Which is the best? TI58C x HP67



#2

			TI58C	HP67
Constant Memory Yes No
Memory (Max) 59 26
Program Steps (Max) 479 224
Card Reader No Yes
Solid State Software Yes No
Speed ? ?
System AOS RPN

I don't want to compare the TI59 with the HP67

Which of the two is faster?
Which is the best?

Edited: 23 Mar 2008, 9:18 a.m. after one or more responses were posted


#3

If you add keyboard quality to the list, the answer is clear.

#4

none can be *best*, comparing two one of them can only be *better* and that is the HP

Edited: 22 Mar 2008, 8:35 p.m.

#5

The TI58C is a great machine (fun to do self-programming programs with). But, not having to key in programs all the time makes the HP-67 a winner (as the "C" in the 58 can only hold that much [479 steps]). That, and the build quality. And the RPN of course :)

#6

When a question such as the one posed in this thread is in a forum such as this one can expect that the responses would favor the HP-67, and they have. Of course, adherents of other product lines may not agree; for example, in April 1977 after expressing disappointment in the newly issued HP-67 Richard Vanderburgh wrote the following comment on machine loyalty in the Volume 2 Number 4 issue of 52 Notes:

Quote:
Unfortunately, excessive loyalty toward one machine makes it difficult if not impossible to compare competing machines impartially, or even correctly. And I suspect that all else being equal, uni-loyalty becomes the strongest toward the most expensive of competing machines; the rationale goes: it just has to be better since I paid more for it.

Richard's disappointment was based on a comparison of the HP-67 with the SR-52. The TI-58/59 machines would not become available for two more months. The participants in this forum may reasonably suggest that Richard's comment is colored by being the editor of a newsletter dedicated to machines from the TI product line. I suggest that a comparison in the article "Large Numbers and Calculators" by Renaud de la Taille which was published in the 12/80 issue of the French journal Science et Vie may be pertinent. The article compared the capability of machines such as the HP-67, TI-58, TI-59, HP-41 and Sharp PC-1211 to calculate many digits of pi. Maurice Swinnen translated the article to English and I obtained permission to publish the translation in TI PPC Notes. Of the HP-67 the article says:
Quote:
...Because the number of registers is modest (26) one cannot do more than 250 digits in the computation of pi, but the HP-67 shows itself to be simple to program while maintaining all the advantages of a scientific calculator. It is rather slow, but one appreciates its solidity and its perfect quality fabrication.

Of the TI-58 the article says:
Quote:
For its price it is the one that offers the greatest possibilities. It has a very large number of registers, it has all the useful mathematical funcions, it has a superb precision, thanks to 13 digits in each register. It is also very complete with respect to programming, having all the tests needed, the control of loops, subroutines, symbolic addressing and direct addressing, user-defined keys, etc.

It also is a subtle machine in which the capabilities surpass the instructions in the manual; it is thus possible to create loops (dsz) with any register, not only with registers 0 therough 9. On top of that, one may use the registers reserved for parentheses by using code 82.


I would add that the TI-58/58C/59 machines provide indirect addressing capability with every data register. I haven't found my HP-67 material. My recollection is that the HP-67 provides only one. Is that correct?
#7

What a throwback to 1977! Then, I had just managed to buy a $400 HP-67 (on a US Navy Lieutenant-Junior Grade annual salary of $8000), when shortly after that the TI-58 and TI-59 appeared. I soon bought a TI-58, then traded it for an HP-35 (red dot model!) and bought a TI-59. I used it and the HP-67 in the 1977 to 1981 era, and did a lot of programming on each.

Truth be told, the TI-59 had much greater speed, accuracy (guard digits), functionality, memory (solid-state and magnetic card), and innovation. It and the TI-58 were the first handhelds to use solid-state application modules. The machines could be mounted on a PC-100 print cradle to get hard copy. The HP-67 was considerably more expensive, yet lacked most of the advantages of the TI-59.

But...TI-58/59 and PC-100 reliabilty was really poor. By 1980, I had gone through five TI-59s and three PC-100s. They would fail with varying symptoms. In addition, programming was much less efficient compared to the HP-67. For example, passing arguments to subroutines was harder, since the TI algebraic stack could not be used like the HP-67. Typically, programs I wrote for the HP-67 were 60 percent the size of the same program for the TI.

I eventually gave away most of my TI stuff and kept the HP-67. I found that when reliability (and quality of "feel") was important, the job had to go to the HP. For just playing around, the TI was usable. Where it really counted, the HP-67 beat the TI-59 (not to mention the 58 and 58C).

Mike


#8

Quote:
What a throwback to 1977! ...

Truth be told, the TI-59 had much greater speed, accuracy (guard digits),


Ah, but were the values in those three guard digits calculated accurately? In the eight-digit 1976 TI-30, the values were often questionable. For example, e^1 was calculated as 2.7182818301 instead of 2.7182818284:

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv016.cgi?read=107358#107358

-- KS


#9

Karl:

Quote:
Ah, but were the values in those three guard digits calculated accurately? In the eight-digit 1976 TI-30, the values were often questionable. For example, e^1 was calculated as 2.7182818301 instead of 2.7182818284:

Many of these simple tests have been proposed as ways to compare the performance of various calculating machines. One which I believe can be traced back to an article in EDN suggests that the operator enter 2 into the machine, take the square root, square the result and see how closely the answer is to 2. If you enter 2 on an HP-67, take the square root and then the square you get 1.999999999 in the Disp 9 format. If you do that again and again you continue to get the same answer. If you enter 2 on a TI-59 take the square root and then the square you see 2 in the display but if you subtract 2 from the displayed value you get -5E-12 . If you do two square root - square sequences you get 2 in the display, but if you subtract 2 you get -8E-12 .If you continue on you will keep seeing 2 in the display but after subtracting 2 you willl see that the error is continuing to grow. Now, some defenders of the HP faith say that proves that the HP-67 is better than the TI-59 because the TI-59 error grows while the HP-67 error does not grow. They forget how much better the TI-59's first answer is relative to the HP-67's answer and how slowly the error in the TI-59's answer grows. How many iterations of square root - square does it take before the TI-59's answer is as bad as that of the HP-67. Make a guess. I'll give the answer in a subsequent transmission.

Palmer


#10

Quote:
Now, some defenders of the HP faith say that proves that the HP-67 is better than the TI-59 because the TI-59 error grows while the HP-67 error does not grow. They forget how much better the TI-59's first answer is relative to the HP-67's answer and how slowly the error in the TI-59's answer grows. How many iterations of square root - square does it take before the TI-59's answer is as bad as that of the HP-67. Make a guess. I'll give the answer in a subsequent transmission.

Palmer


Hi, Palmer --

Well, I have neither an HP-67 nor a TI-59. It appears, however, that the math routines of the TI-59 are different (and sharper) than those of the 1976 LED TI-30, because the TI-59's "forensic result" is 9.000004661314, versus the TI-30's considerably-coarser 9.1770871. The HP-67's result is 9.000417403 -- just like my HP-11C.

The 24-digit result for sqrt(2) is 1.41421356237309504880169.

The HP-11C (or HP-67) gives the correct 1.414213562, which can be "squared and square-rooted" ad infinitum without further degradation of accuracy. The TI-30 also gives 1.414213562 initially, but only one x2 and sqrt yields a one-ULP error of 1.414213561. Error in the first square is -1.9E-09. Given the vast difference between the respective forensics results of the TI's, though, I'd expect that the TI-59 should take many more pairs of repetitions of x2 and sqrt to yield an error exceeding 1E-09 from the square of exactly 2. How about -- 30 squares?

-- KS


Edited: 24 Mar 2008, 3:23 a.m.


#11

Karl:

Quote:
... I'd expect that the TI-59 should take many more pairs of repetitions of x2 and sqrt to yield an error exceeding 1E-09 from the square of exactly 2. How about -- 30 squares?

The answer is 196. It should not be too surprising that the number of iterations required is different for different input integers. A short table follows
Integer        Iterations

2 196
3 116
5 178
6 84
7 0
8 72
10 63

where the zero for an input integer of 7 suggests that there is more to the story, and there is. If you scan what the HP-67 delivers for the square root - square of an integer test you will find that it yields exactly the input integer for many input integers; i.e., for the range of input integers from 2 through 100 one gets a return different from the input in only 28 cases which are 2, 3, 5, 6, 8, 10, 32, 33, 41, 42, 45, 47, 51, 54, 56, 60, 66, 70, 73, 77, 78, 79, 85, 86, 88, 90, 92, and 98.

I think that in general you will find that the TI-59 does fairly well on square roots. For example, for 2:

For 2:

Exact 1.4142 13562 37309 504...
HP-67 1.4142 13562
TI-59 display 1.4142
TI-59 internal 1.4142 13562 373

For 3:

Exact 1.7320 50807 56887 729...
HP-67 1.7320 50808
TI-59 display 1.7320 50808
TI-59 internal 1.7320 50807 568

My recollection is that if you want to make the HP-67 look better than the TI-59 you will have better luck working with the logarithm and y^x functions than with the square root function.

While the HP arithmetic seems to have been relatively stable from the HP-35 through the HP-41 the TI arithmetic was evolving. For example, the TI-59 had the well-known non-commutative multiply. The TI-66 had the non-commutative multiply fixed but still delivered answers TRUNCATED to thirteen digits. The TI-95 delivered answers ROUNDED to thirteen digits. May I remind you that the internal arithmetic that I used with the HP-41 to get more accurate quadratic solutions yields thirteen TRUNCATED digits.

Palmer


#12

Karl:

Back in 1963 we were working with the square root squared test and decided to see what would happen if we would run five square roots followed by five squares. The results for integers from two through twenty-five can be found on V8N3P14 of TI PPC Notes. A sample from the larger table is

Integer             HP Family                  TI-59 Display            TI-59 Internal

2 2.000000022. .2. 1.999999999917
3 2.999999991 3. 2.999999999806
5 4.999999931 4.999999999 4.999999999240
6 5.999999931 6. 5.999999999641
7 7.000000143 6.999999999 6.999999998864
11 11.00000003 11. 10.99999999866
18 17.99999980 18. 17.99999999665
25 24.99999998 25. 24.99999999863

The results for the HP machines are sometimes higher and sometimes lower than the input integers. The results for the TI-59 are always lower than the input integer. I think that the difference is related to the rounding used in the HP machines and the truncation used with the TI-59.

When I originally developed the data for TI PPC Notes I did not have an HP machine. I got the expected HP results by doing EE INV EE after each calculation on my TI-59. Subscribers who had HP machines confirmed that my results were correct. That is one of the advantages of a machine with a longer word length. It is possible to relatively easily simulate the results from a machine with shorter word length, at least for some cases..

Palmer


#13

I'm confused about the results shown for the "HP Family".

None of my "Classic" HP calcultors (HP-35, HP-45) produce anything like the results in your table. And my HP-67 gives similar results.

As stated in one of your earlier posts, the HP claim to fame was: "If you enter 2 on an HP-67, take the square root and then the square you get 1.999999999 in the Disp 9 format. If you do that again and again you continue to get the same answer." Indeed I confirm that is the case.

The date you cite was 1963, well before the selling date of the TI-59 or HP-67 calculators (or the classics). Perhaps you should clarify what the data presented should mean?


#14

Quote:
I'm confused about the results shown for the "HP Family".

The date was a misprint. It should have been 1983.

The test is five square roots in a row followed by five squares in a row, not five cycles of square root - square. I just did several on my HP-41 and got the results in the table. I have also done it in the past with an HP-11. Just now I also did it with an HP-27 and got the table results.

I trust that this will help.


#15

Thanks, that does make sense.

I should have read the text closer.

#16

Hi, Palmer --

Quote:
While the HP arithmetic seems to have been relatively stable from the HP-35 through the HP-41 the TI arithmetic was evolving.

In view of what I've posted previously about the 1976 LED TI-30's sloppy math, I would certainly hope so. There was plenty of room for evolution there.

However, it is noted that the HP-67's math also was improved from that of the "Classic" HP-35 and other early models, due mainly to diminished ROM-size constraints. Please see "The New Accuracy: Making 23 = 8", a sidebar in the November 1976 article in Hewlett-Packard Journal about the HP-67. Few if any changes were made in the math routines from that point through the Voyager series until the HP-71B was developed in 1983.

There aren't any real surprises in your table, given that the HP's are retaining 10-digit results, while the TI-59 was retaining 11-digit results. One could extend the tables to the Saturn-processor HP's, which retain 12-digit results. For a given starting number, compute the common logarithm of the ratio of the aggregate roundoff error from an old HP against a TI-59, or from a TI-59 against a Saturn-processor HP, and you'll generally get a result between 0.5 and 1.5. In other words, each additional significant digit reduces the error by roughly an order of magnitude.

I still like HP's approach: Calculate the most accurate result possible, then display all the significant digits that are retained. Sure, inverting 7 twice doesn't yield exactly 7, but most users choose to display a limited number of significant or decimal digits (FIX/SCI/ENG). Conversely, TI and Casio tend to use "Standard" display mode as the default; this requires use of hidden guard digits to eliminate those annoying inaccuracies due to roundoff.

-- KS


Edited: 27 Mar 2008, 4:58 a.m.


#17

[Karl:

Quote:
There aren't any real surprises in your table, given that the HP's are retaining 10-digit results, while the TI-59 was retaining 11-digit results.

The TI-59 retains thirteen digits not eleven.
Quote:
I still like HP's approach: Calculate the most accurate result possible, then display all the significant digits that are retained. Sure, inverting 7 twice doesn't yield exactly 7, but most users choose to display a limited number of significant or decimal digits (FIX/SCI/ENG). Conversely, TI and Casio tend to use "Standard" display mode as the default; this requires use of hidden guard digits to eliminate those annoying inaccuracies due to roundoff.

Those "most users" who choose to display a limited number of digits are choosing to operate in a mode which is essentially the same as that in which the TI machines normally operate; i.e., displaying fewer digits than the number of digits internal to the machine. Pages 31-32 of the HP-41C Owner's Handbook and User's Guide addresses the situation with the following statement
Quote:
No matter which format or how many digits you choose, display control alters only the manner in which a nmber is displayed. The actual number itself is not altered by any of the display control functions.

My recollection is that one of the examples for the matrix routines in the HP-41 Advantage Pak operates with a reduced number of digits in the display while it inverts the input matrix and then inverts the inverted matrix. The displayed results suggest that the input matrix was recovered exactly; however, change to Fix 9 mode will show that the input matrix was only approximately recovered. The UNDO function in a machine such as the HP-28S will recover the exact input matrix..

I did some of the five square roots followed by five squares testing with my TI-66 (non-commutative multiply fixed) and TI-95 (a rounded thirteenth digit). The results including those for the TI-59 are presented below. As I had suggested earlier in this thread the TI-95 results are similar to those of the HP family in that some of the resullts are greater than the input integer and some are less.

Integer            TI-59 Internal            TI-66 Internal             TI-95 Internal

2 1.99999 99999 17 1.99999 99999 20 1.99999 99999 83
3 2.99999 99998 06 2.99999 99998 93 3.00000 00000 04.
5 4.99999 99992 40 4.99999 99997 95 4.99999 99999 70
6 5.99999 99996 41 5.99999 99997 14 5.99999 99999 94
7 6.99999 99988 64 6.99999 99997 05 7.00000 00000 71
8 7.99999 99993 51 7.99999 99996 48 8.00000 00000 50
10 9.99999 99987 48 9.99999 99999 47 .9.99999 99999 09
11 10.99999 99998 66 10.99999 99995 6 10.99999 99999 9
12 11.99999 99998 07 11.99999 99996 0 12.00000 00001 3

I think I'll stop running any more tests for a little while. Typing all of the digits strains these old eye-balls.

Palmer


#18

Hi Palmer,

Quote:
I did some of the five square roots followed by five squares testing with my TI-66 (non-commutative multiply fixed)

The TI-66 shares the programming model of the TI-59 but the machine has very little in common with its predecessor. It is a complete redesign made by Toshiba. So it doesn't surprise that the results are different. It's not just the multiplication that has been "fixed".

My main concerns about the 66 are that it is much slower than the older machines, there is no module and the standard library isn't included, and last not least, I could never find the PC-200 printer on the Net.

Has anybody ever hacked the protocol of the TI-66 printer port?

Marcus


#19

Marcus:

Quote:
... and last not least, I could never find the PC-200 printer on the Net.

I purchased mine very soon after they became available, either from EduCalc or Elek-Tek, I can't remember which. I don't think that they were available very long, but I do recall that both the TI-66 and the PC-200 were later available at substantially reduced prices at one of those facilities that sells discontinued material.
Quote:
Has anybody ever hacked the protocol of the TI-66 printer port?

I don't recall seeing anything about that or seeing a patent.

I agree that the TI-66 was a much different machine from what one might have expected from a follow-on to the TI-59. Of course, the TI-88 was the machine which was intended to be the follow-on to the TI-59 and a competitor for the HP-41.

Elsewhere in this thread I commented on the uneasiness at TI over some of the things we were able to do with the TI-59. That uneasiness led to the following statment on page F-3 of the TI-66 manual:

Quote:
There are no HIR commands or other hidden features on the TI-66 that you may have accessed on the TI-58/58C/59 through illegal key sequences.

That statement was like running a red flag in front of a bull as far as TI users were concerned. A substantial number of special features were found. In fact, the issue of TI PPC Notes which introduced the TI-66 to it's readers included some special features identified by Dave Leising who had receivd an engineering model.

Palmer

#20

Palmer --

Quote:
The TI-59 retains thirteen digits not eleven.

I got my info from www.rskey.org, which listed the following for the TI-59:

Display size: 10(8+2) digits

which I took to mean an 8-digit mantissa and 2-digit exponent (plus three undisplayed gudard digits). Perhaps it displays 10 digits without an exponent.

Quote:
Those "most users" who choose to display a limited number of digits are choosing to operate in a mode which is essentially the same as that in which the TI machines normally operate; i.e., displaying fewer digits than the number of digits internal to the machine.

Yes, of course, but my point -- 'Conversely, TI and Casio tend to use "Standard" display mode as the default; this requires use of hidden guard digits to eliminate those annoying inaccuracies due to roundoff.' -- was that at least one undisplayed guard digit is necessary for a calculator to be able display "exact" rounded results in the default display mode favored by TI, which is known as Standard on the HP-48G and successors. Standard display mode strips off trailing zeroes to utilize the minimum number of digits required. The HP-48G in Standard mode shows "7.00000000001" instead of "7." for 7[1/x][1/x], due to a lack of hidden guard digits.

In practical applications, results of floating-point calculations are usually not tidy integers or short-fractional numbers such as 17.25; other simple calculations such as 1/3 will produce a long annoying string of digits. So, the Standard display mode is of limited value, in my estimation. However, the market seems to expect it.

-- KS


Edited: 29 Mar 2008, 3:18 p.m.


#21

Karl:

Quote:
I got my info from www.rskey.org, which listed the following for the TI-59:

Display size: 10(8+2) digits

which I took to mean an 8-digit mantissa and 2-digit exponent (plus three undisplayed gudard digits). Perhaps it displays 10 digits without an exponent.


I went to the site. I agree that the display is as you describe it. There is an entry designated as precision which say 13 digits. The TI-59 really does maintain thirteen digits inside the machine.

The machine does display ten digits in the turnon mode. If the user then switches to scientific (EE) mode the machine displays an eight digit mantoissa and a two digit exponent; however, if the user switches the machine to another fix mode, say Fix 4, either before or after setting scientific mode then the scientific display will have a five digit mantissa(four after the decimal point). That response is very similar to setting SCI 4 on an HP-41.

Palmer

#22

Quote:
I don't want to compare the TI59 with the HP67

Why not? The two machines came out within a few months of each other in early 1977.

Some minor corrections: The TI-58 has a maximum of 60 not 59 data registers, and has a maximum of 480 not 479 program steps.

#23

Hello!

Quote:
I don't want to compare the TI59 with the HP67

Why not? Those two were the direct competitors then.

For me personally, the Ti59 is the winner, because my parents could afford to buy me one around 1978 that helped me through my last two years of school and all of univerity. The HP67 (and its successor, the '41) were off the price scale by more than a factor of two for the educational market. The better calculator in many respects, but not affordable for the people who needed it most and therefore completely useless. Zero points to the '67 so to say...

Today, 30 years later, from the collectors point of view, the '67 is the better calculator, simply because most of them still work and many '59s don't, mostly due to keyboard failure.

The RPN capability of the HP-67 is no factor for me, because when I most needed a calculator, RPN machines were unaffordable for me, and I simply had to get along with AOS...

Which of the two would I keep if I had to sell one? Difficult decision, but out of gratitude and nostalgia, I think I would keep the Ti.

Greetings, Max

#24

I think you can't really compare one with the other. From a perusal of various web sites, I've uncovered the following chronology:

HP-65    ???-1974
SR-52 Sep-1975
HP-67 ???-1976
TI-58/59 May-1977
TI-58C ???-1979
HP-41C ???-1979

It seems that TI and HP kept leap-frogging one another. First HP started it all, with the HP-65 which had 100 unmerged program steps.

Then the TI SR-52 came out with 224 unmerged program steps. The HP-67 "responded" with 224 merged program steps (which is effectively a lot more "steps"). So I think the 67 should be compared to the 52.

Then the TI-59 came out with 960 steps (merged or unmerged?) and lots of registers. I think the one to compare with the TI-59 is the HP-41C. It came with about half the memory of the TI-59, but could be expanded to 2.5 times the memory of the 59. And of course, it too had "solid state software modules".

So, each new calculator was "better" than the one before it. By the way, of the ones in the list, I've owned an SR-52, HP-67, and HP-41C. You can definitely do more with the 224 steps of the 67 than the 224 steps of the 52, and you don't have nonsense like having to start any computational subroutine with "(" and end it with ")" in case you call it from within a larger expression.

One thing I liked about the 52 is you could use "any" key as a label. For instance, you could label your sinh program with "LBL sin".

Stefan Vorkoetter
http://www.stefanv.com/calculators


#25

Quote:
Then the TI SR-52 came out with 224 unmerged program steps. The HP-67 "responded" with 224 merged program steps (which is effectively a lot more "steps"). So I think the 67 should be compared to the 52.

Another entry in this thread states:
Quote:
Typically, programs I wrote for the HP-67 were 60% the size of the same programs for the T.I.

Back in December 1977 Richard Vanderburgh reported on V2N12p4 of 52 Notes that:
Quote:
At leaat one HP-25 user has conceded that 100 unmerged SR-56 steps amount to more effective memory than 49 merged HP-25 steps.

For those who are not familiar with the old merged/unmerged debate I note that a transfer command such as RCL 01 requires two steps in machines such as the TI-59 which use unmerged code but only one step in machines such as the HP-67 which use merged code. That would suggest a two to one ratio except that the algebraic codes require only one step in unmerged systems. That would suggest a ratio slightly less than two except that the comparison codes and in an unmerged system typically require three steps and a Dsz code with direct addressing (used to increase speed) requires four steps in an unmerged system. So the ratio is probably in the general range of two with the exact ratio depending on the mix of instructions.

Differences between machines such as unmerged code versus merged code were among the reasons that Richard Vanderburgh and Richard Nelson agreed back in the olden days to compare machines on the basis of performance of agreed on programming problems rather than on a comparison of a list of machine characteristics. For the HP-67 and the TI-59 there were some interesting results:

The HP-67 was faster at finding the prime factors of an integer.

The TI-59 with the Master Library installed was a much more powerful matrix solver. It could calculate the determinant and inverse of a 9x9 matrix and solve an 8x8 simuiltaneous equation. In an earlier thread on the history of the matrix manipulation competition I wrote that the best that the HP-67 could do with simulltaneous equations was 5x5. Valentin Albillo responded with a note that the HP-67 could do 7x7. That was several years ago. I have yet to get a copy of the program to try on my HP-67. Using the TI-59 program which was the basis of my sixth and eighth order simultaneous equation solvers for the hp-33s the TI-59 can solve a 16x16 system, but at the expense of a considerable reduction in user friendliness and some reduction in accuracy.

The TI-59 before fast mode could print a calendar in about two and one half minutes. Of course the HP-67 couldn't print a calendar but my recollection is that there was an HP-97 program which could. I don't know how long it took.

The Science et Vie article noted that the HP-67 could calculate 250 digits of pi and the TI-59 could calculate 1287 digits of pi.

Before anytone screams too loudly I admit that the advent of the HP-41 tipped the advantage to the HP product line; however, I suggest that the HP-41 tecnology was not at all contemporary with that of the HP-67 and TI-59 since it came two years later.


#26

I think the "MERGE" feature of the HP-67 gives it a powerful advantage besides everything else that has been mentioned.

tm

#27

Ah, these past arguments are usually so much fun.

Compared to some here, I might be in a good position to discuss these things. I was a very committed TI fan throughout the 1970s. Why, in particular? Memory power and cost.

The HP products were just so much more expensive to middle school / high school student than the TI products, that there just wasn't any real way they were attainable.

And, for much of the period in the 1970s, the TI products, at least on paper, seemed better technologically.

Consider from the timeline Stefan posted:

HP-65    ???-1974
SR-52 Sep-1975
HP-67 ???-1976
TI-58/59 May-1977
TI-58C ???-1979
HP-41C ???-1979

The date of the comparison is incredibly important. In Late 1977, I would argue for the capabilities of the TI-58/59 over the HP67. In late 1979, I would argue for the HP 41 over the TI-58/59. I don't think a comparison without specifying a date is really fair. :-)

For the two years or so that the TI-58/59 were out and the HP competitors were the HP 67/97, the TI-59 was roughly 1/2 the cost of the HP67 yet had almost 4X the memories. On the sliding scale of memory allocation, the TI-59 could have 30 memories and . When turned on, it had 480 program steps and 60 registers. That's a lot to work with. 2X the steps and registers at 1/2 the price. Then again, the TI-58/59 had the plug in rom modules that had no comparison in the HP world at the time. 5000 program steps to call as subroutines or use as main programs. Incredible. As much as I love my HP 67 today, at the time the price/capability ratio was probably in the TI-59 corner, IMO of course.

Yes, I know the TI-59 mechanically was much inferior. Sure thing. :-)

Then the TI-58c came out at an even cheaper price. In early 1979, I bought a TI-58c for about $90 at the time the HP 67 was being sold still for $400 or so. That was an incredible amount of power for $90. . . and it finally did not lose its memory contents when you turned it off! :-)

Of course, even as Palmer admits, the tables turned quite a bit when the 41c was announced...they certainly changed for me. I got a brochure for the 41c in the mail (I still have it) showing the 41c being held in someone's hand and it had letters showing in the display. And you could add modules like the TI calculators had been doing for years!

I bought one when I graduated from High School, joined PPC and changed camps.

But I do have a difficult time looking back at the early days and saying that a 26 register machine at 2X the price was really a great deal better than a 100 register machine (that still had 160 program steps when a full 100 registers were allocated AND a 5000 step rom module installed).

Isn't it great that we're all individuals who make up our own minds? Big :-) here.

#28

Hi, Palmer O:

Palmer O wrote:

    "In an earlier thread on the history of the matrix manipulation competition I wrote that the best that the HP-67 could do with simulltaneous equations was 5x5. Valentin Albillo responded with a note that the HP-67 could do 7x7. That was several years ago. I have yet to get a copy of the program to try on my HP-67."
     
      I take it that you're not trying to imply that I was lying when I stated that I have such a 7x7 program for the HP-67. Because if you are, then I'll take offence.

      The one and only reason you haven't got a copy of it is because I have neither the time nor the inclination to type and post a complex 224-step program blindly, with no physical or emulated HP-67 to try it on first if only to check that I've typed it in accurately and provide some sample runs.

      If you're interested in keeping civil conversation with me, think twice before suggesting I'm bluffing because I never do that. What I do, I do remarkably well. What I don't, I humbly recognize I'm a miserable failure at it. But I never bluff. Keep that in mind.

    Enough said.
Best regards from V.

#29

Valentin:

Quote:
If you're interested in keeping civil conversation with me, think twice before suggesting I'm bluffing because I never do that. What I do, I do remarkably well. What I don't, I humbly recognize I'm a miserable failure at it. But I never bluff. Keep that in mind.

I'm in a state of shock from what looks to me like an overreaction. When I recover I'll answer in detail.

Palmer


#30

Hi, Palmer:

    No need to be in any state of shock or whatever.

    You say you think I overreacted because you hadn't the slightest intention of suggesting that I was bluffing (or lying) ? That's enough for me, pal.

    The problem with postings and e-mails and such is that body language is absent and emoticons do not really convey the mood and I personally don't like them very much so I tend not to include them.

    In a nutshell: I wasn't nearly as upset and overreacting as it might seem; had I told you those same words in person over a beer we would both have had a good laugh.

    That's my curse. I sound much more serious than I really am. Matter of fact, I'm joking most of the time.

Best regards from V.

#31

Hi both;)

Though I wish to avoid to be involved into any quarrelling here, I do think that the following information may be helpful.

There is a program which solves linear equations on a HP67/HP97 of 7x7 size. I still have its documentation with the listing in my files. It is the program #50684 (User's Program Library Europe) entitled "Linear Systems k.k (k=1, 2, ... , 7)" written by Raymond Broeckx from Wilrijk, Belgium.

Hopefully, this proves that the program really exists;)

"Those days" I have used the program many times with my HP67. The program occupies a single card only. My magnetic card with the program is still readable. It works flawlessly, but it really takes some time for a 7x7 system.

All the matrix coefficients, as well as the right-hand-side vector components are to be entered only once (row-by-row) and in a straightforward manner. However, owing to the "shortage of storage places", not all the coefficients are entered at the beginning of the calculation. You enter the first equation, after some time the second, etc. Obviously, the reduction is made as soon as the equation is entrered. The original matrix is reduced into identity matrix.

If someone wishes to obtain the program listing I can fax it (yes, FAX it, this would be easiest) to him/her. You may contact me through the Museum.


#32

Thanks, Nenad. The one I have internally and externally works as you described. Matter of fact, that's the only way to do it with just 25 data storage registers plus a single index register.

By the way, the 7x7 program I have is *not* my creation, it was originally written more than 30 years ago by my friend Fernando del Rey, who right now is planning to send an article for publication in the next Datafile issue including a relatively short program to solve NxN linear systems (where N can be pretty huge) in the new HP35s.

Best regards from V.

#33

Quote:
If someone wishes to obtain the program listing I can fax it (yes, FAX it, this would be easiest) to him/her. You may contact me through the Museum.

I would like to have the program but do not have a fax of my own. I will find one that is available and give you the necessary information through the museum.

Your description of the operation of the program suggests that it operates in a similar manner to one that Valentin wrote for the HP-41. It also sounds similar to a program for the TI-59 which was published in the Swedish newsletter Programbiten. On the TI-59 the program expands the linear solution capability from 8x8 with the Master Library program to 16x16, but at the cost of a less convenient entry mode and somewhat decreased accuracy. I look forward to working with your program.


#34

If he gets a chance, i can turn the fax into a PDF and make it available to all.

Gene

Edited: 28 Mar 2008, 6:46 p.m. after one or more responses were posted


#35

Yes, yes yes! Thanks!

-- Antonio


#36

Well, don't thank me yet. He hasn't had a chance to fax it so far. :-)

Thanks Nenad when you do! Even bigger :-)

Edited: 28 Mar 2008, 6:47 p.m.


#37

Gene, you have mail! I did not receive your message yet.

I have scanned the program listing myself, even put it into a single .pdf file, but did not find out how to make this pdf publically available to the moHP.

My proposal is that I send the .pdf to Gene and he will certainly know how to do this. So, I gave up faxing, though it was the easiest way for me, because "don't ask what the moHP community can do for you, but ask what you can do for the moHP community".

Best regards to all from Split


#38

You have mail Nenad!

As soon as I get the PDF I'll put it on some webspace (or two or three places) and post the link here.

Thanks!


#39

HP67 7x7 Linear Systems PDF

here's the program. Thanks Nenad.

Anyone who wishes to take this file and mirror it anywhere else, please do!


#40

I don't have a hp-67, so I'm trying to convert the listing to the HP-41. I have several questions about some of the instructions:

1.  Several Program Steps are just C or B.
I'm assuming that this would be GSB C or GSB B?

2. Step number 70 is =0.
Is this X=0?

Same with step 124 - Would this be X#0?

Would Step 150 be X>Y?

3. And what is Step 184? -x-

I have a program I wrote several years ago that attempts to take a souce listing for HP-67 and convert it to HP-41 (with a Card Reader Module installed). Thought I'd test it with this program.

Thanks for your help.

Bill


#41

1) Yes, I think so. C should be GSB C.

2) Yes, I think so. Line 70 is X=0? Line 124 is X NE 0? and Line 150 should be X > Y?

Shortcuts that seem obvious at the time can be tough to remember later.

3) Line 184 -x- was a PRINT X command. This could be replaced by a PAUSE or two.

You didn't mention it, but some of the other steps need deciphering too.

Line 16 is probably X = Y?

Line 11 is STO i ... What is the : there for?

Line 50 is probably X<>Y

Lines 172 and 177 are probably X > Y?


#42

Hi Gene,

Thanks.

Quote:
Line 11 is STO i ... What is the : there for?

This one also had me confused, but the comments to the right is

ai := ai/ay

so I thought the colon might be divide?

Take a look at line 89 and its comment.

I think divide might be it.

You're right, shortcuts don't always translate very well.

Has anyone tried entering it on a real 67?

Thanks,

Bill


#43

As it might be visible from my previous posts, I have the program ready on a magnetic card. Last night I was about to resolve these issues in a practical way: read the card on my HP67, then SST to the dubious steps and finally post the results here. The machine read the first side and then...

Houston, we have a problem here!

The outcome: gummy wheel problem AGAIN! I had this problem solved yet in 1999, by my friend (I was not brave enough to do this myself). I am not sure if he had used o-rings or silicon tubing, but it seems that the "thing" desingrated, this time after 9 years.

This time I have a well documented repair procedure available, a piece of silicon tubing (thanks, Katie) but I am not sure if I got some additional courage meanwhile, though I have gone through a similar job on a HP41 card reader.

Is it normal that the repaired wheel lasts not more than 9 years?

#44

Don't forget the SR-56. It was a reduced-function version of the SR-52 (minus card reader and memory) that was the size of the SR-50 and SR-51.

It was my first progammable calculator, bought in 1976, but stolen in 1977.

Mike

#45

Quote:
HP-65    ???-1974
SR-52 Sep-1975
HP-67 ???-1976
TI-58/59 May-1977
TI-58C ???-1979
HP-41C ???-1979
It seems that TI and HP kept leap-frogging one another. ... So I think the 67 should be compared to the 52.

IMHO a *fair* comparison would be to the *mean* of 52 and 58/9. Else the follower has an advantage always 8)

I remember that calculator race in the seventies being a hard temptation everytime a new HP or TI appeared on the market. Being a student with little money to spend, there was a lot of window shopping :) In Germany, you always got "more calc power for the money" buying TI then. After some years, however, it turned out the reliability or long-time quality of TI's calcs was lower than HP's of that time (particularily remember having seen some TI57 degrading). With the appearance of LCD calcs the race ended - a period of HP dominance began, lasting some years. Of course, all this is my personal memory only.

Edited: 24 Mar 2008, 6:23 a.m.


#46

I don't have a TI59 but I have 2 TI58C's and 1 TI58 since they were new. I also have the PC100 printer. I don't have the HP67 but I have the HP97 for about 5 years. I think I would like the 67 better than the 58c or 59. But I know that I am bias toward the HP calculator and RPN.

#47

I think the SR-52 and HP67 were contemporaries for about 10 months.

The TI-59 and HP67 were contemporaries for well over 2 years, about 26 months.

Hence, the thought of comparing the TI-59 and HP67/97 models.


#48

Gene-

I'm not all familiar with Texas Instruments products. Did any of their calcs have a simular "MERGE" feature of the HP-67?

tm


#49

Quote:
I'm not all familiar with Texas Instruments products. Did any of their calcs have a simular "MERGE" feature of the HP-67?

I haven't used my HP-67 very much. I did take a quick look at the manual and think that I understand the MERGE concept.

On the TI-59 I do not think there is a way to duplicate the MERGE function by adding the contents of a second card at the end of what is already in the machine.

What can I do with the TI-59? I can replace any 240 step segment of the memory, and I can do that under program control with the INV 2nd Write sequence. I have used that feature. There is a trick by which I can use up to eight (maybe it's seven, I'm a little rusty in this area) of the instructions from the overwritten program after I have read the new card under program control.

Can an HP-67 user use merge under program control?


#50

Quote:

Can an HP-67 user use merge under program control?


Yes. Programs may span over several cards. At the end of each program segment, instruction MERGE may be used to read the new card, where the code on this card replaces the code presently in the program memory and the program continues its execution. As I remember, a typical example was a program (from User's Program Library Europe) for polynomial approximations that resided on 5 or 6 cards.

MERGE statement was also used to merge data registers saved on cards, but that is another story.


#51

I had done some additional reading in my HP-67 manual and had found the section "Pausing to Read a Card" on pages 292-294. Page 294 says in part

Quote:
...The automatic card read during PAUSE allows the programmer to even be absent when the card is read!

That is exactly the reason that I used the INV 2nd Write sequence to change programs part way through the calculations in my "Polynomial Curve Fit with Errors" program for the TI-59. You can see that program in the TI section of the Library section of Viktor Toth's site. Page 10 of the program documentation states in part
Quote:
... The TI-59 transfers blocks of eight instructions to a buffer register for execution. If the Write instruction of an INV 2nd Write sequence is placed in the program as a location divisible by eight then up to seven additional instructions from the originally loaded program may be executed AFTER the new program is loaded in memory. ...

Can the HP-67 do something like that?

Finally, if you look at the first page of the Volume 8 Number 2 (March/April 1983) issue of TI PPC Notes in Viktor Toth's site you will see two examples of the capability of the high resolution graphics mode with the TI-59/PC-100 combination. I recognize that the HP-67 does not have a printer interface. Can the HP-97 provide a graphics capability similar to that of the TI-59/PC-100? I apologize for not presenting an illustration of high resolution graphics here. I am having problems with that part of my computer.

Palmer


#52

The MERGE function on the 67 and 97 will not execute any further instructions of the "old" program before executing the MERGEd code. Of course, the MERGEd code can GSB or GTO any code in the "old" program before the MERGE point.

The 97 doesn't officially support any kind of graphics. Some things can be done using NNNs, but it is also possible to burn out the printhead, depending on whether the PICK chip in your 97 was made by Mostek or AMI. With the AMI PICK (IIRC), attempting to print some NNNs leaves one or more print elements turned on continuously, and they weren't designed for that.

There was a lot of coverage of NNNs, 97 printer graphics and smoking the printheads in the PPC Journal.

The 41C family with the 82143A printer had fully supported graphics printing capability, though the print buffer size precluded printing completely arbitrary graphics.


#53

Quote:
The 97 doesn't officially support any kind of graphics. Some things can be done using NNNs, but it is also possible to burn out the printhead, depending on whether the PICK chip in your 97 was made by Mostek or AMI. With the AMI PICK (IIRC), attempting to print some NNNs leaves one or more print elements turned on continuously, and they weren't designed for that.

There was a lot of coverage of NNNs, 97 printer graphics and smoking the printheads in the PPC Journal.


TI's design didn't support any graphics with the TI-58/59 and PC-100 combination other than printing full characters. The high resolution graphics mode yields much finer resolution graphics because it prints individual dots.

Back when we were doing fast mode and high resolution graphics on the 58/59 TI was very reluctant to publicize it. I was told that they were afraid that the techniques might somehow damage the calculator or printer and that a problem that occurred with the HP-97 was the basis for their concern.

#54

Especially in Germany, didn't SHARP and CASIO have significant market presence in programmables?


#55

Don't know about Germany, but in Italy it's still true nowadays.
SHARPs and CASIOs are sold even in supermarkets....

-- Antonio

#56

IIRC, Sharp & Casio didn't play a major role then - they came later and are well present nowadays - but Commodore was an important competitor for some years. In the very beginning, there were a lot of companies in addition, offering very few, maybe just one electronic calculator. You also found Aristo (the slide rule company) as one major player on the market, even with pretty simple calcs (from today's point of view) with slide rules on their back for the more complicated scientific calculations ;) Pioneer times!


#57

Aristo made some quite advanced scientific calcs for that time (1972-1978) incl. hyperlogs, although none 'with slide rules on their backs'!
That is a confusion with Faber-Castell.


#58

Ooops! Of course you are right! I got these two slide rule manufacturers mixed up. Thanks for your correction.


Possibly Related Threads...
Thread Author Replies Views Last Post
  HP67 eBay auction Cristian Arezzini 3 248 12-08-2013, 03:47 PM
Last Post: Raymond Del Tondo
  HP67/97 M.E. Pac cover Ignacio Sánchez 2 202 11-22-2013, 04:04 AM
Last Post: Ignacio Sánchez Reig
  TI58C By Pierre Houbert Updated Pierre 13 484 09-03-2012, 11:36 AM
Last Post: Pierre
  TI58C By Pierre Houbert Updated Namir 3 170 07-13-2012, 12:03 AM
Last Post: Matt Agajanian
  HP67 motor in HP65 card reader - too slow Harald 4 265 05-21-2012, 06:52 PM
Last Post: Dave Shaffer (Arizona)
  Updated TI58C Emulator by Pierre Houbert Namir 14 433 03-12-2012, 12:29 PM
Last Post: Jeff Johnson
  Android HP67 - v1.1 posted MikeO 14 519 02-08-2012, 07:29 AM
Last Post: Thomas Okken
  HP67 Emulator for Android MikeO 23 771 01-28-2012, 07:06 AM
Last Post: Ingo Paschke
  HP67 for the Android MikeO 17 567 05-24-2011, 10:05 AM
Last Post: Vincent Weber
  HP65 vs HP67 card reader Alberto Fenini 2 151 04-17-2011, 05:09 PM
Last Post: Alberto Fenini

Forum Jump: