HP-48 FRAM Cards

[snrowe]

(07-23-2016 01:40 AM)Dave Frederickson Wrote:  

(07-23-2016 01:04 AM)snrowe Wrote:  Quote from article on page 3:
"A read operation in a FRAM is destructive because it requires switching the polarization state in order to sense its state. The read operation has to restore the polarization to its original state after the initial read which adds cycle time to the read operation."

Also page 6 the table states FRAM has a destructive read.

I see, in this case, "destructive read" has two different meanings. Magnetic core memory and FRAM employ a destructive read architecture where it's necessary to change the state of a bit in order to read the value. The original value then needed to be rewritten back. This is destructive in that the data is lost when read, but does not wear out the device. Core memory had no such characteristic. With respect to endurance, reading and writing reduces the lifetime of FRAM and flash, so one could say that FRAM is more destructive than flash because of the write-after-read architecture, but FRAM still has much better endurance. Does that help?

Just to clarify things a little more, it's not necessary to change the state of all the bits in order to read them. With core memory, and forgive me if I get the polarity incorrect as it's been a while, to read the data all ones are written to a register. A sense line then detects a pulse from the bits which were zeros and changed state. Then the zeros are written back.

Dave

Hi Dave,

The source of information on FRAM I read up on suggested the switching polarization of the FRAM bit was a wear out mechanism due to the motion of atoms required in doing so. I could not find any references that suggest what you're saying is correct, could you point me to a source that explains the read process doesn't "wear out" the memory?

Thanks in advance,
Scott

[Sylvain Cote]

TI has a small FAQ on FRAM here: PDF

[JDW]

Scott, I just reviewed the TI PDF posted by Sylvain and I must say that nothing I've seen thus far has led me to believe FRAM will wear out fast, on HP calculators or on any other computing device, unless perhaps that device exceeds 10 trillion read and/or write operations, which isn't going to happen on an HP calc, I wouldn't think.

But the fact remains these cards are overpriced. That almost makes our entire dialog moot seeing fewer people will buy them for that reason, and for the reason you also need a 48GX too.

[snrowe]

(07-23-2016 02:57 AM)JDW Wrote:  Scott, I just reviewed the TI PDF posted by Sylvain and I must say that nothing I've seen thus far has led me to believe FRAM will wear out fast, on HP calculators or on any other computing device, unless perhaps that device exceeds 10 trillion read and/or write operations, which isn't going to happen on an HP calc, I wouldn't think.

But the fact remains these cards are overpriced. That almost makes our entire dialog moot seeing fewer people will buy them for that reason, and for the reason you also need a 48GX too.

Hi,

The link Sylvain posted agrees with other sources I read up on FRAM has a read/write limit somewhere between 1-10 trillion cycles. While this is certainly a lot of cycles, when this is applied to a program constantly running off the card It could theoretically become an issue in a couple of years depending on duty cycle of memory access.

I intended to use these cards in my HP48SX running a low power ham radio monitoring station on a continuous basis (hence my concern for a large amount of read cycles). In this regard the HP48SX is better suited than the HP48GX due to it's lower power consumption (longer battery life). I do have an HP48GX, however I've grown to like the SX better. For heavy calculations I use the 50G

[JDW]

(07-23-2016 03:14 AM)snrowe Wrote:  I intended to use these cards in my HP48SX running a low power ham radio monitoring station on a continuous basis...

Now I'd love to see a video of that setup! Wow!

[Dave Frederickson]

Scott,

I agree that read/write cycles wear out FRAM, but there's a difference. Read cycles wear out a device, so that can be called "destructive read". FRAM also employs a write-after-read architecture which could be called a "destructive read architecture". It could be called this because in order to read the device it must erase the data first, hence "destructive read" architecture. Because FRAM uses a destructive read architecture is no grounds for dismissing the technology as inferior. On the contrary, even though a write-after-read architecture has more read/write cycles per data access compared to flash, it wears out at a much lower rate. FRAM can endure more than 10 billion times the number of write cycles as flash. Plus, FRAM is nonvolatile.

So because of the architecture you find FRAM less desirable for the 48SX/GX, despite the much greater endurance and not requiring a battery.

Personally, the architecture doesn't bother me.

(07-23-2016 03:14 AM)snrowe Wrote:  The link Sylvain posted agrees with other sources I read up on FRAM has a read/write limit somewhere between 1-10 trillion cycles. While this is certainly a lot of cycles, when this is applied to a program constantly running off the card It could theoretically become an issue in a couple of years depending on duty cycle of memory access.

You call 150,000 years of continuous data logging "a couple"? Compare that to less than 7 minutes for flash.

Dave

[matthiaspaul]

(07-23-2016 12:58 AM)JDW Wrote:  The data shows FRAM to have greater endurance than Flash, and most computers use SSDs now. So in effect, you're safer using FRAM in a 48GX than a computer is using its SSD! How is that a knock against FRAM in a 48GX?

The point is that SSDs and flash cards are used for data storage (replacing hard disks etc.), not to replace a computer's RAM (SRAM or DRAM), where many orders more read-write-cycles will occur.

As I wrote already, FRAM is a great idea for cards used in slot 2 (where the memory is used for storage). However, I consider it pointless to expand the calculator's RAM using a 128 KB FRAM card in slot 1, not only because of the somewhat limited life-time of the FRAM card in this configuration, but also because the other 128KB half of the calculator's memory remains SRAM - that is, the calculator would still loose its memory contents, when the batteries are exhausted. This drawback could be eliminated only by replacing the internal memory by FRAM as well...

Greetings,

Matthias

[JDW]

(07-23-2016 05:10 PM)matthiaspaul Wrote:  The point is that SSDs and flash cards are used for data storage (replacing hard disks etc.), not to replace a computer's RAM (SRAM or DRAM), where many orders more read-write-cycles will occur. ...limited life-time of the FRAM card in this configuration, but also because the other 128KB half of the calculator's memory remains SRAM - that is, the calculator would still loose its memory contents, when the batteries are exhausted. This drawback could be eliminated only by replacing the internal memory by FRAM as well...

You speak of limited life of FRAM when used as memory, then to cite the benefit of replacing memory with FRAM. :-)

By the way, has anyone ever swapped the RAM of a 48GX with FRAM?

[matthiaspaul]

(07-23-2016 10:56 PM)JDW Wrote:  

(07-23-2016 05:10 PM)matthiaspaul Wrote:  The point is that SSDs and flash cards are used for data storage (replacing hard disks etc.), not to replace a computer's RAM (SRAM or DRAM), where many orders more read-write-cycles will occur. ...limited life-time of the FRAM card in this configuration, but also because the other 128KB half of the calculator's memory remains SRAM - that is, the calculator would still loose its memory contents, when the batteries are exhausted. This drawback could be eliminated only by replacing the internal memory by FRAM as well...

You speak of limited life of FRAM when used as memory, then to cite the benefit of replacing memory with FRAM. :-)

Yes, it might look like a contradiction, but I was just trying to answer several potential questions at once. It all depends on the goal someone wants to achieve:

If an extremely long operational life is paramount, this rules out the idea of using an FRAM card as memory expansion in slot 1.

If, however, someone can accept a somewhat limited (but probably still long) operational life, but wants to have a calculator with persistent memory for hassle-free long-time storage of the calculator, so that he can continue calculating exactly where he left the calculator ten years earlier, he might again consider FRAM. This won't work, though, because the other part of the memory would still be battery-powered, so the calculator would have to be reset when that battery is exhausted. Ihis idea would work only after replacing the internal SRAM by FRAM as well.

This isn't an issue, however, if the FRAM card is only used for storage, as much fewer cycles will occur in this configuration. So, using an FRAM card in slot 2 sounds like a good idea and it would save the user from the ever repeating the error-prone procedure of replacing the batteries. Storing the calculator away for ten years, the user would lose the calculator's RAM contents, but not the programs and data on the FRAM card, so he would be back in "business" very soon.

Greetings,

Matthias

[Dave Frederickson]

(07-24-2016 02:28 AM)matthiaspaul Wrote:  If an extremely long operational life is paramount, this rules out the idea of using an FRAM card as memory expansion in slot 1.

Matthias,

What do you consider an extremely long time?

If we're to believe the information in the articles above, FRAM has endurance similar to that of SRAM. According to one of the articles, FRAM can endure 150,000 years of continuous data logging.

100 trillion write cycles in 150,000 years works out to around 21.1 writes/sec. If we expected the FRAM to endure for only 50 years, we could write data at 665 million writes/sec, continuously. That's a lot faster than the 48GX, so even if you could write data continuously you could do so for a lot more than 50 years.

Dave

[matthiaspaul]

(07-24-2016 04:29 PM)Dave Frederickson Wrote:  What do you consider an extremely long time?
[..]
If we expected the FRAM to endure for only 50 years, we could write data at 665 million writes/sec, continuously. That's a lot faster than the 48GX, so even if you could write data continuously you could do so for a lot more than 50 years.

That's a good argument, I didn't made that calculation myself.

Thanks and greetings,

Matthias

[JDW]

I just wanted to add to this thread by letting everyone know that I wrote the EBAY seller, slvr00gt, about why he no longer is listing the 128k FRAM cards for the 48GX and he replied that he will list more around Aug. 24, 2016 (2 weeks from now).

I also asked him about performance. He said that while FRAM is MUCH faster than the old SRAM cards that need batteries, the 48GX limits performance such that the FRAM card is only about 20-30% faster than the SRAM cards. But honestly, that's not bad at all; and to top it off, you never need a battery for the FRAM card.

The 128k size is ideal for merging with main RAM (you can merge a card larger than 128k). My intention would be to use the FRAM card to store Meta Kernel.

[JDW]

I wrote the EBAY seller of the FRAM cards, since he is the only person on this planet who appears to sell them (and who appears to be the designer of the cards too). I asked him when the 128k FRAM cards would be available. One week ago, he told me 2 weeks, so that means they will be ready for sale 1 week hence. But last night I asked him why he isn't selling 1MB or larger FRAM cards. (He sells 1MB and even 4MB SRAM cards, which require batteries.) He said that he would need 10 or more orders for the 1MB FRAM cards at $200 to make the changes necessary to create them.

My thinking is, it would be useful to have a 128k FRAM card in Slot 1 of the 48GX to merge with system RAM, and it would be useful to have larger memory (1MB will do, I guess) in Slot 2.

It's all very pricey for an old calc like the 48GX, I know, but this is FRAM we're talking about, which never needs a battery and which is actually faster than the SRAM cards of old, albeit not as fast as the FRAM cards could be due to the speed limitations imposed by the GX itself.

So I am mentioning this in this thread so as to determine if there is sufficient interest. Obviously, the seller isn't going to make a board design change just to accommodate me and sell me a single 1MB card. Nor do I need ten 1MB cards for myself. But if there is interest from 10 of you, then it becomes a possibility. So please let me know your interest here.

I am not advertising for a seller (although perhaps I am indirectly). I am just being selfish and wanting a 1MB FRAM card. :-) But in sharing this info maybe I am not being too selfish. So please let your interest be known. Thanks.

[Ron Ross]

Why not just buy two or three of these FRAM cards? Use the extra cards in your 48GX and swap out as you move from one large program to another if you really have that much software that you run from your GX.
.
I have do have that specific RAM card combination in my Hp 48GX (128K and 1 Meg) that I bought from Koltz when he still made cards. Yes, it is a nice combo, but even when I used my Hp 48GX, I really just used the 128K (merged) and loaded some other software in the upper banks that I actually never really used. Yeah, I got bragging rights, but I didn't actually need or use that extra 1 Meg of RAM ie another 128K card would have been enough. I had a better system than the FHBB Hp 49G, but after the Hp 49G+ came on the scene, I never used my Hp 48GX all that much (not that I use my 49G+ extensively, although I did wear out its buttons). Today I do keep a lower tier Hp 48G+ in my desk for some higher end problems that I occasionally have (which has only 128K available)
.
I am actually biased towards a good pocket scientific and hope the new third party DM42L becomes available. As far as Hp, the last respectable scientific was their Hp 32sii.

[JDW]

(08-16-2016 01:04 PM)Ron Ross Wrote:  Why not just buy two or three of these FRAM cards?

Well, each of those 128k FRAM cards sells for $138 (not including shipping). So while buying 2 at that price might be justifiable, buying 3 or more does not.

But basically what you are saying is that a 128k FRAM card merged in Slot 1 and another 128k FRAM card sitting in Slot 2 is all that would be needed on the 48GX?

The way I look at it is this. If you put Meta Kernel on the Slot 1 card (128k), that entire card is used up for MK alone. Then you only have the 128k in Slot 2 to store libraries. The same would be true if using SpeedUI instead of Meta Kernel, if you install the full versions of SpeedUI, and then you add ALG48 and an Equation Writer (on your Slot 1 card). You'd really only have the card in Slot 2 for your other libraries.

So do I understand correctly that you have a 1MB FRAM card? (I ask because I don't know precisely what Oliver Klotz made in the past. All his EBAY listings are defunct now.) If it is a 1MB FRAM card and if you don't need it anymore, I might know of a good home for it. :-)

[Ron Ross]

JDW's Quote " But basically what you are saying is that a 128k FRAM card merged in Slot 1 and another 128k FRAM card sitting in Slot 2 is all that would be needed on the 48GX?"
.
Well, it would have been ALL I would have needed. No, my cards are NOT FRAM as they are about 15 years old and use a battery (I bought in addition to the Hp 49G FHBB calculator, the one with Meta Kernel built in) so that I could make my Hp 48GX as good (actually Better) than the Hp 49G.
.
A quick ebay search revealed that 512 KB FRAM cards are currently available for the Hp 48Gx. These would allow you 128Kb ports 3-6 which I suspect would be plenty unless you are a student taking different types of coursework. And if you were such a student, I would strongly suggest you migrate over the Hp 50G and consider a student edition of a math package instead.
.
I also saw a 1 Meg RAM card very similar to mine for a similar price AND it had room for 3 more 1 MEG chips, so that looked to be a prospective 4 MEG card if you have any skill at soldering.
.
Your Hp 48Gx is a nice calculator and may last another decade or even two, but the typical Hp 48Gs that fail, usually suffer from foam compression failure that can affect even a new in the box calculator, so for anyone to actually support this line is putting faith in a calculator with a KNOWN service flaw that may affect 100% of these calculators.
.
Don't misunderstand me, I use an Hp 48G+ as one of my two calculators at work because I like it more than the Hp 50G for work and day to day usage. But that is because I am a Calculator snob used to the older Hp calculator buttons (not only because of the ENTER KEY location) as the Hp 50G doesn't quite feel as good and I don't use the extra features of the Hp 50G (and as I have said, I seldom use the advanced features of the Hp 48G).

[JDW]

Thank you for your reply, Ron.

I am not a student. I graduated from engineering school back in 1994, and I purchased my first HP calc, a 28S, in 1990. I've been using that same 28S for the last 26 years. But recently it has developed problems with a couple keys, such that I must press down on the top of the key as normal, but I must also press against the back of the calc too, in order to get those keys to respond. That problem prompted me to investigate the purchase of a replacement 28S, since I am not into drilling holes in my existing pristine 28S (which would make it look like a horrid monster).

I found a new-in-box 28S on EBAY, which included the manuals. I wanted the manuals too because I lost mine years ago. So I used BUY IN NOW and won. I then sent Paypal payment to the seller only seconds after I won. But after several days the seller never accepted my payment. Knowing his GMAIL address, I put him into my circle on Google+ and found out his real name, then I proceeded to Google him and found he was arrested earlier this year. So I stopped my payment and just waited out the EBAY 32 days for the auction to be auto-terminated. I guess the fellow was back in jail because he never responded to any of my emails, nor my contact on FaceBook (yes, I even found his FB page).

So I then searched Craig's Lists nationwide (there are search engines for that) and found one seller of a new-in-box 28S with manuals. This fellow is a collector of all things HP! He's only selling because, apparently, he is in deep financial straights now, having been out of work a long time. I even called him on the phone and talked an hour. But in the end, he never sold it to me because he said his car broke down on the way to an interview and now he lacks the means to even make a trip to the post office!

So I took this two cases to heart and said to myself, "a replacement 28S in good condition is not meant to be." I wanted a new or near-new calc so I could be assured of many more years of use before the keys developed problems.

And so, I went back to EBAY, fingers crossed, and found a 48GX seller. I sniped the auction (naughty me) and won, paying $123.85 with shipping. I then heard good things about the 50g, so I bought a refurbished unit for $51 with shipping. I also bought Thomas Barber's "Definitive User's Guide to the HP 48g/49g/50g" from Samson Cables, although I've found it pretty dry reading and lacking in graphical illustrations. For that category of book, Mier Jedrzejowicz's "Customize Your Hp-28" tome is much better. Anyway, I then purchased a wickedly expensive $40 USB cable for the 48GX (which included $10 in shipping) from myteknovo on EBAY. I've been using that to transfer libraries from my iMac to the 48GX via the excellent KERMIT tool Hoppi.

I love the 48GX because it reminds me of the good build quality of the 28S. The 50g isn't bad, but it doesn't "feel" the same. Old timers should know what I mean by that. But boy is the 50g fast. And newRPL by Claudio looks to make that speed demon even faster. To me, that's exciting.

So, as you can see, I am now in the HP "enthusiast" category, although I wouldn't necessarily consider myself a "collector" insofar as I intend to use these devices and pass them to my children as they enter advanced math in upper high school.

I've been participating in the forums here since my 48GX and 50g purchase, and it was here I read about FRAM tech. To me, that is exciting because you don't need to worry about batteries. Perhaps a battery matters little when you use the calc all the time, but what if many months go by and you don't use it? So you see the benefit to FRAM. It completely eliminates worry about data loss.

I've not decided what I will ultimately do, but when the EBAY seller comes out with his new lot of 128k FRAM cards, I will likely buy one. Thank you for alerting me to the 512k FRAM card. It's from the same seller.

Anyway, the fact remains that if 10 people want a 1MB FRAM card, that same seller will modify his PCB layout and make them for sale. I just want to mention it to see if there is any interest, despite the fact the 48GX is old and perhaps not too worthy of expensive add-ons like that anymore.

Best wishes.

[matthiaspaul]

(08-10-2016 01:54 PM)JDW Wrote:  I also asked him about performance. He said that while FRAM is MUCH faster than the old SRAM cards that need batteries, the 48GX limits performance such that the FRAM card is only about 20-30% faster than the SRAM cards. But honestly, that's not bad at all; and to top it off, you never need a battery for the FRAM card.

I haven't looked this up, but AFAIR the HP48 uses parallel asynchronous SRAM. If so, there's no feedback from the memory chip to the processor to indicate that the requested data is available or has been written successfully. Instead, the processor uses a fixed timing to access the memory and the memory has to be chosen just to be fast enough to fit into that timing. If it is too slow, obviously bad things will happen, but if the memory is faster than necessary, the processor will never know and therefore cannot take advantage of it.

You could take advantage of it, if you'd overclock the processor, but this will throw over the timing with other peripherals.

If the processor has configuration registers for the memory timing, it might be possible to reprogram the timing to become faster, but the processor won't do this by itself.

Greetings,

Matthias

[matthiaspaul]

(08-16-2016 01:56 AM)JDW Wrote:  So I am mentioning this in this thread so as to determine if there is sufficient interest. Obviously, the seller isn't going to make a board design change just to accommodate me and sell me a single 1MB card. Nor do I need ten 1MB cards for myself. But if there is interest from 10 of you, then it becomes a possibility. So please let me know your interest here.

I would be interested in a 2 MB (or even a 4 MB) FRAM card for the HP48, but not at this price point, I'm afraid.

Also, I find the design and finish of that seller's cards somewhat distracting. There's so much that could be done better that I'm considering making my own design... (If I only had the time... ;-)

Greetings,

Matthias

[JDW]

(08-18-2016 09:56 AM)matthiaspaul Wrote:  I find the design and finish of that seller's cards somewhat distracting. There's so much that could be done better that I'm considering making my own design... (If I only had the time... ;-)

Maybe I should hold out for your edition then? :-)

Seriously, from day one I was concerned about that FRAM card design (even the SRAM cards from that same seller) simply due to "static discharge." As you know, all the old HP and third party cards were made very well in that they were completely covered. Not only did that protect the expensive circuitry from death-by-static, but it also covered the PCB so your eyes are never distracted!

So if anyone would kindly sell us FRAM cards at half the price of that EBAY seller, I think we could sell quite a few, with myself being first in line!