Hexaphonic Pickup Project

[Steve Conner]

Question:
When doing these kind of comparisons, does one typically state the difference in ratios, or do you consider that a 10dB increase is perceived as about twice as loud? In other words, do you say "50% greater" or 20log(60/40)=3.52dB= "barely perceptible"?

You can state noise performance any way you like, but the most useful way depends on the application.

For instance, for my ultrasonic preamp I borrowed a competitor's product and took it apart. This gave about 10uV RMS noise in a 500kHz bandwidth (when measuring white noise, the RMS voltage is proportional to the square root of the bandwidth, so you must always state the bandwidth) and the appearance of it was mostly low frequencies. So I took that as my target to equal or exceed.

I tried a half-dozen different circuits, and a simple 2-transistor discrete design based on an old Douglas Self phono preamp came out on top. This was quite counterintuitive because piezos are normally thought of as high-impedance sources requiring a FET buffer, but at ultrasonic frequencies the impedance is actually very low and can soak up lots of current noise. It also falls with frequency, so the current noise gets low-pass filtered.

The NE5534 sucked in this application. It works great over the audio band but kind of falls apart outside it. I think this is because there are so many carefully-tuned feedback loops inside the thing.

My preamp was going to be used with a sonar-type instrument that would plot the results on a screen. For audio, the end goal is to have someone listen to it. So you would measure the noise in dB with A or CCIR weighting, and then apply the perceptual rules of thumb, 10dB "twice as loud" and so on.

For 16-bit digital audio you would ideally like a signal-to-noise ratio (SNR) of 1 part in 2 to the power 16, which is 96dB. 24-bit audio is about 120dB, but I don't think anyone ever achieves this at a system level. Maybe Audio Precision do in their test sets. I would count myself lucky if I got 80dB SNR in a hi-fi or home recording context.

[cycfi]

It's gated out by the noise gate :-)
Nods. I guess it shouldn't matter that much. There are lots of things in the signal chain with more noise anyway. Not to mention that 50s style Hi-Z pickups are very noisy to begin with.

This discusion makes me wonder now if the modern OPA314 would be a good (sufficient) fit with my current design: Low IQ: 150 µA/ch, Low Noise: 14 nV/√Hz at 1 kHz, Rail-to-rail input and output swings, Internal RF/EMI Filter, no phase reversal in overdrive conditions. Again with a single-stage discrete design, I will end up sacrificing linearity, THD, and more importantly superb PSRR and independence from temperature, just to name a few.

[Steve Conner]

This discusion makes me wonder now if the modern OPA314 would be a good (sufficient) fit with my current design: Low IQ: 150 µA/ch, Low Noise: 14 nV/√Hz at 1 kHz, Rail-to-rail input and output swings, Internal RF/EMI Filter, no phase reversal in overdrive conditions.

Looks pretty good, a LM4250 for the 21st century. The THD performance is very good for such a low power consumption. Mark Hammer suggested lithium coin cells inserted in a slot in the pickguard, and I think this op amp would be a great match for that power source.

[Mike Sulzer]

Not to mention that 50s style Hi-Z pickups are very noisy to begin with.

That is something that needs to be discussed a bit. I was going to do it elsewhere, but why not here since you brought it up? (See here * for those who need a description of noise basics, a bit down from the top of the reference.) A useful fact is that a 1K resistor has a noise density (nV/rtHz) of 4.1. Actually, sqrt(4*k*300*1000) = 4.0702825454751906e-09, but who needs to be too accurate?

A typical 12AX7 is equivalent to a 1.56K resistor or 5.12 = 4.1*sqrt(1.56/1.). (The square of the noise voltage goes linearly with the resistance.) The noise resistance of a triode is given by about 2.5/gm at best, gm in ohms/volt, the transcondance, which is usually given micromhos, so you need a factor of a million in there.

The noise from a pickup and the noise from the tube are independent, and so the voltages add in this fashion: Vtotal = sqrt(Vresistor**2 + Vtube**2), where ** means raise to the power. So if we have an 8K PAF, its noise is 4.1*sqrt(8./1.) = 11.60. Vtotal = sqrt(11.60**2 + 5.12**2) = 12.68. This is not much greater than the pickup alone! So the 50s guitar plugged into the fifties amp did not really have significant noise from the active electronics, just from the pickup, in effect. This is something I forget every few years and have to relearn, since with ss you are always worried about the noise of the active electronics, unless you can afford the current to run quiet high current devices.

Then along came those little ss boxes, and hiss became noticeable. But most people did not notice since hum was usually so much greater, and the gain is usually reduced a lot at the higher guitar frequencies where the human ear/brain would notice it most. And then lots of folks lost their most of hearing anyway.

So, even using the cheapest tube, rather than a higher gm quieter triode, early guitar amps (and current ones) were essentially perfect in this regard.

*Stuart Yaniger's description of his latest tube phono preamp:
His Master's Noise: A Thoroughly Modern Tube Phono Preamp - diyAudio

[Mark Hammer]

Looks pretty good, a LM4250 for the 21st century. The THD performance is very good for such a low power consumption. Mark Hammer suggested lithium coin cells inserted in a slot in the pickguard, and I think this op amp would be a great match for that power source.

I did? To quote Shakespeare's beloved Touchstone character: "Nay I wouldst not know mine own wit lest I break my shin upon it".

[Steve Conner]

I think there's a chicken/egg dilemma here. I always thought guitar pickups were designed with so many turns of fine wire so they would be a good match to the noise performance of 12AX7s. Bipolar transistors perform better with lower source impedances that can soak up their current noise.

If transistors had come first, our pickups would have fewer turns of thicker wire, and it would be tubes that would look noisier.

[Mike Sulzer]

By the way, amplifier designs using large resistors in series with the grid of the first stage throw away much of this noise advantage of the tube. 10K is too large; with a low gm tube such as the 12AX7, you can use a lot less with careful layout.

What about the 500K guitar volume control? If it is up all the way, its noise is mostly shorted out by the pickup. If it is part way up, the situation is complicated, because it is parallel combination of the two legs that matters, and there is a partial shorting out by the cable capacitance. So the nice simple situation described in the previous post is not always the whole story.

[Mike Sulzer]

I think there's a chicken/egg dilemma here. I always thought guitar pickups were designed with so many turns of fine wire so they would be a good match to the noise performance of 12AX7s. Bipolar transistors perform better with lower source impedances that can soak up their current noise.

If transistors had come first, our pickups would have fewer turns of thicker wire, and it would be tubes that would look noisier.

I know, that is what I think much of the time except when I remember that there is more to it than that, and so I thought I would write down this time. Guitar pickups really do not need so many turns to have good noise performance with a 12AX7. The actual reasons must be more subtle. Some might be:

1. Tubes were initially larger and more expensive than the minatures we use. Increasing the pickup output meant that you did not need as much gain.

2. Louder pickups sound better to many people. In comparing A to B, the louder thing almost always sounds better. That is an old stereo salesman trick, but it might apply to pickups as well in a more subtle way.

3. Once overdrive became the thing in rock, a pickup with a higher output has a real advantage.

4. Once the "sound" of the electric guitar began to evolve, even before distortion, the resonant boost applied to the upper midrange/low highs that occurred with sufficient turns became a real advantage. Lots of styles (but not all) require that "bite". For example, can you imagine the early tele (broadcaster) becoming a country standard without at least some of the sizzling twang?

There's probably more.

[Steve Conner]

Yes, there certainly is more. A pickup might have a DCR of say 8k, and that determines its Johnson noise contribution. But it is also highly inductive, so its source impedance is considerably higher than its DCR over most of the audio band. High enough that it doesn't completely shunt the noise contribution of the volume pot, and the usual 68k grid stopper doesn't matter that much.

[Mike Sulzer]

Yes, there certainly is more. A pickup might have a DCR of say 8k, and that determines its Johnson noise contribution. But it is also highly inductive, so its source impedance is considerably higher than its DCR over most of the audio band. High enough that it doesn't completely shunt the noise contribution of the volume pot, and the usual 68k grid stopper doesn't matter that much.

Well, a two Henry inductor is about 38K at 3000Hz, so it still gets nearly all of the noise from the pot, especially with the cable cap to get the higher frequencies.

A 68K carbon comp. grid stopper is a true abomination, although I recall as well, the old Fender 4x10 bassman used to allow you to add two inputs in front of the first stage through 100K resistors. If I remember right, it had the decency to short out the resistor with a jack switch when you were using only on input.

[salvarsan]

I recently compared the NE5532 against its young upstart cousins the OPA2134 and OPA2604. The old guy actually came out on top with lower noise and distortion, provided the source impedance was low enough, as it was in my application.

The LM4250 isn't exactly what I would call an audio-grade op amp. Musical instrument grade, maybe.

In his op amp cookbook, Walt Jung has a method for estimating optimum input resistor values according to the op amp current and voltage noise curves. While this book is out of print, Walt still distributes a revised 2002 seminar version for free at his web site

10 years ago, I heard too many people professing the superiority of the JRC 4558 for Tube Screamer designs. Rather than wax into a diatribe about snake oil and stupid people, I point out that the 4559 variant sounded a lot better without costing like the Burr-Brown OPA2604. Both of those were designed to drive a 600 ohm load in addition to simply spec'ing better on noise+bandwidth+slew. Noise isn't the only criterion; distortion matters.

Back when I could get the OPA2604 in quantity for less the a testicle, I put it in everything audio-related because it consistently sounded better than stock. For repair+refurb of rack audio gear, it's an easy choice. The 741-outfitted Furman spring reverb improved amazingly well when I swapped in the 2604, and anyone who cares about their analog crossovers should take note.

Stomp boxes and guitar electronics require good battery life.
Your criteria become noise + bandwidth + slew + distortion + current draw + operating voltage, a mix that excludes a lot of HiFi op amps of yesteryear.

Whichever op amps you decide on, the 741 and 4558 should be excluded from your list.

[David Schwab]

...

Back when I could get the OPA2604 in quantity for less the a testicle, I put it in everything audio-related because it consistently sounded better than stock. For repair+refurb of rack audio gear, it's an easy choice. The 741-outfitted Furman spring reverb improved amazingly well when I swapped in the 2604, and anyone who cares about their analog crossovers should take note.

Stomp boxes and guitar electronics require good battery life.
Your criteria become noise + bandwidth + slew + distortion + current draw + operating voltage, a mix that excludes a lot of HiFi op amps of yesteryear.

Whichever op amps you decide on, the 741 and 4558 should be excluded from your list.

I think anything would sound better when replacing a 741! But here's a funny example; back in the early 80s I built someone a clone of an MXR Distortion + (looking back on it, I can't imaging why anyone would want to use that pedal!). The original used a 741, as did many effects pedals back then (like the Phase 90, which had a slew of 741s).

So I made it with a socket, and instead of a 741 I used a TL074. Surprisingly the 741 sounded "better". This is because what makes it a lesser quality op amp, worked well in a distortion pedal. Except it was a bit noisy. I used the TL074 anyway, because it was quieter, but it had more top end, so it was more hi-fi sounding. I think this is true when using a 4558, etc. I've done those Tube Screamer mods for people. Sometimes you lose certain a characteristic with cleaner sounding op amps. Sometimes it's better. Reminds me of when Leo Fender told Jensen (I think) that their speakers were too clean sounding. He wanted something grungier.

But if you change some caps and stuff you can fine tune these distortion pedals. I did that with a Boss ODB-3 Bass Overdrive that was too fizzy sounding.

The placing the 741s in a Phase 90s with TL074s yielded much cleaner tone.

[cycfi]

Stomp boxes and guitar electronics require good battery life.
Your criteria become noise + bandwidth + slew + distortion + current draw + operating voltage, a mix that excludes a lot of HiFi op amps of yesteryear.

That's exactly the criteria I follow in my Op Amp Shootout page comparing mostly modern and some old Op Amps. If anyone has something to add to the list, I'll be very happy to include them. The 4559 is a good candidate. Alas, its quiescent current is 3.3mA (not exactly low power), and it does not list THD in its specs. Ditto for the 4558 with its 1.25mA quiescent current and no THD specs (it is in the list only to compare against). Modern low power Op Amps have quiescent currents in the uA range.

Another criterion, of course is price per Op Amp.

[cycfi]

Hexaphonic Pickup v1.1

Here’s the V1.1 prototype based on the V1.0 design and specification. The version is bumped to 1.1 since there are slight modifications from the original 1.0 design.

Follow this link for more information…

Many thanks to people in this forum. I learn a lot from you guys! I tried to capture some of the ideas in the link (with proper credit of course). Scroll down to the "Future Directions: Towards 1.2" section.


Fully Assembled Hex Pickup



Neo-2 PCB with components



Neo-2 Fully Assembled

[rjb]

Many thanks to people in this forum....Scroll down to the "Future Directions: Towards 1.2" section.

... and rjb (can’t get his real name right now).

Ralph Jude Barthine
(No connection to the Beatles song; Jude is the patron saint of lost causes. )