Max, the largest single that has worked well on the 4bt is the Super HX30W. That one has a 46mm inducer and a 6cc turbine housing. The standard 44mm HX30W also works well.Even the little HE221W has been a good choice but is a little harder to adapt. We've had members try just about every other larger Holset commonly known. Hx35W, HY35W, and HE351ve. The results tended to be the similar. Boost was slow to come on, adding more fuel to try to boost quicker usually caused EGT's to rise, and the power was mostly concentrated at the higher RPM level.
When the 351ve was tried, was it controlled electronically or pneumatically? Pneumatic control (i.e. using a wastegate actuator connected to the VGT control arm) just doesn't give the kind of control needed for rapid spool without either excess restriction, overspeed, or control overshoot. Control needs to be done electronically using throttle position, boost pressure, drive pressure, atmospheric pressure, and turbine RPM as inputs. Its also desirable to have intake air temp and manifold air temp in the algorithm, but not absolutely necessary. Really, you can get by with just throttle position, boost pressure and turbine RPM, but the extra input variables allow more refinement in the algorithm. Even with all that it is still possible that a 4BT simply cannot generate enough gas to rapidly spool the 351ve at it's smallest usable nozzle size. Below a certain nozzle size it simply doesn't have enough turbine surface exposed to the exhaust stream to drive the compressor faster against it's load, and it just acts as an exhaust brake. If that is the case, then the HE200vg would definitely be the VGT turbo to use, although it won't be capable of making as much power. If thats the case, then one could add an HX40 as a LP turbo in a twin setup.
Both BW and Garrett gave variable geometry turbos as well, using moveable vanes instead of a sliding nozzle. They may have VNT models better suited to a 3.9L diesel vs Holset. I feel Holset VGT's have the edge on durability over Garrett and BW though.
The member who worked with the HE series is Steed. Even though some aftermarket companies had controllers for the HE turbo, he built his own and it is probably light years beyond the other guys. The guy is just a computer expert and he figured out what makes the control part of that turbo function. I think he also has a variable control for the P7100. We haven't heard too much from him lately. Being in the military they get top priority. His more recent build is an ISBe 4bt. Not sure if he's done with that or not.
I've checked out his stuff. He's done a good job at figuring out how to control the turbo and seems to know what he's doing, controller-wise. If not for my background in microcontroller, DSP, and analog systems (silicon-level, but had to do board level design and software by necessity) I would probably use his controller. My current experimental controller is ARM based, and while it will initially be focused on running the turbo, I am working on full blown ECM capabilities. The eventual ideal goal is to have a 2-lane FADEC with complete electronic control over the P7100, although I will probably just punt and control everything electronically except the throttle (no drive-by-wire, and hence not full-authority) to eliminate the need for electronic redundancy while still keeping the system fail-safe. In the latter case the controller could go up in smoke and the engine would still run, and I would still have full mechanical throttle control. I would just lose control of electronic fueling, turbo control, etc. So, it would essentially be a limp mode at that point.
The other thing the ECM would be able to do, either in conjunction with engine control or as a stand-alone interface between a Cummins/Bosch ECM running the J1939 network, is be able to interface with an Allison ECM using GM software and GMLAN CAN. That way I could run an Allison TCM with a Duramax cal in it that requires GMLAN messages for operation, rather than my previous method of using an 8.1L cal that allowed PWM control. This is necessary to run 2010-up Allison transmissions that have variable line pressure, as those TCM cals were only used in DMax installs with GMLAN, and the 8.1 TCM can't run the variable line pressure valve body.
I wish some manufacturer made an automotive/industrial version of the ESA's LEON processor. LEON is a SPARC-based processor, but with built in fault tolerance and error checking and correcting features designed to be fail-operable rather than just fail-safe. It was designed for space applications (satellites, probes) and so is made on SOI and put in special packaging to make it rad-hard. This, plus the necessary validation for a space part and the absolutely tiny volume that is actually manufactured, also makes the available space-rated versions VERY expensive! A non-rad-hard, non-space validated part with an automotive or industrial temp rating would be outstanding for low-cost fail-operable embedded controllers. Not to mention that I prefer SPARC over ARM, MIPS, and even RISC-V, but that's a whole 'nother discussion. I think I've hijacked this thread bad enough