QCW build thread (inc. PCB documentation)

Hydron, Tue Jan 17 2017, 03:55PM

In contrast to my normal method (build something without notes or photographs, get confused later), I've decided to have a go at documenting my QCW project as I build it. Hopefully this will both help others with their builds, and me when I come back and try and figure out what the heck I did!

Basic primary/secondary specs:
- Secondary 110mm dia, 285mm wound length, ~800 turns (guess!) of 0.315mm wire
- 300x75mm spun toroid
- Primary of 9.5mm dia tube, 10 turns, "solenoid" wound on 160mm dia. former. With strike rail, but may remove this later.
- Coupling of ~0.4-0.5 (some adjustment possible)
- 24.8nF primary cap (gives relatively low tank impedance, hopefully works out OK!)

With the above setup (shown below with original ducting toroid, subsequently upgraded to spun), we get the following:
- Primary and secondary resonance both ~260-280kHz
- Upper pole of coupled coils lies at ~360kHz (see graph for measurement of primary impedance with secondary present). If this is not high enough once the coil is loaded by streamers, I will need to change the primary/MMC or secondary (or both) to raise the frequency.




Drive details:
- Initially planning to use phase-shift modulation, will switch to buck modulator if unsuccessful or if I need to push more power than hard-switching will allow.
- MMC split into 8 sections, 4 in parallel on each side of primary coil (see MSPAINT drawing, but ignore the incorrect MMC spec and CT location)
- Each MMC section driven by an independent half bridge. This guarantees reasonable sharing of current and switching losses between parallel bridges.
- 8 half bridge modules on individual PCBs, each with 2x FGH40N60SMD IGBT. May upgrade to FGH60N60SMD (mainly for significantly lower junction-case thermal impedance).
- Each IGBT on an individual heatsink to eliminate insulation pads and their relatively large thermal impedance. Main reason for this is to lower the transient thermal impedance of the system, and thus the delta-temperature seen by the IGBT die during each pulse.




MMC details:
- 24.8nF formed of 8 sections of 12.4nF (2s4p). Each section on individual PCB.
- Each section uses 32x 6.2nF/1600V B32672 TDK EPCOS capacitors in 8p4s configuration for 12.4nF/6.4kV
- 12.8kV DC rating (AC peak rating at low duty cycle is similar - caps are specified for greater than nominal voltage operation) for full MMC
- Current rated at >30A RMS, >600Apk @ ~350kHz
- 256 capacitors total, with 512 bleed resistors (_NOT_ looking forward to soldering this). Caps were cheap though, <40 GBP total.

Gate drive and control:
- Will use a gate drive PCB located under the half bridge modules
- This PCB takes logic-level gate drive signals (will upgrade to RS485 for noise immunity) over shielded ethernet cables, and drives 2x gate drive transformers located on the PCB.
- These transformers have a low inductance connection to the half-bridge PCBs via ribbon cable
- Also takes in outputs from 50:1 commercial CTs located on half of the bridge PCBs (CT outputs connected in parallel to add current together). These are run into a second set of 3 CTs (in series) to give an overall 1000:1 reduction with 3 isolated outputs (for OCD, feedback and monitoring). This is experimental, and may be too high a step-down ratio for a QCW design with it's relatively low current, but is the best I can do with commercial CTs. By using 4 separate CTs and keeping them on the bridge outputs rather than one CT on a primary coil connection we avoid the need to insulate the CT from high voltage (present on both ends of the primary in this design).
- For driving 8 half-bridges, requires a second "slave" PCB with 2 more GDTs on it. This is simply a sub-equiped version of the first PCB, with connection between the two via ribbon cable.
- Gate drive control via a separate PCB, un-designed at this point, but will likely use a combination of a microcontroller and CPLD/FPGA, or a PSoc.
- Driver feature wish-list (i.e. will likely never get around to a lot of these!):
* Phase shift or buck modulator support
* Freewheeling (pulse-skipping) support
* Softstart/phase controlled rectifier support
* PFC voltage/current setpoint control (if I ever build a PFC)
* Control via optical/bluetooth serial link/protocol, rather than the simple interrupter type control
* Readback of values to handheld controller, e.g. peak current, bus voltage etc
* Some sort of MIDI/music support to piss people off with Nyan Cat midis

PCBs:
To do all of the stuff above I intend to build most of the electronics on properly made PCBs. I've documented and release the first 3 that I've designed in the following github repositories: .
- Each has a readme file that should be looked at, and schematics/renders in the "Reference Outputs" folder
- I have concerns over the clearance distances, especially on the GDT PCB. I take no reponsibility if anyone copies what I've done and doesn't treat everything (even low voltage stuff) as live!
- The GDT PCB in particular needs a new revision to be made.
- All 3 are being fabbed right now, so are completely untested as of now!
- Also see thread here: for an earlier revision of the IGBT bridge PCB (probably more generally useful, as allows for more heatsink options).

Bridge PCB renders:



Gate drive PCB renders:



MMC PCB render: