Subject:Frequency Sweep? & Aliasing? in Synthesis

Posted by: Rednroll

Date:1/22/2002 8:46:01 AM

I was trying to compile a CD of numerous tracks consisting of Sine Waves and Frequency Sweeps to create a Test tone CD. In particular I was trying to create a Track which was 2 minutes long and did a frequency Sweep from 10Hz to 22Khz, to use as a test disk to find frequency holes in a sound system. For the life of me I couldn't figure out how to do this using the "Tools>Synthesis>FM". I figured I would be able to start with a preset like the "4.5Khz to 22Khz test tone sweep" and edit it to include the the lower spectrum from 10Hz to 4.5Khz. First problem is that the maximum length you can generate in the Synthesis is 60 Seconds. Why is this? So I figured I will have to do 2 seperate sweeps and then edit them together to get my full 2 minutes that I desire. Can someone describe what the parameters would have to be to get a 10Hz to 22Khz frequency sweep to get a 1 minute sweep? I can figure out how to do the rest once I'm able to figure out how I get my start and stop frequencies of the sweep.

Also, while using the "Simple" sine wave generator, I noticed lower frequencies being added to the sine wave of anything I generated above 20Khz. Note I'm using 44.1Khz 16 bit. When you generate anything above 20Khz you can see the meters in sound forge start to bounce up and down, because a lower oscillation frequency has been added to the sinewave. When I create a 19Khz sine wave @0dB the sinewave is pure and there is no fluctuations in the meter and I can barely hear the sound wearing headphones. When I then create 21Khz sinewave @0dB I can hear a lower frequency being oscillated along with the 21Khz and the meters fluctuate with this oscillation. Thus I don't believe my ears are more sensitive at 21Khz than 19Khz, but I'm definitely hearing more sound.

BTW, I've tried this in v4.5 and v5.0 and both have this same problem.

And one other thing. If my sampling frequency is 44.1Khz and simple synthesis let's me make a setting to make a sinewave at 22,050Hz then why doesn't anything get created? I must set it one knotch below the Nyquist frequency for something to be generated.

A little help would be greatly appreciated,
Brian Franz

Subject:RE: Frequency Sweep? & Aliasing? in Synthesis

Reply by: oconned

Date:1/22/2002 3:04:37 PM

get a life...

Subject:RE: Frequency Sweep? & Aliasing? in Synthesis

Reply by: nlamartina

Date:1/22/2002 4:33:17 PM

Okay, how about a mature response,

Anyway Brian, I have two suggestions for you (and neither is to get a life). While I haven't a lot of experience with the FM synthesis part of Sound Forge, I have a couple ideas that could help you out:

1. Purchase a test CD, like Lasertrak.

Now if you're a do-it-yourself kinda guy, or if you're looking to compile a CD to sell, this might not be what you're looking for. But then again, it might be. At the very least, it could give you a starting point, or it could just save you the trouble. Regardless, it's an option. You can pick up one of the CD's from the manufacturer for about $20, or you could get one for about $9 at uBid, plus shipping.

Factory Direct:
http://www.testdisc.com/
(This site is also developing a program specifically designed for audio equipment calibration. Could be useful.)

uBid:
http://www.ubid.com/actn/opn/getpage.asp?AuctionId=27410150
(Or just do a search for "LaserTrak")

2. Use the test tones included with Acoustic Mirror.

As a quick test, I took the 48 second sweep (which covers the sonic spectrum, per your specifications), trimmed the "click" data, and time stretched it to 2 minutes. Not a perfect tone, but certainly a start. It warbles just a tad (for mathematical reasons I assume), but I'm sure a bit of experimentation and some careful calculations could yield a very nice result. The tones are found at this address:

http://www.sonicfoundry.com/download/Step3.asp?URLID=889&SerialNum=&EntryPoint=

Now I have a couple questions concerning your other observations. First, concerning the low frequency data being placed in the +22,000 Hz sine waves… When I synthesize the data, I see the "beating" effect you cite on the meters, but I'm not hearing the low noise you speak of, nor do I notice the woofers of my speakers moving during this time (which would suggest subsonic data, even if very little). However, I can notice a high-pitched electrical hum that pulsates along the same visual representation. This only occurs when the signal exceeds 22,040 Hz. Now I know for a fact that the hum I'm hearing is not the sine wave. My hearing drops out around 19,600 Hz, and then the sound card induces noise at around 22,030 (aliasing perhaps?). Since we both have similar results, but different noises, it suggests to me that what you're experiencing is a hardware issue, and not software. However, I can't be sure of that without more research. What's the frequency response of both your speakers and your sound board?

Again, that's only a hypothesis. I'll keep investigating the matter, and let you know if I find anything out.

Keep in touch,
Nick LaMartina

Subject:RE: Frequency Sweep? & Aliasing? in Synthesis

Reply by: Rednroll

Date:1/22/2002 5:48:44 PM

Thanks for the response Nick. Not quite what I'm looking for. I have a lot of test CD's already a couple made by Abex and another made by ford, which fits their specifications for automotive testing. Both are rather expensive CDs, but I find myself always flipping between 5 different of these test CD's and wanted to make a compulation CD that fits all my needs for analyzing a car audio system.

I am an engineer that works for a small international audio company that happens to own JBL, Lexicon, AKG, DBX, Infinity and Mark Levinson, and about 10 other highly respected audio brand names. I happen to help set up and analyze problems in the protype and post production phase of designing surround sound systems in automobiles. So I'm trying to create 1 CD that will fit all my needs in finding holes and phase problems in the cars audio system. Most analysis CDs don't make their tracks long enough to do a good analysis. So Mr. "occoned" I have a life......can you get me another order of fries please?
Also, the frequency and sweeps must be very accurate, because I am doing very audiophile type of work that requires it, so the timestretch option just won't work. I just hate having to drag a frequency generator out to the car, ripping the dash apart and plugging it into an aux input, when I can just throw a CD into the radio that will do the same thing. I wish someone here knew how to work the FM synthesis, I looked in the help file, but their wasn't much there that explained the parameters. I may have found another solution, Spetral Lab has a frequency generator in it that does the sweep function also, in 20 second intervals where you can give the start and stop frequency. I'm currently doing it that way, and then editing it together in Sound Forge, but it would be much easier to be able to do this in SF, which I know is possible. I just might record my frequency generator and call it a day, but I can't do a smooth sweep that way either.

As far as the aliasing noise, it's the pumping sound that I'm talking about. For me it starts to happen around 20Khz and get's worse as I increase the frequency. If you zoom in on the waveform you can see a "modulated" frequency which is causing the meters to pump and cause the additional sound, so I don't believe this is hardware related, because I can actually see it, and the sinewave is no longer a smooth even waveform.

Thanks again, I still need some help,
Brian

Subject:could be hardware

Reply by: Rednroll

Date:1/22/2002 6:33:43 PM

Maybe it is hardware related, I just did the same test on my labtop and no such pumping noise at 21K or 22K.....hmmmm?

Subject:RE: could be hardware

Reply by: Chienworks

Date:1/22/2002 8:32:00 PM

Just an observation, but it seems to me that since the sample rate is
44.1KHz, by the time you get above 5Khz or so your wave form isn't
going to be much like a sine wave in any case. There just won't be
enough resolution to accurately reproduce the wave. As you go higher
than that you'll hit all sorts of phasing errors as the frequency moves in
and out of sub-harmonic bands of 44.1KHz. Depending on how well your
playback hardware smoothly reproduces the audio spectrum, this could
lead to the beats you're hearing.

I have noticed that Sound Forge does an excellent job of resampling
down to lower sample rates. It might be worth the experiment to try
creating your file at a much higher bitrate (maybe even 176.4KHz) and
then resample it down to 44.1KHz. You might get smoother results. Then
again you might not. ;-)

We've got some good B&K test equipment at work. I'll see if i can talk
one of the engineers into setting up an oscillator to do the 2 minute
10 -- 22K sweep and i'll record it into Sound Forge. If it comes out well
i'll upload it to my server for you.

Subject:RE: Frequency Sweep? & Aliasing? in Synthesis

Reply by: herschel

Date:1/23/2002 9:42:55 AM

If you got a few hours to kill you can do all the synthesis you like with a program called CSound. It can be got from ftp://ftp.maths.bath.ac.uk/pub/dream/newest/
called 'csound_win.zip'. There are lots of tutorials and online manuals available on the web. Starts off complicated but once you get the basics it should be ideal for your needs.

Subject:RE: could be hardware

Reply by: Rednroll

Date:1/23/2002 12:41:21 PM

Thanks for the offer, but don't go through so much trouble for me. Actually I have a B&K signal generator sitting next to me right now at work and can set it up and do exactly what you're going to ask your engineer to do. I was just hoping for an easier way to do it in Sound Forge that gave me a linear sweep time. BTW, with a 44.1Khz sampling rate I should be able to accurately reproduce sinewaves of up to 22,050 Hz (Nyquist Thereom). Although my hardware (ie the soundcard) seems to be the trouble of the aliasing, This should not mater if I produce the sinewave in Soundforge and then burn it to a CDR.

Thanks for all the help and suggestions,
Brian Franz

Subject:RE: could be hardware

Reply by: Chienworks

Date:1/23/2002 1:14:55 PM

Well, the Nyquist Theorem states that you can reproduce frequencies at
up to half the sample rate. But it doesn't mean that you'll accurately
reproduce the waveform's shape. A 22,050Hz wave at a 44,100Hz
sample rate will be a square wave at best. A 22,000Hz wave will end up
as a square wave modulated from 0 to full volume by a 50Hz sine wave
(think about where the peaks will hit the sampling position).

When dealing with typical musical sounds, this isn't much of an issue
because the wave forms tend to be somewhat random in the upper
frequencies, and the upper frequencies tend to be a very small portion
of the total volume anyway. But with a pure sinewave, these problems
become quite apparent.

Subject:RE: oconned response

Reply by: MyST

Date:1/24/2002 9:10:53 AM

Here we have a perfect example of expressing one's thoughts before taking the time to completely think things through to make sure they make sense.
Or, as I like to call it... Premature ejaculation of the brain!

Later.

Subject:RE: CSound

Reply by: johannbad

Date:1/24/2002 6:56:08 PM

oh my god... haven't even thought about CSound since college... that's what i started doing computer music with. not a synth, not midi, stoopid CSound and CommonLisp. haha... sorry to be off topic.. but wow that's one messed up program (altho you can't match the level of synthesis you can get out of it, very powerfull indeed)