electrical engineering questions - cutterhead related

[Bahndahn]

I am looking for some insight from those who are equipped in electrical engineering applications on the following roadblocks I have come to:

Q.1
I need to make a drive coil [electromagnet] for my cutterhead and I only have 28AWG enamelled wire. The amplifier I will use with it requires resistance from 4Ω to 16Ω. I would like to make a coil of 2Ω resistance for minimal added mass given the gauge of wire:

If I put a 2Ω resistor in series with the 2Ω cutterhead coil, will it function properly in terms of frequency/phase/load on the amplifier? If so, what type of resistor should be used [tolerance, wattage, etc.]? Will the series-orientation of the resistor have an effect [pre-coil, post-coil]?

Q.2
I am looking to construct an ultra-simple feedback handling circuit. It does not need to have onboard RIAA. Most simply, an inverting opamp amplifier could allow adjustment of feedback level into an 'mixer' opamp to meet with the pure signal. Professional cutterhead feedback circuits have some other elements processing the feedback signal before integration:

What sort of bare-bare bones considerations should I make in this feedback circuit design? Has anyone experiences or schematics they wish to elaborate on?

Thank you!

[opcode66]

A. 1
For simple coils use this turn calculator or something similar.
http://www.daycounter.com/Calculators/Coil-Physical-Properties-Calculator.phtml

A. 2
Not counting the power supply you are talking about maybe 5 components and a dual opamp for each channel. I can slice out this part of Flo's Caruos schematic and post it here later tonight. Very simple my friend. It is leveraging the inverting input of an opamp in a rather devilishly simple way. I'm working on another project right now. Give me until the morning.

You will also need to source a symmetric (+ and -) power supply. These work very well. One board will power Left and Right channel easily. Especially if all you are doing is negative summing. However, I really think you will want to add the section that boosts the feedback signal. As with a microphone, the signal is very week and needs extra boosting in order to be negatively summed with the input signal. That adds another opamp of a greater gain and higher fidelity as well as a couple components aroudn it (per channel of course). Even with the full Caruso boards, 1 amp symmetric is fine though.
http://www.vellemanusa.com/products/view/?id=522032

k8042.jpg

Paired with one of these you can get solid symmetric 15V DC 1 Amp.
http://www.digikey.com/product-search/en?keywords=vpt24-1040

VPT24-1040_sml.jpg

Use fuses on each rail. Half amp will be fine for each rail in a pre-amp. Use shielded cabling on everything.

I am happy to answer any question you have about feedback electronics or ways to generate feedback. However, and I hope you understand, I can't answer specific question about how I've implemented feedback. Outside of that, I'm happy to share my wisdom and experiential data. I think my latest cuts demonstrate that I have this confidently sorted out and am not pontificating. Ask me here or via pm or my email address. Always happy to help as much as I can.

[dubcutter89]

A1: Yes, you can use a resistor in series to load your amp properly, but you will also loose 50% of your power - so if your amp puts out 50W there will 25W to the resistor and the other 25W to the coil moving the stylus...not very efficient!
The frequency response/phase etc. will be affected, but if your using a moving coil design with low inductance then the effect will be small. No difference where you put the resistor (pre, post). For testing ok, but I would recommend to use other wire or more turns...

A2: Anything that can mix 2 signals without any phase shift, amplitude distortion should work. This can be as simple as 2 resistors or a complex control loop. If you have a small mixer with line in and mic in you can try this. Better would be a stand alone version with some opamps or if you don't mind spending some $ get a caruso pre.

Lukas

[opcode66]

Guess I didn't really read Q 1. I saw ohms and just assumed you were asking about how to determine how many turns would result in a desired value given the coil specs. My bad.

dubcutter89 is correct on both counts. Add a resistor in series is no problem as long as total resistance does not exceed the specs for your amplifier. This is actually suggested by Flozki. He suggests to add a resistor/capacitor pair (wired in parallel with each other). The r/c parallel pair is wired in series with the head as in this picture from the man himself.

head_rc_network.jpg

He was kind enough to remint me of the section in the AES Disc Recording Anthology that explains the AC Bridge circuit that was implemented in the Neumann SAL and VG racks in the SEL Circuit Breaker modules. The AC Bridge has two points which can be tapped to get a voltage that is then fed into a scaling circuit and displayed on the panel in the temperature meter. The voltage is also used to trigger the relays to open and disconnect audio to the cutterhead (electrically controlled fuse) when the voltage exceeds a user selectable value. Therefore, you should never have a coil burn out as long as the SEL is functioning appropriately.

But, it is not failsafe unlike a simple fuse. It is powered by two fuses. If one blows, the SEL relay can get stuck open. If you were to then power cycle the amp rack without replacing the fuse, you would blow the coil connected to that channel in the cutterhead. So, it is a very good system, but not as failsafe as a fuse. It lets you be much more discrete about the cutoff point than a fuse ever will. Since fuses aren't made in every possible value, you have to go with what is close to what you need. Usually a little more protection than you need... So, there are advantages and disadvantages.

Back to the point. As Flozki suggested, as Neumann implemented, as is the case in a number of other audio drive systems, having the RC parallel pair in series with the element being driven (transducer in a cutterhead or a speaker element) helps to balance the load. The way Flo explained it, it helps to amplifier drive the impedance of the drive coil in the cutterhead.

Bonus: if you implement this early on, adding the additional components to form an AC Bridge that can be used for coil temperature measurement is a breeze. The circuitry for switching a relay based on the temperature voltage value is very easy. Very few components. To scale the value for a physical meter readout is also not that hard. However, finding an analog meter that has a scale in C or F is not easy. I've been hunting for a while. If any has any suggestions, help a fellow cutter out, please and thank you.

Here is an ac bridge. R1/C1 are for the load balancing. R2 is the drive coil in the cutterhead. The values of C2, R3 and C3 have to be arrived at though testing via a test rig which is similar to this that is performed on the top half. Once you figure out he values, you build out the bottom half. When there is a balance, A -> B reads zero volts. When the cutterhead coil gets hot and therefore increases in resistance, the bridge becomes unbalanced top half to bottom half and voltage is measured between A -> B.

schering.jpg

I can give you a schematic that would be a simplified Caruso. You can build it on perf board for probably $25 per channel. When mixing audio I prefer to use OpAmps for buffering. So, if you are doing resister summing, you still likely want to buffer the inputs to the summer.

By the way, 2 Ohms might not be what you want. A 2 ohm coil will not produce as much EM Flux as a 4 or 8 Ohm. Additionally, the lower resistance means that when you drive it hard to get a hot cut, it will get really hot really fast. So, it will need protection if you are cutting loud. As I discussed above, there are really only two ways to do the head protection: a fuse in series or an electrically based circuit breaker that flips a relay. If you winding onto a plastic bobbin, you will definitely want something higher than 4.7 Ohms. Otherise, you will have a lot of bobbins get malformed due to the heat, the malleability of the part at that temp, and that fact that it is being jerked back and forth by the flipping EM Field/Fixed Magnet combo. Once it gets pulled and cools that way, it simply wont function satisfactorily and you'll have to wind it again. Take or leave my advice. But, I can say, the extra wire is an acceptable trade-off when working with plastic bobbins/push rods. At least in my experience. Having to stop testing in order to build up a new transducer becomes tedious fast.

[mossboss]

Hey Todd
Whats wrong with any meter that has the the coil ma suits you,
Scale is whatever you want, C or F printed on paper on a good quality printer than glue it on the face of the meter
It is easy enough to remove the front perspex molded plate on any meter
If you print on any paper fairly large than transfer by reducing it to scale via a photocopier, to photo type paper, you get good definition, it works quite well
Ive done it before
There is enough space between the pointer and the face to slip it in and a few drops of carefully placed glue at a few accessible spots will keep it there for ever
Give it a try its not that costly
Best

[markrob]

Hi,

You are asking some complicated questions. If you are stuck with 28 ga. wire, you have to deal with its physical characteristics. That may or may not work out to a good design. If you look at a wire gauge chart

http://www.powerstream.com/Wire_Size.htm

you can see that 28 ga. has a resistance of 212 ohms per km. So, you need to wind 10 meters of wire to achieve a 2 ohm DC resistance. Can you get that much wire into your coil form? I calculate that the weight of this much copper to be in the 7 gram range given the density of copper at 8960 Kg/m^3 and the diameter of the wire at .00032004 m. That seems like too much mass.

That aside, the DC resistance is only part (and the bad part at that) of the equation. The force you generate from your coil is directly proportional to the length of wire in the magnetic field and the current through the wire. The resistance is an unwanted loss element and the main generator of heat. The loss is given by I^2. As far as you amp is concerned, the AC impedance is what you need to look at. This depends on the DC resistance, inductance, and frequency. The inductance causes current in the coil to drop as you increase frequency. Its easy to get high currents at small voltage swings from your power amp at low frequencies since the DC resistance is the limiting factor. At 20 Khz, the inductance becomes dominant and you need to swing larger and larger voltages to force the same current through the coil. On top of that, you have to fight the back EMF from the head motion and the increased need for more current to accelerate the mass once you are on the other side of the system resonance. That's why you need a high powered amp capable of driving current into an inductive load at high frequencies.

The bottom line here is that to design a good working transducer from scratch, requires quite a bit of finesse. You have to juggle a number of parameters and pick a set of trade offs that are sometimes at odds. It seems like a bad idea to limit yourself to a fixed wire ga. size from the get go since its pretty easy to purchase any gauge you need.

Mark

[Bahndahn]

Hey folks,

Thank you for your replies– this info is important to me.

you can see that 28 ga. has a resistance of 212 ohms per km. So, you need to wind 10 meters of wire to achieve a 2 ohm DC resistance. Can you get that much wire into your coil form? I calculate that the weight of this much copper to be in the 7 gram range given the density of copper at 8960 Kg/m^3 and the diameter of the wire at .00032004 m. That seems like too much mass.

Nice to hear from you Mark. It's becoming clear that I have obtained some rather unsuitable enameled wire. I will certainly consider obtaining something like 32AWG. I am beginning to understand some of the tradeoffs involved. Now a next logical progression will be to being informing myself of the functional parameters of inductance.

suggested by Flozki. He suggests to add a resistor/capacitor pair (wired in parallel with each other). The r/c parallel pair is wired in series with the head as in this picture from the man himself.

Thanks for the details Todd! I will have to dissect the info you have provided over the next few days. Regarding this quote, any advice on choosing the capacitor value?

Amongst the tradeoffs I will need to consider, I am at least well equipped with a 700W amplifier to deliver 350W per channel hopefully not in a coil-frying manner.

On the topic of feedback summing:

This most basic form of summing, using opamps to buffer, is something I am certainly prepared to implement. I am curious, however, about what sort of signal processing is most common on the feedback singal [pre-summing], and what conditions may demand some sort of 'compensation' here. I vaguely recall reading that the Caruso board has a 'bass boost' implemented on the feedback signal [pre-summing]. What other compensation processes are common here? Filtering of any sort?

Thanks again, all, for helping out!

[markrob]

Hi,

I suggest getting an open loop design up and running first. This will give you insight into the physics and help you to understand all of the factors that go into a working system. Once you have that going, look to add the feedback. As far as processing goes, you need to first establish that your feedback element truly represents the motion of the head. Then you need to examine the open loop gain and phase response to see if you can successfully close the loop. It will also help you to identify the required loop compensation required for stable operation. Its no different than any closed loop servo system. A text on servo control theory will be helpful. For a start, look at a PID controller to get an idea of the sort of math tools
you'll need.

https://en.wikipedia.org/wiki/PID_controller

Mark

[opcode66]

Hey Todd
Whats wrong with any meter that has the the coil ma suits you,
Scale is whatever you want, C or F printed on paper on a good quality printer than glue it on the face of the meter
It is easy enough to remove the front perspex molded plate on any meter
If you print on any paper fairly large than transfer by reducing it to scale via a photocopier, to photo type paper, you get good definition, it works quite well
Ive done it before
There is enough space between the pointer and the face to slip it in and a few drops of carefully placed glue at a few accessible spots will keep it there for ever
Give it a try its not that costly
Best

I want a very professional look. So, if I give up searching, I'm simply going to have a run of custom analog meters made by a Chinese company. I want to establish a relationship with them regardless. I'll be having them make my LPI meters for the VST based USB 3.0 P/D Computer. That's why. Naturally, I am familiar with the handy life hack you've detailed.

[opcode66]

At 20 Khz, the inductance becomes dominant and you need to swing larger and larger voltages to force the same current through the coil. On top of that, you have to fight the back EMF from the head motion and the increased need for more current to accelerate the mass once you are on the other side of the system resonance. That's why you need a high powered amp capable of driving current into an inductive load at high frequencies.

You've actually just given me an idea. Thanks man. Cheers!

[EpicenterBryan]

Todd, I'm going to throw this into the mix as well.
That RC network between the amp and the head also does a few things that are not obvious but useful.
The RC combination adds a passive HF boost. The usual values result in 3db-6db boost in the 10K-20K range. But more importantly, the RC combo adds positive phase shift in the same range. That may be just enough to close the loop in the HF if you run into phase issues up there.

Bryan

[opcode66]

I have no issues closing the loop. I've demonstrated that.

Besides, any engineer will tell you that there are two principal functions for this part of the circuit in the Neumann implementation which is what Flo is basing his work off. Load balancing and it is part of the AC bridge used for heat detection that I functionally explained at length. What you are describing is a side effect of the implementation.

In some senses the boost is helping the amp drive the high frequencies, therefore load balancing because the transducer itself is inefficient (most especially in the high end). Any boost that is not being negated by the natural inefficiencies of the transducer itself are negated by negative feedback to a great extent. So, any minimal residual effect in the actual traced signal is irrelevant. Ask Flo. This is both the what and more critically the why and the how of things which I clearly comprehend.

Back to the original question. It is to add resistance. It is done normally.

[opcode66]

Paired with one of these you can get solid symmetric 15V DC 1 Amp.
http://www.digikey.com/product-search/en?keywords=vpt24-1040

Correction. You actually want a 48-520 for 15V 1 Amp symmetric. Sorry.

[EpicenterBryan]

I have no issues closing the loop. I've demonstrated that.

I'll try not to use "Todd" and "you" in the same sentence. The "you" was intended for the general reader / experimenter.

Bryan

[opcode66]

The "you" was intended for the general reader / experimenter.

Ahhh. Ok. Well, the rest of my post is really what I think is important to focus on here. So, in short, the answer to if adding resistance is ok is Yes. And, resistance with capacitance in parallel to the resistance, also ok.

[Bahndahn]

Hi,

I suggest getting an open loop design up and running first. This will give you insight into the physics and help you to understand all of the factors that go into a working system. Once you have that going, look to add the feedback. As far as processing goes, you need to first establish that your feedback element truly represents the motion of the head. Then you need to examine the open loop gain and phase response to see if you can successfully close the loop. It will also help you to identify the required loop compensation required for stable operation. Its no different than any closed loop servo system. A text on servo control theory will be helpful. For a start, look at a PID controller to get an idea of the sort of math tools
you'll need.

https://en.wikipedia.org/wiki/PID_controller

Mark

For a good month or so I was playing around with some open-loop designs in the embossing of CDs on my lathe. The process was simultaneously serving as a test run for the body of the lathe itself, so my focus on the cutterhead was variable. Without a proper stylus or the confidence to invest in one, I wasn't able to make disc recordings worth deriving measurements from. I perhaps prematurely rendered this process sufficient as a dynamic system play-period and now want to get 'to the point' and start working towards a feedback system.

I am using the same drivers as EpicenterBryan and therefore am getting confident in giving this a go. I have retained the coils from the stock driver and may be best-off just working with them.

So for now I think I will proceed by:

-re-mounting the stock coils on a bobbin that is appropriately shaped to reach a feedback coil section
-wind a feedback coil
-take some measurements!

I will report back soon!

[EpicenterBryan]

I am using the same drivers as EpicenterBryan and therefore am getting confident in giving this a go.

Bennett, let's do a Skype session this weekend. There have been some interesting developments with Groove Scribe components over the last month in between things, and I haven't had time to share. I'm very involved in John's Grampian project right now as you might have guessed. But I would be happy to give you a sneak preview. I hope to have time in a week to do an update over on the Groove Scribe thread.

Bryan

[Bahndahn]

Bennett, let's do a Skype session this weekend. There have been some interesting developments with Groove Scribe components over the last month in between things, and I haven't had time to share. I'm very involved in John's Grampian project right now as you might have guessed. But I would be happy to give you a sneak preview. I hope to have time in a week to do an update over on the Groove Scribe thread.

Bryan

Absolutely, Bryan. My semester is wrapping up so I'm back with some spare time in hand to work on this.

That aside, the DC resistance is only part (and the bad part at that) of the equation. The force you generate from your coil is directly proportional to the length of wire in the magnetic field and the current through the wire. The resistance is an unwanted loss element and the main generator of heat. The loss is given by I^2. As far as you amp is concerned, the AC impedance is what you need to look at. This depends on the DC resistance, inductance, and frequency. The inductance causes current in the coil to drop as you increase frequency. Its easy to get high currents at small voltage swings from your power amp at low frequencies since the DC resistance is the limiting factor. At 20 Khz, the inductance becomes dominant and you need to swing larger and larger voltages to force the same current through the coil. On top of that, you have to fight the back EMF from the head motion and the increased need for more current to accelerate the mass once you are on the other side of the system resonance. That's why you need a high powered amp capable of driving current into an inductive load at high frequencies.

Wow Mark, I've read this a few times now and man, this is interesting and sensible! I've recall you mentioning constant current amplification somewhere amongst this forum. That seems like a reasonable consideration given these statements. Is that possible with any ole' amp?

Is there a way to measure back EMF?

[markrob]

Hi,

You can add a secondary feedback loop around a standard power amp to make it run in constant current mode. It requires adding a small current sense resistor (something like .01 to .1 ohms) in series with the ground lead of the drive coil. You would need to add a gain stage via an op amp and another summing junction to close the current feedback loop. Then the drive amp becomes a voltage to current source rather than voltage to voltage. The effect of this is to kill the L/R time constant and its associated phase shift from the head. If you don't want to go to that trouble, you can just add some forward loop compensation in the form of a a lead-lag network to compensate for the phase lag of the L/R time constant.

You can measure the back emf of a driver by using the driver as microphone moving the driver voice coil at a known velocity while measuring the voltage produced. That's easier said than done.

You can calculate indirectly if you measure the DC resistance of the head and calculate the blocked inductance (voice coil held in place) based on the measured resistance and the current flow at a known AC drive voltage and frequency (remember the voltage and current are not in phase when you do that calculations). You can solve for the back emf by observing the blocked current vs. the free current given the other known parameters and conditions.

One other method takes advantage of the fact that a DC permanent magnet motor has the same EMF constant as its force constant if the units are measured in SI units. In this case, the the torque constant (Nm/A) and voltage constant (V/rad/s) are equal. Since the driver is a linear voice coil motor, the basic equations are the same with some minor units changes. So, N/A is the BL product of the driver and this will equal the voltage constant (V/m/s). Its easy to measure the driver BL product. Just determine the amount of force in newtons required to cause the voice coil to deflect a know distance (say 150 um). Then, find the DC current needed to cause the same defection. The BL product is then this F/A and this will also be the back emf constant in V/m/s.

Hope that makes some sense.

Mark