EBMX X-9000 copper copper copper EVERYWHERE!!!!

This thing is so gorgeous! GREAT JOB EBMX!!!

[@de-bodgery]

Marshall Scholz stated this and I'm pinning it here after a little editing for readability of his points. First - the GD32F405 is functionally equivalent to the stm32f405. it has 1mb of flash and will run lisp and all the other functionalities of vesc just fine. The difference comes in that only half the 1mb of flash can be accessed fast for machine code use. Since vesc only uses the first 512mb for that anyway this configuration is fine and will have absolutely no impact on the functionality of this vesc. *** - This bulk capacitance configuration is fine. It is close to the tightest hot loop you can get with this configuration and far better than many other controllers. with only a 15khz pwm frequency the hot loop doesn't need to be particularly tight anyways. This configuration is very common on other high power motor controllers and is present on kelly, siemens and ABB motor drives. (That i know of) The amount of bulk capacitance looks more than adequate as well. With 5 paralell mosfets this controller also probably switches fairly slow. which limits the need to a tight hot loop and a lot of ceramic capacitors. *** - The battery current will ALWAYS be less than the motor phase current. the battery current being limited by the input current path and capacitance. Since ripple current is worst at 50% duty cycle, the Capacitive ripple current is worst at 50% duty cycle when you have 450A battery with 900A phase out. This is likely the source of the 400A battery input limitation. This controller can definitely put out 900 phase amps peak with only cooling limiting what it can do continuous. *** Trampa had the ability to produce a controller like this if they wanted. But instead they don't have anything waterproof or anywhere near this amount of power. My only gripe about their advertising is their amount of voltage margin spec'd vs what they have in the parts. Otherwise this is a truly innovative controller that will hopefully give people some of the best ebikes they have ever ridden.

[@armyca]

Good review

[@MrDehicka]

Thanks for the teardown! Here is my little grains of salt on the design: 1. They went with ST Chinese knockoff MC (Gigadevice GD32F405). It is not fully compatible with original ST's STM32F405. They have different timings e.t.c. Unmodified VESC compiled for original STM32F405 will not run on it smoothly without porting the firmware specifically for Gigadevice part. So no opensource VESC firmware on it until they release their source code. 2. They took known Chinese design power board with all those nicely soldered and screwed copper bus bars, but added phase power bridges all the way across the controller. Which limited the space for the bulk caps and added an extra screw connection on each phase just for the benefit to have all the terminals on one side. And those phase bridges are right over and close to low voltage logic. 3. There are some minor errors with thermistors placement and gate drive routing I can see which may (or may not) cause problems at the peak current/power levels. 4. Their specs claim it is 100V input rated, which is misleading. Mosfets itself are 100V abs max! If you connect it to 100V battery the ripple will kill it quite fast! With that little bulk capacitors I won't try it on anything higher than 75-80V. BTW, Flipsky has the same chinease fets and at least they specs it 84V max.

[@christianwagner6244]

nice attention to detail! How much do you think is saved by using a different uC? The VESC-tool shown in EBMX´s dyno run seemed like original, so there is definitely some sort of vesc-app integration. On their website, I think I also read the 84V max or 72V rated battery.

[@de-bodgery]

It is a second party CPU. It did need porting of the VESC code base to run VESC on it, BUT NOW IT DOES!!! And now people have a second CPU they will eventually get to use for VESC platforms. This is good, not bad! It will get to eventually be open source. I've been pushing for that since I first met the EBMX people a little over a month ago. Give it time. They did NOT take down a Chinese design! Good greif! THIS IS NOT A ZSX!!!!! People have been spewing that BS since forever! NEVER HAPPENED. I ASKED! Yes, the ZSX looks similar. Go LOOK at the ZSX power stage instead of making sweeping statements! They are similar in some ways and NOT a copy!!! There are dozens of differences! It is pure coincidence that they are similar. I've got a VESC design I'm a part of. NO connection in any way to EBMX and yet, our power stage is pretty damned similar as well! I know for fact we copied no one. This is spurious BS! I do agree that it needs more ceramic caps. I forgot to say so in the final cut. So thank you for the reminder. What are these minor errors with thermistors and gate drive? 100v max, not possible to run it at 100v. No one claims you can run it at 100v. The mosfet used is also used by Shul, KO and others. It's a decent Chinese mosfet. Not a western part which I'd definitely prefer, but not a crappy Chinese one either. This video deliberately didn't cover the mosfets. Notice the title about copper. I wanted to focus on that and did get off track a little by talking about gate control and thermistors. It happens. The mosfets will get covered, but they are not terrible. EBMX has ran a lot of dyno tests at 37kw on 82v. Never had one die. Still...I'd like more ceramic caps on the power stage.

[@MrDehicka]

​@@christianwagner6244 VESC tool will run without a problem with carefully ported and tested firmware. But you will stuck with the firmware updates from EBMX until they release the port source code with the changes. I use exact same STM32F405 in 64-pin package on some of my board designs at my work. Price is not really a factor. There is bad shortage of ST microcontrollers for almost 2 years and still going.

[@MrDehicka]

@@de-bodgery I Never said mosfets are bad. Just do not throw 100V input voltage in the specs for the controller, it is pure marketing BS. Power board is only slightly modified "ZSX" board I have seen. All the mosfets, ceramic caps, phase terminals and the gate driving BB connectors are on the same exact places! I do PCB design and EMBED programming as my main work. There is zero chance it is a coincidence.

[@frankdrenckhahn9521]

@@de-bodgery When you hold a controller in your hands, you should get the sources if you want to have them. If you paid for the thing and received a unit, the source code offer is nothing you need to beg for. Just request it and you should get the code within a reasonable time frame. It's not Open Source to be hidden away. We often have our code published before we get stock and start selling.

[@helios_ja]

I thought it was stamped aluminium piece. I didnt know it was plated. That's amazing.

[@TudiMPowerService]

Will you make video from the main board or we just we can only see just the FET board? I am interested about the main board. Because we saw the problems on the "competition's" main board too. And maybe will be interesting to teardown a RUSH controller to, because i think the RUSH is this X-9000's equivalent. Oh and i saw the not to nice soldering on the pin connectors. And i also seen a soldering junk too at 10:30 on the video.On the pin connector's bottom left side's first leg.

[@de-bodgery]

Oh foir sure! I'm far from done with the X-9000 tear down. I have many things going at the same time. I'm doing some work for Tronic right now. I need to unsolder the logic board from the power stage of the X-9000 before I do the teardown for that. More things are coming. Rush...lol...working on getting one of those too! I'm already pretty sure it's more of the same nonsense from KO. That middle aluminum plate is another MCPCB with more mosfets on it. They did nothing to improve the existing design and it's poor real estate use, just added more mosfets to an already bad design.

[@TudiMPowerService]

@@de-bodgery I tought that's all we could see from the X-9000. (Maybe about copyright or any reason.) I don't know about RUSH. I never saw any picture from the inside of RUSH. Now i am very interested to see, what is inside, after saw PRO controller. Because 1800A is not a game. So it must be a massive things inside. Not just more mosfets what can heat up foil more faster :D

[@de-bodgery]

@@TudiMPowerService NOPE! More on the X-9000 is coming. Everything you see I showed about the KO pro, that level of exposure, you'll see that for the X-9000 as well...just with way less issues found. Its a matter of time to get the controller apart and do the video, research parts, etc. *** RUSH is going to be shit...or at least that is my prediction. KO doesn't take feedback, they just throw garbage out there and then hype the daylights out of it! They don't dyno anything. They made a big deal about "Starting to dyno", but you see no content on this. That's all the proof I needed that they pull amperage and wattage specs out of their collective asses!

[@de-bodgery]

Comment by @marshall scholz Marshall Scholz (This comment has been copied to the top level thread at @de-bodgery's request to be able to pin it.) There are so many things wrong with this comment. Which for those not in the know, is by Frank of Trampa who is known for bullying other manufacturers of vesc boards. It is also very technically misinformed and incompetent. Representing a fundamental amount of missing knowledge for dealing with motor drives. First: - the GD32F405 is functionally equivalent to the stm32f405. it has 1mb of flash and will run lisp and all the other functionalities of vesc just fine. The difference comes in that only half the 1mb of flash can be accessed fast for machine code use. Since vesc only uses the first 512mb for that anyway this configuration is fine and will have absolutely no impact on the functionality of this vesc. - This bulk capacitance configuration is fine. It is close to the tightest hot loop you can get with this configuration and far better than many other controllers. with only a 15khz pwm frequency the hot loop doesn't need to be particularly tight anyways. This configuration is very common on other high power motor controllers and is present on kelly, siemens and ABB motor drives. (That i know of) The amount of bulk capacitance looks more than adequate as well. With 5 paralell mosfets this controller also probably switches fairly slow. which limits the need to a tight hot loop and a lot of ceramic capacitors. - The battery current will ALWAYS be less than the motor phase current. the battery current being limited by the input current path and capacitance. Since ripple current is worst at 50% duty cycle, the Capacitive ripple current is worst at 50% duty cycle when you have 450A battery with 900A phase out. This is likely the source of the 400A battery input limitation. This controller can definitely put out 900 phase amps peak with only cooling limiting what it can do continuous. Trampa had the ability to produce a controller like this if they wanted. But instead they don't have anything waterproof or anywhere near this amount of power. My only gripe about their advertising is their amount of voltage margin spec'd vs what they have in the parts. Otherwise this is a truly innovative controller that will hopefully give people some of the best ebikes they have ever ridden

[@kennsebesta8464]

You mention the copper blocks are machined from a solid billet. How sure are you on this? There are a couple of things which would surprise me if this were the case. 1) Milling copper is challenging, as it's gummy and ductile. It wears out the tools quickly, and there's a high failure rate. It's also extremely expensive to waste all that copper as chips. 2) It's tough to see from the camera angle, but I don't see fillets between the vertical tubes and horizontal planes. If it were indeed from a solid billet, that's bad machining practices as it creates an unnecessary stress riser. And when I zoom in on some views I swear I can see a faint parting line, so from here I don't see strong evidence of solid billet machining. However, another explanation might be a better fit for what I think I see: they press-fit two separate parts together. The press-fit would be just as good electrically and mechanically, as with the high pressures of a proper press-fit the copper will literally cold-weld itself back together in a short period of time. So the resulting joint is molecularly indistinguishable from the identical milled dimensions, but a whole lot easier and cheaper to make. Could you take a closer look at that joint and let us know if you're positive it's from a single billet?

[@de-bodgery]

As a hobby level machinist, I've machined copper, it's not fun, but doable for the very reasons you mention. Since the vertical sections don't see side to side motion, I doubt those sharp corners will be a problem as stress risers. If this was seeing lateral loading sure that's a problem, but it's not. Never mind the potting and the shell lid also supporting them from lateral loads. I looked pretty close too as I thought maybe there's a parting line between the vertical sections and the blocks. Just now, I looked at each "joint" with 35X magnification. If they are soldered together, they did literally perfect soldering! There is no over flow anywhere, no blobs, no extra solder, no bridging between the 2 parts, nothing! I've been soldering for 30 years. I consider myself to be pretty good at it. I can't solder that perfectly...not ever! IMHO, there are a few places where there "might be" a joint line, but it is seriously subtle! If they are soldered sections or press fit, I'd be astonished! Press fit would leave a joint line. There isn't one. Copper because it's so ductile doesn't press fit very easily. Even with lots of heat on the block and cold on the cylinder, I think there would still be a joint line and I see nothing anywhere. I seriously doubt they are press fit either. I still think they are solid blocks of copper.

[@kennsebesta8464]

@@de-bodgery Thanks for the deeper dive! It's a little strange to me. When I CNC mill something, I fillet the corner. It's effectively free in CAD, and it's almost free in manufacturing costs as well. Using the 3mm BEM to fillet the three tubes would only add 60 seconds per part, but maybe they're machining on a mill without a tool-changer? That would make sense if they're doing this in-house. It'd be very easy to do an interference fit with the tube frozen in dry ice or LN2 and the block heated in an oven. They'd slide in with no effort whatsoever, and would compression weld in place one they warmed up. But I think that would leave a parting line, no matter what, that's not what you are seeing. One thing which would be conclusive is to look at the machining marks on the tube. If there are cutting marks, then it must have been milled round, as a normal copper tube would have extruded. And if there are stretch marks, then it was extruded. And if there are no visible marks, well, then I guess it doesn't say anything one way or the other!

[@de-bodgery]

@@kennsebesta8464 I have a thought and it would totally explain no soldering or a parting line. The parts are solid copper and then plated. This could completely obscure any line between the parts. If that's the case, then the vertical section and the block could be separate parts and press fit together. The cylindrical portions do have radial markings on them like possibly turned on a lathe. It doesn't look like side cutting milling marks to me, but I could be wrong about that.

[@kennsebesta8464]

@@de-bodgery I think you're onto something there. The plating would have occurred after any manufacturing, and it would hide parting lines (as well as machining marks). Any way you do it, it's well built! Thanks for letting us peer inside.

[@de-bodgery]

@@kennsebesta8464 More "peering" is coming! I held off doing more video until I had a gymbol. I was not liking the swaying and other movement and never have! Anyway, I got a DJI Ronin SC and it arrived last night. I figured out how to make it work, changed settings so it operated like I want and then did several test videos with it. MUCH BETTER! Time for more videos! Only thing I don't like about it is I can't fixture the gymbol for macro video...or at least haven't figured out how yet. I may continue doing that on a standard tripod for now. I'll be working on some Tronic and EBMX content over the next few evenings!

[@paulagibmad4807]

And no you not sponsored by ebmx... all controllers are bad exept ebmx. xddd

[@de-bodgery]

Did you watch the video? And NO I am NOT sponsored by EBMX in any way. I bought an X-9000!!! I DID, MY MONEY! They did lend me 2 incomplete controllers to do these tear downs, but then they go back to EBMX. That's as close to "sponsored" as it gets!

[@de-bodgery]

Did you watch my tear down of the ASI BAC8000? How about the Nucular 24 fet? How about the Trampa 100-250? DID I say they are bad? If something is shit, I say so. If some is good, I say so! END OF STORY!

[@smelker]

Did you look at the videos at all? He's just explaining exactly what he looks at in an easy explanation. Think he does it really well from an electrical engineer perspective.

[@joem8601]

@@de-bodgery In my experience the Nucular 24f performance is crap compared to the BAC4000 Why does the BAC4000 perform much better than most controllers?

[@de-bodgery]

@@joem8601 ASI has been doing FOC for a long time now. They have it worked out pretty well. Nukular has been around since 2016 and released their first product in 2018. They are fairly new to FOC. They also had some sort of internal issues where they did no new development from October 2019 until about 15 months later. whatever it was that stalled firmware clearly slowed them down for a while. Since then, they have been pretty good at releasing firmware updates every few months. They have already come a looooong way and done it far faster than ASI. Assuming they survive all the trade restrictions on Russian products, Nucular will only get better and better. Meanwhile, ASI isn't really doing much that is new. There's rumors of a 150v controller form them, but that has been a rumor for a long time. Also ASI is a bunch of control freaks. You buy into their products with the payload of the ASI overlords! I believe the BAC4000 is 24 fet and so is the Nuk 24 fet. They ought to be pretty close to the same power wise. I can't speak for your tune on the Nucular controller. That can cause poor performance. I've messed with PIDs on them to improve motor function.