DT 40 Pro set or......

[jian725]

Hi Bill,

Hi Jian,

What all of the big scanner companies do these days is quite easy -- don't turn it down!! Just come straight out of the bearing journal with the full shaft size, unmodified. No extra machining steps, no extra tolerance, no wasted parts, no additional cost.

For me also,that is easy!!! Don't need additional any cost.
Small ship easy to turn around (Chinese idiom)

BUT, the stiffness is increased by (2 / 1.95) ^ 4 = 11% !! THAT is my point .

Oh,I real don't know that proportion.Now I know how to do it.
Thanks
Jian

[JoJo]

Bill,

I think Jian turns down the shaft, because a shaft will never fit in the bearing when at exactly 2mm. Bearings normally have an inner tolerance of H6. If you use a shaft of H7, for example, you could just move the bearing along the shaft by using very high power. This can damage the bearing.
If you use g6 for the shaft (means about a few 1/100mm less diameter), you can move the bearing to its final mounting position.
The part where the bearing is located, is thicker, so the bearing can be pressed on the shaft.
I am quite shure that also CT and others do this.

About your tests:
Quite interesting! It is the same result, I also had. I also slowered down the scanners a bit to get a nicer looking result of the image.
I also did some tests with notch filters.
The tuning of the filter is not that difficult. The filter has a center frequency of about 10KHz and a bandwidth of about 10KHz. So it starts limiting the signal from 5KHz. There "frequency" adjustment of the filter nearly has no effect. You can replace it by a fixed value. Because of the large bandwidth, it makes no big change if the center frequency is at 9500 or 10500 Hz.
The gain adjustment is the only important one. There is only a small tuning area where the thing is stable.
Naturally, when increasing filter gain, all other parameters of the driver are changed, because of the increasing delay (phase) of the signal.
For this reason, the filter works properly in a small tuning gap only.
The driver first should be tuned without filter to get the basic settings.
Then the filter is added and tuned to the optimal value.
It works.
But as you told: It is not necessary to use the filter in any case.
If there is no mirror resonance at all, it is better to go without the filter.

Joachim

[Pangolin]

I think Jian turns down the shaft, because a shaft will never fit in the bearing when at exactly 2mm.

Nope, that's not the case. Or at least, that's not my point... Jian (and a few other Chinese manufacturers) come through the bearing with a full-size shaft, and then turn down the shaft after they are through the bearing. Cambridge did that on the original (circa 1992) 6800, but then later learned that it is smarter to just come out of the bearing and leave the shaft alone.

There is an intuition that smaller shafts are better because they are "lighter". There is also an intuition that thinner mirrors are better because they are "lighter". Unfortunately, these are instances where intuition leads you down the wrong path... Dynamic stiffness is often far more important than "light"-ness...

I also did some tests with notch filters. The tuning of the filter is not that difficult.

From reading your entire post above, the only thing that I can say is that your own tests with notch filters must be flawed.

A few points:

First, it's not difficult to tune "wrong"... I will post a few pictures of the right way and wrong way to tune a notch filter. The wrong way is to adjust notch depth for maximum effectiveness. This can be done "not that difficult" with a scope. But unfortunately this will be wrong in most cases...

A proper notch filter for a servo system has both frequency and notch depth adjustments. The notch depth adjustment is put there to match the amplitude of the resonance. If the notch frequency and depth are set correctly, then there is no "bump" in servo response, both in terms of phase and gain. As far as I have been able to figure out for myself, a notch filter can only be adjusted *properly* with a dynamic signal analyzer.

Second, in your example above, the notch does very little at 5kHz. The "bandwidth" you speak about is the -3dB bandwidth. BUT, if the resonance is 60dB, then -3dB doesn't buy you much...

Also, you are considering "bandwidth" above, but really you should be considering the "Q" of the filter... (similar, but different way of looking at it) The Q is set relatively low (typically between 1 and 4) to avoid phase loss inside the servo loop. Also, ideally the Q is set to roughly match the resonance properties of the rotor/mirror system, including both frequency and Q. If the notch frequency was off by the amount you suggest above, it would be, essentially, ineffective. Again, I will post a few pictures to illustrate these facts.

Third, Cambridge wouldn't put both a frequency and notch depth adjustment on a notch filter module, if both were not absolutely necessary to tune. So with this being the case, intuition should lead you to believe that smart people would not put adjustable controls where they are not needed. Cambridge are smart people... (and this is one case where intuition WOULD lead you down the right path...)

But as you told: It is not necessary to use the filter in any case.
If there is no mirror resonance at all, it is better to go without the filter.

Well, there is always a mirror resonance. Most often, this is a longitudinal resonance which is well within the bandwidth of the system. In addition to this, there is always an anti-resonance which occurs at a frequency related to the shaft thickness and mirror inertia, and a resonance which is related to the shaft thickness, mirror inertia and rotor inertia (which is why it is important to make the shaft as thick as practical). There may also be other resonances related to the position sensor and other rotor components. If you look closely at the plots in the document I posted earlier (http://www.pangolin.com/DT40Testing.pdf), you will see that this scanning system has an anti-resonance, then a resonance, then another resonance, then another anti-resonance. These resonances are ALWAYS there. The question is, will they erode gain margin yes or no? For scanners that were designed to operate at 30K, and then tuned to 30K, the amplitude of these resonances will not erode gain margin, so you don't need a notch filter.

But that's one of my key points! Some people try take a scanner that was designed to go 30K (6800 and its derivatives) and make them go faster than 30K. This is never a good idea for a variety of reasons. As proof, I fall back on the fact that all people at FLEM reported BETTER looking images after their scanners were re-tuned (down) to 30K.

Best regards,

William Benner

[steve-o]

This is an interesting conversation--I have a semi-related question for you experts. What does 30K equivalate to in Hz? That is; if you put a square wave into the amplifier to obtain full angular deflection, what is the maximum frequency that the galvos would be able to handle before catastrophic damage (mirror or mechanical)
From your oscilloscope pictures it looks to be around 5 KHz.

[The_Doctor]

A square wave, though used to form the inner square on the ILDA test pattern, is probably not the best wave to try to equate to a frequency, because it involves very sharp transients. For scanning, only the lines that are blanked while reaching new draws will be traced with a transient that sharp. I'm not sure what regular wave is best for testing to a maximum frequncy to equate to a point rate in some way, but I think the triangle might be it, because that way you can test linearity and overshoot. Sine waves might be useful, but not as useful as triangles.

Edit: Even with triangle waves, you can get steeper sides if you add dwell time at each peak, making trapezoid waves. What this means that maximum safe scan frequency of a periodic wave is very dependent on the waveshape, which is why specifying in Hz is not as practical as it sounds. It took me some time (and attempts to program scan patterns) to understand that. I also saw that it depends on the scanners. WideMoves have a smaller variation in step speed per unit angle that most other types, so the dwell time could be made constant for most angular changes of drawn lines, only changing for speed or overall scan angle. Scanners with more step speed change per angle change would complicate any attempt to convert point rate to a frequency.

[steve-o]

Well, let's just say that you wanted to put a wave into a scanner to see how fast it'll go before coming apart. Any ideas?

[The_Doctor]

Start slow. Keep scan angle constant, and speed it up gently. If you have a synthesizer or wave generator, either will probably do it. Look for small errors growing in the scan shape. Just take care, cos if you use a sine, you can go a LOT faster than with other waves, so if you change the wave but forget to reset slow, you might really damage something. Listen closely to the scanners too. Their sound can be a cue to upper thresholds for safety, but only as a secondary guide, and only if you can hear them...

More.. If you have Sound Forge or Cool Edit or GoldWave, draw a trapezoid wave (four steps, slope up, hold, slope down, hold, each step equal duration) at around 10 Hz in one channel, do same in the other channel, 90 degrees phase shifted. This will scan a square. Duplicate both waveshapes to form a long continuation to about three minutes worth, then put an upward pitch bend on the whole thing to around 5 KHz (up 9 octaves from 10 Hz). Play it and pause the wave the moment you see anything start to warp, then look at the editor to work out the frequency. This test won't give you a simple useable answer, but once you try it with a few waveshapes it will give you a strong sense of what scanners do with different inputs. It's also cool what happens if you put a very small detune between each channel before or after the pitch bend.

[steve-o]

Mis -communication i think. I'm not testing scanners, I'm just curious. (I'm fooling around with very small magnets and coils in different configurations) Just wondering how fast those things go (in Hz instead of points.)

[The_Doctor]

You just said you were testing scanners, to see how fast they could go before something fell apart. Not sure how else I can interpret a question like that. I answered all I can. It depends on lots of stuff. I described something you can do. If you try it, you'll soon figure out why there's no simple answer, no matter mow much one might be wanted.

[steve-o]

I understand, perhaps I wasn't clear. Just as a motor has a max RPM, I thought scanners on a set single input wave would have a maximum scan back-and-forth rate. I guess not. Apples and oranges I suppose.