EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: John_ITIC on April 28, 2015, 11:21:36 pm
-
I've been stuck with non-working TUSB1310A USB 3.0 PHY chips from TI. They don't reset correctly. The online support is utterly useless since they have no in-depth knowledge about the parts they supposedly provide support for. It appears they have some kids that are working this support - essentially the only thing that comes back is excerpts from datasheets and specs rather than internal information from design engineers. As things stand now, the part doesn't work and i'm at loss how to proceed...
How does one get more useful support when one is "insignificant" to the chip vendors?
-
I have had a pinch of luck with the TI forum https://e2e.ti.com/ (https://e2e.ti.com/).
There are some TI employees that could lend a hand with certain issues but it appears that the forum is rather low on the priority list. In the end, the big problems I solved myself and the easy ones got an answer on the forum. This is similar to what I seem to get with FAE's that work in the distribution chain. The company touts they have an engineering staff to help, but I rarely get real help. One experience with an unnamed company the FAE spent considerable time telling me that the part of my circuit that was working was all wrong, and that the part that did NOT work was just fine.
I am sure that TI has plenty of good people, but I have not found them to be available to the smaller designers like me. Not sure where you fit in terms of volume, but that is probably a driving force. Calling Maxim is hilarious, they want an annual forecast of volume right away. When I explain that I am prototyping and essentially fishing for options - they ask again what is my expected annual volume.
If you find a channel at TI, I look forward to using it.
-
I have had a pinch of luck with the TI forum https://e2e.ti.com/ (https://e2e.ti.com/).
There are some TI employees that could lend a hand with certain issues
Indeed, i posted there two times and wrote one direct email.
They were very helpfull and from the in-depth knowledge about the product they could as well have designed it.
Never had a company help me that quickly and that competent to find a solution for my undertaking.
-
I have been communicating on the TI forum the last month but no real solutions or in-depth information has been provided.
https://e2e.ti.com/support/interface/digital_interface/f/130/t/411316 (https://e2e.ti.com/support/interface/digital_interface/f/130/t/411316)
It is quite obvious that the TI people "in the know" is nowhere near to the people working the support. The volume is only in the hundreds per year so I doubt they would spend money on providing better support. As always, one has to figure it out oneself...
/John.
-
I am also in the low-volume territory but would be happy to pay for support to make up for that. Wish that was an option.
-
100s a year is not that low volume. Plus you can spin it that you are just doing a prototype build and production will ramp up shortly. It's not the current order they are concerned with, it's the potential order to come.
I've always gotten a hold of good TI tech support through our distributor rep (Avnet I think). They bring in a TI technical field services guy who goes out and finds the internal TI engineer guy to answer my questions. If you are indeed selling product, and aren't still buying all your parts from Digikey!, I would say talk to a distributor rep and see if they can line up something with a TI field apps person.
-
100s a year is not that low volume. Plus you can spin it that you are just doing a prototype build and production will ramp up shortly. It's not the current order they are concerned with, it's the potential order to come.
I've always gotten a hold of good TI tech support through our distributor rep (Avnet I think). They bring in a TI technical field services guy who goes out and finds the internal TI engineer guy to answer my questions. If you are indeed selling product, and aren't still buying all your parts from Digikey!, I would say talk to a distributor rep and see if they can line up something with a TI field apps person.
Hundreds a year is abolutely NOTHING. I have a friend who designs products which have volumes in 1-2 millions of units per order, and that is not enough volume to order directly from TI. Have to order through a dealer. To give some reference.
-
If you purchase a partial reel per year.....it's small beans. That is me.
-
Back in the late 1990s where I worked we bought excess stock so that the vendors would support us. Say we had a 5000 part run, we'd buy 20000 parts, keep 1000 spare, nick a hundred for personal use and write the rest off. Just to get past the sales cretin.
-
We had some pretty catastrophic esd failures of some ti parts we used in a design. Our volume was pretty low (a few reels) but we had purchased them from a large distributor where we had direct contact to a sales engineer. TI went to some pretty extraordinary lengths to help us out and prepared a report to show us what had happened and how it was our fault. As it turned out it was actually our fault but it was nice to have some feedback from someone who had access to the cadence design files and tell us what we had done wrong!
My advice would be to talk to a top tier supplier (Arrow, Avnet etc) where you have access to a sales engineer and field application engineer and promise to buy from them in the future. You might get some good contacts and hopefully they will send an email to the local TI FAE to get some assistance for you.
-
If you are really, really sure about your case, calling or emailing a company with "The part does not meet specs" message usually gets attention.
-
Thanks all for your suggestions. I have decided to build a separate HSMC add-on card for an Altera development board to further investigate the behavior of the TUSB1310A PHY. My PHY usage scenario is quite different from the ordinary one (RX-only, hard-strapped P0) so the TI support seems to focus on that rather than on the observable symptoms and try to look backwards into the chip to figure out what could cause such symptoms. In other words, no in-depth analysis is attempted but rather random references to datasheets are made.
I have spent the last 20 years doing mixed software and hardware. For the hard bugs, I have never been helped by support but had to figure it out on my own. I expect this will be the same. It is expected given the complexity of modern technology. I agree that support expectations these days are non-existing except if one happen to work for a major corporation where the chip vendors see $$$ in their eyes. Especially these days with outsourcing, vendors don't have the money for supporting minor customers.
-
I used to have an annual chat with the TI UK marketing manager. 20 mins or so on the phone, but if you wanted anything like proper technical support, or an account, then your spend had to be in the millions a year. They have plenty of customers who make the 300 million televisions that are made each year, and there are far more phones made than that. Just imagine having one of your chips in the television and another in a product that sells a 300 a year. If you're a business owner where do you put your effort?
-
I used to have an annual chat with the TI UK marketing manager. 20 mins or so on the phone, but if you wanted anything like proper technical support, or an account, then your spend had to be in the millions a year. They have plenty of customers who make the 300 million televisions that are made each year, and there are far more phones made than that. Just imagine having one of your chips in the television and another in a product that sells a 300 a year. If you're a business owner where do you put your effort?
Historically yes, but the your effort should be expended on your most vocal customers if you want to run a successful business. All it takes is one person to say something bad in 2015 and bad news travels as fast as packets do between routers.
-
Every company started out with only 300 units at some point.
-
For good support you need to design in over $1M of their parts in any financial year, luckily for me that happened within January this year.
Bruce
-
Every company started out with only 300 units at some point.
I think this is the crux of the problem. Speaking from a US perspective, our markets nowadays seem rigged toward preserving the dominance of large incumbents, at the expense of innovation. A bit off-topic, I know.
I remember calling into Motorola's university support hotline as a teenager, getting free samples and a suitcase worth of free literature. Of course, you can download that stuff now, and samples are still more-or-less available, but I cannot imagine a company devoting that kind of time and attention to a hobbyist anymore.
-
As the discussion is becoming a bit general, let me state from the "dark site or the power", the supporter's point of view, as I am working as a supporter (app eng, product specialist) for semiconductor suppliers since a long time.
The rules havn't changed in all the years: observe the priority of the business case (it's not just the volume), satisfy the customer, find the root cause and solve the problem fast and final, promote other products in your portfolio.
Note: volume is not the real key factor. There are very cheap and very expensive parts out there. If a 100k per days production stopps for 1 days just because of a 1 cent part, you can imagine the pressure. Also you can in case a space transporter can't lift off because of a $500 Raid Hard FET.
But high volume gives a better statistical indication for the root cause of the problem. Rule of thumb: less than 1% failures, a good chance, the limits of the part a slightly exceeded time to time, more than 10% - seems to be a general R&D design failure. Is the problem related to a specific assembly lot (lot&date code) it is an indication, there was a drop of production quality (TI guys may tell some nice stories when they aquired Burr Brown and tried to move the production).
What has changed is the business environment.
Increasing suppliers for the same kind of product (stronger competition), constantly reduced man power (cost savings, "support is just a cost factor"). In my first years we have been 12 supporter, supporting ~1.000 customers and ~50 products. Today we are 3, supporting ~10.000 customers and ~12.000 products. This reduces the time by case drastically, there is less and less time to support "no value problems". Sure the modern tools help a lot to be more efficiant in communication, but the time to understand a problem and to offer a suitable solution, has fairly not changed. Instead you need to be up-to-date with more and more complex products (higher integration) and continuesly increasing numbers and complexity of applications (every day there are new, thats engineering), which generate their specific problems. Beside that the situation on the customers site is exactly in the same direction. This causes a steadily reduced level of universal knowledge, engineers becoming more and more "focussed" and incresing time pressure.
What hasn't changed (yet): there are still humans working out here and they love electronics, thats why they took this profession.
So I would recommend to contact the TI support line in a friendly and open/honourst style. I bet they will try to help you, if there is any chance.
BUT first you should check yourselve if you did no real simple error. Check your design against the circuits given in the data sheet and/or against existing reference designs.
Take a look to your problem out of the view of the supplier - he is selling millions of that part into hundreds of perfect working applications.
Thus the probabilty is quite low, the device causes the problem. The chance of misuse or a defect/damaged part is quite high.
I do not recommend to raise a problem as "a quality issue" just to get noticed. Its not a clever way and will have the opposite effect. You can bed, if there would be a quality problem, the supplier is well informed about already.
-
Pelule.... GREAT post!
"...He is selling millions of that part into hundreds of perfect working applications.
Thus the probabilty is quite low, the device causes the problem...."
Love that part. I need to get you to come have a talk with some engineers I know...
-
I pinged a friend at TI but no dice. Sorry.
I know it IS possible to do pcbs for this part on 4layer OSHPark rules. (5mil trace/space, 10mil/18mil drill/diameter). I did not break out 100% of the I/O when I did that, maybe it would be more possible ignoring the TX channel.
-
I pinged a friend at TI but no dice. Sorry.
I know it IS possible to do pcbs for this part on 4layer OSHPark rules. (5mil trace/space, 10mil/18mil drill/diameter). I did not break out 100% of the I/O when I did that, maybe it would be more possible ignoring the TX channel.
Thanks Marshall.
The main TI PHY debug areas are whether there is a problem using an external clock oscillator rather than a crystal. I'm also wondering whether there is an issue with hard-coding the power state to P0 rather than P2 like in your Daisho design. I'm half-way into designing a HSMC board for my Altera Cyclone IV GX dev kit - it will allow me to get access to all TUSB1310A interface signals as well as experiment with various pull-up and pull-down values.
I'll see how few layers I can get away with for this HSMC board. I'm leaning towards 6 layers so I get some decent power planes too...
/John.
-
For reference; I have determined that the TUSB1310A IC is broken with regards to the below functionality; perhaps someone can be helped by this and avoid getting stuck at the same issues:
1) The TUSB1310A DOES NOT work with an external reference clock, only when using an external crystal. PHY_STATUS de-assertion is very unreliable when using an external oscillator.
2) The internal 100 ohm termination CANNOT be disabled, regardless of the state of the RX_TERMINATION pin state.
I had to re-spin my board to work around these issues.
-
my shot : (some of these may sound stupid, but i do not know what you have tried and not tried, i always start with the simplest things first )
there is an errata notice that external clocking does not work if spread spectrum is enabled ...
http://www.ti.com/lit/er/sllz063/sllz063.pdf (http://www.ti.com/lit/er/sllz063/sllz063.pdf)
I took a quick look at the datasheet. For external clocking they want a 1.8 volt vpp signal. check your signal. especially ground bounce. ideally this signal should be a sinewave. some xtal oscialltors do dc-level restitution. Try coupling in your clock capacitively.
when using an external clock then VSSosc MUST be tied to ground. when using crystal it must be left floating. ( they most likely drive VSSosc to 'drift' the crystal for spread spectrum. they probably have a mechanism that senses this node and disables the actual oscillator block and puts it into bypass so you can externally drive it. )
what silicon rev do you have in hand ? there are at least SIX revision of this thing. Which means there are SIX errata notices too !! the errata notice for rev F does not list what was fixed in rev E...
-
Another question: how stable is the external clock? I have seen issues with an USB phy which had an unstable clock (too much frequency wander).
-
I have spent MONTHS trying to get to the bottom of this. The listed features are simply broken, TI support has been asked about it but not been able to confirm or give any additional information. The external clock has been checked with regards to jitter and is good. The SI is good too. And I'm using the latest Rev. Finally, I had to give up and re-spin the board - I should have done so sooner but stubbornly wanted to track down the root issue. This never happened. I even built a complete debugging board, where every single strapping could be altered via zero ohm resistors and where every single signal could be scoped. I never got these two features to work (external oscillator and disabling termination resistor) so must conclude broken chip (nothing in Errata).
Note: My new board spin works with regards to PHY clocking (external XTAL). I will not spend more time trying to solve the original issue so no need to give suggestions. It would be interesting if somebody could confirm these findings, though.
Edit: My external clock was a square-wave output from a microcontroller.