Hello All,
this is becoming even more exciting!!
@Sumidero:
"Question: Is it possible that some 'comb filter'-like effect appears when reproducing a mono signal through multiple loudspeakers for acoustic measurements?"
Assuming you are in an Airport departure concourse.
You have a big multichannel PA system that should provide voice messages to all the people in the concourse.
You will have a guy speaking to a mic and says something like "Flight xxx will depart from gate yyy".
This is a mono signal reproduced by a multi-channel system.
Comb filtering are likely to happen but this is what you would like to measure to understand if the system would work and what are the downturns on having such a big PA.
Sumidero wrote: "I saw, many years ago, a PC-XP based ISA card for acoustic measurement of loudspeakers that instead of using a sine sweep it sent short trains of frequency fixed sines, opening the record just when the wave train was passing through the microphone."
is this system you are referring to?
http://www.mlssa.com/
IF yes, MLS (Maximum Lenght Signal) measurement system have been the industry standard until the sweep sine method comes out (expecially the log swept sine).
According to my knowledge, It is a very good way of measuring IRs but it has some problems in handling the non-linearities of the reproduction and recording chain.
@Katja
I will have a look to your patch and post some comments later, thank you very much.
I had a read through the 'Hamburg' article and he actually give us the exact function for the inverse swept sine which includes the amplitude inversion (or amplitude reduction, depends how you call it).
The way it is written is a bit confusing, I know, but after a bit of thinking you can see it like that:
assuming your funstion has a development in time (t) that goes between 0 and L-1.
the inverse function including time inversal and amplitude scaling is:
x^-1(t)= L-1-t)*(w2/w1)^(-t/L-1)
so if you want to write as written on physics books where your time variable t goes between 0 and T, it is like that:
x^-1(t)= T-t)*(w2/w1)^(-t/T) where 0<t<T
the function T-t) is basically the expression of the test signal (that you find previously in the paper) where instead of n (or t) you got L-1-n (or T-t).
this T-t) is expressed like that:
T-t)= sin(((w1-T)/log(w2/w1))*((e^((T-t)/T)*log(w2/w1)))-1)
I hope it is much more understandable now; I tried to use the same way of writing it in excel.
If you look at the xls file in the zip file i posted previously the formula is written properly and there is a graph showing the result.
This is very time consuming I know but there is no rush in finishing it as it could be an important tool in the free software community, I reckon.
Have a nice weekend guys
Bassik