowing him to do use the RTA functions and simple FFT analyses.
Asked if there was provision for Playbook, Arthur was not sure.
A final question concerned achieving optimum curves and whether they've evolved or changed over the last two decades. Arthur wasn't sure if it really changed or not, but stated he doesn't subscribe to different or optimum curves, preferring a smoothed out transfer function of the actual response of the room as his goal, letting the high frequencies die off in the speakers naturally, and get vocal intelligibility as flat and smooth as possible. His philosophy toward sound systems is to do as little as possible to make it sound good.
He closed by stating that what Smaart gives us is a powerful tool to smooth out anomalies in system response, but also is able shape things so that it sounds pleasing to the ear, which is the final judge.
He thanked everyone again for attending.
Sy also thanked everyone and apologized for the cold coffee!
