Moderator: (Dave Dermont) Good evening and welcome to our chat with Jamie Anderson of SIA Software Company. Jamie, could you give us some background about what you did before SIA, for those of us too lazy to read the bio?
Jamie: I’ve got an EE (electrical engineering) degree from Worcester Polytechnic Institute (WPI), a Masters of Fine Arts (MFA) from Yale Department of Drama, and then spent six years at Meyer Sound as SIM and technical support manager. This was followed by three years as an independent engineer, working with k.d. lang and Dave Matthews Band, and now I’m with SIA.
Harry: How about starting with the basics of how the program works and your goals? I find a lot of visiting engineers don’t have any clue what the Transform Function is about, and simply want to see the RTA (real-time analyzer) display.
Jamie: SMAART is pretty simple. You get the reference signal, you get the measurement signal, and you compare them. If you have a problem, it is in one of those steps. One of the truths of computers is that they do what you tell them to do, not what you want them to do. The problem quite often with SMAART is that people confuse what they want to measure with what they are measuring.
There are two basic types of measurement that SMAART does: single channel and dual channel. Single channel measurements (RTA, Spectrograph, SPL) tell you about a signal at one point in the system, while the dual channel measurement (Transfer Function and Impulse Response) uses a comparison of what went into a system, to what comes out, to measure what a system did to the signal passing through it. This is what makes SMAART so powerful - you just need to make sure you have chosen the appropriate two signals.
Ken: Why should you use SMAART instead of a RTA?
Jamie: SMAART can be a RTA, which is just a measurement tool. SMAART is extremely useful for mixers and it gives great tonal content information. The RTA just isn’t the right tool for measuring system response. It can’t give you any information about timing, or whether the information it is getting is signal or noise.
Take the example of a system with two drivers, 100 feet apart. You could tonally balance the two drivers so that they read flat on an RTA with pink noise, but they would be a mess to listen to if they don’t get properly time alignment. That is what SMAART’s Transfer Function with Phase Response, as well as the Impulse Response measurements, is for. It is really a case of choosing the proper tool/measurement for the job at hand.
Harry: Can you explain proper connections and then tell us how to set the various parameters? Also Level Setting, Delay Locator, and what to look for when using the Transfer Function.
Jamie: The proper connections are those that deliver the two measurement signals you need to the computer. I’m not trying to be cagey; it’s just that there are so many different ways of managing your measurement signals. The key is that you want to get the reference and measurement signals you select into SMAART, and at a good level. That would be somewhere around -15 to -12 on the input meters.
Pink noise is easier to use because it has a low crest factor - it isn’t really dynamic or “peaky”. With music, it’s harder to get the signal levels to stay at the optimum levels. When setting your delay, you have to keep in mind how much HF (high frequency) is in your source, and how far away you are from it. You want your delay measurement window to be at least three or four times the delay you’re expecting to see. The reason I am concerned about the amount of HF? If you only have LF (low frequency), as in trying to measure a delay on a sub, the Auto Delay Function is basically useless.
Harry: How do you go about aligning them (two speakers, 100ft apart)? I have been using the Impulse Response for one, then the other, and then doing the math… Is there a better method?
Jamie: First, pick the position you want to align them at. This is key. Remember, one delay time does not work everywhere. Think of two pieces of string attached to the speakers. As you move around, one goes slack while the other stays tight. After choosing your alignment point, measure the arrival time of the first speaker. I use the Impulse Response measurement directly for this.
The Auto Delay Locate Function just does an Impulse Response measurement and chooses the tallest peak. I like to see the entire response to make sure I’m not being fooled, and then I turn the first speaker off, turn the second on, and measure its arrival. Whether you do the math in your head, or on a napkin, or using SMAART’s Compare Delay Function is up to you.
Devin DeVore: What are your thoughts on applying Fletcher-Munson curves to SMAART measurements? SMAART is linear and our ears aren’t.
Jamie: Fletcher-Munson (F-M) is definitely important - I find it comes into play when looking at responses of different areas of your system, such as down front versus mix position. Regardless of what curve you end up with at the mix, you’re going to want to take F-M into consideration when looking at the response up front.
It will most likely be louder down there, so you want to allow for more LF to deal with the increased level. This is also a good rationalization for dealing with LF buildup in front of the subs… :>) To be honest, I think the key here is paying attention to level variations throughout the room and show and account for that in your system response curves. The starting point is kind of arbitrary.
Devin DeVore: Think we’ll ever see a ‘reference level’ standard in AES like the projection world has?
Jamie: Maybe, but I’m not sure how useful it would be. Sorry about the lame answer.
Moderator: Since the F-M curves flatten out at higher SPL, the area down front can sometimes sound “over-hyped”. These are the kinds of decisions we make as humans. It’s your gig as the sound guy… Am I close to correct?
Jamie: I guess this speaks to our goal of uniformity as system engineers. We have to balance measured uniformity with perceived uniformity. Plus, we often fail to take level differences into consideration when comparing response curves. Level uniformity can go a long way to providing an excellent environment for mixers and artists. And yeah, we always have to be taking the human into consideration when making alignment decisions.
This is why auto-align is a pipe dream.
Devin DeVore: Funny you say that ... I learned the most about ‘mix energy’ and frequency masking from mastering engineers, and I’ve been applying that knowledge to system tuning. Anything thoughts on frequency masking and SMAART?
Jamie: Yes, but I’m not sure I can type out a sufficient answer. In general, though, systems that are smooth in response, not “peaky”, tend to be far more useable. They don’t need to be “flat”. But having one range of frequencies as a stand out makes for a bumpy ride as a mixer. On that subject, the RTA works as a great mix tool for doing channel EQ (equalization) for just this reason.
If you do a long average of a signal - say a guitar - the curve won’t be flat, but it probably shouldn’t be peaky. An engineer I worked with - Grant MacAree on k.d. lang - showed me this technique for sound check. He’d have each person play their instrument nice and evenly, and we’d do a long average. And wouldn’t you know it, those pesky peaks that you try to use your channel EQ to tame show right up on the RTA trace.
Joel: During a live performance, how can we adjust the analyzer to consider the sound coming from the stage as noise, and try to eliminate it from the readings?
Jamie: First rule during a show is measure twice, cut once. In other words, be VERY careful. Watch your measurement for areas of stability and instability. The key to getting a good, usable measurement during a show is getting a mic position that is dominated by the speaker(s) you are trying to measure. Often the mix position is the worst place cause it has left, right and stage showing up.
Use lots of averaging, and use your ears. To be honest, I sometimes turn Transfer Function off during a show if I can’t get a stable measurement. It’s better to do nothing than make random adjustments based on bad measurements.
Bob Capotosto: Can you cover the relationship between the Transfer Function and coherence traces?
Jamie: Coherence is basically looking at how consistent your data is. In a perfect world, your measurement should be dead-on for each consecutive measurement, such as when you are measuring an electronic device like an equalizer: no noise, high repeatability, and good coherence.
Coherence can be affected by noise, reverberance (late arriving info), bad measurement delay setting, time misalignment, distortion… To figure out if the bad coherence is being caused by noise, simply turn your measurement noise up. As the level goes up, coherence should get better. When you hit the point that it no longer improves from turning the noise up, you have reached the point where the poor coherence is not due to room noise.
What you are left with then is the product of reverberance, cancellations, distortion, etc. Keep in mind that you shouldn’t expect to see great coherence in a medium-sized room. Remember to increase your averages to stabilize your measurements and give you better s/n (signal to noise ratio). You won’t see the coherence trace go up, but you will be able to trust your data more. With 64 averages, I trust data at 20 percent Coh (Coherence) as long as it appears stable.
Harry: Does this apply to 3.5?
Jamie: Since 3.5 doesn’t use overlapping FFT’s, you are basically looking at more data in fewer averages. In SIM, we used the following guidelines, and this would certainly be comparable to 3.5. At two averages, we wanted 90 percent or better; at four averages we wanted 60 percent or better; at eight averages, we wanted 40 percent or better; and at 16 averages, we wanted 20 percent coherence or better. I guess the only drag with 3.5 and coherence is that it is not live Coh.
AlanH: Reference mics… even Behringer is making one. Audix, Earthworks… How do the lower priced ones stack up?
Jamie: As you pay more for a mic, you get: flatter response, extended range, lower self-noise, and better consistency from mic to mic. There’s a lot of mics that will work well with SMAART. The question is: how flat do you need it? What frequency range do you need? What mic-to-mic consistency?
Jamie: To be honest, most live measurement situations do not need a B&K (Bruel & Kjaer) 4007, although it is a great mic. The key: do you need a hyper-flat response above 16 kHz? What is it worth to you? And, will you freak out if that mic is stolen or bashed with a chair cart?
Chris Kathman: The chair cart raises its ugly head!
Jamie: That is probably why SIA sells the Earthworks mic. It is a great balance between performance and price, a great value. We also sell the Audix, which is also a good measurement mic at a great price. I guess it just depends what you need out of your mic.
Larry Elliott: There was some thought a while back that there would be a (Apple) Macintosh version of SMAART. Is this still being considered?
Jamie: The short answer is no, not really. The problem is that we don’t have the resources to go multi-platform. It’s an economic issue. For the price of developing and maintaining the code, we would have to crank the price way up. Sorry. And I am/was a pretty devoted “Mac Guy”. It just isn’t in the cards right now.
Larry Elliott: I seem to get “adequate” performance on a G4 Powerbook with Virtual PC Version 4.
Jamie: That’s what we’ve been hearing. I think that’s great news. Unfortunately, that also makes justifying doing a Mac version harder.
SRS Audio: Jamie, I was in your SMAART class held at NSCA. My boss asked me the other day if I could precisely measure reverb time with SMAART?
Jamie: The NSCA class was way over-booked. My bad, I guess. Anyway. I guess the key here is “precise”. There are questions of how you excite the room, what frequencies you are specifically talking about? The T-60 can be amazingly different at differing frequencies.
SRS Audio: Speech range, mostly.
Jamie: You would need to break it down into bands, and even then, it’s a bit of a subjective call in selecting a starting and stopping point. I guess I’m the wrong guy to be asking about this, in a way, because I do not work as an acoustician on a regular basis.
There seems to be a fuzzy art/science line in these measurements.
Harry: Many people say the great thing about SMAART is you can use the show as a source. How can this be done when both sides are on, and usually different distances from the mic?
Jamie: I guess it all has to do with how isolated you can make your measurement. You can also use a mono source and place the mic in the exact center. The key is NOT TO MAKE DECISIONS ON A BAD MEASUREMENT. I just wanted to say that loud to remind people that a squiggly line doesn’t mean you have a useful measurement.
Chris Kathman: Yeah, people! What he said!
Jamie: I also believe that, if you do your job setting up, you can generally make adjustments by ear during the show. I regularly turn off TF (Transfer Function) mode and switch people over to RTA during a show. Put the RTA off the Cue buss, and let the mixer use it to police his signal. RIGHT ARM! FAR EAST!
Devin DeVore: What about measuring distortion? It would be good information to know how the distortion changes with different SPL and different array set-ups, Also, Dr. Don preaches well about critical distance.
Jamie: Distortion is a hard thing to measure with SMAART. So many elements go into your measurement, and there are so many things that could be affecting your coherence, which is the main indication of distortion in our measurements. Looking at distortion versus level is probably a measurement best suited for a lab, and maybe an AP.
But that is not to say it isn’t a critical factor in our listening experience, just that it is something that we probably won’t be using a SMAART measurement during a show to figure out. And, you can bet that if the good Dr. is preaching about it, then it is definitely a key issue. I guess one of the things we are talking about here is measuring at a distance that isn’t in the high SPL of the horn throat, and not being back buried in the reverberant field.
For the purposes of our measurements, let’s stay in the linear range of the devices when we measure them. Yeah, we might take them up to their limits as the show goes on. But remember that SMAART is expecting a linear system when it makes it’s measurement, so non-linearities like clipping and the limiters that are there to protect us from clipping are best kept out of your alignment measurements.
Devin DeVore: Any suggestions on phase steering with SMAART and EQ’ing for phase? Can you or Dr. Don add some case studies, or another SMAART class with a focus on phase steering?
Jamie: Phase determines how two signals will add. The whole game of system engineering and system alignment is controlling and managing the interactions between elements of your system. The key ingredients in phase are polarity, phase shift (from EQ’s, crossovers, etc.), and timing. Control those and you control how and where your system adds and subtracts.
Take the example of a stack of subwoofers. If you simply put them in a nice vertical stack, they will tend to narrow their pattern in the vertical. If you want to affect this pattern, say open the coverage up vertically, you can delay subs incrementally as you go up, effectively acting like arcing your subs in an upward direction.
Jamie: While not perfect, this will open up the coverage because you are shifting the equal time arrival point upward. The same effect can be applied to two stereo stacks of subs to de-emphasize the big LF build-up in the center of the room, which is the equal time arrival point. You can play with delaying one side of the subs to push that equal time arrival point - the focus - to the side.
Jamie: T he thing to keep in mind, always, with these types of things is that they are imperfect substitutions for true physical solutions to your problems. Steering is taking control of the elements that determine how a system interacts. Level and phase.
Moderator: Thank you, Jamie.
Jamie: Thanks. Let’s do this again. Maybe I’ll be a little less shell-shocked. Hasta and peace.