Somebody needs to get this kid into lasers

[buffo]

Carmangary;

If you have a sampling rate of 48Khz, you need a minimum of two points to recreate a 24Khz signal. One sample will be the high point of the wave, and the other will be the low point. Of course, you'll end up with tonal distortion in the resulting audio, since if you've only got two points you don't have a sine wave, you have a triangle wave. This is actually an extreme form of aliasing distortion, one of the two primary sources of distortion inherent in digital audio. (The other source being sample error or digitization distortion.)

As a general rule, you need a minimum of 5 samples to recreate a sine wave with enough resolution that it will sound acceptable to the human ear. Now, most high end sound cards support a sampling rate of 96Khz, so if you divide that rate by 5 you get 19.2 Khz, which is roughly 20Khz. Thus, in audio circles, 20Khz output is considered to be the maximum for most of today's sound cards. (This also happens to be the upper frequency limit for human hearing - barring those select few that can hear ultrasonics. There are also some specialized cases - children and pregnant women - where the hearing range can also exceed the 20Khz limit.)

Now, having said all that, you need to realize that 20Khz is *NOT* the same as 20Kpps. Not even close. In reality, a signal of 20Khz would be the equivalent of 240Kpps! When we talk about signal bandwidth (measured in Khz), we're talking about the maximum frequency signal that we can accurately reproduce to 70% of original size (or greater) using the galvos. Now, remember that the circle in the ILDA test pattern is the point where the scanners are ballistic; they're accelerating as fast as they possibly can, because they're trying to trace out a twelve-sided polygon that lies outside the square, but in actuality we see it as a smooth circle inside the square. The circle is smaller than the polygon because the scanners never get to the current corner point outside the square before the next corner on the polygon is sent. They're always behind... (They're ballistic.)

Now, to draw a circle takes one complete sine wave on each galvo. (We'll ignore the 90 degree phase shift for a moment.) Thus, you've got 12 points being sent to the galvos to make that one wave. One wave per second = 1 Hz. So if you're sending at 30Kpps, you divide by the number of points in the circle (12) and you get a small signal bandwidth of 2.5Khz. Make sense? And note that this is the *small* signal bandwidth. Larger steps (say, full deflection from right to left) will have a lower bandwidth.

If you're still confused, may I refer you to this thread for more details on small signal bandwidth, with my thanks to Bill Benner for originally explaining it to us. (The meat of the discussion starts around reply # 30 in the thread.) But the bottom line is that no matter how great your sound card is, there's no way you can effectively use more than 2.5Khz of bandwidth - assuming you have your scanners tuned to 30Kpps.

As to the difference between the output of a laser show controller (DAC) vs the output of a sound card when viewed on an oscilloscope, remember that the controller is trying to move the galvos to specific points. This creates step-changes that at times can look like square waves. But square waves are horrible for audio. This is why sound cards have filters on the output to smooth out step-changes in voltage. It helps minimize the effects of the aliasing distortion that is inherent in digital music. But those filters also change the waveform. That's good if you're trying to recreate an analog wave of a piece of music (which will be rich with sine waves), but not so good when you're trying to re-create a vector graphic drawing that may well need those square wave jumps to accurately draw the image.

And finally, about your comment that oscilloscopes being orders of magnitude faster than galvos - yeah, that was my point above. The guy that did the demo on the oscilloscope was probably feeding the 'scope the maximum bandwidth his sound card could output so he could get more points on the screen at once. But if he gets into laser-projected vector art, he's going to have to live with a lot fewer points because of the limited bandwidth of the galvos.

Adam

[The_Doctor]

Nice! More proof that soundcards make really good DACs.

Very true.
I don't know if 48 KHz always equates to 48 kpps. I thought it would but was told in response that a good laser controller has a variable clock rate. But then again, so does the Layla 24 under WXP with the WDM driver, so I rest my case. It's good for 96 kpps too.

As to the signal being linearly swept between each value, that ought to be good for scanners. As the sound card can still sweep full scale in the time required, the output filter can be considered a slew rate limiter that prevents the galvos from seeing a rate change any faster than the currently selected sample rate on the sound card. Could be very convenient as a way to limit speed of step change for a given galvo, just set the max sample rate equal to the galvo's max point rate. Still got to watch scan angle though.

[buffo]

just set the max sample rate equal to the galvo's max point rate.

No, this isn't right either. If you set your max sample rate at 40Khz and you're running a set of DT-40's tuned to 40Kpps, by your example you'd think everything would be fine. But it's not the sample rate that you're worried about, it's the frequency of the fastest signal you send with those samples.

So depending on the type of signal (in other words, the waveform specifics), you could be under the small signal bandwidth even when your sample rate is at 96Khz (because the waveform you're sending has lots of tiny points with almost no change between them), while in other cases you could be way over the small signal bandwidth with the sample rate set at only 2Khz, because you're sending square waves to the galvo at 1Khz. (Remember that the 2.5Khz limit is for *small* steps. Full deflection is at least an order of magnitude slower.)

Case in point: Bill Benner said that when he tunes up galvos using his equipment at Pangolin, he feeds them square waves at 50 Hz to check for critical damping. 50 Hz... Not 500 Hz, not 5Khz, 50 Hz.

Bottom line is this: The Kpps rating is complicated and confusing. It's really only relevent when you're tuned to the ILDA test pattern at the proper specs. (A given angle, a given speed, and certain features of the picture in the right place.) Worse, the rating makes galvos seem to be a lot faster than they really are.

Adam

[The_Doctor]

When Bill does that I think it's because he's testing the fast response, and a fast repeat rate. The actual frequency of that square probably isn't so important, it's probably chosen as representative of what people might actually ask a scanner to do.

When I said set the sample rate equal to the point rate I didn't mean they were the same, I meant only that by doing this you invoke a sound card's filter at a specific frquency related to the sample rate chosen. That sets a maximum slew rate. It's faster than any signal you'll send to those galvos, but slower than the one-shot full range signal that might theoretically occur and damage a galvo if that slew rate limit was not there. In a good scanner driver this would be irrelevant as it would have its own slew rate limiter, but it's nice insurance, and harmless to whatever you're trying to render.

[carmangary]

Buffo, not sure what you are trying to say. I just know that a 48K sound card means 48K samples/sec and when using that as a DAC it equates to 48Kpps. You average sound card only does 44,000 samples / second and sounds just fine.

[The_Doctor]

I think he means that as you draw a scan you are using a rate of change designed to work with your galvos, or at least with an ILDA tuning so other correctly set galvos will scan right, and that the curves and steps will be well within the capability of a slower actual sample rate for a DAC than 48 or 96 KHz. I found that when using the WideMoves, many basic shapes would be fine at 11,025 KHz, the fast step response is the first thing you notice losing out if you drop the sound card sample rate.

My point was looking at it the other way round. Even if you had a low sample rate, but from a DAC that didn't filter the output like a sound card does, a large step could still exceed a safe slew rate for a galvo. In practise, lowering the rate to as low as will let you draw things properly will bring the filter on a sound card DAC down to a frequency that gives you some extra protection.

[buffo]

Buffo, not sure what you are trying to say. I just know that a 48K sound card means 48K samples/sec and when using that as a DAC it equates to 48Kpps.

No, it doesn't. Read my post above again, or better yet have a look at the thread that I linked to which contains Bill Benners original post on the subject.

Kpps is *NOT* just how many samples per second you are sending. Kpps is how many points per second you are sending that are ALSO within the small signal bandwidth for the scanners you are using. This means that the actual bandwidth will change based on what galvos you are using and how fast you tune them.

You keep trying to equate galvo performance to that of a loudspeaker or an oscilloscope. You can't. Galvo's are an order of magnitude slower than a speaker. (a couple orders of magnitude slower if we're talking about tweeters) Also, the bandwidth limit of a galvo changes depending on the amplitude of the signal you send it. (If you absolutely must have an audio analogy, think of a galvo as a device that has the fequency response of a subwoofer when played loud, but the response of a midrange driver when played soft...) It's a lousy analogy, but that just proves the point that scanners are not the same as loudspeakers.

When you are talking about audio it makes sense to talk about samples per second and maximum frequency response. But those terms do not directly translate to galvo speed or to points per second. Points per second is a derived measurement based on the performance of the scanners when displaying the ILDA test pattern at a given optical angle.

Like I said earlier, a rough rule of thumb is to take your scanner speed (in Kpps) and divide by 12 to get the maximum small signal bandwidth (in Khz) that you can send to the galvos. If you must have an audio analogy, then think of the small signal bandwidth as the highest musical note (pitch) that you can reproduce. Note that it doesn't matter if you send more samples... You can only move the galvos about 2500 times per second at maximum, and this assumes that you're only moving them a tiny bit each way.

Sending full-amplitude square waves (which would result in full deflection of the galvos from stop to stop) will reduce your signal bandwidth to around 200 Hz. That is, you are only going to be able to move the galvos from full left to full right about 200 times per second. And as it is, if you do that you'll get a shitload of ringing in the displayed image.

Now, think about that example for a second. You can be sending these square waves at *ANY* speed from 200 points per second all the way up to 30Kpps, and you *still* won't be able to exceed the 200 Hz bandwidth because you're moving the galvos from full left to full right. Faster scan speeds can't overcome the fundamental limits of the galvos themselves.

This is what the tuning process is all about. You set your software to display points at a given rate, and then you display a standard test pattern that has the points evenly spaced throughout the image. Then you adjust the scanner amps so that the galvos display the image as accurately as possible. This is what Kpps is all about. Tuning a set of scanners so that a given image will look the same on my projector as it does on yours.

You average sound card only does 44,000 samples / second and sounds just fine.

Of course it does. That's because speaker systems have a frequency response that runs all the way up to 20Khz. Galvos don't. You need to throw away the audio analogies, because they don't apply to galvos. There aren't any galvos on the planet that can accurately reproduce a 20Khz signal. That's not the same as saying there are no 20 Kpps galvos.

Adam

[buffo]

I stopped reading after your first sentence because I don't think you get it. When using a soundcard as a DAC you program the WAV file so that each sample equals one point to send to the galvos. If you want 30kpps as your speed then you set up a WAV file with 30k samples/sec.

Yes, I do get it. But I don't think *you* get it. Just because you use the sample rate to set the Kpps speed when using a sound card as a DAC *doesn't* mean that 30Kpps (or 30,000 samples per second) is the same as having 30Khz of bandwidth response from the scanners.

You're keying in on a procedural step when using a sound card as a DAC and assuming that it means something other than it does. I'll say it again: Kpps does *NOT* equal bandwidth. The ratio of bandwidth to Kpps *changes* depending on the amplitude of the signal.

Or, to put it another way - just because you set your .WAV file to output at 30,000 samples per second doesn't mean that you are sending a 30Khz signal, or even a 15Khz signal. The actual maximum freqency that you send will depend on the shape of the waveform. (The spacing of the points, if you will.) And that is crucial; because although you have the *ablility* to send a 15Khz signal (albeit a triangle wave) at that sample rate, your galvos simply cannot display it. More likely that you'll be sending a very slow signal (around 2Khz) with *lots* of points to fill in the gaps.

I originally assumed that you were interested in the physics of what was going on with respect to the relationship between Kpps and absolute scanner speed. If you're not, then we can probably stop right here.

But if you are genuinely interested, then I suggest you read more than the first sentence in my post above, because the rest of the post explains the misconception you still seem to have regarding signal bandwidth and Kpps speed.

Adam

[carmangary]

Originally Posted by carmangary
Buffo, not sure what you are trying to say. I just know that a 48K sound card means 48K samples/sec and when using that as a DAC it equates to 48Kpps.
No, it doesn't. Read my post above again, or better yet have a look at the thread that I linked to which contains Bill Benners original post on the subject.

--------------------------------------
Yes it does. And if you don't understand that all I can say is look at the format of a WAV file and then study the APIs for sending data to a soundcard. I am not saying the scanners will be able to react that fast but that has nothing to do with what is coming out of the DAC(soundcard).

[buffo]

You know, it's consdered polite to at least read a person's entire post before replying... I fully understand the .wav file format and the operation of a sound card. (I did graduate Summa Cum Laude with a BS in computer science, after all.)

My point (which you continue to miss) is that there is more to the speed setting of 30Kpps that just sending 30,000 individual points each second. It's also a TUNING standard that relates to the performance of the galvos.

If it was just a speed setting, then in theory any 30Kpps galvos should be able to accept any waveform I send at them at a sample rate of 30,000 samples per second. Now, ask yourself: Can your galvos accept *any* signal that I send to them at a rate of 30,000 samples per second? No? All right then - now we're getting somewhere.

Adam