Does the new 15c ce utilise IEEE754 floats?

[Commie]

Hi Gents,

I have just received my brand new hp15c ce yesterday and I have been getting acquainted with it. My question is, does the 15c ce use IEEE754 floating point format or bcd floating point.? I think the original hp15c calculators would have used bcd floating point and for good reason.

BCD flpt is much more accurate than IEEE754, I notice SwissMicro's use 128bit IEEE754 flpoint and its not good.

With my hp15c ce sin(x) pi gives a small number instead of zero, a tell tale sign of IEEE754 float usage.

[Nigel (UK)]

IEEE754 supports both binary and decimal bases; Free42 on the DM42 is the decimal version, so it should be all right.

On a non-symbolic calculator, sin pi won’t be exactly zero, because 3.141492654 isn’t exactly pi. So a non-zero answer isn’t necessarily a problem.

I don’t have a HP15-ce, but I believe it runs the original HP15c microcode, which has an excellent reputation.

Nigel (UK)

[Werner]

(11-09-2023 07:08 PM)Commie Wrote:  Hi Gents,

I have just received my brand new hp15c ce yesterday and I have been getting acquainted with it. My question is, does the 15c ce use IEEE754 floating point format or bcd floating point.? I think the original hp15c calculators would have used bcd floating point and for good reason.

The 15CE uses essentially the same ROM as the original 15C, with a few bugs fixed, but running on new hardware, many times faster. All algorithms and number formats etc are the same.

Quote:With my hp15c ce sin(x) pi gives a small number instead of zero, a tell tale sign of IEEE754 float usage.

As has been explained many times before, Pi cannot be represented exactly in finite floating-point (be it BCD or IEEE), and so the Pi button gives you the 10-digit approximation of Pi. Its sine is NOT zero.

Cheers, Werner

[Commie]

(11-09-2023 07:42 PM)Werner Wrote:  The 15CE uses essentially the same ROM as the original 15C, with a few bugs fixed, but running on new hardware, many times faster. All algorithms and number formats etc are the same.

In that case, there doesn't appear to be a problem, it's just that I read some where that swissmicros were using the 128 bit IEEE745 fl.pnt and masking of the lower digits so that the numerical errors are not seen. The reason bcd is used is because it is transparent to i/o i.e., the keyboard input and the output display, require no coding or rom tables to make an approximation.

Assuming that the ARM cpu is emulating the NUT cpu and running original code, I agree, there should be no problem so long as the emulation is good.

Next up, the keyboard, non responsive and double key clicks are noticeable, this is not due to a poor quality keyboard but rather the software debounce code is set too fast.

And lastly, I purchased a ti30 pro mathprint calculator with my hp15c ce and the numerical integrator and solver are just as good as the hp but the ti 30 costs 1/6th the price of the hp 15c ce.True, if it had been the year 1982 the hp15c would have been a game changer, but it is 2023 now and the competition has caught up.

[carey]

(11-12-2023 09:54 PM)Commie Wrote:  And lastly, I purchased a ti30 pro mathprint calculator with my hp15c ce and the numerical integrator and solver are just as good as the hp but the ti 30 costs 1/6th the price of the hp 15c ce.True, if it had been the year 1982 the hp15c would have been a game changer, but it is 2023 now and the competition has caught up.

I too have the ti30 pro mathprint — nice calculator, but it’s non-programmable while the hp15c is programmable, so it’s comparing apples to oranges.

Re: the claim that the ti-30 “numerical integrator and solver are just as good,” you’re invited to try using the ti-30 numerical integrator to solve the contour integral in the hp15c Advanced Functions manual (a current thread topic in this forum). Also, the hp15c accepts complex arguments for all functions. “The competition has caught up”? Maybe in some areas, yes, while in others, it seems to have regressed IMHO.

[Werner]

(11-12-2023 09:54 PM)Commie Wrote:  I read some where that swissmicros were using the 128 bit IEEE745 fl.pnt and masking of the lower digits so that the numerical errors are not seen.

Not sure what you mean here. The DM42 (and DM32) use the Intel Decimal floating-point library, offering 34 digits of precision. There's no 'masking of lower digits' there.

Werner

[ebs]

(11-12-2023 09:54 PM)Commie Wrote:  And lastly, I purchased a ti30 pro mathprint calculator with my hp15c ce and the numerical integrator and solver are just as good as the hp but the ti 30 costs 1/6th the price of the hp 15c ce.True, if it had been the year 1982 the hp15c would have been a game changer, but it is 2023 now and the competition has caught up.

Technical aspects apart, they are completely different products, not comparable at all, the HP15C CE is a reedition of a classic (high-end at its time) with limited production (CE=Collector's Edition) while the TI30 pro M is an entry-level or mid-range (much more recent) consumer calculator.