Cutterhead feedback from Hall Sensor

[electronrancher]

I had been playing with some cheap Allegro hall effect sensors when I began considering if they could be used for position feedback in a cutterhead. Has anyone tried using Hall sensors instead of auxiliary windings? The ebay ones I got have 30kHz GBW, so it seems reasonable that they could linearize the cutter response over the audio range. So pre-RIAA your audio input, then just let the cutter feedback do the rest.

[opcode66]

It wouldn't be shielded from the alternating electromagnetic field generated by a drive coil. It would likely be less convenient to mount to the piston of a drive element in a cutterhead than a coil. And, finally, I don't think that you would get the same level of accuracy as you would from a coil. Accuracy is paramount for feedback to work correctly.

[electronrancher]

Good points. I can't answer to the accuracy, as I don't know the field strengths typically inside the head. But since we know the coil drive current, it's field strength can be subtracted from our hall signal to give position with respect to the permanent magnet only. Then the feedback loop just solves the problem of "Regardless of the drag presented by cutting various materials, drive the coil until the bobbin in in position x."

I realize the coil feedback is the tried and true standard, but it seems that it will have complicated AC parameters that would make the result of flat response over frequency difficult. It's essentially a high pass feedback, isn't it?

[opcode66]

Moving coil feedback is identical to a moving coil phono pickup. Dont overthink it. Feedback flattens drive response by attenuating the input signal to account for the physical resonant frequency of the moving parts.

[EpicenterBryan]

I had been playing with some cheap Allegro hall effect sensors when I began considering if they could be used for position feedback in a cutterhead.

The Allegro hall effect sensor will not work for feedback.

The big issue is that a Hall Effect sensor relies on changes in a magnetic field to generate a voltage. If the sensor is mounted anywhere near a driver, it will measure the drive signal via the generated magnetic field in the driver. So assume you can totally isolate the magnetic field (via Mu-Metal or by extreme distance), you still need a way to connect to the actual movement to measure it, and a way to increase distance from the magnetic source of movement or your signal is bogus.

Bryan

[opcode66]

A hall sensor will allow a constant voltage for a constant static magnetic field.

But, as also stated, is affected by alternating fields as well.

Essentially, anything that can be attached to the moving element without disrupting the movement of that element which can generate a very accurate signal representing the movement can be used. To date, the only effective systems I've seen are a coil on the moving element, or AM radio receiver transmitter with a reflector on the moving element.

[electronrancher]

All good points. If the setup were moving magnet, I think it would be much simpler. I'll keep thinking about it - this idea started after building one of those magnetic levitators that it balances coil field with regards to permanent magnet field. The hall sensor measures the sum (difference) in the fields, and the loop drives the coil to match that to an input - vary the input, and the regulated distance changes.

I'll let you know if I end up doing any feasibility tests that show promise.

Also - thanks to both of you from a newbie! Long time lurker, so your contributions here on the board are well known to me, and are much appreciated.

[markrob]

I had been playing with some cheap Allegro hall effect sensors when I began considering if they could be used for position feedback in a cutterhead. Has anyone tried using Hall sensors instead of auxiliary windings? The ebay ones I got have 30kHz GBW, so it seems reasonable that they could linearize the cutter response over the audio range. So pre-RIAA your audio input, then just let the cutter feedback do the rest.

Hi,

This might work, but there are several issues to think about:

1. Noise - These devices are pretty noisy from my experience. Especially as the bandwidth is increased. You need a very large dynamic range and the change in field strength can be quite low at the high frequencies.
2. Velocity vs. position feedback - The Hall sensor will be measuring change in position rather than change in velocity that a moving coil would provide. If you close the loop as a position control, you will have some loop compensation to do as the ideal open loop position response of a moving coil head is second order. You can also convert the loop to velocity if you add a differentiator inside the loop, but this adds more noise to the system.
3. Stray pickup - You have to make sure that you are actually responding to changes in position, not the sensing of the field generated by the drive coil. This gets to be a big problem at high frequencies since the motion of the head is so small.
4. Added mass - Since the Hall sensor must be coupled to the moving elements of the head, you want to make sure the the added mass doesn't affect the performance of the head. As mass is added more drive current is required to reach cutting level (F= MA).

What device are you looking at?

Mark

[electronrancher]

2. Velocity vs. position feedback - The Hall sensor will be measuring change in position rather than change in velocity that a moving coil would provide. If you close the loop as a position control, you will have some loop compensation to do as the ideal open loop position response of a moving coil head is second order. You can also convert the loop to velocity if you add a differentiator inside the loop, but this adds more noise to the system.
3. Stray pickup - You have to make sure that you are actually responding to changes in position, not the sensing of the field generated by the drive coil. This gets to be a big problem at high frequencies since the motion of the head is so small.
4. Added mass - Since the Hall sensor must be coupled to the moving elements of the head, you want to make sure the the added mass doesn't affect the performance of the head. As mass is added more drive current is required to reach cutting level (F= MA).

Thanks Mark. I don't know about noise vs frequency - that's a good point that I should look into. I believe the devices I got were A1305, but I'm out of town and can't find that part number on the web. Here's A1301, which is functionally similar.
http://pdf1.alldatasheet.com/datasheet-pdf/view/120794/ALLEGRO/A1301.html

2) Position - Don't we want to control position? I can certainly see how the loop can be closed using velocity (ramp up the drive until measured velocity is equal to target).. But at the end of the day, we want to take the shape of the audio waveform and put that shape in plastic. Controlling position sounds like a straightforward way to do it.

Question: I'm not familiar with the 2nd order nature of MC. Is it force and inertia/damping/spring response as the terms? If you wouldn't mind giving me some insight here it would be appreciated (not a mechanical guy, derp).

3) Stray Pickup - I think the coil field can be cancelled since we know (or can measure) the drive current to the coil. The coil field should be directly proportional to coil current, so my hope was to just null it out. But I agree - coil field is the variable we will control so it will vary depending on whether the material being cut is hard or soft. It's value is arbitrary and ideally, we don't want it in our signal.

Side note: Isn't there some coupling when a feedback coil is used? Maybe not with some mu-metal in there, but from what I can tell they are often non-isolated magnetically and rather quite close. I'd expect some coupling. Wonder how it works there...?

4) Mass - yup, the mass and the additional wiring would be an added pain in the rear when trying to mount a hall sensor to a coil. Moving magnet would not require the hall sensor to ride on the coil, but that sounds like even more mass to fling around.

All in all - great stuff and enjoying the brainstorming here! Let's keep it up - there are real challenges to it, and it may not even be practical. But it's always better to get the scope of problems ahead of time, so all feedback (har) is appreciated.

[markrob]

Hi,

You need to get up to speed on basic disk recording. The audio waveform is presented to the head post RIAA. The head response is normally assumed to equalized to be flat vs velocity (typical magnetic pickups are also velocity responsive). You'll see cutting reference levels given in terms of recorded velocity (e.g. 5cm/sec RMS at 1Khz). This can be converted to excursion using the basic physics relationships between position/velocity/acceleration.

The theoretical open loop response of a moving coil head is flat vs excursion (position) from DC up to the system resonance (typically in the 1 Khz range). This is the compliance controlled region. Beyond resonance, the response falls at -12/oct making it a second order system. It considered mass controlled in this region. The phase shift approaches -180 degrees as you move past resonance. Closing the loop around this will result in unstable operation (-360 degrees of total phase shift). So you need some sort of loop compensation (lead-lag for example). If you start with a velocity feedback approach, there is an inherent zero in the open loop response (a differentiator). This makes the open loop velocity response rising +6db/oct below resonance (no response at DC since velocity is zero) and falling -6db beyond resonance. The phase shift in this region only -90 degrees. This makes feedback easy since there is plenty of phase margin. The above does not take into account any other destabilizing factors that add extra phase shift (secondary resonances, transport lag, electrical phase shift, etc.). This can make it more difficult to close the loop.

Even before you think about isolation or nulling of the driver fields, it would be good to consider the change in position you need to sense. At the low frequencies, you are talking about something on the order of +/-.010" max with a 50 to 60db dynamic range. If you consider the excursion for the same velocity at 10 Khz and the same dynamic range, you into some really small numbers. In a 1T static magnetic field, How much voltage will the Hall senor generate when the sensor is moved +/- .010"? It gets scary very quickly.

Mark

[electronrancher]

Hi Mark,

Good notes - thank you. And I do need to learn more on mastering, definitely. I've got Read's book here and am comparing your notes to it as we speak.

At least for the case of the coil response, I believe it is a double pole only when driven by a voltage source. It should still act as a single pole response when driven by a current. Actually, this case happens a lot when driving an L or LC tank while controlling voltage - and yes, it's ugly but even that case can be made manageable using type 3 compensation in the loop. (or as you mention, a very low frequency zero). Let me look in to this more, but I expect that if it's current controlled the double pole will be managed.

Last, I did a quick check of gauss per distance for an estimate of a small Nd magnet whose pole is 50 mils from the sensor. The difference in field for the excursion of 10 mils is about 400 gauss, which would give about 600mV total swing from the hall sensor. That seems like plenty. For reference, how much signal do you get from your typical feedback coil?

[markrob]

Hi,

You seem to be on the right track. If you really get 600mv of output with a 10 mil excursion, you may be in good shape. Keep in mind that you will need to have sensing down into the 100 nm range at the lowest levels and upper frequencies. You should be able to quickly determine if the number for output you calculate is correct with a simple bench setup. If you do this, please post the results.

The double pole in the open loop response is strictly due to the mechanical part of the system. Its a spring mass system and second order by nature. The L/R pole of the drive coil due under voltage drive adds a third pole in the open loop response (not a good thing at all). As you point out, going to current drive will knock out this pole. This is easily done by adding a second feedback loop around the drive amp via a small sense resistor. In the Cook patent (US 2,522,567) for his moving iron mono feedback head, he does just that. Not sure if Neumann or Westrex did the same.

BTW, Fairchild used a novel RF eddy current sensing method for their feedback system (which was position controlled). Its described in good detail in a paper found in the AES 2 vol Disc Recording Anthology. If you don't have this, I'd highly recommend getting a copy. You sound like someone that would really benefit from this resource.

Good luck in you experiments.

Mark