Something stirs in the Undergrowth

[g3dahl]

I use Audacity on my Mac. In the current version, importing is done by going to File > Import > Audio, then browsing and selecting the desired file. In the previous version, it was under Project > Import Audio.

Are you able to import a file, but only with truncated bandwidth? I'm not quite sure what you are running up against.

Gary Dahl

[IslandPink]

Thanks Gary - that sounds worth trying . I'm not able to import the file ( yet ) that's the problem . The truncated bandwidth was seen on an analysis using 'Audacity' . I wanted to check the same wav file using AudioTester .
Now however, AudioTester won't load and shows lots of error messages before crashing .
I had some of this when I shut it down last night .
Looks like I'll have to reinstall with some different choices.
It seems to be getting all screwed-up about where to locate and call files from .

[IslandPink]

Just re-installed the program .
'File' has no 'import' option in my drop-downs . Have checked elsewhere, can't see anything ; and the help routine returns no matches for 'import' .

[IslandPink]

OK, I've got the equipment set up for test again . Just had the white noise playing into the microphone input, and it's showing on the screen on a nice frequency graph . What it seems to be doing at present is showing a real-time FFT that is all dancing about and noisy . I'm wondering next how to capture a load of this raw data and then produce a better FFT curve from it ?

[IslandPink]

OK, I sussed something about the 'average' window . If I rack the number up from '1' to about '250' , it gathers data for about 25 seconds and produces a much better plot . I've no idea which sort of the many transforms is best , but here is the default (Rife Vince 3) . It still cuts off at 8kHz, but there's enough data for me to make a quick adjustment of the feed resistor to the tweeter to bring down the level above 5k . I'll do as much as I can in the next half an hour then pack up . Any hints/tips gratefully received !

[IslandPink]

Ah well, have to admit defeat for now . The above plot is not meaningful, as the response doesn't change when I disconnect the tweeter . If I capture data while the sound is not playing, I get a different curve, but only a little lower , around the -110dB level on the above pic. So whatever I'm feeding the soundcard is barely above the noise floor, for some reason . ( Yes, I have the mic pre-amp switched on ! ) . Only the area above about 4k is really peeking above the noise floor .
I'll have to come back to this in about 10 days time .

Edit : Oh fer 'cks sake, the connector into the microphone wasn't quite plugged-in fully !
Help, need to eat something before I faint !

[IslandPink]

Ok, before I finally stopped, I managed to take about 4 more measurements, this time with the input level much higher - ie. the microphone input now working !
See below , with crossovers , 6.8R input resistor to tweeter crossover on this .
This seemed very hopefully at first . I then took further measurements , a second - just a repeat ; then with the tweeter connected without the 6.8k in series to check the level above 5 k would rise ; then a test without the tweeter, but with just the Fostex and its LP crossover ; then a test with the Fostex alone without its LP crossover . The plots are all the same
So, I'm now really fed-up !
(ps. yes, all of these four set-ups sound different to the ears )

[g3dahl]

Hi Mark,

Try this help page. Looks like you need to use File > Open.

Gary Dahl

[IslandPink]

Yes, but that's Audacity, Gary
I think I got past the problems with Audiotester by doing the measurements again and analysing them real-time, but the bigger problem now is what on earth happened on the last four plots I captured, as the data seems to be the same, to all intents and purposes. Also the soundcard now seems to be limited to 8KHz ... maybe I changed it some time after 2009 ? - can't remember .

[g3dahl]

I misunderstood you, then; I thought you were having trouble importing audio into Audacity. I do hope you can get your sound card squared away--optimizing a speaker design is challenging enough when the test equipment is behaving properly.

Gary Dahl

[IslandPink]

Thanks . I had some email discussion with Nick today. Hoping to get a better soundcard for the PC .

[slowmotion]

But how does the music sound?

[IslandPink]

The music sounds great at times, but not as universally great as it did on Sunday !
There has been some run-in of drivers and components, I suspect, plus the usual passage of time making me more critical .
So, I'm not sure at present what is causing the rough edges . The crossover was designed using some sims on LT Spice plus some guesswork . A basic FR plot from axis and about 10 degrees was my first step in de-bugging this situation .
At present there are a few possibilities -
Treble level from tweeter too high - increase series resistor
Fostex 'crud' in area from 5k up still too high in the mix - in which case use a bigger LP choke and/or a zobel to the Fostex.
Tweeter HP set too low - adding a hump in FR around 5kHz .
Time alignment to tweeter not quite right - it is set back by around 20mm at the moment .

Anyway, I think the time-alignment is pretty good, as the tests on Sunday did get a very nice level of integration and 'space' developing around the baffle . Guitars sound very good . Vocals sound pretty realistic - female ones at least. I need more output below 250Hz to get the male vocals right - that comes later . Tone is strong, but a bit too bright now in the treble generally .

If I can sort out whether the tweeter series cap needs to change ( to alter the ~5kHz HP upslope ) then I can buy a better quality cap for this item , which will clean-up the HF a bit more .

[pre65]

I got a "better" sound balance on my baffles by padding the fostex with a 2R resistor, but mine is a series crossover if that makes a difference.

[IslandPink]

It's a different situation Phil . I have the Fostex at about 95/96dB where it crosses to the G3 ribbon at ~100db/w . The tweeter needs to come down . 4.7R sounds roughly right so far.