Wavy-line distortion (resonance?) with DT-40pro scanners at 18Kpps speed...

Hi Everyone!

For those of you who were at FLEM on June 22nd & 23rd, you probably saw David and I messing around with a set of DT-40 Pro scanners that would produce some odd wavy lines at a specific range of scan speeds. I've been trying to chase down the cause of the problem ever since. I reached out to Bill Benner a couple days ago, because I knew that about 10 years ago or so he had done a bunch of work on the DT-40 Pro scanner amp in an attempt to improve their performance. I also knew that for a while the DT-40 pro scanners were using the new design that Bill recommended, but the scanners we were experimenting with in Florida have the legacy scanner amp design. (It seems that they only used the new design for a short while before going back to the old design, although I have no idea why.)

Anyway, Bill felt that this topic would be a good one to post on the forum so everyone could benefit from the discussion. He asked me to post my initial question on both the Pangolin Scanner-Max forum and here on PhotonLexicon, and he would compose a reply and post it to both places to spur further discussion. Here is the link to my post on the scanner-max forum: https://forums.pangolin.com/threads/...-speeds.13150/

And here is my initial question that I send to Bill:

=====
Hi Bill! We ran into an interesting problem at the recent Florida Laser Enthusiast's Meeting, and I'm hoping you can shed some light on the issue.

We had a set of DT-40 pro scanners set up, and we noticed that at a very narrow range of scan speeds (from about 17,300 points/sec to about 18,500 points/sec) there was a noticeable "wobble" in the scanned lines, as if the scanners were trying to draw a wavy line - like a sine wave. It didn't always appear though. It was most prevalent on the laser media test pattern and the "TOP" pattern, but it showed up on other frames as well, but not nearly as often.

The effect was subtle... The wobble was maybe a half centimeter (peak-to-peak) on a pattern that was a meter and a half wide on the wall. And it was most prominent on long diagonal lines, although some of the long horizontal and vertical lines would also exhibit the effect.

We tried several different controllers (including one of my QM2000s and one of my FB3s), and the problem was consistent across controllers, so we're sure it's related to the scanners. Then we tried different scanner power supplies, and even hooked up a second set of scanners outside of the projector (just sitting on the table with a bare green module pointed at them) and the problem was still evident.

I'm pretty convinced that this is some kind of resonance with the scanner, either in the rotor itself, or more likely in the mirrors. Also, these DT-40 pros had the original style scanner amps that had both low and high frequency damping pots and the notch-filter daughter board. I remember that you did a bunch of work with Jian to improve the design of this amp, and for a while they were using it, but it appears that they've gone back to the old design again.

Anyway, the problem only happens at a narrow range of scan speeds, so it's not really a huge problem, but to eliminate it my thought was to add a drop of epoxy to the back of the mirror. That would both stiffen the mirror and also add a bit of mass to the entire rotor, which should change the resonance...

But then I wondered if it might be better to try adjusting the notch filter first, to see if that would help. However, I've never actually messed with that daughter board, and if I remember correctly, you had said that the purpose of that notch filter was to correct for a problem with the early position sensors, and since no one uses that old position sensor design any more, the notch filter is actually useless.

So before I recommend anything be done to these scanners, I thought it would be best to contact you first and see what your recommendation would be. There are a total of 3 identical sets of these DT-40 pro scanners, and they all exhibit the same wavy-line resonance problem.
=====

And as I already posted on the scanner-max forum, shortly after I sent the above message to Bill I received a picture that shows the effect I've been trying to troubleshoot. I've attached that picture to this message to help illustrate the problem. Note that in this shot it's hard to see the wavy-line effect on the vertical lines, but it is present. It's much more visible on the horizontal and diagonal lines though.

Adam

https://i.imgur.com/R3Yl9FI.jpg

HI Adam,

You raise some very interesting and important questions. To aid in understanding, see the screen shot linked below.

http://www.pangolin.com/_Files/Reson...Discussion.jpg

This particular screen shot is from an analysis we did in October of 2013. That's when we first started using Ansys.

(For those who don't know, Ansys is among the most powerful, most capable, and most expensive analysis packages in the world...)

This screen shot is of a 6800-like scanner we were analyzing, while trying to understand our competitive advantages. The screen shot shows the resonance "modes", and also particularly what is doing the "bending" and at what frequencies. Since the DT40 is a kind of copy of the original Cambridge model 6800, this particular screen shot / analysis should be relevant.

It shows that the lowest frequency is the first bending-mode resonance (frequency of 5017.1 Hz). This first bending mode also shows that it is almost exclusively the mirror itself that does the bending.

It shows the next-highest frequency is the second bending-mode resonance (frequency of 8036.1 Hz). It shows that in this particular mode, the mirror itself doesn't bend much, and it is mostly the bending of the magnet, shafts, and shaft interfaces (i.e the "glue").

It also shows the torsional resonance (which in this case is the eighth-highest resonance) at 24003Hz. It also shows how the mirror and position sensor vane are "twisting" at that frequency.

(By the way, this analysis was done when we first started using the program. Since then we have refined our material properties and done other "calibration" to the computer model. I can tell you from experience that all of these numbers are pretty optimistic for a Cambridge model 6800. Nevertheless, the screen shot and numbers can still be instructive...)

Now, as it relates to "wavy lines", here's the deal. The wavy lines are most certainly caused by resonances in the system. Sometimes it is hard to tell whether the wavy lines are caused by the "bending mode" resonances, or the "torsional" resonances. However, I do believe that the particular picture you are showing there is because of "torsional" resonances.

Torsional resonances would be caused by the mirror -- or the shaft being too thin. You mentioning that this was a DT40 scanner, they probably use a 0.85mm thick mirror (or maybe even 0.8 or 0.75), and definitely use only a 2mm thick shaft.

As I have discussed in my videos, and also in my book, stiffness of a shaft goes as the FOURTH POWER of diameter. So if we compare DT40 2mm shaft to the shaft used on a Cambridge model 6800 (and 6210 and 6215), the comparison is (2mm / 2.38mm) ^ 4 = the DT40 shaft is LESS THAN HALF THE STIFFNESS of a Cambridge model 6800. Since Frequency is a square-law (not linear law), the half stiffness does not mean torsional resonant frequency is half as high, but rather 70% as high, or around 16.8kHz according to this analysis (but as I said, I think these numbers started out being optimistic in the first place).

In a similar vane, longitudinal (i.e. "bending" stiffness of a 2mm shaft -- and 0.85mm thick mirror) are also much lower than the 2.38mm thick shaft and 1mm thick mirror used by Cambridge.

Torsional resonances can be "notched" using a notch filter. But this adds at least one more adjustment to the servo driver (usually two more adjustments). Moreover, to PROPERLY adjust the notch filter, it requires a special piece of test equipment called a Dynamic Signal Analyzer, which surely the DT folks do not have. Therefore even if they happened to have a notch filter on the driver board itself, the likelihood of it being adjusted properly is... well... I'll let you do the math on that one ;-)

If the resonance is torsional in nature, then those wavy lines would be exacerbated when the point output rate of the software EQUALS the torsional resonance of the scanner. So for example, in this particular screen shot, it shows the torsional resonant frequency is around 24K. So if things were really bad in the system, those wavy lines would be most exaggerated if our point output rate is also 24K. You might notice that even small changes above or below that point output rate will get rid of these particular wavy lines.

My guess is that you probably had your point output rate set to 16-18K. Is that right?

By the way, sadly I do not have particular images that would be a good visual aid for my next topic, but you can do some experimentations there and capture images. The image that you will be looking for is just the normal ILDA test pattern. Bring that up and look at the CENTRAL VERTICAL FEATURES (for example, the lines at the top of the pattern used for blanking, lines at the top/middle, and lines at the bottom/middle). While displaying that pattern increase and decrease your point output rate. Of course the "circle in the square" will distort, but for the purposes of this discussion, ignore that. What you're looking for is for those CENTRAL VERTICAL FEATURES to either remain straight (at all point output rates) or all of the sudden become bent / distorted / or wavy. When you see THIS KIND of wavy line, it is because of the First Bending Mode frequency of the system (i.e. the mirror resonance).

I've forgotten what those numbers are for Cambridge model 6800, but for the Cambridge model 6215 with a 1mm-thick 3mm Y-axis mirror, the first bending-mode resonance happens at around 3.75kHz (which is around 45K PPS on the ILDA test pattern) and their first torsional resonance is around 24K. For sure I've seen many Cambridge 6215 where the central vertical features of the ILDA test pattern becomes highly distorted when the ILDA Test Pattern is projected at 45K and above.

For comparison, using the same mirror our Saturn 1 (which has the smallest / "weakest" rotor in the ScannerMAX repertoire) has a first bending-mode resonance above 5kHz (above 60K) and first torsional resonance is around 65kHz (around 2.7 times the industry average). Because of this, we rarely have "wavy line" problems. We also are very careful during mirror mounting to make sure everything remains balanced, in which case the bending-mode resonances won't even be excited.

Anyway, I hope that the screen shot along with this brief explanation is instructive. Questions are welcome.

Best regards,

William Benner

Bill's probably right... but first pull those notch filters... There's a good chance they're amplifying errors.

If you have the cambridge tuning docs and a scope, I've attached some notes I made when they came out so you can verify exactly where the problem sits.

When they cloned the cambridge amps, they chose one that drove one of the bigger galvo sets, an as such used notch filter components for the huge mirror galvos... so unless they changed the components there's no way to tune out the resonant frequencies you'd actually encounter on the DT40s with range available on the stock DT 40filters. You need to jumper the two pins closest to the edge once the notch filter is gone. If you want to try tuning the notch filter, my old notes have the formula you'd need.

I actually really liked their original amp since they're a 99% cambridge clones (even the component placements are dang near identical), you only have to do a little bit of work to turn them into 99.9% clones. And then you just tune to the 67x docs and you're golden. IIRC you need to replace 3 resistors and a handful of caps and bang, you've got a shitty but economical cambridge.

I'm attaching the notes that I made when the DT40s came out, IIRC Mr Benner had a similar one that he made that he might still have around. My notes just bring the amp to cambridge-like specs and Bill's removes some party-fouls and also simplifies tuning by removing the damping. (Of course, personally, I think the amp could use another half dozen pots for more tunability)

P.S. If you look at company name on the docs, we sold "Photon Dream" maaany moons ago, so thats not me and crew anymore.

Attachment 54282

Adam, I have Cambridge's old procedure for finding the first shaft resonance and designing/tuning the notch filter components without a Dynamic Analysis machine. I won't post it, on-line, but PM me with an email and I'll send you a copy.
~
It is one of those procedures I do not post because if done by the "wrong" immature person with the wrong equipment, then posted on line in a FAQ or Youtube video as "Gospel" , it could do a lot of damage. Especially as each notch is a bit different from galvo to galvo, model by model, and mirror by mirror, and in some cases from production run to production run... Cookbook designs for high Q (Quality Factor) notch filters are not a good thing.. While doing the test procedure manually, it is really easy to miss the correct resonance while watching the oscilloscope and twiddling the knobs on the signal generator.
~
Cambridge stresses that a notch filter is the last thing you touch while working on a galvo amp, nand really stresses the point about calling the factory before doing so.
~
There are two procedures I wont post, and one of them, position sensor calibration, requires you to know a constant from the factory a priori, and the second is the notch.
~
As the clone design will never have the same photodiodes or vanes in the position sensor used by the mother factory, using the published Cambridge numbers while bringing a new, untuned, or repaired board up from scratch can set you up for real bad juju... A bad sensor constant really ripples through the whole analog PID portion of the amplifier.
~
Tuning /finding the notch needs serious bench experience and strong technical skills. It is far easier and safer to use a Dynamic Analyzer . PS, before some one chimes in and says I could just run a "FFT" on a sound card while the galvo is running closed loop, well try it some time. HA! Good luck reading that picture!
~

Usually on a on clone amp notch board there are five pins. +15V in, -15V in, Ground, Signal in, Signal out. Each brand is different, and you HAVE to find the pins by inspection. You can find the power and ground leads with an Ohm Meter, measuring from the 7815 and 7915 output pins. Ignore them. The remaining two pins are notch in/out and in many cases can just be jumped, for testing, but you may have to retune. IN some cases a commercial system Could oscillate if the notch is jumped. REMOVING the notch without installing a jumper will result in a mirror that does not move. Jumping V or V- to the output stage by mistake will fry something, usually the galvo.
~
Keep in mind a notch board may have more then one notch (N=3 some times), and there will be trim pots for notch center and notch depth for each notch.
~
NOT something to play with unless you have a spare scanner system.
~
Steve










.

I've learned that even cambridge got pissed at that notch filter and yanked it for their "laserist" line of scanamps... But I don't think its obsolete, just a PITA... that probably led to more people calling them after turning that knob than anything else...
But I think the common suggestion of "well don't do that then" when you run into resonance is also a PITA ;)

Personally I wish they splashed the medialas catamp IV vs the 678x... There was some magic in that amp that made everything a wee bit nicer looking... but then again, getting any useful technical docs out of Kirk/Dirk (don't remember his name off the top of my head as the last time I talked with him was 20 years ago?) was always like pulling teeth.

There is nothing wrong with having a good notch in the system.

Later found out that there was ONE engineer doing most of the contracted out German designs. It is his policy to never release anything. Dirk may simply not have had the docs to begin with.

Steve

Wow! Great responses folks!

I remember Bill talking about shaft and mirror resonance at a Laser Enthusiast's Meeting years ago. (Don't remember if it was at SELEM or FLEM though.) But the numbers make sense, and the descriptions match what we saw on the wall. I don't have a picture of the ILDA test pattern being displayed at 18K, but we definitely tried that in Florida and I remember those top lines being "bent" at the tips, so it all fits.

Correct me if I'm wrong, but to properly adjust the notch filter you need both an oscilloscope and a signal generator, right? In that case it might be easier to pull the notch filters off (and short pins 1 and 2 on the connector, as Yadda posted) just to see how they look first. Worst case, you could simply avoid the 18K scan speed...

That being said, I'd be willing to try adjusting the filter if it came to that.

Adam

It takes a bit more work then just aligning them with a signal generator. You do have to tune them in circuit and set the depth.
You also have to use the amp's position sensor front end as the probe for the scope.
~
Usually I just jump them out on low end galvos till I'm comfortable. But I do not connect the jumper right away. In fact I install a sacrificial 6810 P or a 6860* and check for oscillation and proper sensor calibration with the jumper out, the output fuse out, or the PTC thermistor unsoldered on one leg. THis is to prevent coil burnout if something is wrong or the amp is biased wrong from the jumper. This also lets me look at servo gain settings. (*Note 6860 coils are tough enough to abuse somewhat.)
~
By definition you are introducing a complex phase shift in the feedback loop if you have the notch installed.
You do have to check the pinouts too. Some of the clones have the connector pin order reversed. My clone amps I use have the jumper on 4&5.
~
To really tune a galvo amp from scratch needs some tooling, you need to able to measure scan angle as well, with the output stage disconnected.
You also have to check out the position sensor.
~
I'll reiterate. Cambridge used to not give you a scratch tuning book unless they watched you play with their demo station hardware at the conference. They used to bring a test bench and the lead amp engineer with them. It was a nice perk for attending.
~
It was, and should be, an instructor lead course when working on a 2000$ galvo pair. Maybe not so much on a 100$ pair. BUT no one publishes the position sensor constant on the 100$ pairs, which makes for a problem.
~
Just knowing how to set up the sensor and set the nonlinearity compensation pot is NOT enough to use any amp with any galvo. It really tires me when I hear that. You have to be able to understand what the PII or PID loop does, and how its transfer function changes the tuning. If you don't know that, messing with the notch(s) is risky. Again, another reason for the ILDA pattern to exist. Three or Four pots to tune is much, much easier t understand then seven or nine. Let the manufacturer set up the difficult stuff.
~
Also you need a scope. Using a multi-meter for initial tuning is going to result in magic smoke release.
~
Again, Adam, I'm not worried about you trying this. As for the average user, the amp should leave the factory properly set up. In a beautiful world, DT and PT would tell you about ECOs, Versions, and Revisions, Matching Serial Numbers, and test for Resonances on each pair. But they don't. I wonder I they even bother with internal "Acceptance Test Reports" on each pair.


Steve

This is a long post...

TLDR; You can tune at home, but you're often better off leaving this buried under the rug and just "don't do that then...."

FYI, cambridge had open loop tuning procedures for when you ghetto-rig a set of unknown mirrors to their galvos to get you "close enough" before doing the normal closed loop tuning. This is what I did when I played with my DT40s. At the time
all of the settings were configured for single-ended signalling so the ISC (Input scale correction) was way off for differential signalling, and unfortunately Cambridge considers ISC as way more than just a volume knob, it sets the basis for the rest of the tuning so even a slight change in ISC requires complete retuning from scratch.

I'll cover 2 ways to set the notch filter... first is what I do in the field with a minimum of instrumentation and second what you could hypothetically do without any fancy tools. If you have real equipment, then you probably know more than I do, so these instructions aren't for you ;)

1st method)

Required tools: You'll need a frequency generator, an dual channel oscilloscope and a frequency counter. (And that handy pdf I gave out a few posts above)...

The frequency counter is optional but very handy as most frequency generators aren't accurate in actual output. In reality I don't use the counter because my generator keeps the last setting and I already have all the bias values etc marked out in sharpie, and that's good enough for me.

You're going to need to discover the first resonant frequency...
To do this hook the frequency counter to TP7 (Notch filter output) and use TP2 as ground, remove the notch filter module (J6) so the servo loop is broken.
Hook a channel of the oscilloscope to TP3 (monitor current)
Hook another probe to TP1 (position signal)

Hook up the frequency generator to the signal input, set output to be <200mV pk-pk. You want just enough voltage to clearly see the signal so keep the voltage as low as possible. If the voltage is too high, you risk burning out the coils,
especially as we increase the frequency. Increase the voltage if you have to, but realize you're on borrowed time.

If waveform from TP1 isn't centered around zero, you'll want to adjust your frequency generator's DC offset/bias to set this voltage as close to zero as possible.

Start the frequency generator at around 1kHz (an arbitrarily low number for the lulz) and slowly sweep up... Watch the current TP3 and position TP1 for peaking (you'll probably hear it on the scanners before you see it on the scope)
As soon as you find that peak, turn the frequency generator off ASAP! Then pat yourself on the back as you've successfully found the first resonant frequency.

My frequency generator remembers the last frequency so I just hook the generator and counter up outside of the scanamp to pull the actual numbers.

Ok now to set the notch filters, on a real cambridge, you'd pick a notch filter that has your resonance frequency near the center of its range. On the DT40s you'll have to calculate the resistor values and replace the resistors on that daughterboard.
(The formula is listed on that pdf)

Last step is to set the notch filter. I short the base of Q4 to ground through a 6k resistor (this simulates the MUTE test point on the real 678) This prevents driving the actual galvos while setting the filter.

Hook the frequency generator to TP6 (Summing amp input) and the counter to TP7 (notch filter output).

If your frequency generator is like mine, it keeps the last set frequency... otherwise put it back to your recorded frequency/bias/amplitude.

Now turn the notch frequency pot until you minimize the waveform on TP7. Then tweak depth to further minimize the output waveform.

2nd method)

Required tools: A PC with sound card line-in and line-out

YMMV with this method as I've never tried it but its completely plausible at least... If I didn't have an oscilloscope, and I had a bug up my ** about needing to tune the notch filter, you should also be able tune using soundcard and PC...
The AC coupling shouldn't affect the results and actually saves us a biasing step.

The basic concept is to replace the signal generator with Line Out and replace the waveform monitoring with Line In... I would skip TP3 (current monitor) as voltage is out of range (10V pk-pk) and its just a visual to compare to the position signal (TP1).
The position signal should return 100mV pk-pk which is definitely audible if low. In a pinch you could use a voltage bridge or opamp to reduce gain (10:1) to visualize TP3, but I don't think its strictly necessary.
The voltage of line out should be approx 1v pk-pk so that's close enough to work as a signal generator.

So in the end, you'll want to hunt down a tone generator and some kind of pc soundcard oscilloscope app... and follow the steps in the first method....

Again this method is hypothetical, attempt it at your own risk.

Also if you don't feel like doing the maths, you can just take any publicly available tunable notch filter circuit designed to filter out somewhere between 10k-20k and plug it into the filter header, then tune and it should work. Keep signal path distances low to keep too much phase shift from creeping in... The header conveniently provides +/-15V to drive most opamps.

This DIY route could also let you increase the performance of the DT40s as you can just make the worlds fanciest multi-stage notch filter and smooth out all the bad resonances in the entire scan range... While this isn't a magic bullet as it'll smooth out frames that exhibit the resonant frequencies, it'll keep the galvos from detonating while you show those frames at least.

P.S. You will always have to retune the scanamps after you change any notch filter settings.... The filter inherently changes the "loop" of the closed loop so if it doesn't change you probably did something wrong :)

P.P.S. You will also probably want to run this from a laptop on batteries or run the signal input as single ended for the first half of the tuning (shouldn't affect the resonant frequency as thats mechanical) for safety...

Lets add the frontispiece warnings from the book chapter to that shall we? I've reworded it as I have no way to OCR the original and tweaked the warning for a clone amp vs a Cambridge.


Warning One. Driving TP7 is directly driving the output stage of the scan amp as a trans-impedance amplifier. As the clone amps do not have a working coil temperature calculator and protect circuit, if you over drive the input while measuring the resonance, you will pop a galvo coil. Much above 200 mV peak to peak of input, you can cook things. If you apply too much DC offset, you will pop a galvo coil before the PTC varistor can trip. You may not totally pop the galvo coil, but you can easily overheat it to the point that it expands, rubs against the rotor, and gets some shorted turns. The coil to rotor distance is a few thousandths of an inch in older designs.


Warning two, if you run the scanner open loop like that for any length of time at or near torsional resonance, you run the risk of frying your bearings. Time here can be measured in seconds if your galvo does not have ceramic bearings. If you have ceramic bearings and overdo it, you have perhaps a few tens of seconds to make your measurement.


And here is the kicker.
You have to retune the amp from scratch following insertion of the notch filter. Otherwise the old tuning will effectively cancel out the notch, period.

Scratch means backing off both HF and LF damping pots, and servo gain to near zero,



------------------------------------------ My comments---------------------------

I have a few Cambridge Galvos I purchased from a Carolina grab bag that I suspect were ran at resonance for a long time at high powerduring tuning. They have good bearings but are trashed electrically. Coil resistance is all over the place from the rubbing damage. Bill's lecture at Selem stressed avoiding overheating for this reason as well.

--------
Cambridge's industrial customers are taught a "clearing move" that redistributes the bearing balls to avoid wear. Pangolin's newer products do this "move" when the system on.
If you have stressed your bearings while tuning, let the Pangolin system do its thing for while.

---------------------

When making the Redboard series of modified Amps for the 506, I found that some clone amps have some series resistor value changes made to change where the potentiometers are setting when tuned. There are a few fried boards setting in a box on my bench, which taught a lesson. Zeroing the three main tuning potentiometers on some clones results in the squeal of amplifier death at power-on. This is why I'm weary of posting the procedures. I finally mapped out the pot settings of a "good" board using an Ohm Meter and I preset the clones before retuning them.

--------------------

Designing a DC accurate narrow notch is not a cookbook exercise. So see attached for the starting point.
~---------------------
Cambridge stock notches (6744-xx) have freqs of

dash -06 12.5 to 20 khz.
dash-05 7-14 Khz
dash-04 5-10 Khz
dash-03 3.7 to 7.5 khz
dash-02 2.2 to 4.5 Khz
dash -01 1.6 to 3.4 Khz.

Source, section 6.2 of factory manual


-------------------------------------------------------

Why Bill uses a Motion Analysis system, and why a sound card will be tricky to read for this process.

While the sound card FFT idea works in theory, a swept sine is the wrong waveform to run the test with. You get really interesting results from mixing products due to how the sampling works in the PC sound card and FFT display. Most sound card based software does not implement an "equal time", swept scan process like a motion analyzer does and thus you can get some really interesting false displays unless you digitize both the excitation wave and the signal from the galvo amp and do some math on them. In fact the better MA will offer the option to use correlated noise as a test signal instead of the sines wave as the better option for identifying the resonance, It will run the correlation and find the peaks, thus telling you what you want to know with no guessing or errors from harmonic products. I know, I tried the sound card thing, and it takes time to learn to read it.

I just bought a cheap HP 35660A MAU on Ebay today to get around the problems of just using a FFT. Should be interesting to see the difference in a 20K and 30K clone.

Steve

No disagreement here. :)

I figured, worst case, letting out the magic smoke on some phoenix dragon tiger galvos is much less painful than the original cambridges... My first set of cambridges cost over $3k! (though it included a 3rd galvo set for blanking)... but at the time I felt everything about them was nice since I no longer had to deal with the stupid torsion bars (yuck)... The fragility of GSIs made me want to guard the stupid scanheads with a ruler and smack every finger that went near the unpowered galvos and shout "No ger-finger-poken!" in a bad german accent. What is with people's fingers and attraction to moving mirrors? I figure there's a phd paper in there somewhere...

IMHO as long as you're sane and follow the instructions exactly, you won't put too much wear on the galvos... In fact the DT-40's have a 3rd hidden protection mechanism... the provided power supplies will let out the magic smoke long before you achieve a full aggressive tune on the scanheads... thus saving you from having to replace the fuse. The newer DT drivers implemented temperature stuff, but I always felt it was a bandaid... if your galvo gets too hot to touch you have bigger problems.
IIRC Cambridge's max rated coil temperature was hotter than boiling water! So my Mk.IV fingers made for a great temp guage...

In all seriousness, If the previous post wasn't clear, the scanhead will audibly complain when it hits the resonant frequency... that point is right when you *really* want to stop scanning... and setting the notch filter should be done with the scanner muted...

Holy shit! Thanks for all the info guys! :)

I had no idea this process was so involved, or so risky. :eek: So basically you drive the scanners open-loop (no feedback) and intentionally put them into resonance, but hopefully with a low enough input signal that they don't run away and destroy themselves before you can kill the signal... Then you mute the output and adjust the notch filter (with the same signal applied) until you get the lowest output...

And this is only to correct the first resonance (which, presumably, is for the entire rotor), correct? So if there is a second resonance in the system (say, of just the mirror, for example), then there's no way to deal with that, is there?

Sheesh! I'm conflicted here. On the one hand, I really want to try this. On the other hand, I'm nervous about blowing up someone else's scanners. (And they are used scanners at that...) And for sure I'd need to buy a good function generator first. (I have a couple dual-channel 'scopes.)

Let me think on this a bit. Maybe it would be something to try at SELEM?

Adam

PS: Yadda mentioned torsion-bar scanners being broken by people playing with them... Interestingly, DZ and I had a discussion about this recently. He's been using G120s with Turbo-Track amps for quite a while now, and the output is flawless (largely due to those awesome amps). But then I asked him what happens when the torsion bar breaks. To find out, he dug out an old set of G120s that had broken torsion bars and hooked them up. Amazingly, they still scanned the ILDA test pattern correctly! I don't remember what the scan angle was (likely less than 8 degrees), but it sure looked good to me... My guess is that at wider scan angles there would be problems though.

PS: Yadda mentioned torsion-bar scanners being broken by people playing with them... Interestingly, DZ and I had a discussion about this recently. He's been using G120s with Turbo-Track amps for quite a while now, and the output is flawless (largely due to those awesome amps). But then I asked him what happens when the torsion bar breaks. To find out, he dug out an old set of G120s that had broken torsion bars and hooked them up. Amazingly, they still scanned the ILDA test pattern correctly! I don't remember what the scan angle was (likely less than 8 degrees), but it sure looked good to me... My guess is that at wider scan angles there would be problems though.[/QUOTE]
~

Yeah, Turbotracks will do that with the G120. Nothing else I'm aware of can.

Torsion bars are nitinol in some cases, and spring steel in others. Galvos used to be so expensive that I would press the G120 shaft out and silver solder or braze a new T-bar in. Repositioning the shaft then takes an hour with the O'scope. Then they redesigned the lower bobbin to make a one piece assembly and that did not work any more.
~
----------------------------------
If you have a proper bench, a good Bench PSU set with programmable current limiting , a scope, a good signal generator, a Pangolin setting there, the manual, and patience, tuning from scratch is not that difficult for a trained ET.
~
Thinking you can just waltz in and twiddle knobs however will result in loosing money. It is like learning to fly a plane. You have to break your pre-conceptions, forget much of what you learned on the PC Flight Simulator, and look at the procedures and instruments. This won't make a good Youtube video either, too easy to loose your place talking and tweaking at the same time.
~
Steve

Yah as easy as that :) "Just watch out for the alligators and you'll be fine... "

If you have a second resonance band, you can either get a cambridge dual-notch filter (6744 series) or home brew your own version. If you have the 678x instruction manual, the schematic should be in the appendix somewhere...

Finding the second resonance is exactly the same as the first... only you start at the previous discovered frequency and sweep up from there.

Just use your ears as a guide and you should stay safe...

Torsion bars (along with tuned weights) were a trick to get higher natural resonant frequencies out of the older galvo technologies... I'd imagine it would work without them, but you'll see the resonance popping up earlier which can kill your overall performance.

P.S. also there was a one-armed bandit's chance the mirror starts up in the wrong place and the whole thing shakes itself to oblivion while trying to stabilize itself... Thats one of the reasons the greek scanners (got branded eyemagics) impressed me... you could clamp
the mirror and shaft live without detonating anything... though I'd imagine you were probably degaussing the *** out of the magnets.

Buffo is ordering the low distortion digital wave gen I suggested. It has the least bells and whistles of anything on the market, but the very nice waveforms and dead simple user interface are worth it.
~
Alligators? More like a dozen pack hunting, hungry Humboldt Squid who are tired of cannibalism.
~
L. Michael Roberts of Laser-FX and Mike Thanos From Cambridge walked me thru all this years ago, Thanks Gents if your reading this. Playing with what was then (still is one of ) the world's fastest analog servomechanisms has served me well in my career.

~
Steve

Lol, don't judge, but I'm still using my "vintage" analog HP generator ever since I got burned by a digital generator I had bought while doing impedance testing many moons ago... I did pick up a fancy looking siglent one a little bit ago but haven't really used it outside of oohing at the pretty graphics on the screen.

Buffo, when you get your new toy, plot a few graphs.
~
Calibrate your test wall with pencil marks (very important tool) for say 5 degrees full angle, 8 degrees, then 10, 15 and 20 degrees.
Hook up your favorite projector.
Ground the Y axis, bias on a laser somehow and bypass the interlocks if installed. Generate a DOT.
Ground the negative side of the x axis on the ILDA connector and hook the sig generator to the positive side.
~

Start at 30 Hz sine wave and adjust the line on the wall for the width of the 8 degree marks.
~
Ramp up the frequency slowly until the line collapses to 50% of its width. Note the frequency. Listen to the scanners for strain.

~
Do the same for each increasing angle.
~
Your measuring the angle specific 3 Db point for your scan pair, and you can plot the response graphs for each starting angle.
~
If your brave switch to square waves. A 30 Hz square wave will teach you much about your damping. Note that square wave performance will be less then 0.7X sine performance.
~
This is a eye opener the first time you do it. Especially when you see the line width go up and then back down some times at modest angles.
Bonus points if you graph the input power while you do this.
~
Don't be like the synthesizer fellow who was on here a few years ago, who fried his galvos pushing the waveforms too high in frequency.. Work in moderation.
~
The ScannerMax folks have a similar plot for each mirror size in the current datasheet for C506. Neat graph to study.

~
Too bad there is no way on Earth to relate this test to KPPS.
Oh wait, the slope of the 3 db points on the right kind of graph.. Hum... Not an exact relationship, but interesting.
~
Steve

The Instek SFG-1003 works fine for me.. Should be good for Buffo.
Old technology now, not that cheap, but is a solid beast that won't die easily.
Not an HP by a long ways.

Steve

Quote:

---

Originally Posted by mixedgas

Don't be like the synthesizer fellow who was on here a few years ago, who fried his galvos pushing the waveforms too high in frequency.. Work in moderation.

Steve

---

Ah... if only I could have collected a dollar for every scanset demolished by people making lissajous... little did they know that bad values of oscillators would treat the galvo mirrors like orbital slingshots...

Quote:

---

Originally Posted by yaddatrance

Ah... if only I could have collected a dollar for every scanset demolished by people making lissajous... little did they know that bad values of oscillators would treat the galvo mirrors like orbital slingshots...

---

You just made me laugh.
~
Ironically he was on here trying to pick our brains trying to make the universal galvo protecting filter block... So he could sell his product as abstract consoles.
Not so easy to instantaneously protect a galvo set and still get good graphics. At least not easy if you do not tap into the feedback signals.
He really did not want to hear about a 1.5 KHz roll off.
~

I could retire early if I had 20$ for every scanset destroyed by ravers on drugs. Some day ask me about repairing scan amps every few weeks for the guy who ALWYS retuned his scanners mid rave on acid.... I finally stopped taking his money, sometimes you need to fire the customer. At least to ease your mind, and for the good of the industry. Ex child actor you would all recognize. He swore up and down that tuning while high made for the best shows ever. That was in the early 2000s...

Steve

To do it right you need one of these:



Bill

I made this video in 2015:
https://youtu.be/Tj69w3QkDDk

You'll note that this is a private (aka unlisted) video.

It was made to compare our system to something developed by another company who had a 10mm scanning system and digital servo driver. From a performance standpoint, our system outperformed theirs in every way, but the good point about their system is that it is self-tuning whereas ours must be tuned manually.

This video is kind of "Part 2". The "Part 1" video showed their system with our test patterns, and showed all of the things that was wrong with their system. I'm not showing Part 1 because it would reveal the name of the company, and it's not a point to publicly call them out...

The purpose of the video was to start a discussion of a possible collaboration. By the way, the discussion went nowhere... Their system is still just as bad as it was back in 2015 although ours has continued to evolve (and we're literally working on self-tuning right now).

Nevertheless, at time 10:20, you'll see the dynamic signal analyzer built into the system and how it works. This is what's really needed to set a notch filter (or two) set correctly. Doing it with a function generator and scope has some follies...

By the way, in the video I mentioned that this servo driver could not drive the C506, but over the past year we've rectified that, with a new PD board for the C506 that is compatible with the Mach-DSP. Some PL folks may even have our C506 with Mach-DSP driver.

Bill

Somewhat offtopic, but related to the c506... I helped some folk here shoehorn a scannermax set with older cambridge scanamps (using your component replacement docs)... Nicely done! Those are some really well designed scanners and the costs were surprisingly less than what the nicer (really nice?) chinese sets go for.

Bill I've been meaning to jump on your forum and ask, whats the max speed I can clock XY2-100 into the Mach-DSP? Is it safe to go over 2Mhz? 20us a point (defined by the spec) is a hair bit slow for what I want.

Quote:

---

Originally Posted by yaddatrance

Bill I've been meaning to jump on your forum and ask, whats the max speed I can clock XY2-100 into the Mach-DSP? Is it safe to go over 2Mhz? 20us a point (defined by the spec) is a hair bit slow for what I want.

---

XY2-100 standard is indeed 2MHz clock rate. It's 20-bits per coordinate. At 2MHz that's a 100kHz sample rate. (That's what the "2" and the "100" mean.) But that's 10uS per coordinate not 20uS.

Theoretically Mach-DSP can accept coordinates with clock rate up to 25MHz, but I've never tested that...

Bill

Quote:

---

Originally Posted by buffo

Sheesh! I'm conflicted here. On the one hand, I really want to try this. On the other hand, I'm nervous about blowing up someone else's scanners.

---

No need to be conflicted, I'll volunteer a set of DT-40's for this. Let's do some destructive testing!

Quote:

---

Originally Posted by araugh

No need to be conflicted, I'll volunteer a set of DT-40's for this. Let's do some destructive testing!

---

Wow! Now that's an attitude I can get behind! :)

After all, what's the worst that could happen, right?

https://i.imgur.com/InbxIxe.gif

This ought to be fun!

Adam

I could not afford the lovely SR785 Scannermax has, But a beat up HP 35660A is winging its way here as we speak. Supposedly it passes its self tests. One of those auctions where the guy thinks it will work, but does not want the hassle Fleabay will put him through if he charges the big bucks and it fails. Not to mention it is boat anchor sized and a Lexel 88 PSU has less mass. :-)
~
So if the 35660 works,
I think we're going to have a teachable moment (maybe an hour) at SELEM if my boss lets me escape. I'll try for Thursday-Friday presence.
~
I'll bring a 6850 Amp, the scope, the arbitrary waveform generator, and some clone boards. The 6850 is the nearly indestructible teaching tool. The clones are bottled magic smoke waiting to happen. Call it the hard core galvo amp setup / scope tuning short course.
~
We can also run 6800 with bad bearings, and 6800 with damaged coil, and 6800 with no mirror. We can also run some 6810s with various mirrors. I have a few Chinese notch boards.
We'll set you up good Adam.
~
Having a DSA laying around is a good thing when you work with mechanisms all day. I can justify it as a personal tool, for what I do at work.
~
No Brad, You do not need your own Dynamic Signal Analysis system at SELEM......Leave the card in the wallet. '-)

~
l'll bring the Turbotrack with 506s on it, too.


Steve

Hi Steve,

I hate to break it to you, but the 35660A does not have "swept sine" measurements. That's one reason why you could find them cheap on eBay. By comparison, the 35670A does have swept sine, as long as it has option 1D2 installed.

There are also older ones, such as 3562 (that's what Cambridge had).

I've supposedly got the software for a form of swept sine in the "Periodic Chirp" setting ????
~
I also have the option for an external sweeper, which I have.
~
Its not going to be perfect then... I read the manual twice before purchasing,. I know it is going to be a bit of a Dog.
Live and Learn.... Should still be a useful tool as a dual SA.

Steve

Wow, this thread rapidly got way above my pay grade! ;)

Quote:

---

Originally Posted by araugh

No need to be conflicted, I'll volunteer a set of DT-40's for this. Let's do some destructive testing!

---

Yeah, I was thinking the path of least resistance for the sets I have exhibiting this behavior was swapping out the mirrors for higher quality 3mm stock mirrors from Pangolin. Otherwise, avoiding 18Kpps scanning would be simple for most people, but I tend to use it for beam alignment, beamshows and glow in the dark art work, so this is how I tripped across the issue.

That said, for those brave enough to attempt it or volunteer their galvos, bravo!

Are we doing this at SELEM? If so, if its successful on the test patient, I'll bring three more sets to adjust.

-David

Quote:

---

Originally Posted by Pangolin

XY2-100 standard is indeed 2MHz clock rate. It's 20-bits per coordinate. At 2MHz that's a 100kHz sample rate. (That's what the "2" and the "100" mean.) But that's 10uS per coordinate not 20uS.

Theoretically Mach-DSP can accept coordinates with clock rate up to 25MHz, but I've never tested that...

Bill

---

Thanks

Yah my memory is never what it never was^H^H^H^H^H^H^Hused to be... I have a customer who has an older device that puts out points at effectively 120kHz (low excursion, low step size) but his analog circuitry is getting noise from somewhere on his circuit and I suggested going digital all the way to the amp but found out that all the digital amps are nominally 100kHz... Was originally just planning to tap his digital signalling and converting, and now it sounds feasible again.

While the 35660A doesn't natively have swept sine, you could get something close on it and other similar units by commanding a sweep through the HPIB/GPIB port. Nowadays a usb to GPIB only costs a few bucks so thats also a route.

You can get the 35660A HPIB programming manual here: https://doc.xdevs.com/doc/HP_Agilent...0Reference.pdf

Hi guys,

By the way, there's no chance of destroying anything. While our Saturn 9 literally have enough torque to break a mirror (and will if the notch filter isn't set correctly), neither the DT40 nor their driver have enough gain to do that. The worst that could happen is that wavy lines get worst. Although I've never tried it, tuning a notch filter "by eye" using a test pattern might not work.

It's like shooting a basketball. The hoop isn't much bigger than the ball, and there's only one placement where the ball goes straight through the hoop. Everywhere else doesn't get you any points... (setting aside it bouncing off the rim a few times then going in, which never happens in electronics)

Bill

Bill, While I have you here, Do you have Co2 mirrors for 506? 3-4 mm beam diameter ?

I imagine everybody wants Saturn for marking, but since I have spare 506s I figured I'd ask.

Steve

HP 35660A Is Here. Both Channels are working just fine. Learning curve for the soft keys is a bit steep, but it will do what I need it to do. Not just with galvos either.

Steve

Instek SFG-1003 showed up on Friday. ("Oooo... Shiny!") So now I have a new toy to play with. :)

We will definitely have a go at adjusting the notch filter at SELEM. I'm also looking forward to demonstrating how a set of G120s will still function even with broken torsion bars. (DZ offered to help out with that!)

Adam

Just remember to pull the attenuator pin on the SFG.