In case you don't realize, the HP48, or its successor the HP39, is NOT a better version of the 38/39G! It is a COMPLETELY different calculator with a totally different operating system. It even uses Reverse Polish rather than D.A.L. in its algebra expressions.  If you are a university student then you may be interested in the HP48 or the HP49, as they are far superior to the HP38/39G even if a lot harder to learn how to use.

The 48 can do also a calculation like this one as it reads by entering it in quotes & using the EVAL command, however this is not the natural way of working on the 48, whereas it is for the 38/39G.  There is no doubt that the 48 is a wonderful calculator - it is the recommended/required machine for many engineering and architecture courses - but don't expect your knowledge of the 38/39G to be of any use whatsoever in learning to use the HP48. The capabilities of the 48 in algebraic manipulation are FAR higher than the 38/39G and, because of this, most high schools ban them from examinations.

P.S.  I had been informed that the other problem with the HP48 (for schools) was that the infra-red link operates over a much greater distance than the 38G, and that with the 48 you can send notes and programs to the other side of the room. However, I have since been told that this is not the case - both of them can send over long distances but their receive capabilities have been cut down to ~10cm.  This is apparently true for both models.

As to any of the other brands, I don't have enough experience on which to base a recommendation: I have specialised in the 38/39G, since that is what is being used exclusively by my students.
 
 

22. Is there a bug in the Sequence aplet?

Yes, as a matter of fact there are three bugs and an annoying lack.

Bug #1: Run the Sequence aplet and move the highlight in the SYMB view to U1(N). Enter the definition of U1(N)=2^N. When you press ENTER the calculator will automatically calculate the missing values of U1(1) and U1(2). Now change the definition of U1(N) to U1(N)=3^N. As you may have noticed, the values of U1(1) and U1(2) are not correctly recalculated! If you change to the NUM view the error is still present. The only answer is to delete U1(1) and U1(2) before entering the new definition.
 

Bug #2: Go to the HOME view and set the value of A to (say) 0.95 using STO and then enter the Sequence aplet and enter the following sequence: U1(1)=10, U1(2)=10*A and U1(N)=U1(N-1)*A. This sequence gives (as it should) the geometric sequence 10, 9.5, 9.025, 8.57375, etc. in the NUMeric view. However if you now go back to the HOME view and STO a new value into A of (say) 0.5 the sequence does not respond properly to the change. If you change to the NUMeric view again you will find that the sequence is now given as 10, 5, 9.025, 8.57375,…. In other words, the sequence has been re-evaluated for the first two terms but not for the remainder.
Note:  This is fixed in the HP39G.

Bug #3: This is not actually a bug... If you enter a sequence into U1 as U1(1)=3, U1(2) = 8 and U1(N)=U1(N-1)+10 then the values for U1(2) onward will be correct but the value for U1(1) doesn't fit the rule. In fact the calculator is only doing what you asked it to and the fact that your values don't fit the general rule for U1(N) is irrelevant.

An annoying lack (on the HP38G only): Suppose I have a geometric sequence (for example 40, 32, 25.6,....) and I want to calculate the series for it (40, 72, 97.6,....) by summing the terms. The simplest way to set up the series for a seqence would be as follows.

• Set up the sequence by entering U1(N)=40*0.8^(N-1). U1(1) and U1(2) are automatically calculated.
• Set up the series by entering U2(1)=U1(1), U2(2)=U1(1)+U1(2) and U2(N)=U1(N)+U2(N-1). Unfortunately, if you try this you will find that the calculator won't accept the defining of U1(1) as U2(1). You can work around this by entering the actual numeric values of 40 and 72 for U2(1) and U2(2) but this means that the values will not be correct if you change the definition of sequence U1. If you're a teacher and you want to run a lesson investigating the sum to n terms and the sum to infinity then this is not very convenient.

There is a way of getting around this problem. It's a bit of a pain to set up but easier to use once it's working.

• Set up the sequence in U1 as before.
• Change to the Function aplet and enter F1(X)=U1(X). The Function aplet doesn't mind referencing another aplet as part of a function definition.
• Change back to the Sequence aplet and enter the U2 sequence as U1(1)=F1(1), U2(2)=F1(1)+F1(2) and U2(N)=U1(N)+U2(N-1). If you now change to the NUM view you will see that the second sequence U2 is the series values for U1. The advantage of doing it this way is that if you change the definition of U1 then U2 will correctly reflect those changes immediately. Don't forget bug #1 - if you change the definition in U1(N) then you'll need to delete the values in U1(1) and U1(2) first.

Update:  On the HP39G this is much easier.  Just enter sequence U1 as U1(1)=A, U1(2)=A*R and U1(N)=U1(N-1)*R.  Now enter U2 as U2(1)=A, U2(2)=A+A*R and U2(N)=U2(N-1)+U1(N).  You can now change the values of A and R in the HOME view and have this mirrored in your sequence and sum of sequence.
 

23. Why does the function PREDX give incorrect answers?

This is a bug that I'm not sure HP are prepared to acknowledge but as a maths teacher I'm going to stick my neck out and tell you "I'm right and they're wrong". Mind you, the problem is not isolated to HP - I have yet to find a calculator which does this correctly. This is a bit involved to explain so bear with me...

Suppose you have a set of bivariate data (as below) where you are going to be predicting y values from x values.
 
 

From this we find a line of regression of y on x of : where the values of a and b are calculated using the formulas and . For the data above the line is y = 0.5x + 2.5 This line can now be used to predict y values from x values. Notice that b depends on the value of the covariance and the variance of the independent variable (the one being predicted from).

However.... if we now change our minds and decide that we want to predict x values from y values then many people (apparently including HP) believe that you can simply reverse the earlier line of regression.

ie. 

If you do it this way then PREDY and PREDX reverse each other. For example PREDY(3) is 4 and PREDX(4) gives 3 back again. This is what happens on the HP38G, and on every other brand I've checked.

This is wrong. The value for the gradient of the new line is given by . The new value of b derives from the variance of the y values (since they are now the independent variable) and this will NOT be the same as value of b = 2 which we obtained above.

If you've entered the data above into columns C1 and C2 then you can check this by changing the SYMB view.

• To obtain the first line you would have set the SYMB view to S1: C1 . . C2
• Reverse the roles of x and y by changing this now to S1: C2 . . C1
• Replot the data and make sure the regression line has been calculated.
• Return to the SYMB view and record the new equation.

You will find that the new line of regression for x on y is x = 0.92857y - 0.5 . The calculator gives it as y = 0.92857x - 0.5 but it doesn't know that we reversed x and y.

As you can see, this is nothing at all like the line we obtained by reversing the original equation.

HP's view on this is that PREDX is not predicting an x value from a y value. It is answering the question "what x value, if entered into PREDY, will give this y value?"   I don't accept this and I feel that any reasonable person would interpret it my way.  However, obviously HP don't agree with me on this since it has not been changed in the HP39.  I did bring it up quite forcefully but they didn't accept my arguement.

24. Is there a version of the Connectivity Kit which runs under Win '95?

The question:
I use the HP38G in my classes but have run into a problem. I download aplets from the internet (such as ones like yours) to supplement my lessons. My connectivity software works only with Windows 3.X. I noticed that you had reference to "The New Mac Connectivity Software has been released" on your web page and wondered if you knew where I could get connectivity software for Windows95. I have searched HP web site, but they give no reference to it.

If you have the old first version (1995) of the connectivity software then it said on the packaging very prominently that it will not run under Win '95.  This is not true. I have yet to find a single case of it not running. I run the CONN38G.EXE program and the ADK under Win '95 with no problems. I suspect that it is a case of a legal disclaimer that got out of control.

A new version of the connectivity kit was released in January 2000 which is designed specifically for Win '95/98 and is much easier to use.  A version for the HP39G was released (in beta) in September, 2000. You can download these from my Utilities page. Unfortunately Mac users are currently out in the cold.  A new version for Mac users is in the pipeline and may be available in 2001.
 

25 & 26. Can I erase the aplets that come with the HP38/39G? Is there a way to upgrade the memory of an HP38G?

The question: I have two questions-

• Can I erase the aplets that came already on the HP?
• Is there a way to upgrade the calculator memory on an HP38G? Like replacing the chip in the HP with one that has more memory.

1. No, the original aplets are built into the calculator's chip so you can't erase them. The worst you can do is lose any data which you have stored in them when you reset the calculator or let the batteries go completely flat.

2. No.  On the HP38G the BIOS was built around the assumption that the memory was 32K and this is hardwired in and cannot be changed. If you hear people say it can then they are probably confusing it with the HP48GX which CAN be upgraded. I know it's a pain - I'm hoping that the next generation will have either more memory or upgradeable memory or both. If someone argues with you about the memory, this is not just me talking - this comes direct from Jim Donnelly who was the head of the software/BIOS design team.

Update 1:    For more up to date information on this see Question 42.

Update 2:    Buy an HP39G!  It has 232Kb of user memory!
 

27. How can I print more useful things using an infra-red printer?

The question:
I am currently using the HP infrared printer and seem to have limited options to print from the 38G. Printing screen snapshots is no problem but printing lists of data in columns similar to what appears in the stats aplet seems not to be supported. Is there an aplet available which extends the printing options from the calculator?

Is there an aplet? Not that I'm aware of but...

Pressing ON + PLOT will store a snapshot of any screen (even one in mid update) to grob G0. You can then paste this into a sketch using VAR, HOME, Graphic, G0, VALUE, ENTER. Once it is stored in a sketch you can edit it using the drawing tools and store it again to G0.

The printer commands PRDISPLAY, PRHISTORY and PRVAR are available in HOME view or in a program. PRDISPLAY & PRHISTORY are probably obvious and you can use PRVAR to print the stored grob by typing PRVAR(G0).

Not having an infra-red printer I have only been able to experiment once but you can also use PRVAR to print lists & matrices. For example if you have data in C1 you can use PRVAR(C1).

Another alternative if you have the Connectivity Kit is to use the SEND command in the List Catalogue to send the column (do C1 [STORE] L1 in HOME first) and then open the resulting file on the PC or Mac with a text editor and print.

28. The COM port needed to use the Connectivity Kit is being used by my mouse.

The question:
I am trying to use the connectivity kit on a Windows computer. The school computers only have one serial port which is used by the mouse. Is there any way around this apart from disabling the mouse?

Yes. This problem is caused by the computer vendors trying to save a bit of money. A mouse can operate from the serial (COM) port (plug) or by having its own separate card built into the computer with its own plug. The first option is a serial mouse, the second is a bus mouse. The advantage of the second option is that it leaves the most commonly needed port (COM1) free for other uses. The disadvantage of course is that a bus mouse is slightly more expensive.

If you look at the COM1 port you will usually find that it is a 9-pin male plug. The cord that comes with the Connectivity Kit has a 9-pin female plug. Most computers also have another serial port (COM2) which is usually a 25-pin male. All you need to do to make the set up work properly is to buy a special converter plug. This plug (a DB25 Female to DB9 Male converter) is very common and quite cheap. Any electronics store will sell them or be able to point you to a store that will stock them.

When you've used this to install the cable, you need to let the software know what you've done. On the face of the window you see when you run the Connectivity Kit you will see a "Serial Port" option. From this choose the COM2 port . You should now find that everything works with no problems.
 

29. Why is the value of the covariance "incorrect" on the HP38G?

Question submitted by: Chris Williams; Bunbury Cathedral Grammar School

Answer:  The value is not actually incorrect at all. The values of the variance, standard deviation and covariance all have two possible values depending on whether you want the population or sample values.

The population value is obtained by dividing by n and is used when the data you are using in your calculation is the entire population. The sample value comes from dividing by n-1 rather than n, and is used when the data you are using in your calculation is small in comparison to the population from which it is drawn.  In this case you are wanting to use the sample to estimate what the values from the entire population would be.

At high school level we only use the population values rather than the sample values. The sample versions may only become important at university level where you might be doing research in which you were generalizing to the whole population from a limited sample.

In the case of the HP38G, both possible values are supplied (as PVar & SVar etc) except in the case of the covariance, where only the sample value is given. In my view it is a strange choice but at least it means that you can easily obtain the "correct" value by multiplying the supplied value by (n-1)/n.

On the HP39G both values are available in the STAT window.