Збирач потоків

Took the laser PCB process a bit further and pushed this one to a fully working board.

The vias are drilled with a sub-mm bit and stitched manually with wire to tie the planes together. It’s basically sewing the board to keep the return path tight.

Main goal here was reducing loop inductance as much as possible since this is driving a SiC switching stage.

Not trying to replace fab boards, but for fast iteration this is actually way more capable than I expected.

Still experimenting with how far this approach can go in terms of switching performance vs a proper manufactured board.

Hey, I built a Bluetooth audio amp based around the TPA3110. The QCC5125 uses differential audio signals for the TPA. I had to cut some ground lanes on the PCB for it to work because those cheap TPA boards use the same ground. USB trigger board for a 12V linear reg to an isolated 5V converter. Works really good; I only hear a quiet noise about 2-3cm in front of the speaker. I have 2x 15W 4-ohm speakers. What do you think? Anything to add, or just finish it with a case?

I wish I had been smart and careful enough not to use a freerouter.

I'm honestly so happy I finally got it to work. I'm new to his but I love figuring out the ins and outs.

Not sure how you all feel about LLMs, but I have been letting Claude get involved in my electronics projects recently.

After a great deal of experimentation using Rigol DS1054Z over LAN with Claude Code, I've built a small MCP server that lets Claude fully drive the scope.

It exposes things like:

• configuring channels / trigger / timebase
• running measurements (Vpp, frequency, etc.)
• downloading waveform data with basic heuristics for Claude to plan around
• capturing screenshots that Claude can view

It's just SCPI with Python, but wrapping it as MCP tools makes it surprisingly usable for iterative diagnosis!

Repo: https://github.com/erebusnz/rigol-mcp

Curious if anyone else is using Claude for their electrical engineering. Claude is surprisingly good.

Example use:

Here's a real example with a frequency generator + LCR connected to Channel 2 but with Channel 1 set up (incorrectly). Claude was able to get the waveform visible and correctly guess what it was seeing.

using my rigol scope figure out how to get a clean measurement

● rigol.idn RIGOL TECHNOLOGIES,DS1104Z,...

● rigol.get_scope_state CH1 active, 10x probe, 1V/div, DC Scope STOPped Trigger level: 0V Channel offset: ~1.68V

→ Trigger mismatch → unstable/dirty capture

→ Running autoscale...

● rigol.autoscale

→ Signal detected on CH2 (not CH1) → Trigger locked (TD) on CH2 ~0V

● rigol.measure(CH2, FREQUENCY) 1.298701e+06 Hz

● rigol.measure(CH2, VPP) 5.08 V

● rigol.measure(CH2, VRMS) 0.723 V

● rigol.screenshot [captured]

✓ Clean measurement achieved

Results: Frequency: ~1.30 MHz Vpp: ~5.08 V Vrms: ~721 mV

Waveform: Step response with damped ringing → square wave exciting an LC resonance

Since i'm planning to build a bigger batch of USBpwrMe i actually need to test each unit in a fast and repeatable way. Therefore i have designed a test jig that will measure all functions.

There are 2 voltage regulators that will supply the test jig itself with 5V but also a 6V regulator to be able to make a test of an over voltage circuit with a threshold of 5.6-5.7V.

INA139 will monitor the current of the DUT thru a shunt of 0.5 ohm or less. This will be optimized depending on what the DUT will actually consume.

On the test jig board a PIC Mcu will control and manage the whole test and test instructions and results will be presented on a 2x16lcd display. The test is not high tech but the DUT must be manipulated with external resistors and voltages to be tested. This is mostly handled by 3 relays.

Connection to the DUT will be easy using the banana connectors and the USB outputs which has corresponding mating connectors on the test jig.

Following steps will be performed

1 It will measure the current consumption of the board to see if there is excessive power consumption

2 It will change polarity on the DUT and measure if there is any voltage on the output.

3 It will will apply resistors on the D+ and D- lines och the USB-A connector and measure so that expected voltage appears.

4 It will apply resistors on the CC1 and CC2 line for the USB-C connector. Vbus1, Vbus2, CC1 and CC2 are measured. If negotiation is correct it will enable Vbus.

5 It will change input voltage from 5V to 6V and test so that the OVP protection works.

6 Finally it will test the OVP mode switch by telling user to turn of OVP. And measures that Vbus goes on.

The test will hopefully test a unit under 5s.

The Gerber files are already sent to manufacturer and are in production. Now you might wonder why a choose a to small board that won't fit the display. Well at first i did. And when i uploaded the gerbers files it was around 40Usd to get it manufactured and shipped. By reducing the height of the board with 3cm the cost was 12Usd. Since it's only a testjigg and will be put into a casing i rather save some money!!!

The PCB has 4 layer stack up. Not really needed but it's much easier to route the signals and takes less time. The schematic and routing took around 5hours.

Funny thing is that the test jig is way more advanced than the product it is itended to test :) :)