MC9S12D32 with mask set 0M89C?

[imajeff]

I got this part the other day from somewhere maybe DigiKey, the MC9S12D32MFUE. When I tried to look up this particular maskset (0M89C) by searching for the D32, I find many many masksets, but none are this. Finally I used Google to eventually find the errata which is on Freescale website, just they don't apparently know about it. Why could I not find it any way but google?

[Alban]

Hello Jeff,

Files are associated to products or other documents. If they are not associated, you can find them through search but not by navigation.
You could have found it by looking under the MC9S12D64, DJ64, DG64 and A64.

Technical Publication has been notified and the file should appear on the D32 page in 24-48 hours, time for info to propagate to all caches.

The file will now also be associated with MC9S12D32.

Thanks for pointing this omission out.
Alban.

[imajeff]

Hey the Main Moderator Man, Thank you for your help with this.

I've dug further into the docs for "D64" which claims to cover D32. It seems to still omit some things. I look to see what ID it should have, and it says the 0M89C is D64 but does not mention D32 or what I should see in the MEMSIZxx registers.

I just hope that FSL is also aware of this, and is working on complete documentation. I understand that this is "preliminary" stuff.

At least it still gives this which is important to me: Figure 1-3 MC9S12D32 Memory Map out of Reset

[imajeff]

Hmmm,

I just found that also pg. 29 (MC9S12D32 Memory Map out of Reset) shows that the part has 4K Bytes RAM, mappable to any 4K boundary.

That's also wrong. Maybe I should just write a book first, and then submit that to FSL

[imajeff]

I said it was wrong to say "4K" when there is only 2K, but later I noticed the other notes. It is actually 4K address space, but it points out that the lower half (0x0000..0x07FF) are not usable. To me that sounds ridiculous. Why did you not just say:

"The D32 has a whopping 64K of RAM. the lower 2K and the upper 60K are not usable"

I actually did find the reason that it says 4K. There are indeed 4K of RAM on my MC9S12D32MFUE, yet the doc says half is not usable. Since it is not usable, why does it work? Sure I didn't do an extensive RAM test on it, but someone has insisted before that FSL would not do this sort of thing where they have a bunch of D64, something doesn't qualify, so they sell it as D32 because it qualifies for that. Sorry if you don't get what I'm saying.


Message Edited by imajeff on 2007-08-07 01:34 PM

[kef]

I agree that "lower half is not usable" is not enough to understand it properly. I guess it means that even with pure D32 maskset (you have D64 marked D32) and only 2k of RAM, RAM block will occupy all 4k of address space. It's important to know how big remappable memory is and what can it overlap. If I understand it correctly D32 maskset will have 2k of RAM but will occupy 4k of address space.

[imajeff]

That's a good try, but the lower 2K of RAM on this D32 is usable. Yep, I have 4K RAM on this D32. I don't know if you noticed that even the same maskset (0M89C) is also a DJ64 device, so naturally it will have to map the RAM the same way it does for a D64.

I'm just wondering if anyone knows why they give me 4K working RAM when it is only documented 2K.

I would guess that it was going to be a '64 but didn't qualify, but people said before that FSL would not do such a silly thing. That leaves no explanation.

[imajeff]

Also confirmed:  I programmed some Flash in ppage 0x3c and 0x3d. The bits stick.

It's what I suspected... this is so far just like the DJ64, and in fact is the same maskset. Of course the D32 costs significanly less 

Howabout this: Maybe since I got these two D32 as samples, they are just telling me to pretend they are D32 but gave me DJ64 with D32 written on it because they had plenty. That's even worse because when I evaluate D32, I want to know just what I would get.

[JimDon]

I would imagine that the 32K part is not longer manufactured as a 32K part.

Any why would they, as by now the the yield from the 64K mask is the same (meaning they cost the same to make).

On the other hand, you do need to pretend, as the next order could be from an older batch.

Not sure who at FSI would have said "we never do that" because it has been standard practice since day 1 in the industry to substitute a higher spec-ed ( but compatible) device as the process improves.

What I am saying is this is not an FSI thing, but an industry thing.

Of course they should say that by identifying the mask set as that for the 32K part....

As for what you get, not to be a wiseguy by stating the obvious, but the spec sheet for the 32K part is what you get.

[imajeff]

Good that you understand that it does not cost more to make a device with "more memory", specifically when they've already made the one with more memory first.

[at the end, I'll explain what's wrong with what they've done]

I don't see that you comprehend that if they were going to make the 32 after making the 64, but then soon start selling 32 but give 64 hardware, then there was no reason in the first place to ever actually make 32 hardware. What a waste of resources that would be. But look, the D32 is brand new. Have they even really started selling it yet? I got samples direct from FSL.

That tells me that it could only be opposite of what you said. They are starting out with using DJ64 to represent D32, in order to get people started. I might even guess that they are using this to get an estimation of how many customers they will have interrested in that small amount of memory, and find out before they have to actually make the hardware have that small amount.

Now I almost forgot to explain what's all wrong with this; here's my background:

I've been using, of course, DP256 first because that's what was intended to get people started. I eventually dared to buy a large batch of DG128B, to save some bucks in production. I found right away that little things can be missed where my application was relying on DP256 resources that it did not need to. I actually had to try it on the DG128 to see that I still needed other things changed such as where RAM starts. Finally I got it, but then I bought DJ64. Before I actually tested it, I found that D32 was finally comming available, so I got samples of that. Now that I'm testing these, I see that I can't really know if I've properly created a memory layout that will work on the D32 because it is not actually a D32, but a '64. Anyway, I can't do anything about it unless I just want to say "forget embedded development" and start another career, you know. Just switching to another manufacturer will not mean I won't have different trouble with them making decisions that mess up customers.

So have fun.


Message Edited by imajeff on 2007-08-11 11:18 AM

[JimDon]

Well, ya the cost is the same but it's all about the yield.
In theory anyway, the 32K part would have a better yield, so that it's final cost would lower.

However I have to believe on a simple, by todays standards, part like this the yield would be about the same.

The real question is this - is the PARTID that of a 32 or 64?

[imajeff]

The PARTID is DJ64 (of course it points to a specific maskset), and then it is also documented elsewhere as being D32. I thought I've already said that is the worst part about it.

Then again, also, the MEMSIZ registers also indicate it is a '64. I suppose that when or if they ever ship something labled "MC9S12D32" which truely is D32, then they would have to change this MEMSIZ information accordingly (which is not documented anywhere I've searched). But this is not my problem, as I'm not doing a runtime check of the device resources.

My problem is just ensuring this release of software will run on both DJ64 and D32, but I have no way to verify it without D32 hardware.

[Nabla69]

Hello JimDon,

I think there's a misunderstanding between what you say is Industry practise and what Imajeff is bringing up.

FSL does not create a device and if it does not pass sell it at a lower grade one. This is not Industry practise.

What is happening is AGAIN about money, at the customer benefit, so there should not be anyone getting over-excited this time.
The customers WANT a 32KB to get a cheaper part.
FSL does not find it economical to create a 32KB YET because of low demand. Therefore they run from the bigger one, here DJ64.

IF it is economically viable, a real D32 will be created.
The additional 32KB on the silicon is costing money to FSL, but is cheaper than creating a whole new set of masks, tests, part numbers...

In the meantime, you have a DJ64 at the price of the D32.
Even if FSL does not guarantee what is not present in the D32. It is most likely to be fully functional.

Alfred.

[imajeff]

That's what I would conclude; thank you Alfred

I'm glad it's not me who was misunderstanding because I make thing more dificult when I do, but I shouldn't get so cranky when another person misunderstands. I'm working on that

So I don't mean to sound upset that I got a better part than I ordered. It's just confusing. I would never know when I suddenly get only half the resources. And maybe they should have a disclaimer somewhere visible, for those who thought their software has qualified already for the D32. Anyway I'll try not to repeat myself, just read my complaints about that again...       Cheers (to getting more than we bargain for)

[Nabla69]

Hello Jeff,

I see several ways to take benefit from that
Order a big batch from the distributor mentioning the MASK SET.
Do a test yourself and pocket the difference.

I agree it's a bet on FSL
I also know you are controversial... even if I don't often participate, I read daily !
Anyway, I strongly want to correct that I am gladly here to help and do not wish to offence anyone in anyway

Al.

[imajeff]

Yes thank you Nabla (I can't call you Al because I reserve that for Alban). I believe your answer is most informed...

But hey, it's not me that's controversial -- I'd be fine if everyone stopped disagreeing with me

But seriously, I try to be complete and reasonable, rather than lie or claim ignorance just to avoid controversy. I just try to help. If I'm wrong, please somebody correct me (and I don't mean give me a bunch of garbage because I'm not stupid). The problem is that I get paranoid thinking nobody cares about accuracy. Well I don't give up until I get it right anyway, even if people are mad at me (and I'll try to be more polite because that's part of it; my mother (bless her heart) taught me to argue and she gets as annoying as anyone gets).

I always say "nobody like a genius", and it's almost everyone who proves that to me every day. They don't care that what I said was correct. I just be myself and point it out if I see error.  Thanks again.