Thank you everyone for this topic and posts, appreciate being able to learn a bunch and see how solution for a prob is being willed into existence via global discussion and collaboration.
For subsequent larger orders to get the overall price down… Do they have a “subscription”/contract/model that splits purchases/payments across multiple batches/deliveries with unit price at rate based on a commitment to purchase some order amount several times within some pre agreed time frame?
PCBway might, I do not Think JLC does, at least not at smaller numbers like this. If it works I will be emailing them about that very thing. There are account managers that might be able to help.
I don’t see any major issues that would prevent ordering.
The bottom side plane is starting to get pretty thoroughly sliced and diced, so for future revisions/work, I’d try to focus on keeping that a little cleaner.
Is there any particular reason you’re routing at 0.254mm? 0.2mm should be well within capabilities and is my ‘standard’ routing size. Typically I’d route with 0.2mm trace width, 0.2mm clearance, 0.25mm plane clearance. If the board fab’s capabilities go lower, I’d consider dropping that further even for things like the links from the pin headers to the ESP32 as there are so many in parallel.
That’s basically a slot under the USB traces which will disrupt the return paths. There’s a via there that creates a pretty reasonable bridge over it, but it’s better to not need to rely on that. This signal is DC, so it’s extremely insensitive to routing. An easy fix could be moving it left by ~5mm so that it’s under the USB connector. You can also easily hop back to the top side to shorten the slotting which won’t matter because of the DC trace:
This is a pretty major slot that also has traces on the top side following it so there will be current crowding around the top and bottom sides of that slot. I’d try to move them apart a little to allow a sliver of ground plane through somewhere near the center, at least.
This kinda thing:
Looking at how this ground pin connects, I’d also try to get as much of these traces onto the top layer as possible and widen this segment, perhaps by bringing the vias across to the left a bit. Mostly this just looks like you’re somewhat restricted to by re-creating the devkit layout, which is already kinda questionable in a few spots. I don’t think this will matter particularly in practice, though as none of the signals off the headers are ‘that’ high frequency and there are other grounds.
Easy enough, I thought about it but that seemed worse in my head. Glad to know it is not.
Let me double check the specs for JLC, I can change it for sure. This has to be a “standard PCB” not “economy” because of the ESP module so I know the specs get a little better.
Okay
I can make a bit of space. UI34-35 really make things a pain here. They were on the bottom so all that was on the top. I flipped them so the voltage regulator would get more heat dissipation from the last review.
I think it does connect to a large return on the top? I will double check this evening.
It does, but just by a slightly weird route. The goal is to try have all the ground connections share the lowest inductance path possible between them so that they can all be considered the same voltage, regardless of the frequency of the signal. It’s a little harder to consider in this case as this is a board that’s going on another board, so it all gets a bit strange and can’t necessarily just be considered in isolation, but I don’t think any of this is actually critical, it’s more just ‘clean up’ kinda stuff. The board will be fine if submitted as is, as best as I can tell.
I have 15 ESPs in my house doing random things. And a couple in my drawer for when inspiration strikes. So far, none of them say V1. I will need to change that.
I will need to figure out how to decrease the price a bit to make it a comfortable price for the seller and the buyers. Who knows, if they do work maybe I do a giant batch and just go for it.
I really respect that you put thought into the price of your products and try to be fair.
That said, you really don’t have to save me from myself .
I know when I buy really cheap crap that I may get parts that don’t work properly.
I’d rather avoid that.
I also generally don’t mind paying more for something I’m pretty sure will work correctly.
It’s also important to me to pay enough that someone taking the time to produce a known good product for me is properly compensated for their time.
I learned another lesson today. I opend a fresh bag of 100 ESP’s from the same source as before. So this is the 4th bag of 100…so far every single one I have pulled out is bad, like actually bad not just sketchy.
That sucks, hopefully that ends up solvable somehow.
So many lower-price supply chains are based around balancing failures/faulty products with the value that the lower price provides. It’s VERY rare that I’ve ever seen someone keep buying from cheap sources long term, eventually they get bitten and move back to the regular suppliers.
Sometimes it’s counterfeit parts (big issue with USB-UART ICs a while back), sometimes it’s fake parts (bunch of SPI RAM out there that was just the packaging, no actual IC inside), sometimes it’s grey shift parts (made after hours using questionable methods/QC and sold to dodgy suppliers), sometimes it’s QC failures (being swiped from the bins and sold on to dodgy suppliers).
The thing that always scares me is that with a lot of those situations, there’s simply no way to know for sure that the parts don’t meet spec without testing them. Maybe something falls over at temperature, maybe some pins have damaged internal pull-ups, maybe a pin has a damaged output driver, maybe the internal reference is inaccurate, maybe an ADC has high leakage current, maybe there are bad sections of flash, etc. etc. and on and on and on. A lot of those issues you might never know about. They might not affect your design at all. They might affect it just enough to cause issues down the line.
Perhaps there’s an argument to be made that outright failures to flash aren’t the worst outcome and are a pretty clear flag that your supply chain has issues?
To be honest, given the amount of time and energy you have to put into these cheaper boards, I kinda feel like you should have 2 shop options.
Jackpot with Genuine board(could be your V1 ones if price comparable), preflashed and ready to go, or
Jackpot without ESP32, and if you want to save $5-$7, you can source your own. With the FluidNC web installer, it’s super easy for anyone to configure on their own.
Option 2 can come with a disclaimer or information and warnings about bad boards and boot problems, etc, but I bet most would choose just buying the higher quality board
I 100% agree. Having a reliable carrier board on there that is tested before shipping is a huge value add. Having the option to BYO ESP32 allows for both the lowest cost-of-entry scenario for those that want that and are willing to trade sourcing/troubleshooting time for lower price. The tested and known-good ESP32 included option is for the rest of us who can’t be bothered mucking around with all of that…
Can take $2 off the price for option 1, to $57.99, and make the genuine with flashing $69.99 if it’s covers the board cost plus time spent flashing, etc
It is usually not this hard for me. I buy some from all of them and it is pretty easy to weed out the “main” supplier and when you start buying a lot they usually start getting better.
Agreed.
Yeah, the time has come. I have the USB-C batch to work through and that should give me enough time to figure out the pricing. V1-esp or Espressif, should be about the same. I love seeing my logo and all but it does take work to order them so if I can not save more than $0.25 a board it is not worth it.
Not sure if I will offer an option without an esp, have to think about it. Maybe just for the option to buy an external antenna version. I have a buy your own board option on jlcpcb and pcbway, so really if people do not like what I offer they can DIY.
The upside to using good boards is I can more easily get them to partner stores around the world. I have been waiting for “all” the bugs to be worked out to place some in other locations. I think this was the last hitch. The Jackpot is validated in my opinion, it turns out it was just a crappy ESP set that has been messing with me. I will be making the header pin holes a bit smaller on the next batch, like Jono said just keep making them smaller until the headers sit a little more uniform, and I do not want to go so small the board house complains.
They are selling well and the better esp’s will boost my confidence.