Any DIY5 pics out there?



#66

I was wondering whether some DIY5 pics have been posted already - for all of us who didnt't go the the last HC meeting? Thanks.


#67

There isn't a DIY5 yet. At HHC 2011 I showed a DIY4 calculator modified to use a graphic display. I refer to that as the DIY4X. The short answer to your question is that I don't have any good photos or video at the moment. We got the graphic display in August, and were in quite a hurry to get it working, so our current display adapter board is somewhat of a kludge, and not very robust. At the moment it is broken, or I'd take some photos and post them.

Longer answer:

When we first got the display this past August, we weren't sure that it would be satisfactory, so rather than actually modifying the calculator, we kludged the display, which uses a SPI (synchronous serial) interface, to plug into the SD card slot.

When we saw how amazing the display is, Richard Ottosen designed an adapter board to plug into the 14-pin connector for the DIY4 character display module. This allowed us to put the calculator into a reasonable package, rather than having the kludge hanging out the SD slot. This is the model I showed at HHC 2011. I don't remember whether I showed the SD slot kludge.

Richard etched the adapter board himself, and there were some problems with the etch, which resulted in the adapter board not being as robust as we would like. Richard warned me that it might not survive being mailed to me, or being taken to the conference. As it turned out, it worked just fine until I took it back to the hotel room after the Sunday conference session was over. whereupon it stopped working.

Rather than fix it or make any more manually etched boards, Richard is going to design a better adapter board and send it out to a PCB fab house. We will use that to update all three existing DIY4 boards. We are considering making a few more DIY4X calculators with the new adapter, but the longer-term plan is to redesign the main board to include the graphic LCD support, as well as some other design changes including USB support. That will be the DIY5.

When we make more units of the DIY4X with the new display adapter, or we make some DIY5s, we intend to make a small number of units available to outside firmware developers. I've already got HP-41 firmware running on it, but we want to support newly-created calculator firmware as well. That development can take place in parallel with our efforts to develop a production-quality industrial design.


#68

Appreciate the explanation - pls. keep us posted on the progress and good luck with the project.

Edited: 13 Dec 2011, 5:06 p.m.

#69

Richard reminded me that he sent me some photos of the DIY4X during and after assembly before he shipped it to me in August: http://www.flickr.com/photos/22368471@N04/sets/72157628413263533/


#70

Eric,

Really great development! Especially the display looks like running circles around everything else I know in the calculator market. Do you plan to allow for softkeys in the DIY5? What are your plans for the function set?


#71

The DIY5 is intended to be available with at least two different keyboard layouts.

One layout will be an -41 style layout, the same as we currently have on the DIY4X, and is not intended for soft keys, although there is nothing in the hardware that would prevent the firmware from offering them. There are only five keys in the top row. The function set of the provided firmware would be similar to the 41CX with some extensions to support the SD card and such. It could support alternate firmware if someone wanted to write some.

The other planned layout is the Pioneer/42s/20b/30b layout, which is intended for soft keys. I believe we can package the DIY5 with this keyboard layout in a package of comparable size to the Pioneer. While it is expected to have a MicroUSB connector (device only, for connection to a PC or Mac), we have not yet decided whether to include a MicroSD card slot. The issue is that SD cards (including Micro) can draw up to 100mA, which is too much for lithium coin cells to reliably provide.[*] We're looking at three possibilities:

  1. No MicroSD card, lithium coin cells, slim packaging comparable to Pioneer
  2. With MicroSD card, two AAA alkaline cells, more wedge-shaped packaging to accomodate the AAAs at the display end
  3. With MicroSD card, color OLED display[**] instead of B&W, lithium polymer rechargeable battery, slim packaging
This model is intended as a firmware development platform, so the early units would be provided to firmware developers with only a hardware test program in the flash memory. It would be up to the firmware developers to decide what functions they want to implement.

If you have a desire for a keyboard layout that doesn't match either the 41 or Pioneer, we'd be interested in trying to accommodate it.

Eric

[*] CR2032 coin cells are only intended to provide a peak current drain (pulsed) of about 6.8 mA. When they are fresh, they can do this with only about 0.3V drop. We've found that they actually can supply enough current for an SD card, but at a higher voltage drop. However, near end of life, when you would most desperately need the SD card to work, so as to save the calculator memory to the SD card before the batteries fail, is exactly when it won't work. Near end of life, the internal resistance of the coin cell is very high, so a current spike from the SD card will drive the already low battery voltage below the level that can keep the processor operating. In other words, you'll lose the data in memory that you were trying to write to the SD card.

[**] Oddly enough, the color OLED display is significantly *less* expensive than the black-and-white display. However, it sucks a lot of power, so I don't think it's practical without a rechargeable lithium polymer battery. Personally, I don't want to have to recharge my calculator frequently, so I prefer the black-and-white display and prefer to focus my development efforts on that.


Edited: 14 Dec 2011, 7:03 p.m.


#72

*Me wantsss!!!* :)
FWIW, I'd strongly prefer option 2 above - not 1 because I like the SD, and not 3 because I strongly dislike rechargeables and am all for standard, easy-to-find batteries. And I don't really need color screens on calculators.

Cristian


#73

The point of option 3, with the OLED display, isn't to make people that want a color display and rechargeable batteries happy. The point is that it costs a *lot* less than the high-quality monochrome LCD. And I really do mean a **lot**. The monochrome LCD costs over $45 each in quantity one, and not much less than that in quantity 1000.

Guess what a calculator has to sell for if has a $45 display. Hint: increasing the component cost of a product by $X usually increases the retail price by at least $3X, and more commonly by $4X.

This means that if the rest of the components in the calculator cost only $1, the retail price would still have to be over $135. Unfortunately, I can't get the rest of the components for $1, or anywhere close to that.

The OLED display, on the other hand, costs under $14 in quantity one.

If cost was not a concern, there is no question that I'd want the monochrome display. It looks absolutely beautiful. At the conference, people thought it was a static mockup until they saw the display change as I pressed the buttons.

In fact, the display looks even better in reality than it does in Richard's photo of the oblique view of the assembled DIY4X, because in that photo there was glare from the mylar covering the display. With a real production case, there won't be mylar over it. The top view with no overlay is closer to reality, if you mentally edit out the diagonal yellow stripes of the protective film. It's too bad that photo doesn't show the four-level stack or annunciators.

Note that this is a graphic display, 400x240 (about 173 dpi or 68 dpcm), showing an open-source font, Liberation Sans, which is derived from Bitstream Vera. This font was used because it was readily available and I was in quite a hurry to get it working in time for the conference, not because I think it is an ideal calculator font. The real -41 uses slashed zeros rather than dotted, and I would prefer that. Since it is a graphic display, you will be able to use any font you like, subject to only a few limitations.

Edited: 15 Dec 2011, 3:17 a.m.

#74

Eric,

Thanks a lot for your elaborate explanation. As Christian, I vote for option 2, for the same reasons. I look forward to see this baby progressing. Wish you all the best. If there's any support I can supply, don't hesitate to contact me.


#75

How strongly do you guys feel about needing the MicroSD slot, given that there will be a MicroUSB (device only) interface, allowing transferring your programs and data to/from a computer?

Would you feel that you needed the MicroSD slot if you had both the USB port, and some hardwired internal flash memory for program and data storage?


#76

I don't like swapping cards much. What about Bluetooth or USB on the Go?


#77

We can support USB device (for connection to a host PC), but the calculator cannot be a USB host because it cannot provide 5V at 100mA. To do OTG with "limited power sourcing" still requires at least 4.4V at 8mA, and even that is difficult and expensive for us to provide. Also, the OTG protocol introduced a whole lot of added complexity beyond the already absurd complexity of normal USB, and I don't have time to develop firmware for that.

What is your proposed usage model for OTG? In other words, what other OTG device would you use with the calculator, and for what purpose?

I've considered Bluetooth as an optional module that could be inserted in the battery compartment, but I'm not convinced that it would be cost-effective. The new "Low-Energy Bluetooth" is appealing, but normal Bluetooth devices such as your existing laptop or cell phone do not support "Low-Energy Bluetooth", so there's a chicken-and-egg problem.


#78

I was thinking of connecting two calculators for data transfer via BT or USB. The same can be achieved with IrDA but modern devices lack the technology in favor of BT.

#79

To me it will be a question of the available internal memory.

The user will want some of course. The firmware will need lots. With a MicroSD the space becomes effectively unlimited without it things are more constrained.


I'd expect a firmware image for a device like this to be much larger than for e.g. the 34S -- not so much for the actual mathematical support which needn't be a lot larger. Rather, all the other extras will consume space.

Fonts e.g. aren't small especially when they are a bit map on a high resolution display. I'd expect a couple of fonts will be required at least or a scalable font.


- Pauli


#80

There will be one megabyte of flash in the microcontroller chip. Presumably any that isn't taken up by firmware (including fonts) can be used for user storage. The erase page size is 4KB and that endurance is 20,000 erase cycles per page. The programming current is 7mA in low-power mode.

What I'm proposing is that we could include some additional nonvolatile memory accessible over an SPI port. In the current DIY4 we have a 128KB FRAM, which is byte-writeable and has effectively no endurance limit. We could instead use a normal SPI flash part, or an SPI PCM (Phase Change Memory).

SPI flash memory costs about $1.40 for a 2MB part with 4K sectors and endurance of 10,000 cycles.

SPI PCM costs about $5 for 16MB, is written in 32-byte pages, and has an endurance of 1,000,000 writes per 32-byte page.

Flash uses up to 30mA, and PCM uses up to 50mA, so either is still enough to be a problem for coin cells, and would require the AAA battery option.

What I was trying to avoid was the need for AAA cells, but since internal flash or PCM would need AAA cells anyhow, it seems like having the MicroSD slot is probably the best option. I was really hoping to find an option that was workable with CR2032 coin cells, in order to have a very slim calculator, but after looking at the data sheets for representative nonvolatile memory chips it seems unlikely.


#81

One megabyte of flash seems adequate but not copious. Assuming 128kb of available non-volatile RAM and 1Mb of flash, the flash disappears quickly:

  • 128kb RAM backup -- this seems useful on the 34S with its 2kb of non-volatile RAM and would only be more useful with more RAM.
  • 128kb fonts -- guessing here but a small vector font is in realm.
  • 128kb or 256kb of user flash -- how much would people want? Don't answer all of it :-)

That's half and we've not started the firmware proper. The 34S firmware has had a lot of size optimisations made many of which really ought to be avoided in future and there are quite a few things I'd do differently a second time around. At a very minimum 256kb and likely more -- people always want more functions which use more space: integrated complex numbers, proper matrix support, units, real fractions, unlimited length integers, interval arithmetic, formulas, ...

Anyway, a built-in flash device seems unnecessary for user storage.

The ability to access a read-only flash device might still be useful. Firmware updates via SD card or even firmware on SD card freeing almost all the CPU's flash for the user. Just random ideas here now.


Is there a reason for not including a FRAM part for user storage?
Is the current draw too high or the cost? Being able to drop down to a zero power standby state would be amazing.

This would also free up all the CPU's internal RAM to be volatile if desired. An amount of volatile RAM is really useful too -- the 34S makes very extensive use of this 4kb. A device with plenty of non-volatile RAM wouldn't need it but quite an amount could end up wasted for housekeeping tasks.


- Pauli


#82

Quote:
One megabyte of flash seems adequate but not copious.

It is *eight* times what you have now, and you're already complaining! :-)

Quote:
how much would people want? Don't answer all of it

Of course people don't want 100% of it. They want 300% of it!

Quote:
Is there a reason for not including a FRAM part for user storage?

The density of FRAM is really low compared to other memory technologies, and the price is high. The current DIY4X has a 128KB FRAM which costs about $10. That same $10 can buy 16MB of flash or 32MB of PCM.

On the other hand, the FRAM does have a few advantages. It doesn't need to be erased in sectors like flash, it has effectively unlimited endurance, and it only requires 2.5 mA, so it can be used from coin cells.

The question is whether having 1MB of internal flash, 128KB of FRAM, and a USB port, would make it satisfactory to not have a MicroSD.


#83

Quote:
It is *eight* times what you have now, and you're already complaining! :-)

Complaints are cheap and easy ;-)

With a pile more flash, we'd (the 34S triad) would give the user a lot more space, we'd not spend the huge amounts of effort trimming the code, we'd add more functions, .... It would be easy to double or quadruple the size of the current firmware -- still only half of the 1MB suggested which is lots of headroom for anything I can imagine a calculator being asked to do by its users.


Quote:
The question is whether having 1MB of internal flash, 128KB of FRAM, and a USB port, would make it satisfactory to not have a MicroSD.

A USB port is essential -- backups and upgrades demand this. Could a slave port be put to other use? No idea. But this one is easy to get out of the way. I/O is a must. Without USB, you have to have the SD or serial.


128KB of user (F)RAM is more than a 100x100 matrix or many tens of thousands of programming steps. Without a large library of routines like the 41CL can call on, I don't see this running out quickly. Then add flash backups and storage to further increase the capabilities. Thus, to my mind at least, 1MB/128kB has to be enough to avoid requiring any SD card, allowing the smaller form factor case.

More interesting, to me at least, would be 128KB of FRAM (almost all for the user) + 128KB of (sometimes battery backed) internal RAM (firmware) + 1MB of internal flash (user + firmware). There are housekeeping operations that will require extra storage. The 34S's main ones are some of the matrix operations and some of the statistical functions. We don't have compacting garbage collection which is a must on an always-on device. The 34S has twice as much volatile as non-volatile RAM and we're using it all or close enough to all of it.


- Pauli


#84

The current plan definitely has USB (device only). No decision has been made on the serial port; the development units will have it but production might not.

The three main options under consideration are:

  1. high quality monochrome LCD, USB device, no MicroSD, 128KB FRAM, two CR2032 coin cells, slim, hideously expensive
  2. high quality monochrome LCD, USB device, MicroSD, two AAA cells, wedge shaped, hideously expensive
  3. color OLED, USB device, MicroSD, lithium polymer rechargeable battery, slim, only moderately expensive

I can predict the response. Everyone wants "high quality monochrome LCD" and "only moderately expensive". Sorry, those two features are mutually exclusive due to the cost of the display.

Edited: 15 Dec 2011, 6:28 a.m.


#85

I'd be happy with option 1, with this level of hardware the 42S can finally be topped as the best RPN keystroke programmable. As I mentioned, I feel that MicroSD is something that can be lived without. I also quite like slim calculators :-)


Or option 4 (you knew it was coming):

colour OLED, USB device, no MicroSD, 128KB FRAM, lithium polymer rechargeable battery, slim, only moderately expensive.

Something must be impossible here :-)


- Pauli


#86

I don't seen any technical or cost problem with your proposed option 4. I just question having expensive FRAM built-in when there's enough power to use cheap MicroSD cards.


#87

Thanks for the clarification, of course you are right -- I'm not thinking entirely straight tonight.


- Pauli

#88

The current DIY4X has 16KB of RAM, all of which is battery-backed (vs. 6KB with only 2KB battery-backed in the Atmel AT91SAM7L128).

The new ones (and any we retrofit) will have 128KB of RAM, all of which is battery-backed.

With either chip it is possible to turn off some of the RAM to save power, but the savings is so small that I haven't bothered to try it.

#89

Right now the DIY4X has only 16KB of RAM, though the next ones we build (or the current ones if we retrofit them) will have 128KB of RAM.

Due to the 16KB RAM limit of the current calculator, I do NOT use a frame buffer in RAM for the display, as that would take about 12KB. The display contains its own internal memory, which cannot be read back. The display can only be written in increments of one full horizontal scan line (400 pixels).

The way my firmware works now is that any update to the display loops over the individual scan lines of the update area, render one line, and sends it to the display. An optimization would be to use DMA for the transfer to the display, and to start composing the next scanline while the current one is being transferred.

With the 128KB RAM, it would be possible to have a full display bitmap kept in RAM, but I don't think it's a good utilization of the memory, unless it is needed for graphing functions.

The current DIY4X firmware contains three sizes of the Deja Vu Sans Mono font. The large one used for the stack and alpha-mode displays contains a superset of the 41 character set, with a total of 116 glyphs. There is no compression, in order to keep display updates fast, so it uses 19KB of flash.

The tiny font used for the annunciators does not have the full set of glyphs, and occupies 1084 bytes of flash.

The third font in an intermediate size used only for showing debug messages, and occupies 2032 bytes of flash.

I don't think it is desirable to render a scalable font in the calculator due to both memory and performance constraints.


#90

Those font sizes are much better than I'd figured. Very easy to work withing those limits at least. The 34S's fonts are 122 bytes for the 7 segment characters and 2152 bytes for the two fonts. This represents a fair potion of available flash and the bitmaps are densely packed.


The 34S doesn't use a frame buffer either, in fact we use the LCD RAM as extra semi-volatile storage between keystrokes. The screen is regenerated after every user action and some commands. Given the space cost of a frame buffer, this seems like a sensible thing to continue.


Anyway, 1Mb of flash seems more than adequate.


- Pauli

#91

Personally I'd say dump the USB (it was not a Microsoft invention and therefore *even now* the support in MS operating systems is flaky).

Transferring/backup/upgrades via MicroSD is good. (SD is all I use on the 50G). These days PCs and laptops have SD slots as standard (Most take MicroSD, but SD converter cards are still relatively easy to come by).

#92

I like Option 2. Standard cells are much preferable to lithium ion packs and the problems they pose for replacement. A colour display might be nice for graphs, but for a calculator it's just not warranted. I'm definitely going to like the alternate 32s/20b/30b/32sii layout.


#93

How do you feel about AAA alkaline vs. CR2032 lithium coin cells?

In particular, ignoring the SD card issue, and all other things being equal, do you prefer a slimmer calculator with coin cells, or one with a more pronounced wedge shape with AAA cells?

How much does battery life enter into your preferences? Bearing in mind that more or larger cells result in a larger calculator, which would you prefer out of the following choices:

  1. one CR2032 lithium coin cell, six month battery life
  2. two CR2032 lithium coin cells, one year battery life
  3. two AAA alkaline cells, two year battery life
  4. three AAA alkaline cells, three year battery life
  5. four AAA alkaline cells, four year battery life

[Note that the battery life figures listed are purely hypothetical, as battery life depends on how much you use the calculator. However, the ratios of battery life in the choices are a reasonable approximation, since a AAA alkaline cell stores roughly twice the energy of a CR2032 lithium coin cell.]


#94

Two 2032 cells with a replacement interval of one year is good. I wouldn't be bothered by the additional thickness required to accomodate AAA cells. Two AAA cells and two years of life is sufficient. More AAA cells would edge over into the realm of "too big". I carried around a big old TI85 since it came out. It would be good to have a powerhouse calculator in a smaller package.

Something else that may need consideration is what sort of non-volatile memory to use. It seems to me that programs are to be stored in flash memory. I think that's not prudent given the limited rewrite cycles of flash cells. I believe supercapacitors are small enough to provide enough juice to keep RAM alive while you change the batteries. This segues to the SD card issue. I'd prefer to keep my programs there instead of in a flash soldered to the calculator's PCB. That way I can pop in a new SD card if things get dodgy. I'm already dealing with this sort of problem with digital cameras. It should be very clear to all of us that we have a tendency like something and stick with it for a very long time. Otherwise we'd be posting at ticalc.org. Projects like this should be planned with this tendency in mind.

#95

Option 2. CR2032 just doesn't have grunt to exploit high processor speeds for lengthly programmes.

Actually AAAA would be an option. I have heard they can be difficult to find, but for me they are available in my local Tesco store. Not only are they cheaper than CR2032, but still have longer life (~3x Ah, 1.5x Wh) and higher instant current drain capability (less than 1/30th internal resistance).

Another option is two N cells, giving 3V in the space of about one AAA cell, but probably still better performance than two CR2032.


#96

We're avoiding both AAAA and N because they are a lot more expensive at retail than AAA. By the time other industrial design considerations are taken into account, AAAA and N don't actually allow me to make the case much slimmer than with AAA.

Your Tesco store either has much lower prices on AAAA, or much higher prices on CR2032, than anything around here. We have CR2032 for less than the price of a AAA, but AAAA for much more than a AAA.


#97

AAA: pack of 4 for £2.00 (cheaper/cell in bigger packs)

AAAA: pack of 2 for £4.00 (haven't seen them in bigger packs)

N: pack of 2 for £4.00 (haven't seen them in bigger packs)

CR2032: pack of 2 for £4.00 (bigger packs only available on the internet)

On the face of it the N and AAAA cells are the same as the 2032's, but in reality two N or AAAA cells in series will still beat the pants off two 2032's in parallel (especially w.r.t. internal resistance, but also more capacity. In reality I think you will get more than double the life because of the significant lower drop in voltage at peak currents).



Addendum: Of course buying on the internet is much cheaper (but AAAA remain indeed reasonably expensive).

Edited: 15 Dec 2011, 12:40 p.m.

#98

If you really want this to be DIY, a good compromise battery option exists. Power it from 2 or 3 AAAA sized cells that users obtain at low cost by opening up a 9 volt alkaline battery. The cells inside are not quite AAAA size but close enough to be interchangeable with store bought AAAA cells. They are around 600mAh and are under 8 mm in diameter. The advantage that these have over CR2032 is a much higher discharge rate without damage to the cell and will easily be able to power a microSD card. The advantage they have over AAA cells is that they are about 2.5mm less diameter.

There are 6 of them inside a 9 volt battery. Amazon sells Energizer 9 Volt industrial batteries for $0.83 each, that's under $0.14 per AAAA cell which makes them about the least expensive option. All you need is a needle nose pliers and 30 seconds of time to get at the cells.


Edited: 15 Dec 2011, 10:11 a.m.


#99

DIY has just been the "code name". Don't read too much into it.

8.3mm max dia still requires more of a wedge shape than the 3.2mm max thickness coin cell, especially since the battery compartment design effectively requires two thicknesses of case plastic for cylindrical cells, while the coin cell only requires one.


I find a wedge preferable to coin cells.

My personal preference would be option 3 or 4. It is more than just a matter of battery life for me. There is also the issue of having the correct cells available. I always have some AAA cells lying around but generally have to go out to buy CR2032's. I know that someone is going to say: "Well what's the bug deal if you only have to do it once a year"? There is some validity to that argument and I agree that it is not a big hardship, but personally I would still prefer the AAA cells. The added size and wedge shape don't concern me too much.

Cheers,

-Marwan

Edited: 15 Dec 2011, 11:29 a.m.

Any of those options would be okay by me. In other words, if you can get 6 months of battery life out of it, I'd be happy.

I'd be more concerned with the affect on form factor of the different battery choices. Consider things like the balance and whether it can be comfortably held in one hand too.


I was initially concerned about balance with two AAAs at the display end of the calculator. Two alkaline AAA cells together are about 23g, for a weight of less than an ounce. I did a very unscientific experiment of putting two AAAs in the display end of a calculator mockup, and didn't find it to be much of a problem. The HP-41 card reader was fairly heavy, and I don't recall much complaint about that. (But lots of complaints about how quickly it drained batteries.)


Also consider disposable lithium AAAs, which have higher capacity, the same voltage and lower weight than alkalines.

Frequent recharging is highly undesirable. Standard rechargeables, like AAAs is what I would prefer. If it works one month or more with moderate use on 2 rechargeable AAAs I would be extremely pleased. One week between charges and heavy use would also be acceptable to me.

Some wedge shape is a small price to pay for standard batteries. If it also gives MicroSD, the better.

I vote for (2).

Put simply, I do not want to recharge often, and I do not want to go out to hunt for new batteries several times per year either.


I'm going to recommend against rechargeable AAA, for the same reason that most calculator vendors do, which is that you won't get much (if any) warning of battery low before it dies. We can set a reasonable battery low threshold voltage detection for alkaline, but not for NiMH or NiCd.

Option 2. For same reasons as I noted above plus I really like the convenience of a Micro SD card for backup and transferring data.

I am not as strongly anti-color/anti-rechargeable as some others. My one concern with that approach would be an easy way to replace the LiPo cell/battery when it gives out which is bound to happen eventually. I hate gadgets that make it hard or impossible to replace a warn out battery.

Cheers,

-Marwan


Any option that involves AAA batteries. Easy to obtain power source plus it makes some of the other options easier to incorporate from a power drain perspective. :)

Quote:
I hate gadgets that make it hard or impossible to replace a warn out battery.

I agree with you completely. Unfortunately, there are good reasons why this is done.

If we did use a lithium polymer cell in the DIY5, we would state that it was not user-serviceable, and we would definitely not encourage end user replacement. However, for improved factory serviceability it would use a connector rather than being soldered in place. The case would use "security screws".

Having given an appropriate warning that the calculator is not user-serviceable, I'm certain that everyone would respect that warning, not buy themselves a set of security bits, not open their calculator, and not replace the lithium polymer cell.


LOL

Yes. For sure. No "programmers with screwdrivers" in this bunch.

BTW: i vote AAA cells, or even Katies AAAA idea. I'd probably use rechargables anyway. I use them for a lot of things. And what my vote lacks in technical knowledge may be partially made up for by being one of the few (2?) people who have actually bought one of your calculators.

Quote:
  • With MicroSD card, two AAA alkaline cells, more wedge-shaped packaging to accomodate the AAAs at the display end
  • With MicroSD card, color OLED display[**] instead of B&W, lithium polymer rechargeable battery, slim packaging
    [/ol]
  • I've grappled with this issue routinely as well.
    Having demountable storage seems really convenient.
    However one question to consider is whether demountable
    media really fits into a typical user's workflow.
    Demountable in the event of near-wearout may be all that is
    needed and in the case of 1e6 cycle endurance devices you may not
    even need to provide for this assuming a write to NV memory
    is a power-off event.

    Accommodating the current draw of an SD card starts to
    push the device outside of a grab/use/drop/forget-till-next-time
    use model if that is your primary target. From personal
    experience I have more high power, high performance devices
    which require daily electron feeding than I really care to
    own. If demountable media is a must-have, SIM cards might
    be an alternative or even a similar custom design approach
    using chip scale EEPROM, etc.. assuming you can live within
    about 128KB capacity.

    Concerning power source, if the footprint is extended to house
    common AAA cells, I don't think anyone will miss losing
    problematic CR2032s. But the impact to the form factor
    may be objectionable to some as they will likely contend for
    z-axis space with the display.

    Quote:
    [**] Oddly enough, the color OLED display is significantly *less* expensive than the black-and-white display.However, it sucks a lot of power, so I don't think it's practical without a rechargeable lithium polymer battery. Personally, I don't want to have to recharge my calculator frequently, so I prefer the black-and-white display and prefer to focus my development efforts on that.

    As long as you're selecting from the cell/media handheld
    component supply pipe, everything is unnaturally cheap.
    But an illuminated display may be functional overkill for
    a calculator as whatever task you're using it to assist, itself
    normally requires illumination.


    Quote:
    Demountable in the event of near-wearout may be all that is needed

    We've actually considered putting the MicroSD card slot in the battery compartment rather than exposed on the outside of the calculator. That would have the advantage of making it less likely for contamination to enter the calculator through the card slot, and less likely for the card to fall out unexpectedly and become lost. (I've had that happen with the 49g+ and 50g.)

    On the other hand, it's less convenient if the user wants to routinely swap memory cards. I don't get the impression that many calculator users expect to swap cards frequently.

    Quote:
    If demountable media is a must-have, SIM cards might be an alternative or even a similar custom design approach using chip scale EEPROM, etc.. assuming you can live within about 128KB capacity.

    Although I briefly considered it, I don't think we want to use any kind of proprietary or uncommon removable media.

    If we were to include a small (e.g., 128KB) built-in nonvolatile memory, it would almost certainly be FRAM, both to avoid endurance limitations and because FRAM requires very little energy to write compared to other nonvolatile memory technologies.

    However, from a cost perspective, it is very difficult to justify including both a MicroSD slot and an internal FRAM. Similarly, although we'd like to include both USB and a serial port, we can't justify the cost, and we expect the majority of users to take advantage of USB, with only a small minority having any use for a serial port.

    Quote:
    Concerning power source, if the footprint is extended to house common AAA cells, I don't think anyone will miss losing problematic CR2032s. But the impact to the form factor may be objectionable to some as they will likely contend for z-axis space with the display.

    I'm not sure which axis you're calling "z", but the plan is to put the AAA cells behind the display, as you look at the calculator from the front. Based on feedback received from questions asked months ago, and some of the comments in this thread, it seems that most people are willing to accept a wedge-shaped calculator to accommodate the batteries, but don't want a calculator that is significantly longer than the 42S/20b/30b case.

    The current DIY4X with a 41-style keyboard is nearly as long as the 41 with a card reader. The only way to make it shorter is to ditch the graphic display, or tighten the keyboard row spacing. The 42S/20b/30b case has two fewer rows of keys, so it will be about the same length as the 42S.

    What I personally don't like about the 42S/20b/30b case design is that it is too wide to comfortably operate one-handed, which is how I often use my 41. This is due to the keyboard having six columns rather than five, and there's no way to fix it other than reducing the keyboard column spacing. However, I haven't heard a single complaint about the width from anyone other than Richard and myself, so apparently this is a non-issue for the majority of potential customers.


    According to Wikipedia, FRAM does indeed have cycle limitations. Maybe have a look at MRAM or PRAM?


    I would rely on the information in the manufacturer's data sheet on the part, rather than on Wikipedia. Wikipedia is correct that there are limits on the endurance, but that isn't the whole story.

    The endurance specification for the part in question is 10^14 cycles, and you only access the FRAM when the user pokes the buttons on the calculator to tell it to do a read or write operation, the endurance is effectively unlimited.

    Even if the user writes a user-code program that just beats on the FRAM mercilessly, 10^14 cycles is going to last a **very** long time. Just how long is left as an exercise for the reader.

    MRAM in great. It's available in slightly higher density than FRAM, but also is more expensive. It uses a lot more energy to write, so it is less desirable for memory for a calculator.


    Quote:
    Just how long is left as an exercise for the reader.

    Couldn't resist :-)...

    Well, using my trusty HP-41CV...

    At 10,000 writes per second the memory would last 317.09 years (if I did the math right). I'd say that for all intents and purposes that is unlimited.

    Cheers,

    -Marwan

    Quote:
    What I personally don't like about the 42S/20b/30b case design is that it is too wide to comfortably operate one-handed, which is how I often use my 41.

    You are not alone, I've thought the same thing :-)

    For hex arithmetic, six keys wide is pretty much required however.


    - Pauli


    While I understand the aesthetic appeal of having A through F in a single row, in practice I don't have that big an issue with putting "F" on the first key of the second row.

    IMHO, six-columns keyboards are a blessing for HEX mode, because you can set the A...F digits entry keys in a (redefined) row; having "A" to "E" on one row and just "F" on other row is not as easy or intuitive...

    I don't know if this has been mentioned elsewhere, but it would be nice if the USB port would power the calculator when plugged in like it does with the 50g.

    Cheers,

    -Marwan


    Quote:
    but it would be nice if the USB port would power the calculator when plugged in

    With disposable batteries (vs. echargeable), it would require a significant change to the calculator power supply to be able to do that. Richard designed the power supply for very high efficiency in the normal mode of operation, which is going from battery voltage of 2.0-3.0V (or slightly higher with fresh alkaline cells) to 3.1V. We'd have to add a buck regulator to the board to deal with input voltage from USB, which is at least 4.4V, because we'd have to step that down. I'm reluctant to add the cost of another regulator and associated components, but that's probably worthwhile enough to justify it.

    There are combined buck-boost regulators, but the boost-mode efficiency isn't as good as our dedicated boost regulator.

    Quote:
    I don't know if this has been mentioned elsewhere, but it would be nice if the USB port would power the calculator when plugged in..

    I think that would be a given, as some return for piercing a
    hole in the enclosure to the outside world. I'm not particularly
    in love with USB, but it is so commonplace and inexpensive
    there really isn't any better choice for simple wired external
    power + communications in this class of device.

    Unfortunately it is somewhat of a weak link in terms of reliability
    for device lifetimes expected here as the connectors are typically rated for a maximum 5K mating cycles
    assuming ideal conditions, considerably less otherwise. Perhaps a
    moulded silicone plug shield would be a reasonable option to
    consider.

    On a different train of thought, I've kicked around the feasibility
    of avoiding external connections altogether, approaching an
    environmentally sealed enclosure.
    IR will get bits in/out however a simple ISM RF link will put
    less of a dent into the energy budget, have essentially no
    orientation issues, and avoid the need for IR window logistics
    in the enclosure.

    Charging current can be induced by magnetic flux. PV cells are another possibility -- perhaps remote due to the logistics
    involved. However the concept of a calc sitting idle on the
    desk harvesting waste energy for later appeals to me. But even
    using primary cells is reasonable. Eg, in the case of
    cylindrical AAA cells it is relatively straightforward
    to create a battery well closed by a round plug/o-ring.


    Powering the DIY5 from USB is most definitely NOT a given. It's a feature we'd like to have, but it's not as easy as one might imagine. We would have to either add a buck regulator and two schottky diodes, in addition to the boost regulator we use from the two AAA cells, or switch to using only a buck regulator with four AAA cells.

    If we stick with the two AAA cells and add a buck regulator for USB, we introduce a schottky diode drop in the battery path, which takes away from the usable battery life.

    The problem with using four AAA cells and only a buck regulator is that the buck regulator has to be on all of the time, even when the calculator is "off", because it has to supply at least 1.85V to the processor to maintain memory and timekeeping. While some buck regulators have a "burst mode" for dealing with low-load conditions, the quiescent current is usually far too high. This afternoon we found a buck regulator that might do the job, but we'll have to evaluate it.

    Now I've started to wonder what power supply chip(s) the 49g+ and 50g power supplies use, and why they switched from three AAAs to four.


    If you're expecting to get at least a year of calculation from two AAA cells, then powering the DIY5 externally doesn't sound very useful. Think about how often you need to change the batteries in a TV remote and recall that they typically don't have recharge jacks. USB should be for communication only. It would be nice to make the jack easily replacable if the need arises. Recessed and under a rubber plug sounds good.

    Concealing the SD card in the battery compartment also sounds like a good idea. I don't anticipate swapping it around a lot. Maybe occasionally when loading a bunch of new programs. In any case, it should be removable without having to first remove the cells.


    A year from two AAA cells is based on assumptions that are not likely to be valid for all users. In particular, not for "power users". As I wrote when explaining the options, really only the relative lives of the different options were approximately accurate, and not the absolute times indicated.

    The original plan with two AAA cells and USB was that the power from the USB would power the USB transceiver, but nothing else in the calculator, such as the processor. As I've described, the reason for that plan was the lack of suitably buck-boost regulators. However, there is a problem with that plan, which is that as long as the calculator is plugged into USB the main oscillator has to run at 48 MHz, so the power consumption is higher than usual. This means that using USB would significantly reduce the battery life.

    I've described some of the alternatives previously, and we're now considering how to proceed.

    I don't think we can make the USB receptacle "easily replaceable". It's soldered in, after all. The only products I can think of that have "easily replaceable" USB receptacles are USB host adapters (e.g., for PCI, CardBus, PCI Express, or ExpressCard), where they are replaced by replacing the entire host adapter.

    We probably can't have the receptacle recessed or under a rubber plug, even though I agree that it's a good idea, because it complicates the industrial design too much. If we were experts at industrial design, or could afford to hire people that were...

    I wanted to put the SD card in the battery compartment, and would certainly prefer that it be replaceable without removing the AAA cells, but doing that requires an extra PCB, two connectors, and a flex jumper, which increase cost, decreases reliability, and makes it harder to do the industrial design.


    Edited: 17 Dec 2011, 12:06 p.m.

    Quote:
    We would have to either add a buck regulator and two schottky diodes, in addition to the boost regulator we use from the two AAA cells, or switch to using only a buck regulator with four AAA cells...
    If we stick with the two AAA cells and add a buck regulator for USB, we introduce a schottky diode drop in the battery path, which takes away from the usable battery life.

    Unless you have a need for efficient conversion off the usb
    power rail, a linear regulator can be used to drop the usb
    voltage. And rather than
    suffering the diode drop, a high side p-channel switch on the internal cell gated from the usb rail would disconnect the
    internal cell when usb power is present.

    I suppose if you wanted to get fancy you could add an inductor
    (and free-wheeling diode to ground)
    between the p-channel switch and internal rechargeable cell(s),
    then PWM the gate to charge the internal cells via buck conversion. When usb power isn't present the inductor DC saturates and
    effectively won't cause a voltage drop.

    Quote:
    The problem with using four AAA cells and only a buck regulator is that the buck regulator has to be on all of the time, even when the calculator is "off", because it has to supply at least 1.85V to the processor to maintain memory and timekeeping. While some buck regulators have a "burst mode" for dealing with low-load conditions, the quiescent current is usually far too high. This afternoon we found a buck regulator that might do the job, but we'll have to evaluate it.

    I faced a similar issue with the prospect of 3V cores powered
    from the legacy voyager 4V5 button stack. The current draw
    of smps control circuitry is going to exceed that of a
    linear regulator at these current budgets. Using a
    low quiescent current linear regulator in tandem with a
    buck converter would have addressed the issue of data retention
    power. The SoC can perform the switch between linear and smps
    as needed for power on/off. The problem I had particular to my
    scenario was finding a
    linear regulator with a quiescent current comparable to a NUT
    in a powoff state in order to approximate the fabled decade+
    shelf life a NUT leveraged from SR44s.
    No regulators really come close but TI has some IIRC with about
    a 1uA Iq. I believe since then I found a linear regulator from
    Analog Devices which can best that.

    Quote:
    Unfortunately [USB] is somewhat of a weak link in terms of reliability
    for device lifetimes expected here as the connectors are typically rated for a maximum 5K mating cycles
    assuming ideal conditions, considerably less otherwise.

    When looking for a reliable USB connector, check female micro-usb (NOT mini-USB). With the female micro-USB being adopted as a standard charge input, high-reliability connectors are available. IIRC, the baseline is 15K cycles, and the 'springs' are on the male part (cable end) which is easily replaceable.


    The plan is to use a Micro-B receptacle.

    If we were going to support USB OTG, we'd use a Micro-AB receptacle. However, OTG presents too many challenges for our power management design, and for firmware development.

    Power management for OTG would be easy if we used a rechargeable lithium-polymer cell, because there are single-chip solutions for that. Implementing OTG using primary cells is much more difficult, especially because we need very high efficiency.


    Quote:
    The plan is to use a Micro-B receptacle.

    If we were going to support USB OTG, we'd use a Micro-AB receptacle. However, OTG presents too many challenges for our power management design, and for firmware development.

    Power management for OTG would be easy if we used a rechargeable lithium-polymer cell, because there are single-chip solutions for that. Implementing OTG using primary cells is much more difficult, especially because we need very high efficiency.


    If power is the only/major reason not to do OTG, leave power out but do the rest. In other words, you cannot call it OTG, but all functionality is present except for power. Then it is easy to use a special cable which activates OTG and supplies power to the USB device, or a special powered USB hub which activates OTG and does not need power from the host.

    (I've seen reference to several devices which do this, but the only one I own is the HP TouchPad.)

    A quick check on the calc's in my desk drawer (HP, TI, Casio, Canon) shows that most have five columns of keys on the numeric keypad and six columns on the function keys above them. Almost all (including HP35s and WP34s ex HP20b) are between 76mm and 82mm wide, the exception being the very petite Casio FX-82 Solar (FX-260 in US) at 70mm without its slide-on case.
    I've sometimes wondered if there is merit in having a slim calculator with coin cells for memory retention and hand-held use, with an AA-battery desk cradle giving a better viewing angle for desktop use. However AFAIK such an item has never been available, so maybe it's not so good an idea....


    Possibly Related Threads...
    Thread Author Replies Views Last Post
      OT--Nat'l Semiconductor NS-7100 pics anywhere? Matt Agajanian 25 1,815 12-15-2013, 08:39 AM
    Last Post: gene wright
      So any pics / videos from HHC 2012 bluesun08 6 349 10-01-2012, 04:49 PM
    Last Post: aurelio
      DIY5 calculator progress Eric Smith 7 303 09-13-2012, 05:01 PM
    Last Post: Walter B
      Restoring an HP65 (trying to .... few pics) Alberto Fenini 31 1,246 06-29-2012, 10:55 AM
    Last Post: Peter HOFMANN
      DIY4X/DIY5 progress! Eric Smith 6 252 03-08-2012, 06:51 PM
    Last Post: Eric Smith
      WP 34S - Pics of blue slant overlay on real calculator Bob Cortopassi 10 503 02-21-2012, 12:39 PM
    Last Post: Dominic Richens
      Seeking DIY5 update and webpage David Griffith 1 153 02-11-2012, 01:27 PM
    Last Post: Eric Smith
      strange classic power supply (pics) Alberto Fenini 4 326 01-16-2012, 02:06 PM
    Last Post: Ethan Conner
      DIY4X, DIY5 update Eric Smith 5 229 01-05-2012, 01:39 AM
    Last Post: Eric Smith
      anyone recognize these pics from old HP manuals? Palmer O. Hanson, Jr. 3 246 12-10-2011, 08:10 AM
    Last Post: Don Shepherd

    Forum Jump: