On Jul 14, 2024, at 05:14, sergio_logic via groups.io <sergiu757@...> wrote:

?My experiments with the JFET-PNP topology (used in non-CORE DPA lavs, as discussed in this thread, and this Audio-Technica ) prompted me to make a replacement PCB (10mm diameter circle) for Shure WL185/184/183 using this topology.

My design goals were: must fit in the original WL185 lavalier body, must have similar or better tolerance to RF interference, must minimize self noise.

The original PCB is a 10mm diameter circle with an outer ring as the ground contact, and a “leaf spring” contact for the capsule. The leaf spring has a large footprint and is mounted thru hole.

<image8.png>

This is the design I settled on.

<image11.png>

Using a SMD pogo pin (Mill-Max 0900-2-15-20-75-14-11-0) instead of the leaf spring frees up a lot of board space, which means I was able to cram everything onto the tiny board. For the JFET I used OnSemi 2SK3557 (almost identical to 2SK2394), and for the PNP I chose OnSemi MMBT5087L. The ferrite beads are both TDK MAF1005GAD352AT000.

<image13.png>

<image9.png>

I ordered the boards both on Elecrow and PCBWay (PCBWay has better quality, but Elecrow has way faster shipping to where I live). The way Elecrow routed out the boards, there is a nub remaining, which I filed down:

<image1.png>

After filing, the board fits the chassis nicely and you can see I was conservative with the size of the outer ring:

<image5.png>

<image14.png>

R2 and R3 must be matched for the JFET, so step one is measuring the JFET that will populate the board in order to select R2 and R3.

<image10.png>

<image3.png>

Alligator clips are the wrong tool for the job, they make intermittent contact and may deform the SMD leads. I should’ve used test hook clips, but since I don’t have any (yet), this got the job done. I measured the voltage between drain and ground with two different value resistors between these two nodes: 10K and 20K. Based on this, I calculated R2=18K for 50uA and R1=16K for 1.7V at the source. Measuring Idss (multimeter in amp mode and no resistor) and VGSoff (multimeter in voltage mode and no resistor, so the resistor is the ~10Meg impedance of the multimeter) and then applying the JFET equation is not as reliable because the exponent in the JFET equation is not really ? (at least not for 2SK3557 anyway). I got better results by actually using values around the expected range, extracting VGSoff and IDSs based on these (they won’t match the directly measured ones) and using these to calculate the resistor.
Once this is done, the top side of the board can be assembled. I used a hot plate (cheap USB Power Delivery one from AliExpress) for reflow, with no solder paste stencil, just using a toothpick to place solder paste on each pad. (Yes, the amount I ended up applying is excessive on some pads.)

<image7.png>

On the back side, I assembled the components one by one with solder paste and a soldering iron. I place a dab of paste on a pad, hold the component with tweezers, solder the pad with the iron. Then I can apply paste to the other pad (or pads, in the case of the two transistors), and finally revisit the first pad if necessary. After cleaning with isopropanol, this is what the board looks like.

<image12.png>

1 Eurocent Coin for scale:

<image4.png>

Adding the wire and wrapping it up again:

<image2.png>

<image6.png>

I need to apologize in advance for the quality of my measurements, I don’t really have a good setup for this. Really the best setup would be using wired preamps into an XLR interface, I don’t have either of those. I mostly use Sony UWP-D transmitters, but measuring noise through a compander is unreliable. I also have a Rode Filmmaker Kit (2.4GHz) in which I’ve rewired the connector to match Sony UWP (because it’s a back-up for those). It’s a digital line with a 24-bit ADC and DAC. Connecting the Rode RX to a Sony A7s III (again, far from ideal), I got:

• 5-6 dB higher signal from the new PCB (likely because of the <100 Ohm output impedance vs. ~1800 Ohm of the original)
• same noise floor after adjusting for the gain difference
• better RF suppression (there is a slight RF demodulation noise at 3.2 kHz with the Rode and the original PCB, it’s absent with the new PCB).

A vanishingly small fraction of the signal gain could also be because the JFET-PNP topology bootstraps both Cgs and Cgd, whereas the original source follower only bootstraps Cgs. I estimate the capsule at around 10pF, and the original transistor probably has <1pF Cgd, which means half a dB of attenuation, which is nearly nothing.

THD is probably better because of gate bias using a 1G resistor instead of the back-to-back diodes of the original JFET (likely Toshiba TTK101MFV).

Is all this worth hand soldering 0402 resistors by hand? Probably not, but it sure was fun. I am happy with the 6dB, they were always a bit quiet as source followers for the Sony UWPs I’m using. I have one wired as a common source instead because of this.

Maybe it would be worth switching to this PCB for applications where you are using these or similar mics at high SPL (I’ve seen people use them as instrument mics) with a two-wire topology transmitter like the Sennheiser EW, which precludes a source follower and forces you to wire the Shure as a common source amplifier (with worse THD at the same SPL).

?