PAFs that sound like sh*t

[Joe Gwinn]

I think what happens is that you're forcing the string to one side, stretching it like a spring, and then releasing it more-or-less instantly. My idea wasn't all that vague - put a guitar pick in a fixture that could pull the string aside by a measured distance, then move the pick perpendicular to the direction stretched until the pick disengaged from the string, ensuring that the pick then moved out of the string's rebound. Since it depends primarily on the tension and stretchiness of the string, I think it's repeatable per string.

But I'd love to hear other suggestions.

The traditional approach was to pull the string sideways a calibrated distance with a cotton thread, and then release the string by burning the thread. This eliminated random mechanical inputs to the string from mechanically cutting the thread.

Impulse response tests were common with speakers at one time for testing frequency response, until they found that you can get the same result with a pseudorandom noise signal, called a Maximal Length Sequence, or MLS. You get the same results as with an impulse, but without some of the difficulties of doing a real impulse test.

However, that test does not generate the instantaneous peak signal that a pickup might do, ...

The MLS approach requires a lot of pretty good equipment to work well.

I find that a 10 microsecond pulse fed to a small drive coil next to the pickup works well. The pulse can be quite powerful, much reducing the effects of noise in the tail of the impulse. One still requires a digitizer to capture the transient. Synchronizing the digitizer to capture the transient is an issue. One can trigger on amplitude, or accept a synch pulse from the generator of the 10 uS drive pulses.

I've measured the impulse response of a number of pickups. The problem has been interpreting the impulse response function - they all look pretty much the same, and the translation into tone is unobvious.

I bet you're right there. Easy to use [BH loop testing] as a go-no-go test.

One thing the BH loop ought to tell us is if the correct magnet and baseplate (et al) have been installed. Don't know how sensitive a test this is.

[R.G.]

I don't know if this helps but harpsichords use delrin (guitar pick material) quills to pluck the strings. You might be able to come up with a variation on the mechanism for plucking a guitar string. Each string could have its own quill for one uniform simultaneous pluck of all strings. Here is a video of replacing a delrin Harpsichord quill. http://www.youtube.com/watch?v=14lDkcogzvk

I'll go look - it sounds promising. My first idea was to force the string a calibrated amount to one side. This may be better.

How about using an EBow to 'pluck' the string? You could sustain the string vibration longer to take steadier measurements. It might make symmetry between tests easier....maybe. Any thoughts on this?
DoctorX

Yes - it's a great idea for plotting steady state stuff. I was going to use an optical pickup to sense string motion and use that signal to drive the string coil. I think I can get an adjustable amount of string motion that way. Great idea!

Personally, I think that a test bed that mechanically excited a guitar string would be letting the numbers of variables under test get out of control, which would bog the experiment down and risk it never producing a meaningful result.

What's more, I'd argue that some of those variables being tested don't properly belong to the pickup, but to the string and the rest of the test bed.

I'd vote for exciting the pickup with a field from an electric current in a test coil. Maybe passing a heavy current from an audio power amp through a copper wire placed where the guitar string normally is would do it. The geometry of the fields generated by this might be wrong, in which case a different shaped coil might be needed. Maybe you need two tiny coils each the size of a single pole screw, with a delay between them to model the speed of sound in a guitar string, or whatever.

Leave it to you to get right to the tough problems!

In my mind, there is a big question about whether a pickup can be tested accurately enough with
(a) a driver coil as you note, perhaps per string position
-or-
(b) an actual moving string
-or-
(c) a simulation of a moving string to isolate X and Y motions

You're exactly correct that you don't want to test the test bed. You want to test the pickup. My approach was to make the test bed massive and non-resonant. I actually considered casting the test bed out of reinforced concrete or drilling nut and bridge supports into a slab of limestone ( I happen to have as much limestone free for the taking as I like).

In fact, it would be nice if the test bed was not needed at all. So my first experiment was to see whether a driver coil produced similar results to the test bed.

I don't think I can get away with never building a test bed. That's because to be taken seriously, this work will have to be in a form where I can demonstrate a direct link to a pickup in a guitar. Otherwise, the nay-sayers will have a field day.

I have a LOT of experience with people's responses to a re-creation of something old. There is always a contingent of people who will respond to the proposal by lying in the weeds and sniping at every possible moment. The arguments are always similar - "That's not original" being the most common for effects, but there's also a lot of commentary on details not working. The reasoning presented in support of the sniping varies from "That's not how they used to do it." to fairly technical arguments about this or that aspect. Sometimes, some of the technical arguments are even right.

But I'll have to be able to demonstrate guitar -> test bet -> final test method, I think, to be convincing. Otherwise the followers of Ned Ludd will be in full cry.

When I did a DIY layout of the Univibe effect back in the mid 90's, I got literally years of a few people sniping at it as not original because I used a three terminal regulator in the power supply instead of a cascaded R-C filter. To me that was a no-brainer. To them it was not "original" and so by definition could not sound as good. They could not express how it sounded worse, but they were sure it did.

[Steve Conner]

Hey, RG,

How would we go about testing whether the driver coil produces a similar response to the string plucking testbed?

I have lots of experience with signal processing and I'd be quite happy drawing frequency response charts or measuring magnetic non-linearities using a driver coil and FFT analysis. I know of free software that will do it with any good soundcard.

You do indeed use a pseudo-random noise source or sine wave rather than an impulse. The information is all there in the impulse response to be sure, but it's buried in non-intuitive ways, like JGundry mentioned. I think response charts and harmonic distortion plots would correlate better with tone, and these are better produced with the MLS or sine-wave excitation.

Finding the most unpleasant problems and attacking them first is my day job.

[R.G.]

The traditional approach was to pull the string sideways a calibrated distance with a cotton thread, and then release the string by burning the thread. This eliminated random mechanical inputs to the string from mechanically cutting the thread.

I bet tying a new cotton thread on is hard to automate!

The release mechanism for a firing pin in a firearm may offer some guidance. That can be standardized pretty well. A rolling-block release is pretty consistent. I'd prefer not to have to go that level of complexity, but I guess it's possible.

The MLS approach requires a lot of pretty good equipment to work well.

Yeah, it certainly did at one time. I've found a couple of cheaper solutions for MLS which use PC soundcard output and input. I don't know if this will be usable or not, but the cost to try is pretty minimal. As show, it needs a PC or laptop and some software. The software feeds an MLS series to the output and records the input. It then does the convolution to back generate the impulse response and frequency spectrum from the recorded data.

May not be sensitive enough. I think it's worth a try. If the results are poor, well, I've wasted another afternoon.

I find that a 10 microsecond pulse fed to a small drive coil next to the pickup works well. The pulse can be quite powerful, much reducing the effects of noise in the tail of the impulse. One still requires a digitizer to capture the transient. Synchronizing the digitizer to capture the transient is an issue. One can trigger on amplitude, or accept a synch pulse from the generator of the 10 uS drive pulses.

That's definitely the straightforward way to do it, OK.

I've measured the impulse response of a number of pickups. The problem has been interpreting the impulse response function - they all look pretty much the same, and the translation into tone is unobvious.

Bummer.

One thing the BH loop ought to tell us is if the correct magnet and baseplate (et al) have been installed. Don't know how sensitive a test this is.

Me neither.

[David Schwab]

This is a great thread.... I just started working freelance doing photo retouching for Bloomingdales... (Xmas is coming!), so I haven't had a chance to read the whole thing yet...

So excuse me if this was brought up already...

Bartolini uses a mechanical picker to test pickups.

This is what they say on their page:

In our testing of every pickup before it is shipped, we use a mechanical "picking tester". This simple device picks a short length of plain string (0.017 inch diameter) in a very repeatable manner. Testing with the "picking tester" has yielded some interesting results.

Mounting the Pickup