HP 20B Prediction weirdness?



#11

Hello,

Can owners of the HP20b confirm this, or at least explain to me where I am wrong - I know I am, because I know HP is not!

1 - Turn off HP20b

2 - Press&Hold [N]+[Amort], press [ON/CE] (this will reset the calc to defaults, not necessary, but to make sure we are starting from scratch)

3 - [fn],[Mode], set to RPN

4 - [ON/CE]

5 - [fn],[Data]

6 - enter the following data:

X(1) = 0

Y(1) = 1

X(2) = 1

Y(2) = 2

X(3) = 2

Y(3) = 4

X(4) = 3

Y(4) = 8

(in other words, Y = 2^X)

7 - [fn],[Stats]

8 - on "2 Vars", scroll down to "Predictions"

9 - [INPUT]

10 - [INPUT] until b*a^X

11 - scroll to check that a = 2 and b = 1 (correct)

12 - scroll to "Pred X", enter "3", [INPUT]

13 - scroll to "Pred Y", [=] (returns 8, good)

14 - now scroll back to "Pred X", [=] (returns 2.81, what the heck?)


2.81 is the predicted X for Y=7. If you go back and forth predicting X and Y, Y goes down in steps of "1"

Same happens with the b*e^(aX).

Linear data (resulting from a a*X+b) works fine (you can go back and forth from predicting X and Y, and the values stay put).

After testing with data that correlates to each of the prediction functions, it looks to me that:

b*a^X : weird

b*e^(aX) : weird

a*X+b : ok

a/X+b : weird

b*X^a : weird

a*ln(X)+b : ok

Where am I messing up? I hope I am, because I have no JTAG cable to update the calc firmware...

Best regards,

Joao S Veiga


#12

This seems to work ok in the very latest version of the firmware, dated 4/5/2010. I'm not sure when the fix was implemented.

-Katie


#13

Ouch.

How do I know the version of the firmware in my 20b? I purchased it just last week from Amazon (but 4/5/2010 is pretty recent, can't blame Amazon...).

Also, is there a release notes for firmware (so I at least know what else is wrong in mine?)

Thanks!

Joao


#14

Quote:
How do I know the version of the firmware in my 20b?

Press ON+PMT, then down arrow until you see "SW Version" scrolling on the display.

Quote:
I purchased it just last week from Amazon (but 4/5/2010 is pretty recent, can't blame Amazon...).

I have no idea of what firmware versions are out there. The HP people here, Cyrille and Tim should know.

Quote:
Also, is there a release notes for firmware (so I at least know what else is wrong in mine?)

As far as I know, this is not available from HP. However I really think that it should be posted on their website. Not necessarily a list of all firmware changes but at least a list of known, important bugs and how to get a replacement calculator if these bugs affect you. This will go a long way towards inspiring confidence in their new calculators and maybe get HP past having to produce the 12C forever.

-Katie


#15

Quote:
This will ... maybe get HP past having to produce the 12C forever.

Katie, that got a laugh out of me!
#16

Get your cold clammy hands off my 12c+.

:-)

Actually... they are now on the HP 30b more often than not!
Gene

#17

Thank you Katie!

Mine is 8/01/2008 :-(

Agree with you, without at least a list of corrected bugs, how can I trust that this is the ONLY bug on my firmware version?

I bought it out of stubbornness, as I was forbidden to use my 49g in a course, but refused to surrender to the 30-year old 12c recommendation.

Got it last week on a trip to the USA (30b is nowhere to be found), and I guess I'm stuck with it, as it is not sold in Brazil, so the local HP shop will probably not replace it nor update it for me.

Cyrille's documentation in the Development kit sounds like I can flash the firmware with a RS232 cable (a JTAG cable costs 3x the price of the calculator). Even then, the "HP20b original ROM.bin" which comes with the Development kit is dated 10/19/2009, so it may still have the bug.

A 28s went to college with me in the late 80's, a 49g is my working companion since the 90's. I could never expect a MUCH simpler HP calculator could get into stores with such an obvious problem. I mean, I found it after sporadically tinkering with it for just a couple of days! It shouldn't have taken one man-hour to test ALL of the 20b features (same cannot be said for the 28s or 49g). Heck, I'd gladly do it for free for HP; just send me calculators, one of each, DDP.

So much for being stubborn. Maybe I can slap the 20b front face on my 49g and get away with it in the course.

Any suggestions are welcome...

Thanks,

Joao


#18

Joao wrote: "It shouldn't have taken one man-hour to test ALL of the 20b features"

Gene: Actually, testing something is much more complicated than most people realize. One man-hour MIGHT be long enough to make sure the SIN function works properly for all arguments between 0 and 0.000001 radians. Well, perhaps not *all* arguments.

Well, you say, if SIN works for 0 and 0.1, it should work for 0.05. I seem to recall a problem with 2.02 on a previous calculator.

It is difficult. Each potential line of code can affect some other line of code.


#19

Quote:
Well, you say, if SIN works for 0 and 0.1, it should work for 0.05. I seem to recall a problem with 2.02 on a previous calculator.

Wlodek's book says that the 2.02 problem was with the e^x function on the early HP-35's where it could be seen by entering in order 2.02 ln e^x and seeing 2 not 2.02 .

I agree with you on the difficulty of finding all the bugs. In my olden days working in inertial navigation we had a rule which said that in any software of any substantial complexity there is always one more bug to be found.

#20

I agree with you in some way Gene in that it is impossible to test all combinations due to the complexities of the software.

However, HP IS letting us down:
1) By not publishing errata sheets - for instance why is it that a Museum member has to keep a comprehensive list of 35s problems? This is a major problem - users rely on these calculators for important calculations daily and they should know what the problems are.
2) I do question how serious HP management take testing and funding for it. Again using the 35s as example, there are some glaring bugs that should have been tested - for instance the program length and checksum, this never works but is a feature that should have been tested as a standard part of a test plan. Do test palns actually exist in HP?

And by the way, those simpler tests as per your example should be easily tested in this current age by automated means.

Anyway, just my 2p worth. And despite my moanings I still love my HPs.


#21

Now please don't anyone take this out of context. I am just going to put on my "if I were a businessman hat" and ask a few questions that I would imagine someone in charge of setting a policy might ask.

This has nothing to do with HP policy, nor any internal discussions I've ever heard. I suspect this type of issue has come up with many varieties of products before though, both inside and out of HP. If it was just software you could send out a fix. Since there is hardware involved, everything is more difficult.

1. Does publishing a list of defects gain me more ($) than not doing so? (<-- that is the golden business question, when in doubt - follow the money!)

2. If I publish known defects, does that simply increase the number of people aware of problems that would have otherwise not known there were problems? I have deemed this to be a very minor issue that affects only a tiny percentage of users.

3. If I publish known issues, does that make me legally liable if I choose not to fix it due to time or cost?

4. If I publish known issues, does it cause the general customer who is not affected by the problem call my support center and ask for a replacement even though they aren't affected at all, simply because they now have the perception of a "bad product"?

There are probably more quesitons, but those were the ones that popped into my head while writing them down.

Of course, everything changes with a product if a failure means someone could die or is seriously injuered. Then from a business standpoint, you want to do everything possible. **NOT** from the "it is the right thing to do" standpoint, but rather the "Oh crap! We will lose so much money if we are sued over this". Again, follow the money.

Now think of this imaginary scenario. Let's say that HP suddenly became aware of 10 serious issues in the 12c. These issues were there from the beginning, and are definitely giving "bad" results. However, they are very obscure and nobody has found these issues since the very beginning and reported them. Are these serious issues that should be published? Would it be better to publish this info and possibly risk a core, longstanding product regarded so highly in the industry? It has been highly regarded for over 25 years. Should a business risk changing that, even thought nothing about the product has changed?

My personal view has, and always will be that transparency is best. However, I do recognize that there are some valid concerns those in charge of this type of thing have to weigh and evaluate.

TW


Edited: 14 May 2010, 9:30 a.m.


#22

Quote:
Of course, everything changes with a product if a failure means someone could die or is seriously injuered. Then from a business standpoint, you want to do everything possible. **NOT** from the "it is the right thing to do" standpoint, but rather the "Oh crap! We will lose so much money if we are sued over this". Again, follow the money.

I don't think companies become great by following the "follow the money" philosophy. "It is the right thing to do" is timeless.


#23

Quote:
I don't think companies become great by following the "follow the money" philosophy. "It is the right thing to do" is timeless.

http://www.hpmuseum.org/hp35.htm

There was a time to tell or not to tell was not a question. Scroll down until Hewlett-Packard Integrity and "The Bug" appears.


#24

Yeah, I remember that. Speaks for itself.

So who's going to send Tim a broken pencil?

: )

#25

Tim,

When I read your 4 point checklist I couldn't help thinking that the executives at Toyota went through the same thought process and decided to keep quiet. There might be a lesson for HP and many other big companies with previously sterling reputations in this.

The old HP did the "right thing" with the optional HP-35 recall and that helped make them the greatest calculator producer ever. It could happen again!

-Katie

#26

Quote:
If it was just software you could send out a fix. Since there is hardware involved, everything is more difficult.

Since this is a hardware (and software) issue, and we all know we can't flash our buggy HP 35s, and since there IS a published list of known issues, is HP going to fix the software/hardware issues and produce more HP 35s? Or is that product dead?

I read the "bugs" article Gene referred to, and I know HP will not replace our HP35s, but I would be glad to buy another one if they fix all/most of the bugs on the hpmuseum bug list.

#27

Hi Tim,

Apart from the ""broken pencil" trust in HP" references, in response to your points:

Quote:
1. Does publishing a list of defects gain me more ($) than not doing so? (<-- that is the golden business question, when in doubt - follow the money!)

It is the difference between an honest and dishonest company - or at least the perception. Particularly important if the products are being sold at higher prices because of "their legendary quality". Part of a good quality policy is the publishing of known non-conformances to the specifications. I work in aerospace, I support the use of ICs from a particular manufacturer because they publish errata sheets. Then we know how to deal with or avoid the bugs!



Quote:
2. If I publish known defects, does that simply increase the number of people aware of problems that would have otherwise not known there were problems? I have deemed this to be a very minor issue that affects only a tiny percentage of users.

The Pentium FDIV bug comes to mind.



Quote:
3. If I publish known issues, does that make me legally liable if I choose not to fix it due to time or cost?

It does not change the fact that you are liable, but it can change the judgement made against you (both the company and individual*) if you are found to be deliberately concealing a known major issue by not publishing anything.
*Note: in some countries only companies and not it's individuals can be held liable - just hope you're in the right one.



Quote:
4. If I publish known issues, does it cause the general customer who is not affected by the problem call my support center and ask for a replacement even though they aren't affected at all, simply because they now have the perception of a "bad product"?

Of course, but your conscience would be clear.



Quote:
Of course, everything changes with a product if a failure means someone could die or is seriously injuered.

Major economical damage counts too, as you say throughout your response "it's about the money".

Quote:
Then from a business standpoint, you want to do everything possible. **NOT** from the "it is the right thing to do" standpoint, but rather the "Oh crap! We will lose so much money if we are sued over this". Again, follow the money.

Businesses have closed down because of this. It is better to loose some money by being honest than loose everything by being caught out.



This is why a good test regime is important (coming back to original concerns of this tread), to minimise this business impact of flaws by catching the worst ones before the product goes out.



All of my responses are things that people instinctively know to be right, but have buried deep in their minds for the sake of money.


Regards,
-B

edit:

Quote:
I am just going to put on my "if I were a businessman hat"...

That's what Morton Thiokol engineers did.

Of course you are just putting out calculating devices and sitting behind disclaimers "we are not liable for consequential damage", or it's use for safety critical applications etc.
However, I need SOMETHING to do my calculations with when evaluating risk etc. I currently use at least 2 calculators from different manufacturers and a PC to cross check. If I loose confidence in one manufacturer, I'll not hesitate to move to another.

Edited: 17 May 2010, 12:06 p.m.

#28

Gene: Actually, testing something is much more complicated than most people realize.

Joao: Yeah, so let's drop even the simple tests, since doing it all is more complicated than most people realize.

Testing ALL possible arguments and sequences for ALL features of a calculator is such a ludicrous idea that I do not consider it possible - not even with automated testing.

Take for example all possible argument values, quantities and permutations for the Cash Flow functionality. Even automated testing would take forever (in the actual meaning of forever).

However, testing if the Cash Flow works at all takes less than a minute.

The problem I found is not a case of specific arguments to a function not working, but a big part of a feature not working correctly, for ANY argument.

So, to be pedantic, I rephrase "test ALL of the 20b features" to "test ALL of the 20b features with one set of arguments to rule out obvious bugs, as some people might not realize that testing everything with all possible combinations/permutations of arguments is not feasible in the real world". Is that Ok now?

Thanks,

Joao

#29

Just wondering...Not really suggesting anything yet, but thought I'd poll the audience...

Would you (or others) find it a worthwhile service if you could send someone your 20b or 30b and for a fee, that someone would reflash it with the most recent version (or a specific version, if requested) and mail it back to you? Figure the price would be about $10, inclusive of the return shipping.

What do you think?

I know that I wish I could always have the most recent ROM version from my HP calcs, but given the equipment and know-how needed, that's not always an option. But if I could send it somewhere and get it back, all pretty and up to date, I think I'd be willing to fork out a few bucks for that. Anyone else?

thanks,

Bruce


#30

In the 20b Dev Kit documentation, Cyrille sounds like it is possible to reprogram the calc with a serial cable (without the expensive JTAG cable).

I emailed him for details (as instructed in the documentation), and also asked if it would be possible to get the latest binary ROM image, as the one in the Dev Kit is dated 2009, thus older than Katie's.

I'll post any progress here.

The HP service centers should do this (reflash, or even replace, buggy firmware calcs), but I don't think the Brazilian centers would do anything about this, as this model is not sold here.

If I manage to do this, I'll gladly do it for free for fellow 20b owners, but they would have to live in Brazil, as exporting/importing them would cost more than a new one :-/

Best regards,

Joao


#31

Hi, João;

good to 'see' someone else from our tropical country out here!

I am in the verge of buying an HP20B from local 'representatives', if you know what I mean, but I was even more interested on the HP30B. But this little beast always seems to be hidden somewhere else...

Instead of contacting Cyrille and ask him for documentation (I am sure he´d not complain, though), would it be OK for you to send me what you get? I´m an electrical engineer myself and I like to handle test leads, soldering irons, PC-aided measuring devices and related software, so I´d not bother you with the completion of the task. In any way we could always trade information, as it happens here very often, then we share with the guys, as you mentioned. If you agree with, please:

linux_lcv (at) yahoo [ponto] com -dot- br

If you like electronic devices' pictures, please:

Pictures of the assembling of my MLDL2000 I really like the HP41 series...

Looking forward for your contact.

Cheers and thank you.

Luiz (Brasil, né?)

Edited: 14 May 2010, 12:20 p.m.


#32

Hi Luiz,

Sure thing; if I get additional information from Cyrille or Tim, and they say it is ok to pass it along, I will send it to you (I'm not sure if they actually want to send the information on a per request basis, otherwise it would all be in the dev kit documentation already).

Um abraço,
Joao


#33

Of course, you are right, I stand corrected. I did not consider it this way. I have already downloaded the dev kit, and your comments are perfectly correct.

Thanks!

Luiz

(P.S. - BTW, where in Brazil are you? I´m in Minas Gerais... uai!)

Edited: 14 May 2010, 9:31 p.m.


Possibly Related Threads…
Thread Author Replies Views Last Post
  HP's thinking behind the 20b/30b? John Ioannidis 3 1,771 09-07-2013, 10:21 AM
Last Post: Tim Wessman
  20b, 30b not in HP's web store Eric Smith 3 1,590 02-08-2013, 11:52 AM
Last Post: Walter B
  A non 20b/30b 34S Matt Agajanian 22 5,307 08-20-2012, 01:06 AM
Last Post: Reth
  HP-97 Weirdness Nick Taylor 1 1,008 07-26-2012, 04:17 PM
Last Post: Randy
  [WP-34s] IR conversion of 20b Alexander Oestert 3 1,592 05-18-2012, 05:38 PM
Last Post: Harald
  Why hp-20b more expensive than 30b? Masaki Adachi 9 3,036 10-14-2011, 12:08 PM
Last Post: megarat
  missing segments on 20b running 34s db (martinez, ca.) 4 1,712 10-05-2011, 01:27 AM
Last Post: Walter B
  Flashing HP 20b for WP 34S Steve Brenton 10 2,877 09-30-2011, 09:03 AM
Last Post: Marcus von Cube, Germany
  Case for 20b/30b/34S? David Hayden 1 1,011 09-29-2011, 03:41 PM
Last Post: Walter B
  First post & a question re:20b/30b E.Lub_EU 10 3,540 09-20-2011, 02:36 PM
Last Post: E.Lub_EU

Forum Jump: