Geddes on Distortion perception

I wanted to start this thread because I find that a lot of people aren't aware of the work that has been done in this area in the last five years or so. Much of this work is my own (our own) but there is also a lot of corroborating research - notably by Alex Voishvillo at JBL among others. I am best equiped to talk about my own so I won't specifically discuss others, but I am aware of it and it virtually all is saying the same thing.

In 2003 my partner and I published two papers on the perception of nonlinear distortion. Much of the results from this work is available here http://gedlee.com/distortion_perception.htm.

Basically through an ellaborate test of some 25 college students we were able to show that THD and IMD are meaningless measurements of distortion as far as perception is concerned. Basically one cannot say that something does or does not sound good based on these measurements. .01% can sound outrageous in some cases and 25% can be inaudible in others. The numbers are meaningless.

This result has been confirmed by several sources and now virtually eveyone in the loudspeaker business is coming to the conclusion that making THD measurements is pointless. Floyd Toole believes that nonlinearties in loudspeakers is irrelavent as evidenced by the fact that his new book contains no discussion of this topic. Lorri Fincham recently remarked at ALMA that THD and IMD were completely meaningless as a judge of sound quality. My own presentation from ALMA (China) last year says the same thing and maybe goes even a bit further.

Basically distortion, as we are used to thinking about it, is completely incorrect. This was further confirmed when we did a study of compression drivers published in JAES. In this study no one of about 30 subjects could hear nonlinear distortion up to the thermal limit of the driver - some 126 dB at the waveguide. This result was surprising and quite controversial, but it is holding firm as quite correct.

There are things that we perceive as distortion-like artifacts, but these are not nonlinearities in the drivers themselves, but are actually nonlinearities in our hearing system. This was brought to like by my partner and I in Oct. 2006 at the AES convention. These diffraction-like artifacts are perceived quite readily by us, but only at higher SPL levels, there are not audible at lower SPLs. These effects are virtually ignored in most loudspeaker designs.

All in all the situation is unfolding quite differently than what has been presumed to be the reality.

Recent studies of mine have clearly shown the human PREFERENCE for distortion of low order or at higher amplitudes. These are viewed subjectively as enriching the sound.

I am putting this topic out so that people can become aware of what is being recognized as the truth about distortion. Its not what we thought it was. Pursuing a loudspeaker design to lower the distortion is a waste of time if its nonlinear distortion that you are trying to lower. It simply doesn't matter.
 
Dr Geddes,

Thanks for starting this thread as I find this topic very interesting and hope other pros and/or knowledgeable DIY'ers will chime in.

I noticed your reference to line arrays on the "beyond the ariel" thread and wondered what your take was on the advantages beyond directivity that you mentioned. I have always thought (and perhaps quite incorrectly) that one of the reasons I have been so impressed by the line arrays I have heard is that with all those drivers sharing the load, the distortion levels were far lower than your typical two or three way speaker with only one driver per crossover region... Is there a "system distortion" of some type that is not THD or IMD that could be in effect here, or am I possibly just responding favorably to the slight time smear and the "tail" you were mentioning that comes with all those drivers producing all those various "arrival times" for each piece of the music signal being reproduced.

Anyways, what I find particularly fascinating is the ongoing struggle to sort out the pros/cons for each of the following approaches (and I guess I already know were you come down on these issues as evidenced by your consistent and well thought comments throughout this forum):

1) Classic 2, 3 or 4 way speaker system with one driver for each crossover region designed to approximate a "point source" (here I think you and Mr Olson are fairly well aligned in that even though you have a few different priorities, you are in the point source camp and looking to optimize the performance of individual drivers that have exclusive freq range to cover).

2) Short line array or focused array of 2,3, or 4 identical drivers sharing the same freq range and designed to approximate a point source at a specified listening distance.

3) True Line Array (in the spirit of Dr Griffins design criteria and recommendations)

I have heard many decent commercial examples of number 1 above, and before reading about your work and the many musings of Mr Olson, I had assumed that the only way to improve upon these designs was through active crossovers to help improve the coupling of amp/driver to lower distortion (maybe not an issue now given your comments here) and improve dynamic range. I have not heard any examples of number 2 above but would like to and will do some DIY experiements soon (but not experiments you would approve, just subjective listening for now...). And the few examples of number 3 above were quite impressive to me. I wonder if just the overall mass of acoustic energy transfered into the room all at once is what I am responding so positvely to when I hear good line arrays such as the stuff Rick Craig is putting out there with Selah Audio...

Anyways, to turn a quick thought into a long post... if not IMD or THD reduction, do you have any theories as to whay a line array sounds so much more cleaner and dynamic that conventional speakers?

And finally, I really have been hoping your work would find some commerical traction and I am sad to hear that you remain frustrated in these endeavors. I know this sounds corny, but I really think you should look at the various network marketing models, finding an advocate in every major city that is interested in becoming a sales/demo/distributor to custom home and commerical sound vendors. Can wave guides be sold like cosmetics, vitamins and jewelry? Maybe not, but if nothing else is working it might be worth exploring.

Thanks for your many posts and best of luck in your ongoing research and commercial ventures.
 
Thanks for the comments.

I am affraid that your question presumes the answer.

"Anyways, to turn a quick thought into a long post... if not IMD or THD reduction, do you have any theories as to whay a line array sounds so much more cleaner and dynamic that conventional speakers?"

You are presuming that your premise "a line array sounds so much more cleaner and dynamic that conventional speakers" is actually true. How do you know that? Because you listened to a couple and came to this conclusion? I'm not trying to be argumentative here, but I can't comment on YOUR perceptions. Show me some real data, or some blind studies that confirm your hypothesis is in fact true and we can move from there. But to assume the unproven as fact is simply not something that I can comment on.

Since it is know that people will tend to prefer distortion in many cases, how do you know that your preference isn't related to a higher level of distortion?

You just can't presume things about "preference".

Now you can talk about "accuracy" and show how frequency response irregularities may make something less "accurate", but "preference" is a big can of worms. Remember in my work I talked about "perception" not "preference". They are certainly NOT the same things.
 
That's quite a radical departure from what we have always been lead to believe. Would this also be true for any nonlinear distortion in amplifiers. Assuming the levels in loudspeakers are considerably higher.

When the change over was made from Alnico to Ferrite in the 70's due to a cobalt shortage there were issues with nonlinear effects from the new ferrite motors. One of the side effects of the new ferrite motors was higher distortion levels. The motors were re-designed to lower these levels and are actually better than the alnico's they replaced in this respect.

Was this redesign effort in vane based on these new findings??

If distortion levels are not a valid measure of performance what parameters should you be looking at when you select drivers for a DIY project??

Rob:)
 
Yes, you are quite right, so let me re-state my question another way (sometimes setting up a broader context just makes a person look like an idiot across a broader set of evaluation criteria, I prove that often on this board and others... ):

Are you aware of any reason, with the emphasis on reasons supported by sound theory, that a tightly clustered (like 9 small mid domes arranged like a tic-tac-toe framework) or linear array (like a classic line array of drivers) of multiple drivers covering the same frequency range would have a physical/electrical/acoustical advantage over a single driver. I am mostly interested in your take on the generally accepted belief that such combinations, though potentially suffering from other performance disadvantages, would present the listener with the advantage of much lower distortion than a single driver. I guess the second part would be, if THD and IMD are meaningless in speaker measurements, could there be any other reasons, in theory, why a group of drivers would be able to outperform a single driver in the execution of converting an electical signal from an appropriate source into acoustical energy in a residential listening environment. Just trying to get a handle on what scientific guidelines I should/could consider when evaluating the three different approaches I mentioned above, beyond the obvious considerations of thinking, building, listening, and if properly equipped, attempting to measure...

Hope I am not stinking up your thread with my obviously newbie questions here. Essentially, you are a formula 1 race car driver and I just learned to ride a bike : ) I am hoping you will share more about how you drive the cool car that makes noice and goes fast.

By the way, if ever a reasonable opportunity to hear for myself, I would love to audition the Summa as now I am fascinated to discover how my own perception of what I hear from your speakers translates to my own set of preferences and how I rationalize it all out.
 

Pan

Member
2002-09-19 3:52 pm
Harmonic and IM measurements are in no way meaningless.

In blindtests one can easily hear the coloration from electronics that has HD in the order of less than 1%.

So how gedlee has come to his conclusion is very strange.

Gedlee, are your goal to produce high spl's with subjectively good sound or are you into hifi as in neutrality and transparency?

Recent studies of mine have clearly shown the human PREFERENCE for distortion of low order or at higher amplitudes. These are viewed subjectively as enriching the sound.

Several blindtests that I have done (and many others) with electronics that have even mostly low order distortion shows that the result is a closed in muffled non engaging sound. The loss of transparency is annoying.

The nice effect that some people experience is something I have not experienced so far. The problem seems to get worse with more complex music where intermodulation products between different instruments are the result.

I find it absurd that 5-25% distortion from your loudspeakers would be inaudible.. and also I'm curious how you did those tests.


Someone mentioned an anecdote about line sources sounding clean and you asked him to do some blindtests and get back. Now.. do you perform all your tests blind? Even with the collegestudents? Also when you perform your tests do you use "controls" so you know what the system and the people that constitutes the listening panels CAN hear?

I thought you felt that HD and IMD from amplifiers could be audible even if it hides way down in the noisefloor. Now I assume that with the amps you use this would mean levels far far below that from the speakers. I can only understand that as that you think electronics need to have low distortion but loudspeakers can be tolerated even if they have thousands times more distortion?

What kind of distortionlevels have you been able to detect in blindtests on amps for example?



/Peter
 
Hi Greggo,

Re your line/single driver enquiry.

For equal cone area a single driver will axially beam more than a line source in the near field; then in the near field that single HF beam will be more LF modulated relative to the distributed HF of a line.

Yet there are other subtle changes in low/high reproduction with distance for a line, and Gedlee explains how he makes effort to match the radiation patterns of drivers across the AF spectrum.

Another aspect is that drivers will distort differently when driven via different impedances, so looking at distortion figures "under one light" cannot reveal all.

Perceived distortion of music reproduction is related to amplitude linearity in music time, which relates to coherence, frequency dependent group delays and interface reactivity induced resonances, not the 'better light' sine based measurements after time dependent responses have settled.

Cheers ............. Graham.
 
Robh3606 said:
That's quite a radical departure from what we have always been lead to believe. Would this also be true for any nonlinear distortion in amplifiers. Assuming the levels in loudspeakers are considerably higher.

When the change over was made from Alnico to Ferrite in the 70's due to a cobalt shortage there were issues with nonlinear effects from the new ferrite motors. One of the side effects of the new ferrite motors was higher distortion levels. The motors were re-designed to lower these levels and are actually better than the alnico's they replaced in this respect.

Was this redesign effort in vane based on these new findings??

If distortion levels are not a valid measure of performance what parameters should you be looking at when you select drivers for a DIY project??

Rob:)


I don't believe that linearity of the voice coil drive is all that important, but certainly any measure using THD to optimize it is a waste of time. Since there must be nonlinearity in the voice coil drive (it must go to zero at the end points) I would optimize it to maximize 2nd and 3rd harmonics and minimize the higher ones.

I have found the system design to be far more important than the driver design so I don't obsess on the drivers. But I do look for good frequency response - sharp resonances are hard to control no matter what you do. I prefer fewer drivers to many because the crossover is a real problem and they are always detrimental, so I don't want any more of them than necessary and I especially don't want ANY above about 1 kHz. The biggest problem that I find in my systems is finding a good woofer that can go to 1 kHz without a serious edge hole or spider resonance. They almost don't exist so you do the best that you can.
 

graaf

Member
2007-06-23 10:31 am
Thank You Dr Geddes for starting this thread! :)

first a comment to some previous posts because I can see a huge misunderstanding in them as to what You claim -

from what You have written here in this thread and elsewhere I understand that harmonic and intermodulation distortions ARE relevant. What is irrelevant is measured "total" THD or IMD figure.
Am I correct?

As to the matter of the discussion -
I have no expert knowledge on the topic, only my own audiophile subjective experience.
But to be somewhat helpful in the discussion I decided to throw some Linkwitz observations into it, because Linkwitz seems to be very keen on the topic of nonlinear distortions

He says:
"Low distortion means increased dynamics and clarity, and the ability to play back at near live levels without listener strain. It brings the naturally occurring distortion mechanisms of the ear into play, which are necessary elements for creating an illusion of reality.”

On the other hand he makes an interesting distinction between "audible effects of distortion" and "being perceived as distortion":
"It should be understood that distortion has audible effects long before it is perceived as distortion. For example, it might enhance sonic detail, analogous to contrast enhancement in a photo."

And if I understand correctly his point of view it’s not harmonic distortion itself that causes audible problems in case of musical signal but non-harmonic intermodulation products of harmonic distortion:

"These intermodulation products are usually higher in amplitude than the related harmonic distortion products. I cannot emphasize enough, that anytime when a device introduces harmonic distortion, it will generate intermodulation distortion, when more than one tone is involved. Some seem to think that a little bit of 2nd harmonic distortion, as often generated by tube equipment, has a pleasing effect, because it enriches the natural even harmonics of acoustic instruments. It will also generate non-harmonic intermodulation, and has little to do with accurate sound reproduction or even generating the illusion of a real sound. Instead it imparts a euphonic sameness to all sounds.
Real sounds contain many tones. A woofer might be tested with a set of 5 tones that are not harmonically related, such as 20, 28, 44, 64, and 92 Hz, but which cover the woofer's intended operating range. Each of the tones is a multiple of 4 Hz, with factors 5, 7, 11, 16, 23, and the resulting distortion products will occur at 4 Hz intervals."

Linkwitz says that his tests with "multitone signals" are useful for sorting and relative ranking of drivers: "Their subjective evaluation when part of a complete speaker correlates closely with the measurements"

If I remember correctly these intermodulation products are more likely to be perceived as distortions if they result from higher order harmonic distortions.

I have a question in that regard. Isn’t it that we have mainly low order harmonic distortions in loudspeakers (up to 5th?) whereas the electronic audio circuit can generate much higher of them even up to 19th harmonic?
Can it be that this is the main reason that loudspeakers seem to be relatively unproblematic element of audio reproduction chain from the perspective of harmonic distortions?

Linkwitz recommends Klippel’s study about distortion in loudspeaker drivers: Wolfgang Klippel, "Loudspeaker Nonlinearities - Causes, Parameters, Symptoms", 119th AES Convention, New York, October 2005

Do You know this paper? What’s Your opinion?

many questions, as usual, not all very wise ;) thanks in advance for Your patience :)

best,
graaf
 
We are talking about loudspeakers here and not electronics. The situation is quite different with electronics.

It's the order of the nonlinearity and where this nonlinearity occurs that is audible. If it occurs near the zero crossing then it is more audible than it occuring near the peaks. If it is high order it is more audible.

Now in a loudspeaker the nonlinearity is mostly low order because higher orders require large accelerations to generate and the relatively heavy mechanical system would require very high forces to generate them. And most loudspeakers are linear for small displacements. This means that they TEND not to have audible nonlinear distortion problems.

Electronics on the other hand is quite different. It is easy for electronics to generate high orders of nonlinearity - they are very wide bandwidth, And it is common for them to have nonlinearity at low signal levels. So the typical kind of electronics distortion - crossover - is by far the most audible and actually quite common.

Haven't you ever wondered how it is that you could hear .1% of electronic distortion through a loudspeaker that is typically 1-5%.

As to the line arrays. I generally don't see an adavtntage to them, except in one aspect. Having many "similar" units - and you only have to look at a production variation to realize that supposed identical units are at best similar - will tend to average out the individual resonances etc. making the net result smoother and more palitable. But I don not think that this benefit is worth the downside. Lower distortion is not the answer. I do think that smoothing out the individual driver and using as few as possible is the better option.
 
graaf said:
Thank You Dr Geddes for starting this thread! :)

from what You have written here in this thread and elsewhere I understand that harmonic and intermodulation distortions ARE relevant. What is irrelevant is measured "total" THD or IMD figure.
Am I correct?

This is partially true. Nonlinearity CAN be a problem, but THD and IMD never tell us when it is or isn't. There are ways to determine if a nonlinearity will be a problem or not, but these are virtually never used.

But, I have found through intensive research that nonlinearity in the types of loudspeaker systems that I design is simply not an issue. It CAN be an issue if the loudspeaker system is not designed properly, but it can just as easily be design so that nonlinearity is not an issue.

Everything depends on the system design.

I have a question in that regard. Isn’t it that we have mainly low order harmonic distortions in loudspeakers (up to 5th?) whereas the electronic audio circuit can generate much higher of them even up to 19th harmonic?
Can it be that this is the main reason that loudspeakers seem to be relatively unproblematic element of audio reproduction chain from the perspective of harmonic distortions?

Linkwitz recommends Klippel’s study about distortion in loudspeaker drivers: Wolfgang Klippel, "Loudspeaker Nonlinearities - Causes, Parameters, Symptoms", 119th AES Convention, New York, October 2005

Do You know this paper? What’s Your opinion?


I am not going to argue with Linkwitz through an intermediary, but I will say that nothing of what you quoted has ever been published in a peer reviewed setting and for the most part his evaluation techniques would never be accepted in this setting. Very very few people take the time and trouble to do subjective tests in a manner that would make them acceptable in the scientific community. Does that make them wrong? No, but they can't claim them to be right either if they are not willing to put them up to the scrutiny of peer review.

Please distinguish between Harmonic order and nonlineaity order as they are not the same thing. Harmonic distortion is a symptom of the problem, which is nonlinearity, it is not the problem in and of itself. In this I agree with Linkwitz about the fact that a nonlinearity causes both THD and IMD because these are simply symptoms of the same problem.

So yes the low orders of nonlinearity in a loudspeaker tend to be far less audible than the high orders typically found in an amplifier. But the "where" is also important. Amps can have increasing nonlinearity with decreasing level - this is VERY audible.

I am very familiar with Wolfgangs papers. They are quite good at explaining where the nonlinearities come from. But they don't do anything to say how audible they are.
 

soongsc

Member
2005-03-26 2:31 pm
Taiwan
gedlee said:



...
The biggest problem that I find in my systems is finding a good woofer that can go to 1 kHz without a serious edge hole or spider resonance. They almost don't exist so you do the best that you can.
I also think this is a problem, and wonder whether this might be a reason Ted Jordan used a different suspension approach to his woofers.

I also think that fewer drivers are better than more if the necessary SPL can be reached.
 
jzagaja said:

Could you explain harmonic versus nonlinear order?

In a system there can be defined what is called the "transfer characteristic". For a linear system this is a straiight line through the origin or mathematically

y(x) = gain*x

If the system is nonlinear then this line is no longer straight and in general can be represented as

y(x) = gain*x + a1* x^2 + a2 * x^3 + ...

The nonlinear order is the term in x^2 or x^3 etc. Hence first order is linear then there is a 2nd order nonlinearity, a 3rd order and so on.

Now if I send a sine wave through this system then what comes out is distorted and has harmonics of the main signal, and we name these harmonics by there "order". There is a relationship between the nonlinear order and the harmonic order, but it is not simply one-to-one, and for multitones it is extremely complex such that there basically isn't a direct relationship between the nonlinear order and the various extraneous tones that are generated from multi-tones.

What is certainly true however is that higher orders of nonlinearity tend to generate higher order harmonics and they tend to generate spuriuos tones in a multi-tone situation that are further away from the input signal tones.

So in a nutshell, talking about harmonic orders quickly becomes less meaningfull while talkig about nonlinear orders is always meaningful. One, the nonlinear orders, is a characteristic of the system, while the other, harmonic orders, is a characteristic of the specific signal being used. One applies to any and all signals the other doesn't.
 

Pan

Member
2002-09-19 3:52 pm
Well the thing is that even though high order harmonics and crossoverdistortion in electronics often is very audible.. a class A amp with dominantly second order (and quickly falling from there) still is audibly at one percent or less. That is a type of nonlinearity that can be compared to the speaker driver distortion at low levels. I write low levels because obviously many drivers produce gross distortion with higher order products when they collapses as the input signal gets higher.

I am very familiar with Wolfgangs papers. They are quite good at explaining where the nonlinearities come from. But they don't do anything to say how audible they are.

Then you should get familiar with the website. There are some tests there that you can do, read about it, do the tests and get back. ;) The results people get from those tests do not suport your claims.


/Peter
 

soongsc

Member
2005-03-26 2:31 pm
Taiwan
Don't forget that various forms of stored energy is also a very audible source of distortion. If we see to systems that have very close frequency response, stored energy might be one of the driving factors determining why they sound very different.

On thing to bear in mind is that our perception of sound balance is very much influenced by continuous frequency content. If energy from past signals mixes back into oncoming signals, then we hear a different tone balance. Not much of this can be seen from normal distortion measurement methods. However, if distortion measurements can use normal music signal to determine distortion levels, then we might be able to see more closer relationship beteen the measured distortion and what is heard.
 
I'm very interested in hearing the storage energy. Here's my transducer with 0.3mg ribbon membrane. When you stretch the foil then you will get 10dB storage energy at circa 800Hz.

BTW: frequency response reflects transformer resonance.
 

Attachments

  • ribbonmeasurement1.gif
    ribbonmeasurement1.gif
    23.1 KB · Views: 1,464

graaf

Member
2007-06-23 10:31 am
Pan said:
Well the thing is that even though high order harmonics and crossoverdistortion in electronics often is very audible.. a class A amp with dominantly second order (and quickly falling from there) still is audibly at one percent or less. That is a type of nonlinearity that can be compared to the speaker driver distortion at low levels. I write low levels because obviously many drivers produce gross distortion with higher order products when they collapses as the input signal gets higher.

/Peter

well, not necessarily, check NRC THD measurements of various loudspeakers:
http://www.soundstageav.com/avreviews_speakers.html

they are measured anechoically from 50Hz to 10kHz at 90 dB at 2 meters (equivalent to 96dB at 1 meter), which is (in their opinion) "very loud and considered far beyond normal listening levels"
96 dB at 1 m continuous definitely is not "a low level", isn't it?

as to "normal listening" I think they know what they are saying, it’s "National Research Council" after all :)

BTW I have read that harmonic distortions are most audible at SPLs around 70 dB average (M. E. Bryan and H. D. Parbrook "Just detectible amplitude of 2nd - 8th harmonic in the presence of a 360 Hz fundamental (f1)"

the fact is that the distortions level measured by NRC for tested loudspeakers very seldom reaches above -45 dB threshold i.e. about 0.6% at 1kHz (amplifiers THD data are for 1 kHz)

it is higher only in case of exotic designs - for instance first order crossover to low for a tweeter (several percent THD in case of some Thiel and Dynaudio loudspeakers at around 1<2 kHz)

even unfiltered (sic!) wide range driver (two-cone, with a "whizzer") applied as a midbass driver (in Zu Druid) has THD figure in this measurement generally below -45 dB (only one spike to -35 dB at around 5 kHz)

some minimonitors also failed in that regard but 96 dB continuous at 1m is perhaps too much for a minimonitor, they are probably not made for that

most of measured loudspeakers distort more at low frequencies, where those distortions are generally less audible, aren’t they? (correct me if I’m wrong)

best,
graaf
 
I think the only way to answer the question is application specific.

I've spent a lot of diy time making comparitive distortion measurements. If used correctly they have value. Can they be misused? Sure.

For example, in the most common diy scenario, a 2 way with a 6.5" woofer, it is quite easy to drive the woofer to audible distortion. Using comparitive distortion measurements is a no brainer. In addition, in this most common scenario, a 1" dome is used. Generally, a lower xover point improves power uniformity. But, how low can you go with a given dome tweeter? Well, looking at comparitive distortion measurements can help quite a bit.

For example, this tweeter is a terrible choice and has obvious strain and distortion if you try a 1.5k cross-

[IMGDEAD]http://www.audioheuristics.org/measurements/Testing/Tweeter5/graphs/LPG_nonlinear_1pt5k.GIF[/IMGDEAD]



On the other hand this tweeter has an outstanding distortion profile and can handle 1.5k with ease (notwithstanding the high spls that horns can generate) Note also the difference is more due to higher order products, not 2nd and 3rd order and related, which are not that dissimilar and would make the bulk of any THD type measurements.

[IMGDEAD]http://www.audioheuristics.org/measurements/Testing/Seas27tbfc/27tbfc_graphs/Seas27tbfc_nonlinear_1pt5k_SL.GIF[/IMGDEAD]



Sure, in a well designed 3 way, or a 2 way horn with a 15" woofer, well, except in the case of a truely poorly engineered setup, nonlinear distortion will probably not be very audible. So arbitrarily picking a driver on the basis of an arbitrarily lower distortion parameter is not the correct one.

Still, many low spl systems that audiophiles have, for example, small woofers crossed low to say ribbon drivers do suffer from audibly high distortion.

Nonlinearity is undesireable. All things being equal, better engineered drivers have lower nonlinearity. Having said that, there is a point at which going below a certain level has no audible benefit. The trick is engineering your system to make sure you're well below the limit.