Update #31: Delay Details #2

[BrockMcKean]

I think that having a narrow list of "recommended sound cards" would be a poor idea.

I think more likely what we're talking about is a short list of sound cards that do not fit the vast and overwhelming majority of sound cards that are in use today. Matching some widely used standard, like USB is quite easy, now. When USB was brand new different products worked differently with different implmentations of the same "usb standard" at the time. Even now there's no way to guarantee that the device you are interfacing with is going to provide some specific minimum amount of current. It depends on what other devices may or may not be connected and how much current the host can source in the first place.

I don't think there's anything wrong with matching compatibility to 99% of sound cards and providing a simple USB solution for ones that require additional and/or different circuitry. Doing so would require a hardware-software interface where the software would have to detect the sound card, identify that it is a problem card, and send a special signal via audio out (or usb?) that toggles the circuit. What's worse, is if you want 100% compatibility with every sound card in existence, that probably would require more than just two circuits to match it perfectly.

Throw in age, oxidation and a bunch of other factors that no one in the entire industry is going to account for and you have plenty of cases where a sound card will just simply not do the job it's intended to do in this system. Heck, if you plug a laptop into a charger most of them will have white gaussian noise as well, distorting the original sound wave. Filtering that and coupling the noise out is not so simple to do for EVERY sound card and every machine in existence.

[Pete]

Sorry for my absence guys (car problems and a downed hard drive scuppered my recent efforts)

I think the original kickstarter campaign aimed to give everyone a DC coupled sound card, and the moment they moved into an AM world the problems began (FM would have been better as the amplitude wouldn't matter, but more expensive). Yes SNR is a factor but by far the most irritating factor is bandwidth roll off which causes the position of an on laser signal to be different to an off or on laser signal. The other is as peachy requires a constant audio signal if there is any drop in the audio buffer then that causes annoying artifacts (the more I work with python the more it seems like a joke language for getting stuff written quickly without caring about processing speed). Even classifying one soundcard would not give you a similar result on an 'identical one' necessarily. There are current fixes for this in the software however the hysteresis issues (known as memory here) need fixes implementing before the success of this can be truly evaluated. Another issue touched upon here is the USB current causing issues with the sound signal, most people are fixing this with floating supply (a phone charger) it is essentially a ground loop error where return current meant for the USB is returned down the path of lest resistance so will then share the ground return up the audio ground, offsetting the signal and adding AC components in the case of the negative rail switcher and the laser controller.


Having said all this, I've been working on my own "USB petechy" and the boards should arrive in the next week or so. With one USB I can get all the power and comms (including drips) and will hopefully solve all of the above issues. I'll post pictures when they arrive, but there's not much to it, it should be ~£5 to build even in low volumes....then I just need to get it interfaced with the software, any volunteers?!

[User_Defined]

That sounds awesome Pete!

As for the hysteresis/memory issues, maybe it can be tuned for in software? Hysteresis usually is predictable, so I'm thinking it would just need to calculate an overshoot based on the anticipated deflection angle of the laser.

Kindof like backlash compensation on CNC systems.

The peachy team would need to get a test setup to collect data on memory vs deflection angle, and then generate a lookup table based on

1) the angle of the axis,

2) the direction that axis is presently moving.

3) the present speed of the axis.

4) the future direction of the axis (acceleration or deceleration)

This is all rather challenging, and I personally would have great difficulty implementing it. Officially it would be called feed forward hysteresis compensation.

I wonder if there is any way to provide inexpensive closed loop feedback from the moving mirrors, because in that case you could get accurate positioning with a basic PID controller.

I do remember that there are magnets on the mirrors... maybe a hall sensor could detect some positioning info? Or maybe if the magnet is moving quickly, a sense coil could get the EMF, send that EMF voltage wave back to the controller, then integrated the wave to provide the positional data (plus a constant) of the system based on Faraday's Law.

Important to note, that this EMF value can be considered as the "I" in the PID.... it may be possible to get a "P" or proportional feedback in some other way.

These are just ideas, so please correct me if Im wrong in any part.

[Pete]

I'd thought of some kind of feed forward controller, this could work, don't know if a software FIR filter would completely solve the problems, it would be really bad for prints to have any overshoot, especially on square corners. In my intial firmware tests I was getting about 50KSamples/s on both channels so should be enough to do stuff like this. I suppose if you were being crazy mental you could interleave every valid point with the laser on with an off point at one of the extents.....hmmm!

The hysteresis fades to almost nothing with the damping system removed, unfortunately then any external vibrations cause significant wobblies and with the current drive system (when the audio buffer underruns) it is unuseable without the damping....a step reponse in drive leads to significant oscillation. I will try adding a LPF to my driver and remove the damping to see what I get.

Adding DC position feedback would be relatively trivial electronics but the mechanicals are probably going to be too precise to get decent results, also adding weight to the rotor....at that stage there's probably an argument for rigid bearings and more drive....then we've made some proper galvos......we should make these....http://elm-chan.org/works/vlp/report_e.html

[Feign]

Having said all this, I've been working on my own "USB petechy" and the boards should arrive in the next week or so. With one USB I can get all the power and comms (including drips) and will hopefully solve all of the above issues. I'll post pictures when they arrive, but there's not much to it, it should be ~£5 to build even in low volumes....then I just need to get it interfaced with the software, any volunteers?!

I for one am incautiously excited. Optimistic even. Now if only we could get a word on this new damping system they're working on. (How bad is the histeresis on the oiled-tab system anyway? I really liked the simplicity of it when I first saw it.)

[User_Defined]

Megneto resistive sensor:

http://www.digikey.ca/product-highli...r-Series/52414


Taken from Wikipedia:

The AMR effect is used in a wide array of sensors for measurement of Earth's magnetic field (electronic compass), for electric current measuring (by measuring the magnetic field created around the conductor), for traffic detection and for linear position and angle sensing. The biggest AMR sensor manufacturers are Honeywell, NXP Semiconductors, and Sensitec GmbH.


So Im sure you can sense the angle of the galvos with it. With something like this you might even be able to electrically dampen the system with a closed loop angle feedback from the magnetic sensor.

For those familiar with control loop theory:

Youd have your input signal with a voltage proportional to what angle the mirror would be.
Next it would go to your amplifier circuit to drive the mirror.
From the mirror you would then have a positional measurement by the magnetoresistive sensor.
The error is defined by where the mirror actually is, and where you wanted it, and this signal is fed back to combine with the original signal to close the loop and subtract away the error. A system like this should auto dampen.

The physical setup would need to be decided... I would imagine a circuit board re-work to get the sensors close to the mirrors' magnets... Or maybe they sell chasis mountable sensors.

Personally I would have just made two PCBs for each mirror axis, combining their motors, sensors, drivers and tuning. Then they would mate with the snap fits. You could have the exact same PCB design for each mirror PCB.

A jumper connector can connect both boards together. When assembling the PCB's youd just populate the computer connections for one board, and leave the second blank (having the jumper wire connector between these two PCB mirror assemblies).

I hope I am being clear, the goal would be to modularize the design, and use a PCB for electronics, AND the mirror support structure. It might hopefully elliminate

1) some wiring to the coils, as they would be now on board
2) the main PCB, making the overall device smaller
3) make it much easier for standardization and mass production.
4) use less plastic.
5) reduce assembly time, and possible wiring mistakes.

Im sure there are more advantages, the one I am most interested in, being an electrical engineer, is the standardizing for mass production.

BUT, All this depends on the accuracy and usability of these types of sensors.

-UD

[Aztecphoenix]

maybe to move into the future, we need to look into the past.

The old ZX Spectrum computers were designed to use any tape deck that you had available rather than a model specific deck and took in the audio signal and converted it into binary code, it also had the ability to compensate for the differences in tape deck output which means using the same technology for the Peachy wouldn't require he AC coupling and all the other stuff they are having to deal with. Schematics for the circutry are appairently online and as it is more than twenty years old any patents have worn off.


maybe looking into this might help with some of the problems the peachy is having and could solve the sound card problems.


to get a better idea of what I am talking about check out The Ben Heck Show episode he released on friday.

[BrockMcKean]

Youd have your input signal with a voltage proportional to what angle the mirror would be.
Next it would go to your amplifier circuit to drive the mirror.
From the mirror you would then have a positional measurement by the magnetoresistive sensor.
The error is defined by where the mirror actually is, and where you wanted it, and this signal is fed back to combine with the original signal to close the loop and subtract away the error. A system like this should auto dampen.

Yes, it would take care of any distortion in actual mirror position due to noise, or vibrations. However, what about the printing material itself? Just because you dampen the mirrors doesn't mean the resin isn't bouncing around. The resin has to be physically dampened somehow. Also, there are plenty of hall effect sensors that can be used in parallel as you are describing and in series with the coils as well. Not entirely sure what's different or special about that particular sensor from Honeywell other than the low power requirements.

The old ZX Spectrum computers were designed to use any tape deck that you had available rather than a model specific deck and took in the audio signal and converted it into binary code

This is possible, but doesn't really make any sense to do. If you have a sound port on a device that is capable of running the peachy's software, it should have a USB port as well. To go from analog to digital and back to analog you're going to need some sort of "processor". Which means that you honestly might as well just use USB. The problems that arise from different sound cards range from different noise levels to different output levels. Analog is precise, meaning the signal must be exact. Digital allows for a lot of wiggle room. There's no way to build a circuit that will always be compatible with every sound card. Plugging a laptop in will usually produce some noise on the line, resulting in a distortion in AM. Some sound cards that are on-board desktop cards have a significant amount of noise on them as well. If you own a phone that wasn't taken care of as well as it probably should have been it will likely not produce sound without some minor distortion that may be inaudible, but will affect AM.

[User_Defined]

Yes, it would take care of any distortion in actual mirror position due to noise, or vibrations. However, what about the printing material itself? Just because you dampen the mirrors doesn't mean the resin isn't bouncing around. The resin has to be physically dampened somehow. Also, there are plenty of hall effect sensors that can be used in parallel as you are describing and in series with the coils as well. Not entirely sure what's different or special about that particular sensor from Honeywell other than the low power requirements.

Adding a closed loop would make the system more reliable and reproducible, along with providing a memoryless dampening of the mirrors (something they claimed to have problems with). The resin bath can be dampened if people want, but it is of less concern than the mirrors. These are two different problems that must be addressed separately.

I linked that sensor as an example of something which can be used with the strength of magnets they chose for their design. Also, it is small, inexpensive, and comes in a standard surface mount package. Basically, I linked something that I felt might work, as opposed to a larger less sensitive model.

- UD

[User_Defined]

Sorry for the double post!

I had a second look at the sensor I linked, and it was not the right type of sensor! The one above is a simple on/off sensor design to replace reed switches in low power applications.

What I meant to link was something like this:
http://www51.honeywell.com/aero/comm...C1501-1512.pdf

This device has angle resolution less than 0.05deg over 90 degrees operating range. However it does have a hysteresis error of 0.017deg (but over the entire range that is less than .02%).
Something like this could be easily used to implement an analog PID positional controller to ensure accuracy in the galvos and dampen the drive.



-UD