Too little memory for SoftDevice 113, no valid QDID

[PKaempf]

Dear Madam / Sir,

we started developing a new electronic device which uses Bluetooth 5.0 four years ago and are now ready to certify it. However, during testing we had to realize that using Nordic's SoftDevice 112 will let us fail five out of 97 tests (see appended test summary). This could be remedied with the use of Nordic's SoftDevice 113, but its memory footprint is too large for the hardware we use (Nordic nRF52805). We expect that we can pass some of those five tests using workarounds, but one test in particular (L2CAP/LE/CID/BI-01-C) needs a credit based connection which is not supported in SD112.

Those tests are in detail:
1. L2CAP/LE/CID/BI-01-C -> Needs Credit Based Connection, which is not supported in SD112.
2. SM/PER/PROT/BV-02-C -> Needs apparently the Security Manager, which was optimized out. Will not fit into memory.
3. GATT/SR/UNS/BI-01-C -> Worked with SD113
4. GATT/SR/GAR/BV-06-C -> Still has issues
5. GAP/BOND/NBON/BV-03-C -> also needs Security Manager
We think that we can pass the GATT and Security Manager-related tests, but this will require workarounds.

Given that we would have passed all tests a year ago and the shifting standards pulled the rug from under our feet, all I need is a still valid component QDID on which to base the qualification. However, all the component QDIDs for the 52805 are expired. Please advise how to get a product using your nRF52805 certified in 2024! Long term we have the possibility to switch to a 54-series chip, but those are commercially not yet available. In other words, if we cannot get this design qualified, we will not be able to continue our work as planned.

Regards,

Peter

[Susheel Nuguru]

Peter,

What do you mean when you say this

Given that we would have passed all tests a year ago and the shifting standards pulled the rug from under our feet,

Do you mean that you are now trying to qualify to a higher bluetooth standard or did the tests in the qualifications changed to include credit based connection? 

If you are trying to get qualified for BLE5.0 for the first time and the credit based connection is requirement to pass it, then suddenly you are adding the requirement on your product that it should pass L2CAP based feature. L2CAP was never supported on S112 so it looks like the requirement on your product changed and with nRF52805 there is little to no maneuvering space with memory for this kind of surprises. 

If you have previously qualified your product already which is based on a component that is expired and trying to requalify, then the rules about "component expiration" are changing as of July 1 - there won't be this expiring concept anymore. So maybe you will solve your problem if you can just wait until then and use the new qualification process. If that is not an option then you might have to work with someone from the BT development team to look at what options are available for the 52805.

[PKaempf]

Dear Susheel,

thank you for taking this up!

We use BLE for streaming the data from inertial sensors, using a custom characteristic. My understanding is that we need to certify this based on an existing and valid QDID. All SoftDevice 112-based QDIDs are older than 3 years and have expired. So we are forced to go with the existing SoftDevice 113 QDIDs and the test plans resulting from them. Those test plans include requirements which only SoftDevice113 supports, but the limited memory of the nRF52805 will not allow to load all functions in combination with our firmware. We skip things like pairing and Security Manager to get something that will fit into the 52805. SoftDevice 113 needs 4 K more memory than SoftDevice 112, so there is no chance in hell that we will get anything usable running on the 52805. Therefore, only a SoftDevice 112 based firmware will fit, but there are no existing and still valid qualified designs left which could be used as the basis for a valid qualification.

What we have qualified before is a similar system using the nRF52832 with integrated antenna. The 52805-based device is new and has not been qualified before. We were planning to go with SoftDevice 112, but had to learn when we started the qualification process a few weeks ago that all QDIDs based on SoftDevice 112 have expired. To get the process running in Launch Studio we can only select SoftDevice 113-based QDIDs, and as you say, this leaves me no maneuvering space.

It's not as if we are adding the credit based connection of our own free will - we are forced to do this in Launch Studio to get anywhere. But this is a dead end, so we are pleading for your help. I cannot imagine that we are alone with that!

By the way, Mattew Parker of Bluetooth SIG support recommended we turn to Nordic with this problem. To ask us to work with BT will not solve this. What I need is not finger pointing but a still valid QDID based on SoftDevice 112.

Regards,
Peter

[Susheel Nuguru]

Hi Peter,

PKaempf said:

All SoftDevice 112-based QDIDs are older than 3 years and have expired. So we are forced to go with the existing SoftDevice 113 QDIDs and the test plans resulting from them

I think there is not much we can do from our end too. But if you can wait until July 1, like I said, then you can use new qualification process that will allow you to use the QDID of the S112 even if you had some component from 5 years ago you could still use it. In your 2nd question there, I think that is the easiest and path for you that does not include increase in BOM cost to upgrade nRF chip with bigger memory to use S113 in the chip.

[PKaempf]

Hi Susheel,

so the answer is more delays and forfeiting our existing qualification ticket. Please be aware that we will need to pay the higher administrative fee anew for the privilege to have to wait after July 1st for starting the qualification all over again. 

Honestly, this is quite disappointing. Nordic let the previous qualifications needlessly expire and left their customers stranded. And Bluetooth.org's reputation also has taken a deep dent by this shady maneuvering. Trying to explain this with the need for better interoperability (which is the stated reason for all that qualification smoke and mirrors) is utterly futile.

At least we can use the existing hardware design, so there is a way out. But I will not be able to logically explain what happened to any halfway intelligent person. 

Regards,
Peter

[PaKa]

So the QDID for the S112 listed in the docs section, 138610, is an end product listing. End product listings do not expire so this is still valid and is the QDID you should reference when using the S112 in a new product. The End Product QDID is valid even if the components it's built up with has expired QDIDs.

Did you use this QDID when trying to register your new design? if not, can you try to use that as that should still be valid and should allow you to register your design with the nRF52805 as is today.

[PKaempf]

Hi PaKa,

of course we tried. But since this QDID is for an end product, no modifications must be made. We, however, need to implement a custom characteristic for our product to function, so we need to start from a component QDID. And Nordic has orphaned those for the nRF52805 (a device they still actively market, btw.).

I wish you could prove me wrong!

Regards,
Peter

[PaKa]

If you mean a "proprietary" characteristic, then you don't need to do anything for that. It is considered outside of BT qualification rules. This would in that case be something that would only affect a profile subsystem, which sits on top of the Nordic end product QDID. So anything you change there would only affect that profile subsystem and not the underlying QDID.

You are allowed to add proprietary characteristics without affecting qualifications.

if this doesn't cover what you are trying to do then please provide more information so we can look into it. Note: we have always told customers to use the End Product QDIDs and almost all designs with nRF devices are using these. If you look at all the products out there you can see clearly that you can use these QDIDs for a wide range of final solutions.