(and long answers to short technical questions about IMP and Liberty AudioSuite)
return to Liberty Instruments Main Page
> I am having problems figuring out how to make accurate phase measurements. I can change Delay, via F9, to get the curve to change wildy. when is it the "real" value?
LONG ANSWER FOLLOWS!:
I think some of the problem is a little misunderstanding of the nature of acoustical phase. The speed of sound being finite, the "real" value is very much a **relative** value depending on where you are in relation to the speaker, and where you define the speaker's spatial reference point to be. There is no one "real" value, unless you define some specific spatial relationship between the mic and the speaker or driver. You can't even say "at the speaker's source", because where is that?-- at the voice coil? cone edge? pole piece? dust cap?
If you move the mic one wavelength further or nearer (at some frequency, say 20kHz, where a wavelength is 0.67") then the phase will roll one additional time between very low frequencies and that frequency. Move further and the phase rolls will change more (and concentrate more as frequency increases). Changing the DCMP (delay, [F9]) is equivalent to moving the mic forward or back, assuming the source acts as a point source. (There is a short article on Time, Frequency, Phase and Delay, which you can find on our web page, discussing this in more detail, if interested).
An amplifier or filter, or other electronic components have very temporally defined output points, so it is easy and practical to define an exact phase response for those. You can measure at that one defined point, and that is the answer, period. But for a speaker, there are an infinite number of output points in 3D space. The response *magnitude* changes for a speaker in ways that are intuitively clear... off axis the highs usually go down, and moving further the whole response magntidude curve goes down in a relatively uniform way. So the spatial relativity of response magnitude is rather innocuous.
But with phase, things get much more messy -- first, the scale is periodic: when the phase drops enough negative (to -180 degrees), it suddenly becomes positive again (since -180 degrees is in *every* way identical to +180 degrees). Second, the phase at high frequencies is more sensitive to position than for low frequencies, simply because wavelength is inversely proportional to frequency. And the shape changes in an apparently wild manner, doesn't just move uniformly up or down as does magnitude. Further, the phase response doesn't affect waveform reproduction integrity in an easily observed way -- uniform (flat) DELAY is perfect for ideal reproduction, but that can look like a really ugly phase response; a really smooth phase response could conceivably result in really distorted waveform reproduction. (Delay is the unrolled phase divided by frequency.)
So, how to handle it in measurement?.... What I do is define my measurement reference plane for each driver as the **mounting baffle plane**. That point is physical and easily determined and does not involve digging into the driver to find the pole piece or voice coil, or whatever, and it puts all drivers measured that way on a uniform basis for use with simulation software. Refer all driver measurements to their mounting planes, or such a common external surface.
To do that:
Make both the Mic measurement and the Cal measurement using MLS (SINE gets strange because of windowing necessities). Before doing the FFT on the Mic measurement, set Time Marker #1 all the way to the left (that is, don't use the time marker to remove any delay) -- this assures that soundcard gains are accurately removed by the cal channel. Then do your FFT of the dual channel data. Phase will look like a mess. Measure the distance = x, in inches, from your mic's tip to the baffle plane. Speed of sound is 13,543 inches/second, so you can assume that there is x/13.543 milliseconds of delay taken by the sound to get from the baffle plane to the mic. Press [F9] and enter this delay value (x/13.543). Now the displayed phase is correct for the driver, as refered to its baffle mounting plane (that is, it is what would ideally be measured if you could put the mic at the driver, at that plane, without messing up the driver's operation or getting into nearfield effects). It won't be a straight line, and shouldn't be.
If you use this data for all drivers, then the simulations should work out. This is all assuming that the design optimization in the simulator is being done on a rational basis -- if one is trying to optimize for "phase=0" or such nonsense, disappointment will surely result-- even if it were possible, the mic would have to be embedded into all the drivers to realize it. The only rational basis for phase optimization of a loudspeaker is for uniform (or some prescribed) DELAY versus frequency, not for any specifc phase shape-- the phase curve, even a perfect ideal one, will always be a wild function of just how far away you are from the speaker. More often, the optimization is for magnitude flatness of a combined crossover, and this technique of normalizing phase will work with that correctly-- it matters only that all drivers are measured under equivalent conditions so that any delays are matched.
> Perhaps a silly question, but what are the
'probes' and what are they used
Not silly at all. The probes are basically just a few 1% resistors mounted inside an RCA inline connector, with some wires coming out that have alligator clips on the end. They are used for sampling electrical signals and attenuating them
so they can be measured by the FIJI DSP card. You can't put high level signals into the card, as it can be damaged by voltages over 5V peak. You can make your own probes if you want to save a few $, they are not difficult.
A speaker converts electrical to acoustical energy. To really measure it correctly, you need to measure both what goes in (electrical) and what comes out (acoustical) and calculate what happens between. The mic does the acoustical pickup, the probe does the electrical pickup. Also, impedance measurements require two electrical measurements, so two probes are used to determine both voltage and current through what is being tested.
>If I understand correctly, I don't need Windows
to run LAUD, do I?
You don't need it, LAUD can be run from DOS only.
But you'd find it much easier to install the Fiji DSP soundcard hardware with Windows.
Getting LAUD to work with DOS-only can be a bit of an ordeal, because the FIJI card's
supplied software, which is Windows based, can't then be used to determine what resources
the FIJI must be set up for. Windows 95 is a much easier environment in which to
set this up. Also, after installation, Win95 will allow you to go between LAUD and a
word processor or other software to move graphs into reports, etc.
Note, btw, that I said "Win95". If you have the choice, get 95, rather than Windows98. Win98 can work ok with LAUD, but you have to be careful to keep Win98 from trying to use the FIJI card after LAUD has rearranged it for its own uses (Win98 locks up if it tries to use sound hardware that doesn't behave as it expects -- not exactly "progress"!). More on Win98/LAUD
>Can LAUD be used with a laptop or notebook computer (for portable operation)?
Sorry, but in general you cannot use LAUD with a laptop/notebook PC. (But you can use Praxis).
LAUD works only with certain specific DSP soundcards (the ones currently available are the Turtle Beach "FIJI" and "PINNACLE"), which require installation into an ISA slot. LAUD relies on synchronous and simultaneous stimulus and acquisition, not possible with the built-in sound functions of laptop PCs. None of the PCMCIA type cards will work at all with LAUD, and neither the MIDI, Joystick, Serial nor Parallel Printer ports are usable to connect to a LAUD compatible card. They lack sufficient bandwidth for LAUD, which must stream digital audio and DSP commands over the bus.
There have in the past been several laptop type computers which either had an internal ISA slot or which could provide an ISA slot via an external expansion chassis. Some machines can be equipped with a "docking station" (such as IBM's DOCK1) having ISA slots within -- in such a case, you may be able to use LAUD in your notebook, while it is in its docking station (making it not quite as portable). If you find such a computer/compatible docking station combination and wish to use it with LAUD, make sure that it has an ISA slot which provides a complete 16-bit full bandwidth slot.
If a portable LAUD system is required, the best choice is to use what is called a "lunchbox" or "portable" computer. These computers use a standard PC motherboard (with ISA and/or PCI slots) and use standard hard drives and other peripherals. The lunchbox computers use a special chassis which includes a handle at the top, a built-in LCD screen and a fold-down keyboard. This makes a very versatile portable system, somewhat larger and heavier than a laptop but much easier to expand or upgrade as it uses standard desktop PC type components. For technical usage, these are much less restrictive than are laptop or notebook computers.
> Is there a way to capture the screen image of
LAUD (such as Frequency Response graph) so that it can be pasted to a WORD document or
HTML file? Can the graph be saved as a file that can
be view at a later time?
Two ways, actually. I'm assuming you are working from Windows.
First (general-purpose) way is to use the key combination [Alt-PrintScreen]. This copies whatever is on the screen (even DOS programs like LAUD) to the Windows "scratchpad", so it can then be pasted into Windows "Paintbrush" (just use Edit Paste when the picture is in your scratchpad). From there you can edit or just save it to a file.
The other (built-in) way is to use LAUD's "Print toFile" option. Get there via the menu sequence [* Display Printplot Config] then choose toFile, and either the Black&White or Color version (B&W is best for printing, full color uses
a lot of ink and doesn't add much info for most plots). When you try to print (using [Shift-F2] or via the [* Display Printplot] menu), you will be prompted for a file name and then the computer will freeze for a few moments while the screen is copied to a bitmap file. You can find the file in your C:\LAUD directory, with the suffix ".BMP" (such as "MYPLOT.BMP"), and the file can be edited with Paintbrush or just inserted into your Word documents.
If you want to do these operation "live" (measure with LAUD, move it into a Windows document), have LAUD running (from within Windows), then use [Alt-Tab] to flip back to Windows -- you can then open Word and flip back and forth between that and LAUD (and/or other programs) using [Alt-Tab] again.
> Can I measure a speaker's anechoic response in
a normal room with just LAUD ?
Yes ...and no. Prepare for the long answer!:
You can use gating in LAUD (in its SINE, or MLS instruments) to get a frequency response measurement equivalent to an anechoic chamber down to a frequency of about 300Hz (it depends on how long of a delay can be arranged between the time a wave arrives directly from speaker to the microphone an when the first echo off a room wall, floor, or ceiling arrives at the microphone -- the lowest anechoic response frequency will be about 1/(that delay time)). Gating basically measures the response of the first-arrival wave, and ignores all waves after the first reflection arrives at the microphone.
Below the lower frequency limit, it is not possible to gate out the reflections from
the room because they arrive before a single cycle of the direct wave has completed.
At low frequencies, the time for one cycle increases, and will at some point
surpass the time between direct wave arrival and arrival of the first reflection.
Also, near this low frequency limit, the resolution becomes very poor. These
limitations are due to physics, NOT to the measuring system: it can not be
overcome by using Sinewaves or Chirps over Impulse or a different brand of MLS, for
instance -- it is simply due to the fact that any surfaces within a wavelength of a
speaker are in fact part of that speaker at that frequency. There is
no anechoic portion of the response, in the room at low frequencies.
At lower frequencies, an anechoic equivalent response measurement can be made by placing the microphone so close to the speaker that echoes are swamped out by the intensity of the measured close-direct wave. This is called a "near-field" technique, and is usually usable up to about 200Hz. The upper limit for this technique is due to the size of the driver being measured and the baffle size being used (the mic cannot be the same distance from all parts of the driver or baffle, so this causes error in the measurement at small wavelengths).
It is common to measure the high frequencies by a gated measurement and the low frequencies separately by a near-field technique. It is possible to join the two measurments into a single graph, though in practice this can be a difficult thing to accomplish accurately (phases and gains must be matched, sensitivities must be equated for the different distances, and a sufficient overlap range must exist). In practice, such a combined full-frequency graph has no real use except for marketing purposes -- for engineering design the two (or more) plots are usually left separate.
There is often a problem with the frequency range between 100Hz and 400Hz, as both of the mentioned techniques may be compromised in that range. Compromise spacing between "nearfield" (less than an inch) and reference distance (usually 1 meter) can often be used, with judgement required in the interpretation. Use of a large measurement room (with mic and speaker placed far from any surfaces) -- or an anechoic chamber, if that is available-- is the only sure practical way to improve the measurement in this problematic range.
Of course another issue is whether getting an anechoic response measurement is really even desirable at the lower and mid frequencies. No speaker is EVER used in an anechoic chamber normally and such a chamber is not at all similar to any room in which a speaker is really used, particularly at low frequencies. For example, a speaker will invariably be placed in use at a consistent distance from a reflecting floor and probably the walls as well, and this will cause a somewhat consistent effect on the frequency response at the low frequencies (usually an energy suckout between 200Hz and 500Hz). A realistic design should probably anticipate this effect because, again, that floor should be considered to be a significant part of the speaker at the low frequencies.
I would recommend making gated measurements in a normal laboratory for the frequency range above 400Hz, as this can be representative of what your customers may hear, and need not assume any particular room configuration. At low frequencies the room is unavoidable, but is somewhat predicatable: use near-field technique to be sure the enclosure is tuned correctly and that sufficient energy is available from the speaker at the low frequencies.. Then make in-room measurements (including all echoes, no gating or nearfield) at a typical floor (and wall) arrangement to adjust driver placement for response below 400Hz-- the woofer location on the baffle will have much effect here --and to adjust perhaps the crossover and equalizer.
Be warned this response will NEVER be flat -- but getting the best balance this way will have much more benefit to the perceived sound of the speaker. More so than would tweaking an irrelevant anechoic-equivalent response at the mid and low frequencies.
>Will LAUD work with Windows 98?
LAUD works fine with 98, but Win98 is not so well behaved with LAUD -- Win98 can lose its wits if it tries to use the Fiji soundcard after LAUD has taken over the Fiji's DSP after LAUD is exited. So please be sure to see at www.libinst.com/Win98.htm on our page for information on getting around this glitch!
>Will LAUD work with Windows NT,
Windows2000, or Windows XP?
No. Sorry, not possible -- LAUD needs direct access to hardware, NT/2000/XP forbids hardware access from DOS applications without OS approval.
>The LAUD manual says I need
a dos mouse driver present. How do I know if
that is there?
You are very unlikely to have this problem. This was mostly an issue with those who were operating Windows 3.1, and some who had installed 95 on top of 3.1. If for some reason the mouse driver is not there, you would be able to tell because the LAUD software would not give you a mouse pointer -- you would have to operate using he keyboard only (which is possible for most operations but not very convenient). If the mouse doesn't work, we can send you a generic mouse driver that will probably do the trick.
>I'm making SPL measurements, but notice that the
results seem to be independent of power level into the speaker.
I expect that when I vary power in that measured SPL should vary. Instead, I
see the calculated SPL vary with distance, but remain constant at a given distance
regardless of input.
LAUD (in the MLS or SINE instruments) does not measure "SPL", but "SPL Sensitivity" -- in other words, the SPL that will result for a given drive level of 2.83Vrms (which is not necessarily the level applied in making the measurement). It will NOT be a function of the drive or output signal level as long as the speaker and system are in their linear operating regions.
Note that the MLS and SINE instruments are LINEAR SYSTEM measurement devices, which assume that what is being measured is linear -- that is, if the drive level increases by X dB, the output level increases by the same X
dB. So it doesn't matter what the actual drive level is for this measurement -- the output is assumed to scale linearly with the input and the relationship (the "SPL Sensitivity") is a constant curve. It is a measure of how sensitive the speaker is (how efficiently it converts electrical voltage signals to acoustical output), NOT how loud it will play!
Measurement of how loud the system will play is a Distortion analyzer (NONLINEAR System) type measurement, so to do that, use LAUD's DISTAN. In that case, how much the signal is distorted, as a function of output level, is the critical parameter (after all, simply "how loud" is meaningless --applying 1000Megawatts to a speaker will make it play quite loudly for an extremely brief and explosive time -- but the result would be a tad ....distorted....!).
>When I'm combining responses from a port and a driver, both measured in the near field, do I have to adjust the levels to get them to combine correctly?
A good article dealing with this (if you have access to AES reprints, or
want to order them from www.aes.org ) is:
"Simulated Free Field Measurements" by Struck and Temme, JAES Vol 42, No. 6, 1994 June.
In a nutshell, assuming that both the port and woofer response are measured with the same gain and both in the nearfield.....
Increase the woofer's response "gain" by the amount (in dB): 10 log[(woofer area)/(port area)].
The reason for this is easy to see, in a thought
At very low (well below box resonance) frequency, the farfield radiated power from the woofer and the port must be equal and opposite. Why? Because since the port just radiates what comes off the back of the cone, the phase is reversed (and at near-DC frequency, the distance and box will have no effect on the phase) . And also we know that at very low frequencies the radiated power of the summed system tends toward zero (woofers don't radiate DC pressure).
So, think of just plain air squirting in or out of these
two different sized apertures, with radiated power being proportional to volume velocity
(gallons/sec, or the like). You can imagine that if the same magnitude power radiation
happens from both (so they can cancel near DC as they have to), then the air must be
squirting out faster across any point of the smaller aperture. Because the same air
*volume* must go through the total smaller aperture as goes through the the larger
aperture, in a given time instant. The way the signals sum at higher (but still longish
wavelength) frequencies is the same way they sum at near DC, except of course the box then
causes them to no longer be equal and opposite.
The nearfield mic will pick up pressure at just a point in an aperture, so it must be reading a **nearfield** power signal, at the larger woofer, that is (woofer area)/(port area) smaller than at the port. The air rushes past faster (i.e., with higher pressure) from the smaller port to arrive at the same volume velocity, for a given radiated power. And, to get a power signal ratio into dB, you take 10* the log of it.... and there's the formula!
>Why does my microphone data file only go down to 1kHz or so?
The mic is not calibrated far below 1kHz for two reasons:
1) The mic capsule is too small and the element is too light to have very complex behavior below this range. Tests indicate that the response is flat (within a dB) from this range down to about 10Hz.
2) Calibration against the reference mic is a large effort to do at very low frequencies, because the freefield method can't be used (room reflections are too severe). To do so would unnecessarily increase the costs of the microphones and the calibrated capsules. Extra calibration at low frequencies can be done (at extra charge) but the results are pretty boring -- the result is invariably more or less just a drawing of a flat line down there.
LAUD and IMP use the lowest frequency value in the correction data file and extrapolate, assuming the same calibration value for all lower frequencies.
>In this system, what frequency is the highest limit to measure for response, 2nd Harmonic Distortion, 3rd HD and total HD ?
The soundcard can only resolve frequencies below 21kHz (because the highest sample rate is 48kHz). So response can be measured to 21kHz, second HD for fundamentals below 10.5kHz, third HD for fundamentals below 7kHz, etc. Total HD measures up to 10.5kHz, but only harmonics up to the 21kHz audio band limit are included -- so, for example, above 7kHz, the "total" includes really just the second harmonic.
> Will there be a Windows or NT version of Laud?
Not of LAUD, but Liberty Instruments' new -praxis- system will fill this requirement.
> The LAUD manual makes frequent reference to the Mic/Probe Preamp. I am using LAUD with a Turtle Beach FIJI or PINNACLE card. Do I need this preamp? Will it enhance my LAUD system?
In general, references to the Mic/Probe preamp can be ignored when using the Fiji card -- for example, if the manual mentions using the Gain switches on the M/P Preamp, you merely don't move any switch (since it is not present or needed).
Using the M/PP is essential with the ECHO or PSA type cards, because the input gain adjustment range was limited and the mic preamp portion of the cards are often noisy and rolled off. But with the FIJI, much better distortion resolution is possible if the M/PP is not used; operation is much simpler without the M/PP; and since the FIJI has a quiet, flat mic input and much gain adjustment range, the M/PP is not needed or recommended.
>What's the best card to use with LAUD- Fiji, Pinnacle (and, with or without Digital I/O)?
The FIJI and PINNACLE are absolutely identical in the portions that are used by LAUD. Pinnacle is a Fiji with a very sophisticated MIDI synthesizer/sampler added, of interest to musicians but not to engineers. For LAUD-only use, the main differences are cost (Pinnacle costs a LOT more), ease of installation (Pinnacle needs more hardware facilties - interrupts, ports, etc., all of which must install in NON-PLUG'N'-PLAY mode), and physical size (Pinnacle is BIG - a FULL size card, rare these days, and won't even fit into some computer enclosures). The FIJI is nearly always the best choice.
There's only one type of Fiji Card - the ones "with Digital I/O" are
identical to the "without" except they come packed with an adaptor that plugs
onto the card. LAUD does not use this Digital I/O adaptor, but it
will not be in the way, either. It may be useful for pro or semi-pro audio editing,
if you do that with your FIJI card. But save your money, don't get it if cost and
complexity are to be minimized.