John,
I too appreciated that BuildYourOwnGuitar page. And I also appreciate the extent and detail of the various modeling approaches.I'll confess that the complexity here is getting a bit beyond my math skills. But as the OP here, I'd just ask if all of this can be boiled down some to tests I can do myself in the situation I described. You can get pickups very cheap sometimes, likely made in China and sold on e-bay or ali-express, often for less then $5 each. Are they crap? Well most would jump to say so. But in actual home brewed guitar projects I've been building, costs are a factor, and I simply can't rule out parts on the basis of assumption that their low price must mean they are garbage. So with that in mind I've set up a variety of test jigs so i can quickly get an idea of each pickups's subjective sound (to my ear), and also have experimented with a variety of front end circuits. One thing became very clear very quickly, and that was that buffer amplifiers (to let the pickup see a non capacitive high Z) could make some very cheap pickups sound a lot better. But I also have noted that some pickups definitely have more inherent high end then others, even given the benefit of o0f buffering.
So the purpose of my original post was to learn whether there were any reasonably simple electrical tests I could put together, to better predict which pickups likely will have better frequency response. I do have some equipment... audio generators, a decent scope and such, and now I've been introduced to the free "ARTA" software that works with your sound card to serve as both a generator and measurement device. So can you offer any advice on how something like this GuitarFreak software could help in the prediction process, and explain what external equipment is needed to make it useful?
Can definitely help.
What kind of pickups? I'm assuming ceramic magnets (grey bar underneath)with steel poles? or can you get alnico for a fiver?
Anything with steel in it seems to have more inherent damping, possible combined with relatively high inductance, leading to some dullness. Reducing load on the pickups can help a lot by reducing damping. Using high value pots (eg, if its a single coil, 500k instead of 250k, then even making the tone pot no-load). Then feed that through a high impedance buffer to negate cable capacitance. You can make that with one JFET if you wish, and set it up with a high input impedance, at least 1M but you can also do twice that. Then test, and maybe add a small cap like 330pF or so across the pickup.
To analyse it, using a SPICE sim, or GuitarFreak, need to work out some parameters. Much can be estimated by, knowing the type of magnet, measuring the dc resistance (not actually of great significance in itself but is easy to measure and gives some clues), and then measuring inductance. Can you do that with your meter?
Then, if you want to go to a much more explicit test, you could consider setting up a test rig. Now although I'm very interested in this and enjoy the results, I haven't done this myself. But what I think it entails is (garage science version):
Generate a sweep frequency output from a signal generator, feed it through a linear amp stage that can drive a 100ohm load - actually that is not very demanding.
Make a test coil with say 100 turns of insulated bell wire on a pickup bobbin, with no core.
Mount it over the pickup
Feed that via a 100Ohm resistor from the sweep signal
Connect the output of the pickup to a buffered circuit with a known high input impedance, then into a pc via whatever interface you have, to create a frequency versus output plot in db
Do that with the unloaded pickup, then add a known resistor and cap across the pickup coil. We are using 470pF and 200k.
You get two plots, each with a peak frequency and height, each with a general upward slope of 6db/octave. I can use those plots to derive parameters, which starts with taking that slope out.
But, you might prefer to bypass all maths, if you have one guitar that you know has a good tone and where you know the circuitry (ie, maybe it is a standard Strat and you know the pot values etc). You'd do the test on that guitar, via say a 10' cord. Then try the cheap PU, using just very short leads into the buffer, and experiment with the load resistor and cap to get the best match. That then determines what the active circuitry has to be for a guitar using the cheap pU that gets as close as possible to the reference.
What kind of pickups? I'm assuming ceramic magnets (grey bar underneath)with steel poles? or can you get alnico for a fiver?
Anything with steel in it seems to have more inherent damping, possible combined with relatively high inductance, leading to some dullness. Reducing load on the pickups can help a lot by reducing damping. Using high value pots (eg, if its a single coil, 500k instead of 250k, then even making the tone pot no-load). Then feed that through a high impedance buffer to negate cable capacitance. You can make that with one JFET if you wish, and set it up with a high input impedance, at least 1M but you can also do twice that. Then test, and maybe add a small cap like 330pF or so across the pickup.
To analyse it, using a SPICE sim, or GuitarFreak, need to work out some parameters. Much can be estimated by, knowing the type of magnet, measuring the dc resistance (not actually of great significance in itself but is easy to measure and gives some clues), and then measuring inductance. Can you do that with your meter?
Then, if you want to go to a much more explicit test, you could consider setting up a test rig. Now although I'm very interested in this and enjoy the results, I haven't done this myself. But what I think it entails is (garage science version):
Generate a sweep frequency output from a signal generator, feed it through a linear amp stage that can drive a 100ohm load - actually that is not very demanding.
Make a test coil with say 100 turns of insulated bell wire on a pickup bobbin, with no core.
Mount it over the pickup
Feed that via a 100Ohm resistor from the sweep signal
Connect the output of the pickup to a buffered circuit with a known high input impedance, then into a pc via whatever interface you have, to create a frequency versus output plot in db
Do that with the unloaded pickup, then add a known resistor and cap across the pickup coil. We are using 470pF and 200k.
You get two plots, each with a peak frequency and height, each with a general upward slope of 6db/octave. I can use those plots to derive parameters, which starts with taking that slope out.
But, you might prefer to bypass all maths, if you have one guitar that you know has a good tone and where you know the circuitry (ie, maybe it is a standard Strat and you know the pot values etc). You'd do the test on that guitar, via say a 10' cord. Then try the cheap PU, using just very short leads into the buffer, and experiment with the load resistor and cap to get the best match. That then determines what the active circuitry has to be for a guitar using the cheap pU that gets as close as possible to the reference.
Interesting. Did you notice the capacitor is in parallel with the inductance in Fig 1?
In Fig 2, Lemme adds an AC voltage source to simulate the pickup's output voltage, and moves one end of the capacitor to ground.
However, the frequency response plot he comes up with is qualitatively identical to the one I got from my LTSpice simulations - using a parallel capacitor, not one to ground.
I think this works because the source impedance of an ideal voltage source is zero. Therefore it doesn't matter which end of the voltage source the capacitor is connected to - the "hot" end, in Lemme's Fig 1, or the grounded end, in his Fig 2.
Which is exactly what happens if the cap is in parallel with the inductance!
I simulated that already - we actually get a slight signal dip at the resonant frequency, even into a 1 megohm load.
These are things which are very relevant if someone is trying to physically model an entire electric guitar, presumably in order to generate a software model of one.
Similar things have been done for pianos several years ago ( https://www.pianoteq.com/ ). The virtual piano in Pianoteq can generate very realistic piano sounds in response to a virtual hammer-strike on the virtual strings.
Well, for starters, thank you for sharing your work!
Much of what you described matches what I've already posted in this thread - the test coil, the series resistance to ensure constant-current operation, the use of an integrator to convert that to constant rate-of-change, if desired.
The truth is, I myself have little interest in simulating an entire guitar in software, so I'll leave that to you, and to others who are excited about the concept of creating a software model of an electric guitar. I guess one day "EZ Drummer" software will be accompanied by "EZ Guitar" software.
As for me, I prefer the real thing (guitars), with the extraordinary amount of tiny real-world subtleties that come with it. For example, if I play guitar with my fingers instead of a pick, with a clean tone and a wide bandwidth amp, I can hear the sound of my fingerprints sliding over the strings - those tiny skin ridges make multiple rapid contacts with the string as your fingertip passes across it, giving you a sort of shimmering treble.
-Gnobuddy
Thanks for looking at that Gnobuddy - and good luck with your explorations.
What I'm doing is not actually intended to lead to simulating a guitar sound, its more about understanding real guitars and how various elements change their tone. Its intended as a design tool for making choices about guitar circuitry, and as more data is added, choosing pickups and how to combine them. The graphs of course, don't let you hear a sound by seeing them, but based on an intent to say, make guitar X sound brighter, or make Y sound more like Z, or what value to use for a bass-cut cap. it can help home in on a selection by understanding what is significant, prior to actually testing by ear.
What I'm doing is not actually intended to lead to simulating a guitar sound, its more about understanding real guitars and how various elements change their tone. Its intended as a design tool for making choices about guitar circuitry, and as more data is added, choosing pickups and how to combine them. The graphs of course, don't let you hear a sound by seeing them, but based on an intent to say, make guitar X sound brighter, or make Y sound more like Z, or what value to use for a bass-cut cap. it can help home in on a selection by understanding what is significant, prior to actually testing by ear.
That sounds like a very worthwhile goal. Good luck with the work!
I took a dead-simple route, one that works with every guitar I own, and every amp I might plug into: I almost always use a 7-band graphic EQ pedal between my guitars and amps. It cannot fix a truly bad pickup, or create treble that isn't there - but, those limits aside, it gives one a lot of flexibility in tone-shaping.
I actually own two of these graphic EQ pedals, because if one fails, I don't want to be without one. (It's also useful to have one to shape the raw guitar tone, and a second to shape the sound of any overdrive or distortion pedal in the chain.)
For me, these EQ pedals are without any doubt the most versatile and most useful guitar effects pedals I own. It still confounds me that they are not part of every plugged-in guitarists rig.
-Gnobuddy
Well thanks for all that. As it turns out, most of what you're describing seems to be missing the point of my original question, but its interesting just the same.
My gut feeling is still that the test-coil + frequency sweep is most likely to reveal what you're looking for, i.e., relatively minor differences between different pickups.
I don't believe our ears are magical, so if we can hear it, we can measure it. We just have to find the proper measurement to do, so that it reveals what our ears are hearing.
If the ARTA (pink noise + FFT) approach (edit: ARTA driving the test-coil, pickup driving the FFT input) isn't giving you enough frequency or amplitude resolution, an old-fashioned slow sinewave sweep should do it. It just takes a little longer to run the measurement.
It may also be necessary to focus on one specific region of the frequency response. I suspect that this will turn out to be not too far from the frequency band where the ear is most sensitive. In other words, I wouldn't (at least at first) waste my time measuring pickups at, say 100 Hz; my guess is that the region from, say, 1 kHz to 6 kHz is where the answers will be found.
Keep in mind that it is known that the human ear can detect quite subtle frequency peaks, if they extend over a relatively wide frequency band (low Q). It may be that a difference of 1 dB in a pickup's peak response is audible, if it spans an octave or more.
I have another suggestion, too: what happens if you drive the test-coil with white noise, and listen to the output of the pickup under test (using something like an op-amp buffer and headphone driver, or flat-response audio amp and speaker)? Maybe your ears will be able to immediately pick out the "good" pickups from the bad, for your purposes.
-Gnobuddy
Thanks! Yes that is exactly the kind of advise I was looking for. The ARTA software works, and do0es seem to point out small but significant differences. But the one thing it doesn't seem to have (unless I haven't discovered it yet) is a SLO sine wave frequency sweep as you've described. I have a home brewed frequency sweeper though. I'll have to think on a way to synchronize them.
Excellent, now we know you're on the right track!
You may not need a very slow sweep - slower sweeps are to reveal finer frequency details, and my first guess is that there won't be any of those in a pickup's response.
My concern was that noise-based measurements, while quick, tend to have a "fuzzy" nature. That will make it hard to accurately compare two curves. A swept sine should give you a crisp, clear frequency response curve.
If ARTA won't do a sine-wave sweep, perhaps RMAA will: mh-audio.nl - Home
There is also STEPS, on the same page as RMAA, a bit further down. It seems to be designed to do exactly what you want, i.e., stepped sine-wave frequency response measurements.
(Full disclosure: I don't use Windows, so I've never used RMAA or STEP. But RMAA, at least, seems to be well endorsed on this forum. And the price is right, too - it's free.)
-Gnobuddy
Last edited:
Thanks! I'll definitely check into RMAA. As the the ARTA noise test, I have mixed feelings too. on the one hand, it does have several averaging options, so if you let it run a minute, you get a pretty smooth representation.
As for the originally suggested test method earlier in this thread, basically feeding the sweep (or noise) to the pickup through a 1 meg resistor, and measuring its response the same way, it did in fact point out that a better sounding pickup had a peak at a slightly higher frequency (7.5khz as opposed to 6khz). But as you've pointed out, its hard to really see the true connection between that test method and the way a pickup actually produces sound from vibrating strings. I've still yet to rewind the test coil you suggested (too many phases of the same project on my bench), but I'm anxious to try it again when i do, because it "seems" closer to a true simulation. Then again, in the back of my mind I'd have the same question.. is an induced magnetic field of audio exactly the same as a non magnetized steel string vibrating.
It doesn't have to be exact - it just has to be good enough for you to get the data you need. The universe is kind that way!
Have you ever heard the joke about the physicist solving the farmer's problem of her cows jumping over the fence, with the catch that the solution only applied to spherical cows in vacuum?
That's just a joke, but scientists do, in fact, extract useful information from simplified models of the (very complex) real world. At first, Isaac Newton only proved his laws of motion for infinitely tiny rigid points, which had finite mass (and therefore, infinite density!) But that very simplified model is in fact so good, and applies to so many things that are not perfect points with mass, that centuries later, NASA can land rovers on Mars using Newton's laws.
In our case, will the test-coil pickup measurement accurately reproduce the effects of, say, one slightly chipped Alnico magnet slug in the G-string hole of a Fender pickup?
Well, no, of course not!
But will it let you see the basic frequency response characteristic of one pickup, and compare it to another?
I'm betting it will.
-Gnobuddy
It doesn't have to be exact - it just has to be good enough for you to get the data you need. The universe is kind that way!
Have you ever heard the joke about the physicist solving the farmer's problem of her cows jumping over the fence, with the catch that the solution only applied to spherical cows in vacuum?
-Gnobuddy
Well as long as this method will at least simulate to sound of strings vibrating in a vacuum, I'm good with it!
Just kidding of course.
And I'll definitely experiment more with this method. Surly you can understand my healthy skepticism in asking these questions. After all, if I had $100 for every time something seemed very logical was proven wrong, I'd have retired early many years ago! The most important thing to avoid the pulling of much hair, is when someone has actually used a method (which you have), and found it successful.Believe it or not, about twenty years ago, I spent a few months babysitting an experiment involving steel strings (music wire) under tension in high vacuum. One end of the wire was anchored to a heavy aluminum end plate, the other end to another heavy mass. Once you "plucked" a string in vacuum (done with a nearby small coil), it would vibrate for minutes on end. (Not seconds, minutes.)
So, if you want to beat a Les Paul guitar's sustain, all you have to do is machine the guitar out of solid 6061-T6 aluminium, and play it in a vacuum!
Nature has a way of making us look utterly silly, especially when we think we're being smart. Somebody said "Many a brilliant hypothesis has been utterly ruined by a single inconvenient fact".
-Gnobuddy
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.