Geofex low power onboard buffer build report

[free electron]

Here is the link to the article and schematic:
http://www.geofex.com/FX_images/Onboard_Preamp.pdf

I decided to give it a go and designed a small compact PCB for it. The PCB is only slightly larger than a dual CR2032 battery socket, most components are SMT (0805 size, SOT23). I added a few extra ground pads, through plated M3 mounting holes (useful to connect the ground to the cavity shielding), separate "Battery -" terminal to use a stereo jack for power switching.

Here's a few pictures from the building process:
1. Solder paste applied to the pads

2. Human Pick&Place machine using vacuum pick up

3. All components populated

4. Reflow soldered using hot air station

5. Through hole components installed...

6. ...and soldered. PCB cleaned with isoprophyl alcohol


A few test results:

Measured current consumption (Fluke 87V in hi-res mode): 112ľA

Then i wanted to see how much headroom it has with the two coin cells 6V supply, my DIY DDS based signal gen (10-100kHz) was able to output 4.8Vpp max and the buffer was asking "Come on, is that all what you've got there?"


Frequency response, measured with a 24bit 192kHz soundcard, i added the response of the card itself for reference (out directly to input):


I'm going to measure the input and output impedances and test the buffer with few guitars in the following days, especially with some killer output humbuckers to see if 6V is enough (most likely is).
Generally it's a very nice, clever and extremely useful (if you need to add a battery powered buffer here and there) circuit! Many thanks RG!
I'll to post the gerber files soon. RG, i still have a few PCBs left. If you'd like to get one for evaluation, let me know!

Piotr
(i hope i got the picture links right, preview somehow doesn't work anymore..)
 

[R.G.]

Very nice work.

I think - you're right, preview doesn't work any more; the pictures didn't come through.  
! They're there now !

But the description is very tantalizing. I cobbled my unit with through-hole stuff on perfboard. An SMD version of this would be very tidy indeed.

It's also very nice to know that the design is robust enough to translate to other parts and an SMD implementation. I always try to design for that, but I've messed up the issue often enough to be happy when things work out the way I thought they ought to.  

[digi2t]

Fantastic write up. On top of that, SMD looks a bit less intimidating now.  

[free electron]

I was trying to be as close as possible to the original schematic when it comes to component selection. Had to order MMBT5088/5087 from the US, we are like more BCxxx-style here in Europe
Here is the PCB layout (37x33mm):



->digi2t
the hardest part in SMD is actually forcing yourself to try it

[R.G.]

One goal of mature engineering practice is to make the designs as tolerant of component variation as possible. I -think- that the design would work with the BC-family components as long as the pinout is the same.

I ... hope...   

[free electron]

There is one way to find out, i thought about using BC550C for the NPNs and BC560C for the PNP which i have plenty of laying around. Pinout is the same.

[R.G.]

Give it a try. I think it will work OK.  Two transistor feedback pairs are usually unconditionally stable unless you do funny things in the feedback path to make them unstable. The details of how much impedance, gain accuracy, frequency response and so on you get may change a little, but not so much that I think a guitar amplifier would notice.

[Kipper4]

Thats super neat, good work mate.
Jelous. I'm tryin to up my game but it looks like i have a ways to go.

[bluebunny]

Neat build.  Are you planning to do a run of (half-)populated boards for SMD cowards?   

[sajy_ho]

Awesome work, Just one question: How did you place it in your guitar? I mean I was interested in this topic but thinking about opening my strat's screws, made me give up on this project.

[free electron]

Thanks guys!
I don't have any plans to sell the half-populated boards, sorry. I'm going to post the gerber files and maybe upload them into the OSHpark library.
I may only sell the rest of the bare pcbs i have left, pm me if you're interested.

->sajy_ho
i haven't installed it into a guitar yet. I just build it and want to do some testing. Strat users generally have more stress with access to the electronics

Ok, time for more testing:
1. Input impedance. I used a common method of building a voltage divider out of a external resistance and the input impedance of the circuit, measuring the amplitude of the signal on the scope:



When the blue signal is 50% of the yellow one, the Rseries (trimpot) will be equal the input impedance. After i found that spot i took the trimpot out and measured it's resistance:



The result was about 824kOhm @ 1kHz (test signal). Nice!

2. Output loading. I wanted to see how does the circuit behave when in typical recording situations, that is driving a typical mixing console line input (~10k Zin) and an instrumental input with higher input impedance (about 300k). I did the test with two amplitude values of the signal: large signal, about 1.7Vrms and small signal about 100mVrms.
Animated gifs show how does the output waveform change when the circuit drives different impedances.
Large signal:


Small signal:


This is actually a very nice experiment that shows how the driving capability depends on the amplitude of the signal.
Next, i took one guitar which has a Dimarzio Evolution humbucker in the bridge position in order to see if there will be any clipping. The output of the buffer was plugged into a Line/Instr input in my mixer. The first part is the 10k Zin line input setting. Clipping the bottom half of the waveform, as seen in the previous test is clearly visible. I switched to 300k Zin in the second part. The signal was nice and clean with no sign of clipping.



And a sound sample: https://soundcloud.com/user8582335/oboard-buffer-test

I think the circuit does the job it was designed for very well:
- provides high input impedance for the pickups,
- provides relatively low, but, perhaps even more important, constant within all the usable frequency range output impedance,
- does that at a very low power consumption.

One thing to watch out is the buffer should be rather plugged into inputs with Zin above ~30k, most of the guitar gear falls into that category. Well, maybe except a few rackmount FX processors which have 10k inputs and are causing problem with tube FX loops too...

I wonder if replacing the battery with two 3.7V Li-Pos, adding some smart energy harvesting circuitry would make it, huh, eternal?

[PRR]

> Animated gifs show how does the output waveform change when the circuit drives different impedances.

Neat GIFs.

Wonder how you got 101mV in 0r?

> inputs with Zin above ~30k

Yes. The output standing current is about 0.1mA. The output voltage is not even 3V peak. 3V/0.1mA is 30K.

There is some White-Cathode-Follower action through the 100r and 100nFd. However I do not think this significantly aids the signal power output (could be wrong).

The "Right Now!" hack to drive 10K is to change 5.1K to like 2.2K or 1.5K. Triples the battery drain. May have other consequences (though I would not expect anything horrible).

[PRR]

> Input impedance. ... measuring the amplitude of the signal on the scope:

1) Your 'scope CH2 input impedance is shunting the buffer input. If 'scope is 10 Megs, the error is small. If 1Meg, it is comparable to the thing you are measuring.

2) Using a "matching" series resistor is great for textbooks but can give trouble on the bench. Many hot amplifiers will go crazy if the input is not held-down by some fairly low source impedance. It is good that you did not find this (R.G. wasn't born yesterday). But in general, it may be better to use a series resistor which gives about 10% drop at mid frequency. This is "low-Z source" for most amplifiers, yet causes enough sag for precision meters to read precisely. In this case, perhaps 100K. Which is also approximately the highest impedance of a naked pickup. Read the volts do the math. Now sweep it. Ideally an input impedance is flat at all frequencies. In this case there will be a shelf below ~~20Hz due to the input cap letting-go. There will always be a drop at high frequency because there is capacitance everywhere. In fact we usually "need" capacitance on a guitar pickup because the winding was tuned for some length of guitar-cord (that's what "Magic loading" box is for). I would think an indicated impedance over 250K at 6KHz would be plenty-enough for guitar. By circuit inspection, it is doubtful this buffer has "any" input capacitance beyond the small strays of the teeny parts and short traces. (There's ~~1000pFd in the first transistor but that boostraps-away by NFB factor or 1,000X smaller, ~~1pFd.)

[free electron]

0R is of course my stupid mistake. It should be infinity, or actually the scope input Z (1Meg in that case).
Input Z was measured with the probes set to x10 and 10M input Z.
I did the test one more time using a 100k resistor in series and a few frequency points between 20Hz and 20kHz:

[Gus]

Nice thread about R.G.'s circuit

Good to read about free electron's build

nice screenshots

[PRR]

The dip at 250Hz and rise at 5KHz are odd.

Since they fairly small, and smaller than typical source impedance, I say "don't care".

Perhaps if R.G. gets curious and has the sim warmed up he might find insight. I would fiddle the cap from upper leg to current source base, see if that forms a low-low-low-Q resonance. (A similar 2-pole in the Dynaco preamp causes a subsonic bump, never seen in days of 20Hz oscillators.)

Capacitive loading is small, entirely moot for guitar, and in-line with expected bradboard strays. An ultra-neat compact installation next to the pickup would show less buffer capacitance, but there's 200pFd-600pFd of stray capacitance in any large winding (pickup) so 20p or 10pFd in the buffer is insignificant.

[sniper1rfa]

For those perusing the internet and stumbling across this, I had a run of these PCB's made. have 125pcs available if anybody wants a couple chucked in the mail for them.



Also, if anybody has any idea what a currently-available BJT might drop in, that would be helpful. I assume the MMBT5088LT1G and  MMBT5087LT1G, but I don't know much about the subject. 

[R.G.]

Surface mount bipolars all seem to have the same pinout. That was one of the pitfalls of substituting through hole transistors. If it's in the MMBT line it's probably OK. The MMBT5088 I found on Mouser does have the same pinout, as does the MMBT5087. The current mirror transistors can be 5088s as well.

Actually getting low noise out of it would also require that the bias resistor at the base of the input transistor have low excess noise, and it's probably a good idea to use low noise types for the resistors associated with the PNP and the output of the current source.

However, the generic MMBT3904/3906 types would probably be good enough for rock n roll.