Hexaphonic Pickup Project

[cycfi]

I think this has already been covered (or, at least, implied) in class.
Presumably, the hex pickup would be used near the bridge to minimize string cross talk.

Yes, the PU is intended to be near the bridge.

However, I think a smart noise-canceller (read: DSP based) can determine the signal levels and apply cancellation only below a certain threshold, so I'd say it can be used at the neck position too.

noise is cancelled when the signals are added; I think that should read if the signals are added.
In the case of a hex fuzz, for example, I would think you'd want to "de-hum" each string before going to the effect.

My plan is to do a (one-time) analysis for each string and across all strings, while muted, to determine the proper noise cancellation for each string. Also related will be another analysis for each string and from adjacent strings, this time with each string being plucked, to determine the crosstalk from one string to another. This information will then be used for proper crosstalk cancellation. AFAICT, noise cancellation and crosstalk cancellation are related problems. The analysis that produce the constants can be obtained at manufacturing time for each PU set (the constants are static). Cancellation can all be done in the DSP immediately after ADC.

Hope that makes sense?

[rjb]

My plan is to do a (one-time) analysis for each string and across all strings, while muted, to determine the proper noise cancellation for each string.... The analysis that produce the constants can be obtained at manufacturing time for each PU set (the constants are static). Cancellation can all be done in the DSP immediately after ADC.

Hope that makes sense?

Dude, most of the discussion in this thread has been above my head. I think you lost me in post #3 with the 6 analog outputs multiplexed over a stereo cable.

I don't understand about static constants to define proper noise cancellation- doesn't the produced noise depend on the environment? I'm not even sure that question made sense.... But don't mind me- I'm sure you know what you're doing.... Carry on- you're doing a great job!

[cycfi]

Dude, most of the discussion in this thread has been above my head. I think you lost me in post #3 with the 6 analog outputs multiplexed over a stereo cable.

I don't understand about static constants to define proper noise cancellation- doesn't the produced noise depend on the environment? I'm not even sure that question made sense.... But don't mind me- I'm sure you know what you're doing.... Carry on- you're doing a great job!

Ignore it, it does not work :-( Mike is right. We need a true humbucker or at least a void coil to pick up the noise (perhaps an SMD inductor would do).

Back to the drawing board and on to version 1.5 ...

You all rock! Thank you very much for your insights!

[rjb]

Mike is right.

Damn, I hate when that happens.

[Mike Sulzer]

Peace, indeed.

Hex fuzz is an interesting case. When you are not playing any string the gains are the same and hum cancellation occurs when the signal are summed. When you play one or more strings, the degree of cancellation is greatly reduced, since the gains are unequal, but the signal hides the hum. At least this is how it works for me.

As for lower cross talk near the bridge, has anyone measured this? I do not think a string moves enough even near the neck so that it approaches a neighboring pole piece, and the other effect of moving closer to the bridge is to reduce both the main and cross talk signals, leaving the ratio unchanged.

I think this has already been covered (or, at least, implied) in class.
Presumably, the hex pickup would be used near the bridge to minimize string cross talk.


I was going to ask teacher about that.
In the assigned reading (you did read the assigned text, didn't you?), he wrote that noise is cancelled when the signals are added; I think that should read if the signals are added.
In the case of a hex fuzz, for example, I would think you'd want to "de-hum" each string before going to the effect.


Zzzzzzzzzzzzzz.

[Mike Sulzer]

Well, I have some concern about a small coil that puts out more than a strat pickup with fewer turns. Is it because that neo magnet is strong enough to cause string pull? I would prefer to keep the signal level up by grabbing the flux from the vibrating string with a permeable material and using a weaker permanent magnetic field.

As for production, if you are going to buy coils, I think it would be better to use a stock part rather than a special order. Easier to get small quantities, easier to find an alternate supplier, etc. Of course, you have to verify that the stock ferrite serves your purpose.

Possibly a trivial problem; still a problem.


To quote a wise man, "But why would you want to?"
Looks to me like you're adding unnecessary (key word "unnecessary") bulk and complexity for the questionable advantage of "a bit more signal".
Referencing the graph in post 48, looks to me like the prototype is already a few dB hotter than a Strat pickup.


Yawn, did you say something?

PS-
Mike, I apologize for the "sleeping in class" crack. I meant to insert a smiley but forgot.
And fwiw, I agree with your assessment of the EMG ckt.

[rjb]

Peace, indeed.

I concur.

Hex fuzz is an interesting case. When you are not playing any string the gains are the same and hum cancellation occurs when the signal are summed. When you play one or more strings, the degree of cancellation is greatly reduced, since the gains are unequal, but the signal hides the hum. At least this is how it works for me.

So, you're saying that, in your experience, hex fuzz is one case where humbucking across string pairs would be adequate.
Guess I picked a bad example.

As for lower cross talk near the bridge, has anyone measured this?

Not me.

I do not think a string moves enough even near the neck so that it approaches a neighboring pole piece,

Even when bending strings? Now I'm really confused.
Unless the string bending is handled with DSP, of course.
Anyone know where I can get a good used DSP B-Bender for cheap?
(Where's the smiley for "groaning at my own weak attempt at humor"?)

and the other effect of moving closer to the bridge is to reduce both the main and cross talk signals, leaving the ratio unchanged.

Here's my guess:
When the pickup is moved closer to the bridge, you can raise it closer to the strings without choking them.
The relative distance between "coil A and string A" versus "coil A and string B" decreases.
Thus, the ratio of main to cross talk signals increases.

[rjb]

Well, I have some concern about a small coil that puts out more than a strat pickup with fewer turns. Is it because that neo magnet is strong enough to cause string pull? I would prefer to keep the signal level up by grabbing the flux from the vibrating string with a permeable material and using a weaker permanent magnetic field.

That's out of my area of expertise. What am I talking about- I don't have an area of expertise!

As for production, if you are going to buy coils, I think it would be better to use a stock part rather than a special order. Easier to get small quantities, easier to find an alternate supplier, etc.

Having a design that requires single-source vendor parts kinda sucks- but it's not as though we're talking about building life support equipment.
Besides- "Use off-the-shelf parts, you get off-the-shelf designs."

Of course, you have to verify that the stock ferrite serves your purpose.

All I know about that is: soft ferrite is usually used for inductor cores and hard ferrite is usually used for magnets.
Ferrite (magnet) - Wikipedia, the free encyclopedia

[cycfi]

Peace, indeed.

Hex fuzz is an interesting case. When you are not playing any string the gains are the same and hum cancellation occurs when the signal are summed. When you play one or more strings, the degree of cancellation is greatly reduced, since the gains are unequal, but the signal hides the hum. At least this is how it works for me.

Right! The main issue is that it only works only when the signals are arithmetically summed.

Basically, you have, where A...F are the individual signals, and N is the noise component (assuming the PUs are perfectly matched):

A + N
B - N
C + N
D - N
E + N
F - N

I thought that I could extract only N and de-noise individual signals before any further processing. I was wrong. There's no way to do that. It's only possible if I have another void coil that picks up just the noise. The question now is how to easily do that? I'm thinking about using a dummy coil (does not have to be the same as the coils) that picks up the noise component. That dummy coil can be placed elsewhere (maybe outside the PU). But I wish to avoid that complexity, if possible. Another idea is to print a long spiral/coil in the PCB (near the edges) that connects to another preamp with the sole purpose of extracting the magnetically induced noise. The signal of that would be very weak, but it can be amplified. Do you think that would work?

As for lower cross talk near the bridge, has anyone measured this? I do not think a string moves enough even near the neck so that it approaches a neighboring pole piece, and the other effect of moving closer to the bridge is to reduce both the main and cross talk signals, leaving the ratio unchanged.

I'll do some measurements soon. I'll post my results.

[cycfi]

Well, I have some concern about a small coil that puts out more than a strat pickup with fewer turns. Is it because that neo magnet is strong enough to cause string pull? I would prefer to keep the signal level up by grabbing the flux from the vibrating string with a permeable material and using a weaker permanent magnetic field.

As for production, if you are going to buy coils, I think it would be better to use a stock part rather than a special order. Easier to get small quantities, easier to find an alternate supplier, etc. Of course, you have to verify that the stock ferrite serves your purpose.

The actual signal strength of the coil alone is around 1/4th of the Strat. Not bad, really. And I did recess the Neo a bit to limit its pull. The preamp has a gain of 5, so I have a hotter signal, somewhere in between a Strat and a LP. A weaker magnet will require a higher gain in the preamp which will multiply the self noise in the electronics. So my goal was to make it as hot as it can go without compromising the frequency response. From my experiments, 2000 turns and a small recessed 5mm X 5mm Neo is just about right.

(Aside: The bobbin is 7mm in height and the Neo is 5mm and recessed 2mm from the top of the coil. Somehow the recessed Neo gives the PU more focus and tighter bass response. Does anyone here know why that is?)

As for production of the coils, I've already invested a lot of effort on the current design. I have a semi-custom solution using stock inductor parts from Chipsen's catalog. I'm having a good relationship with the company and they are willing to produce what I need. Small quantities is out of the question, but I am willing to shoulder the cost.

[Mike Sulzer]

Another idea is to print a long spiral/coil in the PCB (near the edges) that connects to another preamp with the sole purpose of extracting the magnetically induced noise. The signal of that would be very weak, but it can be amplified. Do you think that would work?

I think you would have to amplify it too much and the resulting "hiss" would be noticeable. You might say, well, let's limit the frequency response of the hum sensor amp, but you really want broadband cancellation because some switching high current things emit a lot of magnetic interference at harmonics in the kHz range.

[Mike Sulzer]

OK, that is the amplified level. That makes me feel better, and with 8nV/rootHz your noise level is low.

Well, I think you kind of answered your own question! You lowered the magnet until it sounded OK, meaning that you removed even the initial effects of string pull, or at least the ones you noticed and did not like.

The actual signal strength of the coil alone is around 1/4th of the Strat. Not bad, really. And I did recess the Neo a bit to limit its pull. The preamp has a gain of 5, so I have a hotter signal, somewhere in between a Strat and a LP. A weaker magnet will require a higher gain in the preamp which will multiply the self noise in the electronics. So my goal was to make it as hot as it can go without compromising the frequency response. From my experiments, 2000 turns and a small recessed 5mm X 5mm Neo is just about right.

(Aside: The bobbin is 7mm in height and the Neo is 5mm and recessed 2mm from the top of the coil. Somehow the recessed Neo gives the PU more focus and tighter bass response. Does anyone here know why that is?)

As for production of the coils, I've already invested a lot of effort on the current design. I have a semi-custom solution using stock inductor parts from Chipsen's catalog. I'm having a good relationship with the company and they are willing to produce what I need. Small quantities is out of the question, but I am willing to shoulder the cost.

[David Schwab]

I guess L1, & L2 is the pickup?
And, what would be a good DCR Range for L1, and L2?
Thanks,
T

Yes, those are the coils.

The notes I have say an an EMG 81 is about 4.2k per coil, or 5500-6000 turns wound with 43 AWG wire.

[David Schwab]

But why would you want to?

What the circuit does right:

It cancels magnetic hum well. The two coils put out hum signals with the same polarity which then subtract because they are connected to inputs with opposite polarity. The fact that the two coils are loaded differently does not affect this since essentially no hum current needs to flow throught the 30K resistor to the inverting input.

What it does wrong:

It loads the two coils differently at high frequencies, causing a drop in the highs from one of the coils.
The signals of the two coils have opposite polarity, and the subtraction of the diff amp makes them the same so that they should add. The voltage on the summing junction (negative input) is determined by the voltage from the coil connected to the positive input and so this is increases the current through the 30K resistor, causing a large loading effect when the impedance of the other coil gets high as it does at high frequencies. If they use medium impedance coils, this would not be too much of a problem. How many turns do they use?

They load the two coils different to voice the pickups. EMG calls it "tone modeling" or "impedance modeling." They are high impedance coils, usually 4,500 to 5,000 turns per coils. Their differential circuit, along with shielding, producing an extremely quiet pickup. These are about the only pickups I have ever used that do not pick up noise from lamp dimmers. They also allow you do disconnect the string ground, which does not add any extra noise. I used them for about 15 years on both guitar and bass, and they are dead quirt in every situation. Other instruments with conventional passive humbuckers still produced noise in the same environments that the EMGs were dead silent.

[Mike Sulzer]

They load the two coils different to voice the pickups. EMG calls it "tone modeling" or "impedance modeling."

Yeah, right. This is from their web page:

One of the most important aspects governing the tone of a pickup is the resonant frequency. EMG Pickups use "Impedance Modeling" to manipulate the two coils. This innovation allows us to shape a mix of the reactive slope and resonance from each of the two coils. The idea is to achieve a complex mixture of each coils phase and frequency response resulting in a richer tone from the pickup. This means the sound is vibrantly alive with more harmonics than from conventional passive pickups.

When you load one of the coils more heavily at high frequencies you do not get more harmonics; you lose high frequencies. They use a cheap noisy op amp and a circuit that does not do what they claim. Let's not mistake sales lingo for engineering.

And where is the resonant frequency? With no capacitance to replace the cable capacitance, and a conventional number of turns, it should be quite a bit higher than a normal humbucker. Maybe this is the justification for the claim of "more harmonics" compared to a passive pickup. But it is not what they say.