room responses etc.

[Magnus Danielson]

From: Martin Czech <czech at Micronas.Com>
Subject: room responses etc.
Date: Tue, 8 Aug 2000 09:13:44 +0200 (MET DST)

Hi Martin,

> I'm taking up the room responses thread which was laying down so long
> due to excessive cycling.

I have said this to you before, all this cycling isn't good for you! ;D

> The correlation technique is now clear to me
> and first computer experiments were successfull.

Ah, nice to hear!

> Further on I plan:
> 
> -a set of portable battery driven devices for recording of such responses
> anywhere I like and I'm allowed to. Battery driven things might be
> psychological better for getting a permission for churches etc. as well.

Also, it takes less time to setup and you get finnished in a short moment.
This can also promote people to allow their rooms to be measured.

> -I want to use a pseudo random noise source, this gives a pulse
> train which should directly drive a loudspeaker via MOSFET push-pull
> tansistors and capacitive coupling. I need only a unipolar supply then
> and the "amplifier" will consume a minimum of power, cause it is really
> switching. The speakers and crossover network should filter out most of
> the hf garbage...
> 
> -I want to use my portable DAT for recording. Professional equipment seems
> to do severall recordings for averaging.  This leads to synchronisation
> problems with the noise source. One idea to get arround is to use the
> optical out as clock source via pll for the pseudo noise gen. This
> locking would also reduce possible alias problems.
> 
> -All this (speaker, battery, microphone, DAT, cables...) should fit into
> a medium size suitcase.
> 
> Problems: 
> 
> -nonlinear characteristic of "power amp". How does it change with
> battery life? 

Depends on your design. One trick is to use a switcher which will ensure the
rigth powerlevels so that you don't have to worry that much. As your battery
power drops, the switcher will just pull more current.

> -hf artefacts? alias ?

Yes, if you do not kill them off sufficiently.

> -how stable will the pll be?

The PLL will track the relatively stable clock as it pops out of your DAT.
You can make your PLL pretty narrowbanded.

> -every now and then the system has to be calibrated, i.e. a recording
> has to be made in an anechoic chamber. Needless to say I have none.
> would it suffice to go outdoors and do a free air recording?
> What about ground reflexions?

The trouble with anechoic chambers is that they are only good downto some
frequency, then the damping doesn't do much and it behaves like a box anyway.

I have done zillions of free air recordings of speakers. Now, this is why the
correlation technique is used! Yes, you will get a ground reflection and
possibly a zillion others. However, the first trick is to cut the responce just
before the first reflection (which should be the ground reflection if you have
done things right). Now, the time length from the begining of the impulse
(as seen by the mike) to the truncated end of the responce will set the lowest
frequency for which the responce says anything. If you have very, very little
energy after the truncation then the error of extending with zeros is not very
large, but otherwise you don't look beyond that point. Also, an FFT should only
be done for the valid area. The second trick there is to lift your speaker and
mike assembly up in the air, at least 3-4 meters. Now, naturally you have a
tower around that allows you to do that. When you do this you effectively
moves the ground reflection away in time (and amplitude!).

The ground reflection is approximatly twice the higth away, so a lift of 4 m
is worth 8 m to the first reflection, so that gives 8 / 343 = 23.3 ms of time
or 42.875 Hz of frequency.

Also, the longer away the first reflection is, the more has the speaker
silenced, so a truncation and then concatenation of zeros provides less error.

> -microphone characteristic, I don't know which char. is best

Well, normal measurement mikes are cardiod, but from room responces in mono
you would probably go for spherical responce. For stereo responces I would use
two cardiod mikes, either at 90 deg or 180 deg of spreading.

With a Soundfield you can select your characteristics back home ;D

Cheers,
Magnus