Attach a Teensy onto a PCB vs raw chips - production?

Status
Not open for further replies.
A

Andy

Member
Hi Everyone

I've built a prototype device using a Teensy 3.0 that I would now like to take to the next stage.

I potentially might need 2000 small PCB boards about 20mm by 60mm so a little big bigger than a Teensy that provide the standard Teensy functionality (IO pins and USB). I additionally am utilising an ATSHA204 chip and a 0.5mm FFC connector.

The perfect solution from what I can tell is putting the raw Teensy chips on the PCB. A mini54 (from Paul) and an MK20DX128VLH5. There are even eagle files to assist in this. I have no background in electronics so I'm slightly out of my depth even with the amazing input the forum gives. Is that all you need? I'm using the USB HID functionality of the Teensy to do key presses. Does this require any more chips? Do you need to add resistors and things like that to the board? Would a standard electronics prototyping company understand all this and with the help of the eagle files be able to create my board and I use it just like a Teensy?

Could any forum member give a shout how much this type of PCB might cost including assembly for A) a prototype and B) 2000 production? I'm based in the UK, has anyone done anything similar over here?

An alternative would be to use a Teensy as it is and attach that to a designed PCB. Would anyone even consider this approach for potentially a 2000 production? I attach my USB cable direct to the pins on the underside would I still be able to use this way of connecting USB instead of the micro USB socket? Can a PCB extract those pins and take the connection away to solder pads? Again I see there are eagle files for this but I don't think they carry the USB pins.

Paul if you are reading, how much would 2000 Mini54 chips cost please? I'm a massive fan of your work and I can't believe how much you manage to help everybody on the forum and solve their problems.

As a final question Paul said on a different thread that the raw Teensy 3.0 chip could be locked so that firmware can't be overwritten:

http://forum.pjrc.com/threads/24365-Protecting-Firmware-Code

Could someone spell it out to me with a general summary on how I could do this?

Regards

Andy
 
As much as I'd like to sell you 2000 Teensys, realistically at that volume it probably makes economic sense to design your own board. Maybe?

Then again, if you don't have some big customer committed to actually buy 2000 pieces, or something like a Kickstarter campaign with 2000 backers, your best path might be to start with a small batch the easy way (soldering a Teensy to an extremely simple board, or whatever else you can do that's quick and easy and temporary to get started). If your product goes like many do, you'll probably get a lot of valuable feedback in those first 100 customers, so it's often a smart move to keep things simple and not get the next 1900 pieces set in stone, so to speak.

The other thing you'll get in the first 100 sales is a good idea of the actual sales trend. Even if you do all the PCB design and other engineering, you'll still be looking at some pretty substantial setup costs and investment in buying materials to make your first large batch. That's a tremendous financial risk to take with zero sales history. But if you've made and sold one or two small batches, then you'll *know* how well they sell. You'll know how much money you'll save with a large batch vs the relatively high cost of making small batches, and you'll have a solid idea how long it will take to recover those costs. Presumably you'll make relatively little profit per piece with the higher cost of making small batches, but if you do net something (actual money... though the experience is probably more valuable), you'll also have a little more capital to spend on the first big batch, plus all the one-time costs associated with it.

In terms of making your own PCB, by far the most popular route is using the free Eagle CAD software and OSH Park to fabricate the prototypes. Eagle is frustrating, but it's free, and because it's free so many people use it (despite all its flaws... did I mention the low price) and there are LOTS of tutorials online about how to use it. Mentally prepare yourself for an aggravating experience. Just keep reminding yourself that it really does work since so many people do use it, and it didn't cost you anything. While you're learning, it's much easier to make simple PCBs, so start small and build up as you gain experience. When you have your PCB made, the good news it'll only cost you about $10 at OSH Park for 3 pieces, if you keep the size to 60x20 mm. it's so very cheap that you can probably afford to make several attempts before getting it perfect.

When you decide to make a first small production run of 100 or so, OSH Park (in the USA) and others like iTead (in China) have services that are slightly less expensive to get you about 100 pieces. Every business is different, but if yours is like most, I'd highly recommend starting with a small production run to get some initial sales and customer feedback. That's usually a much more successful path than trying to go directly from a great idea instantly to a large production run.
 
Last edited:
Oh, also, the free version of Eagle is legally licensed only for non-commercial usage. At some point, probably before you start selling products, you ought to look at the licensing and decide what to buy. As CAD programs go, Eagle is pretty cheap.

Standard disclaimer: none of this is "legal advice". That can only come from your own attorney.

There's also an open source CAD program called KiCad. This next bit is probably going to touch off a flame war, but so far the only positive things I've heard regarding KiCad are related to the fact it's open source or free to use without any limitations (like restricted board sizes in Eagle). Then again, people who use Eagle don't tend to love it either. Apparently KiCad is improving quite a bit. Many people have actually used it for successful projects, so KiCad might be another option?

Beware of PCB vendor-supplied software. A few of them provide free programs that are actually pretty simple and easy to use, at least for relatively simple PCBs. But those programs produce encrypted files that can only be read by that one PCB vendor. Learning Eagle or KiCad or some other CAD program might be more work, but later you'll be very glad your files are usable by any PCB vendor.
 
Last edited:
Regarding PCB software, I would suggest DesignSpark PCB, it is free and I don't think it has a non-commercial clause. It supports import from Eagle, and I much prefer it's UI over Eagle (I really can't stand Eagle).
 
A couple of thoughts here...

For one, consider how crammed a Teensy 3 is already... little room left over! You are asking to make something smaller yet use a larger bootloader chip than what the Teensy 3 currently uses (the Mini54 is sold to us is significantly larger than the equivalent (BGA?) bootloader chip on the Teensy 3 series.)

That in turn suggests you having to mount SMD components on both sides sides of a PCB and the very likely need for a 4-layer board if you want equivalent performance (especially from the ADC). Few PCB layout programs offer a 'free' version that can do 4 layer boards.

I'm not saying your project is impossible, but you likely need professional help from someone with the skills of Paul to design the board for you. So that adds cost and complexity up front.

Online vendors like SEEEDstudio, etc. will sell you PCBs, stencils, and/or assemble your devices too. Most offer a pretty comprehensive calculator so you know what you're in for. IIRC, when I priced it out once, the assembly fees came out to about $15 for a more complex board than the Teensy 3 (part count: 100) At higher volumes, I bet your assembly costs would drop pretty dramatically, especially if you only use SMD components.

I've made peace with Eagle's quirks and non-intuitive UI design decisions. If I ever have a commercial use for my stuff, then I'll buy a license and enjoy much larger PCB playgrounds as a bonus.
 
Hi Chaps

Thanks for the replies. I should have said in my original post that this is to provide a solution to an internal problem for the company I work for. Obviously we would get a proper production prototype made but then hopefully get sign off for the large quantity required.

The board can be a little bit bigger as the final size is not yet agreed.

I've built the prototype using the teensy but it might be time to pay for assistance in making a production prototype.

It sounds like it might be easier adding a complete teensy to a custom pcb plus my ffc connector and ATSHA204. Its just would prefer a specialist to assist at this point.

Anyone in the UK done something like this?
 
Constantin said:
the Mini54 is sold to us is significantly larger than the equivalent (BGA?) bootloader chip on the Teensy 3 series.
Click to expand...

Actually, all 3 sizes are now available, including the smallest one which PJRC uses on Teensy 3.0 & 3.1.

http://www.pjrc.com/store/ic_mini54_tqfp.html

It was a challenge (and expensive) to get a ZIF socket compatible with that size. But we've heard quite a number of requests to make the smaller sizes available, so we did. It just took a while. When you try working with that incredibly small chip, I believe you'll understand why...
 
I find Paul's advice spot on. Before committing to a series of 2000 boards I'd run a small series of 50-100 boards to see what comes up in feedback. This is independent from any financial and risk issues but generally a good practice when developing products of any kind.

Secondly I can hands down comment OSH park PCB. Stunning quality and just simply beautiful to look at. I would definitely not recommend ITead for boards as intricate as the Teensys . I've ordered 10 prototype boards from them and the drill accuracy in the many 12 mil vias on my LED driver board was pretty bad. While the 6 prototypes that I built all worked I'd certainly assume a significant failure rate on a lot of 100 or even worse 2000.
For my 100 production boards I ordered boards from Hackvana and while these also are not as super precise as OSH park boards, they are much better than the Itead boards. Also if you needed different that standard PCB thickness to 2oz copper e.g. for better thermal performance hackvana.com can provide that while ITead and OSH park cannot.

Having said the above in your case I'd look for a local provider that can deliver a professional turnkey solution. I know of no such provider in the UK but I know a company in Germany. We had them quote us on the LED driver board and they were helpful and competent. Companies like those provide a full manufacturing turnkey solution, meaning, you send them the prototyped and proven design files and they send you populated PC boards. They provide the full sourcing off components and in our case suggested a better alternative for an obsolete component.
In the end it did not work out in our case for a number of reasons ( logistics difficulties, customs uncertainties, financial risks) but due to your proximity and much larger number of boards you may want to consider them
 
Last edited:
PaulStoffregen said:
Actually, all 3 sizes are now available, including the smallest one which PJRC uses on Teensy 3.0 & 3.1...When you try working with that incredibly small chip, I believe you'll understand why...
Click to expand...

Apologies, Paul, I hadn't seen that update yet and thank you for offering the more compact versions! If it's OK with you, I'll add them as options to the Eagle libraries I have been working on?
 
Thanks everyone for your input. I will definately take it on board. I would create a test production first of about 50 and doing some testing and analysis before commiting to the full quota. I think I would create an FFC connector on a PCB board with the ATSHA204 chip that can join to the full Teensy board.

In regards to the other part of my question:

As a final question Paul said on a different thread that the raw Teensy 3.0 chip could be locked so that firmware can't be overwritten:

http://forum.pjrc.com/threads/24365-...-Firmware-Code

Could someone spell it out to me with a general summary on how I could do this?

Anyone get any ideas on this?

Thanks guys, keep up the good work.

Andy
 
I’ve done both custom teensy boards and made boards that the premade teensy plugs into.

I feel the custom board looks better and more professional, but plugging in a premade teensy is quicker and easier and fits into our business model better.

My question is... if these are going into an industrial piece of equipment do you think anyone besides me really cares or notices if it’s a full custom teensy or a custom pcb with a teensy plugged in?
 
Yet another 'it depends'.

Have two model series for my side business.

For simple low-end stuff (atmel 328s, TI 430s, etc), have never used any of the canned proto boards. But this is to mostly to control component quality, EMI, and environmental ratings.

For the hi-end stuff have used T3.1s and T3.2s for several years but never bothered to do a 'custom' processor PCB. PJRC does an decent job of cranking out a consistent product, so have never bothered to do a custom ARM cortex processor board. And none of my customers (mostly agricultural businesses) have commented on my component selection. In any case, large-scale production is not a requirement because never have sold more than a few units of hi-end stuff in a single month.

1. The T3 board is always soldered
2. Have not ever sent firmware updates to customer.
3. Control your product's design. If is does the intended job well, the customer will not care about internals.
 
Hello All,
Going through this now..... using Teensy 3.2 soldered on to a carrier PCB. Got here through: https://www.pjrc.com/store/ic_mkl02.html
But what I was hoping to find when I came across the MK04 bootloader was an idea of the error checking the MK04 implemented. We're uploading the HEX file to a Ubuntu Linux based SBC, and have successfully got the teensy command line loader working. It's been suggested `md5sum [file]` will be OK to check the HEX file integrity after transfer to the Linux SBC. And, I think it's a pretty safe bet that there is a CRC scheme running between the teensy loader and the MK04, but I really need to confirm that before we get deploying. Has anyone seen where that's written up (other than the source code on the Teensy loader)?
 
Can u tell us a bit more about your project and what it does ?

I am about to design a multi teensy processor board and am considering the mounting of a teensy to a board vs mounting the chip

It seems eagle is worth the effort !
 
amensch said:
And, I think it's a pretty safe bet that there is a CRC scheme running between the teensy loader and the MK04, but I really need to confirm that before we get deploying. Has anyone seen where that's written up (other than the source code on the Teensy loader)?
Click to expand...

As far as I know the USB protocol itself does CRC Checks. No need for a additional CRC, at least for the packets.
 
Last edited:
Teensy Loader doesn't do a CRC check.

If you really need a CRC check, the best way is to build it into your application. Frank's FastCRC library is the easiest (and fastest) way. Just have the library compute a CRC starting at address 0 and do something like print it to the USB serial or one of the hardware serial ports.
 
Status
Not open for further replies.
Back
Top