4 Layers is Better than Two: Another ELS Control Panel Prototype

[@johnsickmann9224]

Hi James, When doing another release of the boards you might want to consider placing large pads on the internal layer where the screws pass through the board. These pads should be large enough to encompass the size of a washer/spring washer/star washer (plus a little more. This will ensure that the mounting fastener when tightened ( which may cut through the pre-preg ) does not create an undesired connection/short to the internal layer. I have been caught by this in the passed. Great work and love your videos.

[@Gaunt596]

that does help, but when mounting PCB board to enclosures, you almost always want at least a flat washer between the mounting fastener and the board, even with the pad, to prevent damage to the solder mask, and to spread out the clamping forces to prevent cracking.

[@dirtdart81]

@@Gaunt596 if you're cracking FR4 with screw pressure you're tightening way too hard

[@johnsickmann9224]

@@dirtdart81 I agreed but you need to allow for the end user you are not in control of (wink wink). This is more a be safe precaution. Even a flat washer can have sharp edges and it is easy to have that side in contact with the PCB. More a suggestion to be aware of this issue was I have been caught-out with this and seen it many times while servicing equipment in the past.

[@Rob_65]

Even better, use plated mounting holes with larger copper on the top and bottom of the board. Sometimes libraries (Altium for sure) have different types of mounting holes that also have a "keep out" area so no traces will be routed close to the mounting hole.

[@orangetruckman]

Just more proof that your dedication to quality is on the high end! Your efforts are appreciated sir.

[@zmast333]

Thank you for the link to Hartley's video. I watched it a year ago and then forgot how it was called. Brilliant introduction to EMC.

[@doulos5322]

dude my respect for you just went up 10 fold for mentioning Rick Hartley EMI Video. I'm glad you caught the caps that was going to be my first suggestion "filtering out the high frequency from the signals and the other was on grounding the incoming cable on the left if it has ground or if the connector shield has a ground with a "360 degree ground" termination. why that don't help with EMI rejection it does help eliminate external factors that can make the EMI worse.

[@TommyBoy7Heads]

Hey, today I learned a new way of thinking about EMI and ground planes! Thanks for the new rabbit hole to go down and explore.

[@HM-Projects]

Howard Johnson's book on high speed digitial design has an excellent section on layer stacking and EMI. Excellent reading, recommend.

[@chrisarmstrong8198]

Rick Hartley's video is terrific. Thanks for linking to it.

[@TyroneDamShewlaces]

Your implementation is pretty awesome in my opinion. I'm a long-time machinist and my machines have no need for this, but my nephew is just now learning a bit of machining and just purchased a small import lathe. I have pointed him to the playlist & wiki and expect him to adopt your version for his machine. From what I've seen, your implementation is hands-down the best design out there. Sure seems like a great way to take a cumbersome lathe and make it a pleasure to use.

[@liampollard2908]

Well done! I'm an electrical engineer specialising in micros, and all this makes sense; my initial training was in radio, and we basically have radio circuiots when dealinjg with micros. My favourite micro is the Rasapberry Pi series, which I have usd in a Railroad track parameter monitor which is used to inspect track and the 5 parameters that are associated with itr. It's all fun!!

[@Milosz_Ostrow]

Having dealt with high speed circuit board and flex circuit design in the disk drive industry for many years, I can say your explanation of the design changes and selection of a multilayer board is all correct. The only other way to fix the radiation problem might have been to install R-C snubbers in the the high current, fast-edge signal lines, which would have greatly increased the component count and might have been a wash with the cost of the multilayer PCB. The actual clock rates in the control panel are probably not very high, but the fast edges produced by the LED driver IC are the problem and that's what makes this a "high speed" circuit.

[@truthkeeper1945]

one of the best creators

[@staviq]

10:20 Those small components will pivot between the tweezers, and the harder you squeeze the worse it gets. I've done a lot of hand soldering including 0201 smd components. This is what you want to do. You want tweezers with a slightly wider tips ( basically barely enough so you can no longer call them pointy ) , and you *always* rotate the board to fit the way you are holding the component, because rotating the component is quite counterproductive as you noticed :) Put the component somewhere close on the board, and then, adjust your grip with the tweezers. Ideally you want to grab the component by it's top half without resting the tweezers on the board itself, because that inverts the pivoting behavior. This alone makes it waaaay easier to solder Some general tips for hand soldering smd: Use way more flux. Don't try making them straight or centered, even if one end ends up hanging in the air a bit, as long as the solder connects it to the pad. The only way to make them straight without loosing your mind is to reflow with hot air later. Trying to get them straight just overheats the pads and the components and causes stress in the joints. There is no such thing as too much flux Don't be afraid of adding too much solder and bridging pads accidentally. More solder distributes heat better and cools slower reducing joint stress, especially when you solder one end first and then the other, more solder wil keep the first end hotter for longer and while doing the other end you make the component hot evenly. You then remove excess with the wick, after things cool down for a moment. There is no such thing as too much flux :) If you cant get a pad done in couple of seconds, stop, wait couple of seconds and add new flux. Working for too long in one spot burns the flux and oxidizes the solder. More flux If you want to bodge a component directly to another, pre tin the end of the new component. Right on the board before positioning it, dip in flux and touch with the iron. If the end of the existing component isn't shiny, touch it up too until it is. Did I mention flux yet ? Also, you can use thicker flux to hold the components down, like a glue. Yeah, that would be it pretty much, the rest is just in developing muscle memory and visual intuition.

[@richardgardiner242]

I have an original board. I noticed flickering - but not by eye, it was only apparent when captured on video. I replaced the cable from the control box with higher quality cable with better shielding. The ELS is now working perfectly👍

[@barryomahony4983]

11:00 Makes me glad I went "old school" with a recent small PCB of mine and used through hole. They do seem to be getting increasingly hard to find, especially IC's. And yes, nut drivers are awesome.

[@RB-yq7qv]

Hi James Interesting comments posted development of any product is time and money demanding and any real assistance will assist us all. It's amazing how many smart people we have that show a real help quality. Happy days ahead for the project

[@floodo1]

Ain’t nothin like RC style nut drivers and Allen wrenches. Really gives the “machine screw” feel

[@joeldriver381]

Looking forward to trying a couple of these. Nice work!

[@MattOGormanSmith]

Are C1 & C2 the decoupling caps for the driver chip? Are they big enough to supply the current spikes for all those LEDS? If not, there may be HF noise induced in the power line which might affect the data lines in the cable. I know I sound like Capt. Obvious but I was surprised not to see an electrolytic as the main tank, paired with a smaller low-ESR one.

[@craigs5212]

That was my thought as well. Those led's draw a lot of switching current which gos through the ground pin on the driver. The charge needs to be replaced from the bypass caps stored so it doesn't get reflected down the input cable. I think he was likely seeing ground bounce on the input reference for those control signals, the 4 layer would help this out. But a nice fat electrolytic at the board edge would likely do some good.

[@TheUncleRuckus]

Agreed Nut Drivers are awesome and a must have for any kind of Electrical work! Great video as always James, thank you for uploading. 👍👍

[@bobuk5722]

Hi James, thanks for the very thorough approach. i seem to remember that once upon a time we just called interference suppression 'screening'! Seems we now have a better idea of why it worked!

[@craigs5212]

FYI You could have placed the caps over top of the pull up resistors. From a signal standpoint it doesn't matter if the cap is connected to VCC or ground . (assuming you have bypassed the supply rail).

[@Thelemorf]

Will place an order as soon as my hobby budget allows it!

[@ahbushnell1]

Did I miss the schematic? On the video I finally saw the chip. The power lead does not have a bypass capacitor. That's board 101. Normally every chip power has a 0.1 uF cap. Good video!!

[@OmegaSparky]

I see a 10uF and a .1uF cap. Since the led driver has pretty large power transients - I would suggest adding a second .1uF and place one each right next to each of the two VDD pins. You can see the voltage transients in the signal waveforms. These caps will probably help with that. But, I don't think you needed to go to a 4 layer board. You had room to run a ground return path on layer 1 next to the SPI bus signals. Also

[@rayoflight62]

The problem with RFI. I always solved those problems with low impedance designs. People think that signal lines work with μA currents - no. That makes the signal lines into receiving antennas, and at first PN junction anywhere in the circuit branch, the RF signal is rectified - and counted...

[@tonyurquhart8278]

"so you get to see what its like messing around with small components, lead free solder & LIMITED SKILLS"... Oh! James you are such a comedian!!! Hahahaha! But seriously, I really admire your ability to clearly communicate complicated subjects in a way that is understood by laymen, & your skills overall are superb. Keep up the comed... I mean 'good work'. :)

[@button-puncher]

Nice work! I love the Milwaukee M4 screwdriver for electronics work. I'm a former electronics tech. It has a great clutch and ergonomics plus swappable batteries. I run a double conversion UPS in my shop dedicated for electronics. Normally they are expensive but they can be found of the Bay of e with bad batteries for pretty cheap. An APC RT, Eaton 5PS, or Sola S4K. The Sola is what I use, only thing that can handle the frequency swings running on my old Onan generator.

[@joemcgarry1106]

I would be happy to have your soldering skills. Mine are awful. As usual another stellar video.

[@LIKEABOSSTOKE]

I suggest that you use flux in the form of a syringe. It usually is "no clean flux", which means you don't need to clean your board after soldering. (You do need to make the flux "flow" completely though, since "unflown" flux can be a little bit conductive). In case you want to clean your board anyway, IPA with an acid brush with short(ened) hair does quick work of it. Note that lead free solder melts about 50 degrees higher than standard SnPb solder. For tests and reworks, it is much easier to work with the latter, unless you use a much hotter soldering iron. Note that the "glue" holding the copper on your PCB melts at around 700 degrees F, which means that heating a pad for too long with an iron hotter than 700 degrees will make the pads come off your board. Always use tweezers to hold parts. When reheating a solder to add a second part, like you did in your rework, you can melt the solder and place the part without caring about the quality of the solder, as long as the part sticks. You can then come back with flux and reflow the solder to make a solid joint. This allows to work around needing 3 hands to hold the iron, the tweezers and the soldering wire. Lastly, flux is like lube in many cases... You can hardly have too much. Hope this helps.

[@Rorschach1024]

I personally detest lead free solder for many reasons but the very high melting point and the narrow liquidus/solubility window makes for very difficult soldering, leading to cold joints. If you get the iron hot enough you pull pads off. Worse yet the solder is brittle and is prone to growing tin whiskers over time. It is crap all the way around.

[@eliduttman315]

"It is crap all the way around." AMEN! A google search will expose my rants against those damned and blasted EU RoHS directives. If your stuff stays out of the EU, stick to 63/37 Sn-Pb eutectic around PCBs. The only use I have for lead free ELECTRICAL solder is 96/04 Sn-Ag in "tinning" bare copper wire and a hot, high wattage, iron is essential for that job. Such an iron WILL damage PCBs.

[@hk74654]

James used those syringes with "no clean flux" earlier in this video series and also showed cleaning them with an ultrasonic bath.

[@danielmatthews8475]

James, I've been watching your channel for a long time, and love your content. I wish I had half the toys in my garage that you have. (my wife already thinks I have too many). Just a few comments on this video. Every thing that you mention about high speed layout in the video is correct. How can you go wrong quoting Eric Bogatin and Rick Hartley. But I'm not sure that layout is the real problem with your initial PCB. The blaringly obvious problem with your first circuit board is that it is missing bypass caps on the power supply. When you hang a switching circuit, like your multiplexed LED driver off the end of several feet of smaller gauge wire, like your wiring harness, there is a significant amount of ground bounce (or call it power bounce) due to the the current pulses that are drawn through the resistance and inductance of the wire as the circuit switches. Since your LEDs are multiplexed they are switching all the time they probably cause quite a bit of noise on IC. The typical way to fix this is to have a 0.1uF cap on every IC's power pins to GND. It is also recommended to have a much larger cap (or two) near the power connector on the remote PCB. probably 10uF to 100uF. This will keep the higher current pulses local to your PCB instead of making them run all the way back to the source for their return. The only high frequency currents you want on the return lines are that of the SPI signals and that current isn't very large. If you have the LED switching currents on those long lines as well (because you didn't have local bypassing) they will interfere with each other. In your original PCB you show the SPI bus lines crossing over a small gap in the ground plane, and while that is not recommended, and the return currents do go around and create inductance, that path is not very long, and the loop isn't very big. If you were to calculate (or simulate) how much inductance that is you would be disappointed. Your SPI frequency isn't very high. From your scope plots it looks like 200 kHz. (which is almost DC) Where these techniques really start to become important is for high speed buses (10MHz and above). Even considering the 10th harmonic of your SPI clock, which would be 2 MHz, what does a 2 MHz antenna look like? They are pretty long. The other thing to consider is how they are measured. I have had so many engineers try to tell me they have signal integrity problems and then they show me scope plots they took with either no ground clip, or a with a 6 inch loop in the ground clip on the scope probe. (The software guys are the worst at this, but I guess they have a lot of other stuff to worry about). A scope with 1GHz of bandwidth will show you everything wrong with your measurement techniques. You don't show your scope probe connections, so I'm not accusing you of this, but it is something to be aware of. I have seen lots of PCB layout issues over the years (30+) as an RF engineer, but I've also seen a lot of problems that were blamed on PCB layout that were not that at all. When I'm in a design review and somewhen says we have a layout issue, my first questions are always "how do you know it's really a layout issue?" and "How did you measure it?" Noise problems are usually pretty hard to figure out and when you are stuck it is tempting to want to blame something that seems esoteric and that can't easily be proven wrong. When you want it to work it is tempting to want to throw every fix you can think of at once just to get rid of the problem. The book "12 Seconds of Silence" by James Holmes talks in depth about the techniques that were used at APL to determine root causes for their design problems. I highly recommend that book to anyone with an engineering mind set. As for 4 layer boards, they are getting so cheap that unless you are making 1000s of something it isn't worth the extra layout time not use use them for any circuit with more than a few parts on it. Also, adding the ESD protection is good practice for any circuit that goes out into the wild. Remember to add a bypass cap across the VCC and GND terminals of that part too. Well, that was a little more than I was intending to write. Again, love your channel, Sunday mornings I usually look for you and Rotary SMP.

[@swirlius]

Great video! I've found Eric Bogatin's books and videos helpful in trying to understand PCB design / signal integrity issues.

[@blahblahblahblah2933]

Your comments re noise filtering around the 16 minute mark are well taken. The next step with signal integrity on your board would be balanced differential signaling and possibly a shielded cable with twisted pairs. But that's a major change, so if you can clean up things with these changes it's a big win. With regard to the TVS, another thing I've seen on designs that have input signals on long lines is serial resistors into the inputs (say 100 ohm) which limit the amount of current that can ultimately enter the device.

[@Rob_65]

I have done this type of rework soldering a lot. 0805 and 0603 are large components and still quite easy to solder for me🤭 Just a few tips: when cleaning up the pads with the soldering wick, I see that you brush the wick over the pad. I use to do this too and ruined a few boards that way because pads from small components or ICs may be pulled from the board. It is better to just put the soldering wick on the pad and brush with the soldering iron over the wick. Those large 0805 pads will most likely not be pulled from the board (unless you heat up the board too much) but that is still the better way to do it. Then, with a clean soldering tip, just touch the pad and you will have a nice flat pad on which you can mount the new component. And when mounting "flying" parts like the 100pF caps, I like to place a bit of kapton tape below the component so I don't risk damaging the soldering mask of the traced below the component. Last tip: before soldering the 0603 capacitor, put some fresh solder on the pad of the capacitor first, that helps a lot because it is now much easier to heat up the pad of the capacitor. Your SMD rework skills are not bad. I have seen worse from both my colleagues and myself 🤫

[@jameslmorehead]

Moving the power planes to the outer layers, and keeping the signal traces to the internal layers, will provide the best EMI shielding you can get without going exotic like a 6 layer board with solid layers just for shielding. I work with power inverters for traction motors in the 20kw to 450kw range, so we see very large EMI problems.

[@feynthefallen]

I'm glad you took my suggestion to put capacitors in. What I'd be reall interested in is, would the capacitors alone have fixed the issue? If you have any original boards left, please do me the favor to test that.

[@rpavlik1]

Really nicely explained! It makes me wonder how on earth Adafruit can make so many high quality gadgets while almost completely avoiding 4 layer boards. Presumably an EE degree actually means something 😅

[@Clough42]

It's totally possible to make great stuff with 2-layer boards. Most common hobby microcontrollers have relatively slow (5ns+) edge rates, so can be pretty forgiving, though many Arduino reference designs are pretty noisy by modern standards. 4-layer is getting so cheap now that it presents interesting opportunities for the hobbyist to make routing easier and avoid a lot of common pitfalls.

[@equi-nox]

Hint: for those 100pF capacitors, it does not matter¹ whether they go to GND or positive voltage; all that matters is that they are between the signal and a constant DC potential. So you can just get physically larger capacitors (same size as the resistors) and directly stack them on top of the resistors. Also if you make the resistors smaller, the capacitors need to get larger, since their effect is determined by the R×C product. ≈4.5× smaller R ⇒ use ≈4.5× larger C (470pF is a standard value.) But NB: both of these changes - smaller R, larger C - make it harder to drive the input. With the traces you're showing, it doesn't seem like that's a problem in your setup though; not yet at least. Also-also the C1/C2 decoupling capacitors need to be as close to the IC's power pins as possible. [EDIT: looks like you already moved them to a better place, I should watch the entire video before commenting :D] The circuit board - and in particular the vias - are only in theory 0Ω 0H 0F. But when you're looking at fast edges, the reality of non-zero µΩ nH pF values makes things harder. [¹] - some might nit-pick that the power-on behavior is different when the cap is to VCC rather than GND. Which does not matter the slightest bit.

[@1kreature]

High current lines going at 90 degrees across signal lines is the best possible crossing for minimum interference. What is more an issue is "loop area" for high currents. Placement of decoupling capacitors however could be a lot better...

[@Mike40M]

Though being a mechanical engineer, I had to design a multiprocessor board, 8 by 8 inches,for some embedded industrial systems back in the eighties. As processor speeds was much lower then, EMI wasn't as big a problem. To make PCB layout easier I went for a 4 layer board. Ground plane bottom, then a "north-south" layer, then an "east-west" layer and a positive layer on top. Hundreds of them made, rarely any problems with them. Seems I wasn't completely wrong.

[@JulieanGalak]

Going to 4 layers is the best solution for this sort of thing, but, especially for folks who don't have the ability to do that, something else to keep in mind: coupling between traces (or any wires) is roughly proportional to the cosine of the angle between them. So parallel traces couple the most, perpendicular traces couple the least. So if you have top layer traces pass near bottom layer traces, it's best to cross at a right angle rather than at a sharper angle. The coupling is also inversely proportional to the square of the distance, so separating traces is best. Thus, it's best to cross traces at a right angle and then move them apart, rather than letting them run parallel near eachother. Obviously, some times there are geometrical constraints that prevent this. James had all this correct on the prior version - the main bundle of traces crossed at right angles.

[@arttanaka]

SMT Rework suggestion: I use a (6x6x2 Granite Check Stand Surface Plate & DIAL INDICATOR Gauge Granite Block) and replace the standard probe tip for one that is more like a pin to hold down SMT parts while soldering. The Dial indicator I have been using has a 1" travel which works well. I suspect you already have everything you need in your shop.

[@Skyliner_369]

So I was thinking about the ELS right? and I thought of a feature for the ELS I wanna call automatic feed stop. good for threading and for precision machining. either by reading DRO values (if any DRO would even let other hardware in on those readings) or by just screw-turn counting.

[@Nf6xNet]

9:28 Your soldering is not bad. Your results look fine, and the difficulty you had with the 0603 caps is just something that goes away with lots and lots of practice, as you develop an intuitive feel for the heat flow and surface tension and learn to fling stuff across the room with the tweezers less often. Try working with some 0402 components… by the time you can work with those without wrecking them and the board, the 0603 ones will seem easy! Out of necessity at work, I got to where I could rework 0201 with a lot of difficulty. But that experience made 0402 seem easy by comparison. I agree with the other commenter who suggested a J shaped tip.

[@johnalexander2349]

Sweet, it's working. Time to add more features. If it ain't broke, fix it till it is. I vote for a spindle angle readout from that fancy high resolution encoder.

[@erikdunn]

Maybe a fun project idea for you, but - I've been looking into possibilities for attaching a small/micro drill chuck to those electric screwdrivers. I think it'd be fantastic for drilling out or threading small 3d printed plastic components very quickly. Beats a pin vise and preserves your wrists, and you don't have to worry about snapping any drills if you're chucking them up in a big 1/2 chuck. Great vid, looking forward to more as always.

[@jBurn_]

Honestly: Just buy a drill that is dedicated for that Job. Maybe I did not understand your idea correct, but I think a proxxon micromot 60/E will do the task way better. Here in Germany they cost around 30 € and you can even choose between a tiny chuck for convenience or tiny collets for precision. Of course, threading does not work. If that is your main task those screwdrivers could indeed be helpful.

[@jerzyszczepanski2518]

Hi James. I love your ELS project and I'm in the process of buying the Kit, but I was hoping for litle upgrade. The problem I have is that my lathe doesn't come with Thread Dial Indicator. I was hoping that you could implement in version 2 of you design spindle and leadscrew sync at set point. For example, set points at begining and the end of the thread. Thank you.

[@jrh_engineer]

I would add small series resistors between your TVS array and the RC snubbers. A 1A ESD pulse will still get to 12.5V before the TVS clamps - you're relying on the capacitor in the snubber to deal with the let-through energy. A series R (100R - 220R, 5% to 10% of the pull-up) will slow the let-through surge and give the capacitor a better chance at filtering the high frequency pulse energy.

[@75keg75]

Could you laminate a lead layer between any layers with potential emi clashes. Effectively shielding (time of posting 4:30) so at 7:30 the ground plane act as a “lead” shield.. interesting…

[@bluedark7724]

The secret sauce that Rick and others I'm trying to get into our head is that with microprocessors and signals it's all about the rise time which is nothing to do with the frequency. Unfortunately it's very hard to find me information on rise time. You might be lucky to find a microcontroller that has a simulation model. I'm playing with esp32 and unfortunately I do not have a simulation model

[@Dazza_Doo]

Rick and others really opened my eyes. Dave from Evvblog also had some hidden gems too.

[@worldmenders]

I've designed a lot of boards, from computers to cell phones, and while going to 4 layers is always nice, this board could have been vastly improved as a 2 layer design. For example, the ground connection between the input connector and the IC is quite poor. Adding a ground trace parallel to the data lines from the ground on the IC to the flood near the connector would be a big help. For this design, I would have ground flooded both sides, and stitched the planes together with vias near the IC. The rest of the PCB isn't really a problem (LEDS and Stitches) if the signal integrity of the interface to the IC is good. I also would have placed the bypass caps as close as possible to Pins 4 and 15 (Vdd pins) with good return to the IC ground pins.

[@MSM5500]

_"stitched the planes together with vias near the IC"_ I'd rather have via stitched both the top & the bottom ground plane polygons wherever possible in order to suppress capacitance which will inevitably build up between them if left not stitched. The more vias the better. JLCPCB don't care of how many holes they drill for you so it will not affect the price at all. There is a dedicated automatic feature for that in Altium Designer. The smallest hole size JLCPCB can make is 0.2mm so several hundreds vias can be placed per one PCB easily. The only thing to consider is to remove a thermal relief from that sort of stitching vias in the Altium DRC. BTW, I have no any degree, I just use a common sense.

[@robbyg3989]

Thanks for the vid. Great work.

[@contentnation]

Hi James, a suggestion for the next iteration or project. Add a schmitt trigger for the inputs. Some 10 Cent 74xx2g17 will do. In combination with your existing diode chip it should make it quite resilient against emi on the input lines.

[@ElectricalInsanity]

If you've taken an electronics class, you might remember that a capacitor is just two conductors separated by an insulator. Here you have two traces separated by the board material. Capacitors block DC/low frequency signals but allow high frequency signals to pass. If you have a square wave, you can do a Fourier transform and see a nice spectrum of high frequency signals. Those high frequency signals that get passed through are your noise. This is just how I think of it, because going into the EMF explanations always goes over my head.

[@Superfungus0]

Just a point for future designs from a many times bitten EE - looks like there's still a slot in your planes created by the line of vias towards the bottom left of the board. If you wanted this to be really bulletproof you could space those out enough to allow flood to get through them.

[@paulypaulypauly8011]

Silly question time - when can we buy them from you? Thanks!

[@martinmckee5333]

I've gone to 4-layer for all of my hobby projects. Between easier routing (as most of my boards are dense) and improved signal integrity, it just isn't worth using 2-layer boards in most cases any longer.

[@LordPhobos6502]

This is something I'd never considered until watching this video... now I have 2 reasons to try 4 layer! Do you just route GND, or power as well?

[@martinmckee5333]

@@LordPhobos6502 most of the time I route power and simply connect down to a ground plane (unless it's an analog design that needs a star ground).

[@jpiccari]

When you say shielded cable on the “VFD leads” do you mean input or output? Common practice to put toroid on output but are you also suggesting the output should be shielded?

[@Niels_Dn]

Interestingly a 2-layer PCB is super difficult to get right. It sounds so easy because it’s only 2 layers, but sorting all return paths properly is a nightmare.

[@Saminvent05]

Hi James, I haven't seen the full schematic for this so don't know but do you have any ESD suppression on those button inputs as it's going to be pretty typical for a statically charged person to end up grounding themself pushing those buttons. The 3D printed plastic parts are going to act as a bit of a barrier but a 10-20kv charge could still jump plastic or capacitively couple, it has to go somewhere.

[@siberx4]

Nut drivers are hugely underrated; they're way easier for all sorts of "small" nuts like this if you want to minimize the amount of cursing, cross-threading, and dropped fasteners. This kind of "sharp edge causing noise and interference" problem is prevalent in a bunch of digital signaling; if you've ever perused the catalogs for things like RS485 drivers, you may have noticed that manufacturers will often sell a few different "versions" of a chip with various rated bitrates/speeds for what is otherwise the same design for basically the same price. You might initially think, "why wouldn't I just buy the fastest one?" but the answer is that the different "speeds" differ largely in how steep their rise and fall ramps are for their drivers. Slower/softer transitions do limit the maximum signaling rate, yes, but it also means you produce less high-frequency harmonics that can cause interference to other circuits so you really want a driver chip that's "just fast enough" for the signaling/baud rate you intend to use. This will provide maximum signal integrity with the minimum potential for introducing extra noise into your systems. That TVS "multipack" is a great component, and I'll definitely be keeping it in mind for my future designs! A lot simpler than including footprints for a separate TVS package on each data line independently.

[@j.douglas5461]

Hi James. Big fan of your work. What ever happened to this? Did this ever go into production? I'm about to buy your ELS kit but would gladly wait if I knew there was an improved version available or coming out soon. Please let us know. Keep up the good work sir!

[@JimN_AustinTx]

James. I had to comment after watching this latest video. You failed to mention the latest version now includes decoupling capacitors. The blatant lack of bulk and high frequency decoupling caps was a fundamental issue from day one. I’d go on to say that most of your issues are power related and not EMI related. Always focus on good power delivery first then signal integrity and lastly EMI if needed. Whoever suggested C1 and C2 is on the right path. However with your VFD nearby I’d add something in the multi-hundreds of uF as well. Why didn’t you mention these? You certainly spent a lot of time on Signal integrity and EMI but you thought the boards performance on the scope looked good. It looked terrible from a power and simultaneous switching perspective. Look at the high frequency bumps on the low signals. That’s just a mirror of the ground supply at the time. Shows a fundamental lack of good power supply decoupling. Similar glitches are seen as the signals were high. All the signal filtering is OK but it’s not addressing the 900 lb Gorilla.

[@Gobhoblin126]

You said that there's no way completely eliminate issues like this but there is one way you can get pretty close, use a differential signaling protocol such as rs485 for the data link. You could even just send the existing signals through bus driver and receiver pairs and keep the signal timing the same.

[@arthurmoore9488]

That was my thought as well. RS-485 transceivers are stupidly cheap. Go with a full duplex one from TI, and an ATTINY202 (< $0.50) and you've got bi-directional communication in 4 wires, while still having a reset pin for the micro-controller. If you're willing to forgo the reset line, you could go with a half-duplex transceiver and set the "reset" pin to act as a TX control.

[@wileecoyoti]

Always great to see a reference to Rick Hartley, between him and Eric Bogota my job got much less mysterious. Any thoughts on moving your power/reference planes to the outside of the PCBs? I found this sometimes makes things easier with predicting standing/interfering waves across my switched power and data lines (which are frequently nanosecond rise times) because it's FRx on both sides of the trace instead of FRx+air. Either way great work!

[@928Livio]

Are any more kits going to be available soon? I have a Busy Bee CX615 Milling/Lathe combination set up I want to convert to ELS.

[@TheArtOfMistake]

Its all about a trace return current path and magnetic coupling. Gnd ref directly below a signal trace all good.

[@thechosensasquatch]

With moving to self-designed boards, have you considered moving the microcontroller directly onto your display board? It would certainly help with EMI hardening.

[@bobuk5722]

Not feasible. For one thing, the circuitry belongs to Texas Instruments. Secondly, have you looked at the complexity of it?

[@hoggif]

My guess would be the 100pF caps make a more differrence than 4-layer board. Too bad both of those were not tested separately to change by only one wire at a time. I'd still guess some random short spikes on signal lines screwing up the signal and making it flicker before. 100pF caps should take the smallest ones out of. Sure, making both 4-layer and adding caps is the best thing which is good for production runs. I think the coupling to high current lines may not be that high as the signals are already routed at almost 90 degree angle to each other. With a 4 layer board like that, there is now not even a doubt of signal coupling that at least excludes the possibility all togeather.

[@ultimateworkshop2000]

I like this ELS very much especially being overkill, I have watched the whole series but just when I decided to source the parts to build mine, I found others that added a second stepper to the X axis and integrated ELS with CNC and even more a single or multiple profiles turning, threading, boring and even curves. In spite of your ELS is better engineered, I can’t resist making a thread with press of a button. Do you have plans for such upgrade?

[@DMike92.]

Honestly, from what I see on the scope, you could use any of those 4 new boards, including the one with no modifications. Maybe you tried it?

[@glendaleclassiccarwashiish4727]

Hi. Are you coming for summer bash this year?

[@timdontwannasay5889]

If really like to buy all the components. I've looked but no "kit" to be found. Went to TI's website and couldn't seem to get the board to the cart.. maybe a "buying for dummys" video?

[@Cybernetic_Systems]

Howdy James, if you want to improve your SMD soldering game, grab a J tip for your soldering iron. They make life much easier and are great for drag soldering fine pitch IC’s too.

[@buidelrat132]

More boards more better! ELS forever!

[@leslierhorer1412]

I can't resist pleading (begging if I have to) with you to *PLEASE* implement the threading / turning to a shoulder feature you discuss in the git hub. There is an unused button which could readily be used to enter and exit shoulder mode and to set the shoulder position.

[@dandecl5843]

Will it be just a remove the original board and replace with the new version. No more programing etc? And then good to go?

[@The411]

ordered a F280049C. I'll watch ebay for your new batch!

[@imajeenyus42]

Hi, could you let me know what Gerber viewer software you’re using there? I’ve been using an ancient version of GCPrevue and it’s rather clunky! I feel ya about the interference issues, had to chase things like this on boards in the past and it can get very frustrating.

[@Clough42]

I'm using CircuitMaker for this, and I hate it. I'm doing everything new in KiCad.

[@imajeenyus42]

@@Clough42 Thanks, I forgot KiCad had a gerber viewer, will check it out. I still use Eagle mostly, having grown up with it and drawn a bunch of libraries ;-)

[@roflchopter11]

So your bench testing said that the datasheet circuit (10k with caps) is marginal on rise time? Did you try to do any kind of bench testing with noise, which is the purpose of reducing the resistor values and adding the caps? That could be quite challenging to reproduce.

[@paulmanhart4481]

Does this mean that we need to buy new boards? When will these be available on eBay? I suppose they don’t need any mods like the previous boards? Paul

[@whitehedr]

Is the control panel PCB going to be available when this board revision his the market?

[@lesthompson5907]

Hi James. now your producing you own boards will you be helping such as me & producing a complete redy to use. mother board operating siystom easily fitted to my lathe at 72+ years of age & full of artistes i can't do the intricate work now haven't the skills to Bild I can fit. plug in box how much would the come to ? i want to fix my old lathe before a kick the bucket. all in one package how much would that be Les.

[@monroegiger]

How do I order your newest upgraded electronics?

[@75keg75]

15:28 One more time, we're gonna celebrate Oh yeah, alright…

[@jhbonarius]

Great approach, but there is some reasoning error in there. The 10k 100pF signal looks fine. Even if some more load and capacitance is added by any long cable, it's still a reasonable pulse shape. I don't know the threshold voltage on the inputs, but you often don't even need full Vcc on the pulses. In high speed signaling, especially in auto equalizing circuits, signals look nothing like square waves anymore, as that would just be a waste of energy and actually be a cause of crosstalk. Remember, going from 10 to 2k, the source drivers have to drive 5 times the amount of current. That could be an issue if you're pushing them to their limits, or they're not properly cooled. Did you check that? Have you also considered using shielded and/ or twisted pair cables?

[@scrout]

Is this TTL? So 5v signal but switching is down at 2.4v? The fast edges are definitely generating emi, the 10k/100pf provide a -3db point at 160khz. The 300 nsec normal risetime seems very fast for these circuits.

[@shockwave77598]

Use Differential transmitters and receivers in very noisy areas. You need more wires, but it's a very robust solution.

[@Clough42]

If I were starting from scratch and building a commercial product, this is likely what I would do. This project, from the beginning, was designed around cheap off-the-shelf maker components, and I didn't find any great bolt-together options for that. I tried a few things, like the TI driver breakouts, but they all had specific compatibility issues that made them unsuitable without spinning all custom hardware.

[@jimihendrix5576]

Thank you sir.

[@PeteBrubaker]

When can we buy these, I want! :)

[@robertpeters9438]

Try running a ground trace on both sides of a critical trace in the same plane.

[@Dazza_Doo]

What your Differential Pairs Wish You Knew with Rick Hartley - AltiumLive Keynote

[@bluedark7724]

Air soldering station ?

[@ramkitty]

If that is the same mini electric screwdriver i got the case is amazing

[@Factory400]

Curious what keeps you from using CAN bus. I use it on many commercial applications on CNC machines that have insane EMI.

[@roflchopter11]

It's overkill for this, could easily add like $10 to BOM cost, more if the servo driver doesn't support CAN. He could borrow ideas from it (and things like RS232) though, like differential signalling.

[@zyeborm]

The display boards are a commonly available general purpose board. If he went to can he'd be rolling his own (more so than just tidying up the existing board) and people would be forced to use his board rather than being able to do everything themselves easily. Can or similar would be a "better" solution but better is often the enemy of good enough.

[@Factory400]

@roflchopter11 RS485 could be a reasonable option if BOM cost is truly a concern. I have never found the cost of CAN on BOM cost to be worse than the alternative....including numerous revisions trying to make incremental improvements in tough EMI environments.

[@joseph9915]

@@roflchopter11 shouldn't it be more like $1.50? Well, just for the board end.

[@roflchopter11]

@@joseph9915 you'd have to add a transceiver and controller to both sides and probably upgrade the MCUs.

[@bking0740]

How do I get my hands on one of these kits?

[@Clough42]

I'm out of stock at the moment, but I will be making more.

[@bking0740]

@Clough42 What do you charge for a complete kit? Do you have a site that you sell these on?

[@ergindemir7366]

100pf or 100nf?

[@danmooney7192]

I'm not going to pretend that I understand an iota of the technical part of this project. My question is there enough call for this system in the after market or are you shooting for selling the system to a lathe manufacturing company?

[@patrickfrantz2466]

It wouldn't hurt to add a 10uf and a 0.1uf cap at the connector between 5v and ground.

[@Clough42]

The bypass caps used to be there, but I moved them to the other side of the driver chip, right next to the power pins.

[@patrickfrantz2466]

@@Clough42 Nice to hear from you. Caps at the part are good too. This descends into personal preferences pretty fast. I spend a lot of (too much) time fighting really tough mil spec conducted and radiated emission spec projects for radio applications. The caps at the connector would help with stuff coming in (susceptibility) on the input power as well as help to keep your locally generated noise from going back out on the cable (not an issue for your project though). Your understanding of signal return path physics was spot on. Many EEs don't ever learn that. Did you notice that when the ground plane signal gets back to the connector it has to open up a little bit to get to the ground pin? That is why on longer connectors, or on really fast edge rate signals most of us eventually learn to space out the ground pins to keep a ground nearby the signal pin. Lots of opinions about cable shields. The shield attenuates E fields but doesn't do a thing for B fields. In order for the cable shield to work it needs to be grounded though. The best theory that I have seen is to only ground one end so as not to have load currents also carried by the shield. B fields are best addressed with phsical separation. They drop off as a cube of distance. E fields only drop off as a square. Also, the same field area thing happens in cables. That is why it is a good idea to run twisted pairs for your signal lines. The signal would be twisted with a return ground. Pat

[@grahameida7163]

Good job…

[@blahorgaslisk7763]

Regarding ESD you often will hear people who claim that they've "worked on electronic circuitry for forty years. Never used any ESD protection and never had ESD kill anything" or something similar. That may be true, but more likely is that they either just have selective memory or they may have damaged a lot of circuitry but not to so bad that it failed immediately. Thing is ESD can damage chips in ways that will shorten the life of the component and/or change the tolerances in both input and output signals. This is nothing you can see with your naked eye so you really don't know if it has happened. So I really recommend that you at least consider the risk of ESD and do what you can to minimize the risk. So at least leave that wool cardigan and the spandex shorts in the closet when you are building something like this...

[@floodo1]

TVS = transient voltage suppression for anyone wondering

[@gymprofessor329]

I would buy some no clean flux in a syringe. If you put a blob down where you are working it makes the solder flow like a dream. If you work under a puddle it basically gives you infinite time with the solder flow able. ESPECIALLY important to use flux with lead free solder.

[@tastenheber]

Agree. I'm not a big fan of the pen type flux, since often, it's just not enough. Also the high quality stuff doesn't even evaporate on lower temperatures, it's like moving a pencil through water. I don't mind washing and rinsing the bord later with some IPA.