QUESTION:
Dear Professor, It would please me much if you could find the patience to describe to me the Transfer Function and its relation to the physical ear. Thankyou much, Thomas (CD 200)
ANSWER:
No problem Thomas - the transfer function is a complete description of the response of a linear system to any input. A linear system obeys the principle of superposition. This means that the response to any input is just a version of that input where each frequency is individually scaled in amplitude and shifted in phase. The transfer function tells us how the amplitude and phase are affected at each frequency.
The ear is not a linear system. For example, when listening to two tones (sinusoids) played simultaneously, you may very often hear a third tone resulting from distortion produced by the inner ear (we'll have more to say about this). This is a violation of the principle of superposition. It makes it very difficult to predict the response of the ear to complex sounds.
Prof L.
QUESTION:
What's the difference between changing pressure and changing power?
ANSWER:
Power is proportional to pressure squared, increasing pressure by a factor of 2 therefore means increasing power by a factor of 2^2=4. Similarly, in dB, doubling pressure means a 6 dB increment, doubling power means a 3 dB increment. Halving pressure is a 6 dB decrement, halving power a 3 dB decrement.
Prof L.
QUESTION:
How many equations will you be asking on the exam, what types of questions will you be asking?
ANSWER:
The short answer is that you are responsible for everything that was covered in the handouts and the lectures. There are very few equations - I won't for instance ask you to derive the wave equation but you should understand what is required to satisfy it and what laws it is based on. I won't ask you to compute a sinusiodal function but you should know what sin, w, A, and the other terms of this equation refer to. You should also be familiar with the short formulae for power, energy, intensity, etc. as they are given on your handout. The kinds of questions I will be asking will be exactly like those you had on the in-class quiz and on the class page.
Prof L.
QUESTION:
I am still confused on spectral analysis. Are we just suppose to know how to get the amplitude and the phase out of the waveform. I feel like I am missing part of the concept. How do you know the frequency.
ANSWER:
OK Tracey - the important thing to remember about spectral analysis is that it is just another way of representing sound. We began with a representation as a time waveform - amplitude of displacement over time. We showed that the time waveform is sinusoidal for a simple system (spring-mass system) and that this sinusoidal motion can be completely specified by its amplitude, frequency (number of cycles per second) and phase. Spectral analysis is just another way to represent this information graphically. We have two plots: one giving the amplitude (y-axis) as a function of frequency (x-axis) and the other giving phase as a function of frequency. Both plots therefore are given by a single vertical line. That's really all there is. For other more complex resonant systems (guitars, horns, voice etc.) there are other modes of resonance (other vertical lines) representing still other frequencies of vibration, but the principle is the same.
QUESTION:
Could you exlain the gate function briefly. I understand it is used to gate sounds on and off but the whole graph thing has thrown me.
ANSWER:
Yes indeed the graph representing the power spectrum of the gate function is quite complex and the relation it bears to the gate is certainly not obvious. Not to worry. I gave the gate function merely as an example - an important one. It shows that when we turn sounds on and off we generate energy at many different frequencies. This is an important point - Our representation of a sinusoid as a single vertical line is a theoretical representation pertaining only to continuous sinsoids having no beginning and no end.
QUESTION:
>Also could you work a problem with the RMS. Thank you
ANSWER:
RMS Problem: Compute the rms of the following instantaneous amplitudes.
2,0,3,-1,-2,0.
Step 1: S - Square all values
4,0,9,1,4,0
Step 2: M - take the Mean
(4+0+9+1+4+0)/6 = 3
Step 3: R - take the square-Root
answer: square root of 3
Hope this helps. Prof. L.
QUESTION:
I was wondering if phase shift increases as the frequency of the driving force decreases. It seemed this way from the computer model in class.
ANSWER:
Good question. - Actually Brad, its just the other way around. When the frequency of the driving force (fd) is small the frequency of oscillation is roughly in phase with the driving force. As fd increases the oscillation becomes more out of phase with the driving force. This stands to reason really - the spring-mass system has inertia so it takes some time to respond to changes in force acting upon it. If the changes in force are very slow (fd is small) then the system can follow or 'keep up' with those changes (response is in phase). If changes occur rapidly (fd is large) then the system begins to lag behind (response out of phase). Ever feel like your lagging behind when so much new material is presented in such a short time?
Prof. L.
QUESTION:
Hi Professor Lutfi - is there any way to do the homework a bit at a time? i.e. set up the mail message, finish some of the definitions, then save it without sending it, come back and finish it later, then send it?
ANSWER:
Yes Lisa - Here's what to do:
Click on the homework assignment in Netscape. Now instead of clicking on MAIL, go to the FILE menu and select SAVE AS. Give the file a short name and select TEXT as format. Finish the assignment using any word processor, just be sure to save your work as a text file. When you are done mail your file in Netscape using the ATTACH document feature. Select ATTACH within mail, find your file using the FILE and BROWSE buttons and then SEND. When I receive your assignment I will send you a reply.
Prof. L.
QUESTION:
I'm still confused about d' the distance between density functions. Is this always the difference between the means of each bell curve.
ANSWER:
d' is the distance between means in standard deviation (z) units. The difference in z scores (which is d') expresses the separation of the two distributions by taking into account both the distance between the actual means and the spread (sigma) of the distributions. It does this by expressing the difference relative to the spread, i.e. z=Mean/sigma.
Does this help?
Prof. L.
QUESTION:
What's duplex theory?
ANSWER:
Duplex theory proposes that their are two cues for localizing sound: interaural (between ear) differences in phase and interaural differences in intensity. Interaural differences in phase are used for sounds having frequencies less than 2000 Hz. THis is because at higher frequencies the wavelength of the sound is smaller than the diameter of the head resulting in interaural phase ambiguities. In other words the sound may pass through several cycles before reaching the other ear. At these high frequencies interaural intensity differences are used. When the wavelength is smaller than the the head sound is reflected and a sound shadow is produced on the opposite side of head. This provides a difference in intensity at the two ears that can be used to locate sounds of high frequencies.
Hope this helps.
Prof. L.
QUESTION:
II've never had a physics course before, including high school, so I hope a strong base knowledge of physics isn't required for CD202. Otherwise, I'm afraid that you'll be seeing too much of me in office hours! I enjoy the class thus far, and hope you keep it entertaining. Any suggestions for studying these graphs? How vital is detail in this class? Are we suppose to get a holistic understanding of these concepts, or would you like us to follow the book's more complex and detailed descriptions of what you give in class? I'll have to admit, this stuff is pretty foreign to me, and I'm not sure how to study it. Please get back to me if you want to devuldge any handy study suggestions. Thank you!
ANSWER:
Thanks Kristine for your message. You're not along in finding much of this material new - be assured though that I am aware of this. There are in fact some tips I can give you for studying the material - I've mentioned them before and I will be mentioning them again.
First - Pay attention to the Handouts. The basic concepts and defs required are all neatly packaged there. Study them BEFORE you attend class on each topic. If you know the material in the handouts you will do well in this class.
Second - Use the lectures to clarify material in the Handouts. If something is unclear during the lecture, ask the question then, or at least right after class.
Third - Use the readings in the assigned text to clarify material in the handouts.
Fourth - don't panic. The first few days are the hardest of the class - believe me it gets better. Also, noone who has been willing to do a little work has done badly in this class. If the palms of your hands still feel swety, then meet with me after class to go over those problem areas.
Hope this helps. We'll be reviewing the material again on Tues. of next week, so if there are any questions you have that would be a good time to ask them.
Prof L.
QUESTION:
I understand that we're supposed to multiply by 20 instead of 10 for pressure-based decimal problems, but where does this idea "power is proportional to pressure squared" come from? I tried squaring the unit for pressure and that didn't come out looking like the units for power.
ANSWER:
Good question. The simple answer is that power involves force exerted over a distance (cm) whereas pressure involves force per unit area (cm^2). If we take the ratio of two powers, we can cancell all other units and be left with a ratio of distances in cm. But if we do the same for pressure we are left with a ratio of two areas in cm^2. That's why power is proportional to pressure squared and why we multiply the log of a pressure ratio by 20 to be consistent with the definition of the dB.
QUESTION:
In your decible cheat sheet, log(2) is 0.3 and log (1/2) is -0.3. If you know the log of any value, can you get the log of its reciprocal just by changing the sign? Or is this just true for 2 vs 1/2 and 10 vs. 1/10 and other powers of 10?
ANSWER:
The log of the reciprocal of any number is just the negative of the log of that number - always. Remember the log function converts the operation of division (e.g. taking the reciprocal) to subtraction. So if log(x) = y then log(1/x) = -y. See,the purpose of logs is to make things simpler, not harder.
QUESTION:
Just to confirm what you did in class today - there's no need for any kind of conversion when your reference value changes from power to intensity? nothing like using 20 instead of 10 when you change from power to pressure?
ANSWER:
No. Because intensity is proportional to power. Intensity has unit of watt/cm^2. But when we take the ratio of two intensities we also have the ratio of two powers since cm^2 cancells out.
QUESTION:
You said that RMS is a pressure-like value - how?
ANSWER:
Remember in class that we described a sound wave as pressure variation over time. Amplitude is a generic term we used to describe the magnitude of the pressure variation - it is therefore proportional to pressure. RMS is the square root of the mean of the squared amplitudes - it is therefore also an amplitude measure and is also proportional to pressure.
QUESTION:
I didn't understand the last problem you talked about in class today where you ended up adding 6 dB if the RMS doubled. How can you relate RMS to decibles?
ANSWER:
In the same way that we can have dB referenced to a sound pressure, we can have dB referenced to an RMS amplitude. By definition dB = 20log(RMS/RMSref). Except for the difference in pressure and power, the unit of measure is of little consequence because it cancells in the ratio.
I hope this helps.
Prof. L.
QUESTION:
Are we responsible for knowing terms that we did not discuss and should we know any more about the terms than their definitions?
ANSWER:
As I mentioned I will not hold you responsible for terms I did not cover in class. If you know the definitions then you should do well on the exam. Knowing the definitions however means more than just being able to recite the words, it means understanding their meaning. If you have questions you can arrange to meet with me or send your questions via email at any time. I will be taking email questions right up to the time of the exam.
Prof. L.
QUESTION:
>Today in class, I was having trouble understanding the concept of >Lateral INhibition. I misunderstood it so well, that my notes don't >even make any sense. If you could enlighten me on the subject, that >would be great! >
ANSWER: Sure - short story is that fibers in the cochlear nucleus (first stage central auditory nervous system) have lateral (side-to-side) connections to neighboring fibers. These connections are inhibitory. That is - they act to suppress excition of the neighboring fibers. When one of these fibers gets really excited, it also tends to reduce the response of its neighbors. This serves to enhance contrast - what starts out as a broad pattern of excitation is reduced to a sharper more local pattern of excitation. In effect lateral inhibition serves to enhance spectral contrasts in the cochlear nucleus.
Hope this helps. - Prof L.
QUESTION:
Could you tell me how signal and noise power affect sensitivity. I wasn't sure when we got done with lecture.
Thanks, Jennifer
ANSWER:
Good Question Jennifer - increasing signal power makes the signal more detectable, hence it increases sensitivity. Increasing noise power makes the signal less detectable and so decreases sensitivity. Generally the ratio of signal-to-noise power is an important factor determining sensitivity, though it is not the only factor that affects performance. For example the cost of making certain types of errors (e.g. false alarm, miss) can also affect performance. If someone threatens to shoot your dog if you miss the signal on a trial, you are likely to make many more positive responses even though you would increase the likelihood of a false alarm. The costs or payoffs associated with certain responses affects your response criterion (response proclivity or bias) but does not affect your sensitivity for detecting the signal. The theory of signal detection allows us to estimate sensitivity free from the effects of response bias.
Does that help?
Prof. L.
Prof. Lutfi:
QUESTION:
I started working on the Homework set and I think I have a good understanding of everything except for what beta represents and what effects d' has on it. I'm guessing that if d' increases, B will decrease and vice versa. Am I on the right track or way off base?
ANSWER:
The sensitivity index d' is independent of criterion B. One does not influence the other. B is affected by the payoffs and costs associated with correct and incorrect responses, and by the probability of the occurance of a signal on a given trial. d' is affected by signal-to-noise ratio and by the sensitivity of the human observer.
Prof. L.
QUESTION:
In general, what happens to the response bias (criterion) in situations where the signal strenght is changed.
ANSWER:
Response bias is not influenced by signal strength, just as sensitivity is not influenced by response bias. The two are independent of one another and are influenced by different factors. Response bias (criterion) is affected by the payoffs and costs associated with making certain types of correct and incorrect responses. The sensitivity index (d') is affected by signal-to-noise ratio.
Prof. L.
QUESTION:
If it is possible, could you provide a definition for the time intensity trade for quiet threshold?
ANSWER:
For durations less than a quarter second threshold depends on the total energy of the signal. Energy = Intensity * Time, so if intensity is halved, duration must be doubled to maintain constant energy for detection. That is what is meant by the time-intensity trade.
QUESTION:
Could you explain 2 interval forced choice task and psychometric function briefly?
ANSWER:
In the standard 2 interval forced choice task there are two intervals of time during which a sound is presented. Both intervals contain noise,but only one contains an added signal. The signal occurs at random in interval one or two with equal probability, and the listener must indicate on each trial which of the two intervals appeared to contain the signal. The psychometric function is a plot of percent correct in this task versus the strength of the signal.
QUESTION:
2. what is the relationship between energy and duration of a sound?
ANSWER:
Energy is the product of intensity (power) and duration. If duration is halved so too is energy.
QUESTION:
Professor Lutfi,
From the Speaks book (p.148) I ascertain that dB and dB SPL are equal. Therefore, -74dB re: 1 ubar is equal to -74 dB SPL ? I am either incredibly inept or this is easier than I thought. Are the two equal?
ANSWER:
The two are not equal since the two have a different reference. Remember, a dB value has no meaning unless the reference is stated explicity. SPL designates a reference of 0.0002 dynes/cm^2, whereas re: 1 ubar designates a reference of 1 ubar (1 ubar being equal to 1 dyne/cm^2). If we express the one reference relative to the other in dB then we have 20log(0.0002/1) = -74 dB re: 1 ubar, or 20log(1/0.0002) = 74 dB SPL.
Hope this helps. QUESTION:
>Hi. I was studying today and something is not clear to me. In my notes i >have that the ear is a linear system, but in another part of my notes, I >have that the ear is a nonlinear system. Is it both, or just one? If one, >which one is right? Thank you
The ear is a nonlinear system, we can in fact here tones, produced by intermodulation distortion in the ear, that aren't present in the external sound. Despite this many scientists study the response of the ear as though it were a linear system. We'll have more to say about this later.
QUESTION:
>What is response proclivity?
ANSWER:
Response proclivity is a tendency, a bias, a predisposition, or inclination for a particular response. If I give you $10 everytime you correctly report the presence of a signal you may be inclined towards a positive response. However, your sensitivity for detecting the signal is not changed by my foolish generosity.
QUESTION:
>I understand how to get d' >and I pretty much understand what d' and false alarms and hits are, but I >am not totally confident on criterion and the relationships between all of >these.
ANSWER:
Ok Leah, here it is in a nutshell - d' (sensitivity) is affected by one thing - signal-to-noise ratio (s/n). Anything that increases s/n increases d', anything that reduces s/n reduces d'. A sensorineural hearing loss is generally believed to decrease s/n ratio and so decreases d'. The cost of a false alarm or the payoff associated with a hit does not affect s/n and so does not affect d'. However these things do affect response criterion. If I give you a dollar for each time you correctly report the presence of the signal (hit) but threaten to shoot your dog for each false alarm, you are likely to think twice before making a positive response. My threat affects your response criterion but it clearly does not affect your sensitivity.
QUESTION:
>Does one ear have certain advantages for sound detection over the other >(right for bass or whatever)? I wanted to ask this in class but I didn't >want to risk sounding like a retard.
ANSWER:
Well Brad there's always that risk, but it shouldn't stop you from asking. As far as I know there are no good studies to suggest that one ear is clearly better than the other in any specific situation. That's not to say that two ears are perfectly identical. For example careful measurements reveal small differences between the ears in sensitivity to individual frequencies, and these differences can become significant as we grow older.
QUESTION
Could you please clarify the defintion of coloration for me? I cannot find a good explanation for it in my notes or in the textbooks. Is it just the effects of interference of a sound in an enclosure? Like how the original sound is altered in the time it takes to travel from your stereo speaker to your ear?
ANSWER
Any sound system, including a room, will color the sound by reinforcing some frequencies and attenuating others. The resultant change in the spectrum of the sound is referred to as coloration.
QUESTION
HI Dr. Lutfi, Thank you for writing back so promptly. I just want to try to get the wording straight in response to this question below, so I can be accurate on the exam. So then, how come when in the homework when "force exerted over an area is F/x" whereas "force exerted over a distance is F times x"? How do I know whether or not to divide or multiply? Excuse me if I'm totally missing something.
ANSWER
OK - Suzanne, I agree the wording is a little confusing. The term 'over' is used here in the physical not the mathematical sense. In the mathematical sense a over b is to be interpreted a/b. In the physical sense exerting a force F over a distance x means that some mass m was moved for a distance of x. Let's take a simpler example. If you travel at a velocity of 60 mph OVER a period of 1 hour, the total distance you've traveled is 60 miles, 60 mph X 1 hour (not 60mph/1hr). Similarly if you exert a force OVER a distance x the total energy required to do this is Fx, not F/x.
QUESTION
>Dr. Lutfi- >a transfer function is not an >output, as far as i have learned. the defn. for transfer function that you >gave us, correct me if i am wrong, is the function of a linear system with >amplitude scaling and phase shifting at each frequency. so how does that >differ from the output of a linear system with sinusoidal driving force >with amplitude scaled and phase shifted version of the input sinusoid,?
ANSWER:
The output is an amp-scaled and phase-shifted version of the input, the transfer function tells you by what amount the amp is scaled and the phase is shifted. Somewhat like the volume control on your stereo - the setting determines the amp of the sound at the output of the speaker, but the setting itself is not an output.
QUESTION
>i am confused at how a delta function as a waveform having positive >ampl. at an instant in time, can have a flat spectrum that indicates a >particular ampl. over multiple freqencies? does the impulse itself which >set up the delta function waveform have multiple freqs?
ANSWER
Yes - it has all frequencies at equal amplitude. The reason for this is certainly not obvious and you are not expected to know why. But, if you're curious you might consider what happens at a particular instant in time when all of the sinusoids add together in pi/2 phase - constructive interference yields a large amplitude at that instant. But, at the very next instant the sinusoids become out of phase with each other (being of different frequencies) and so cancell. In fact they never get back in phase with one another again and so the result is zero for all remaining time.
QUESTION
>I am lost concerning what is 'pure tone quiet-threshold, Speech' and how >doe s this apply to the concha and the external auditory meatus? Please >explain. Thanks.
ANSWER
The pure-tone quiet threshold is the level at which a tone first becomes audible in quiet. The resonant properties of the concha and external auditory meatus affect this threshold, causing it to be lowest in the region between about 2000-5000 Hz. THis also happens to be the frequency region where much of the relevant information in speech is contained.
QUESTION
>While going over notes for the exam I came across something about resonance >in my notes about Interval Histograms. Unfortunatly I took down no >clarifications. Does this have to do with the most frequently occuring >time intervals being integer multiples (as scaled like the amp. and phase) >of the period of a sinusoid? Any help would be appreciated.
ANSWER
The most frequently occuring time intervals are indeed integer multiples of the period, which means that the spike responses are synchonized to the cycles of the sinusoid. This in turn means that the spikes occur with the same frequency as the sinusoid. There is a connection then to the principle of resonance which says that the frequency of the response (spikes in this case) is always equal to the frequency of the input sinusoid.
QUESTION
>What is the difference betweens Fechner's pyschophysical law and Stevens'?
ANSWER
Fechner - Sum of jnds = klog(I), Stevens - magnitude estimate) = I^p
QUESTION
>How does suppression actually enhance spectral contrasts, if you supress >both the noise and the signal?
ANSWER
When the signal is separated in time from the masker as in forward masking, only the masker is suppressed.
QUESTION >What makes informational masking any different from other masking?
ANSWER
Its due to uncertainty regarding the masker, hence masker energy need not fall within the critical band. QUESTION
>What exactly is the definition of pitch masking?
ANSWER
pitch masking refers to the experiment wherein an intense masking noise at the fundamental freq fails to mask the perceived pitch at this frequency.
QUESTION
>I have a >question about my hearing. I work at a local coffee shop and working >there entails steaming lots of milk for lattes. The noise from the >steam going into the milk is extremely loud and I have noticed that >after a 6 or 8 hour shift, some times even after couple of hours, my >ears have a "full" feeling, which I am sure can't be a good thing. Do >you think that I experience a temporary hearing loss after work? Do you >think that I should wear earplugs or maybe even resign if this >persists? Am I being a little paranoid? Anyway, I've only worked at >this shop for about 6 months, so I'm sure that if there is any damage it >isn't too bad--I don't notice any changes in my hearing, anyway.
ANSWER
Exposure to intense sound (particulary high-frequency sound as produced by coffee steamers) can indeed cause temporary shifts in hearing threshold - and if the problem persists for long periods of time may produce permanent hearing loss. Its not likely you'll go deaf making coffee, but if the sound is unusually intense and you have to listen to it for long periods of time it might be wise to wear hearing plugs. I'd be particulary concerned if your ears feel 'full' or fatigued at the end of the day.