An HP-35 Production Overview

[Didier Lachieze]

The HP-35 is celebrating it’s 50th birthday in 2022. This is the right time to look back at this amazing calculator, the first personal disruptive device which did put an end to the Slide Rule industry.

I’ve gathered over the years a virtual collection of ~1650 HP-35, mostly from auction sites, but also from different websites or forums where pictures were posted. All of them are identified by their serial number and they also almost all have detailed pictures showing the different elements and accessories that have changed during the production lifetime. Having access to a large sample of units helps bringing some light on the HP-35 production and versions.

Here is a preview of a document I'm working on.

Many details below come from Remembering the HP-35A by Richard Nelson, published for the 35th anniversary of the HP-35 in 2007 and packed with very interesting facts and reference documents about the HP-35.

Production highlights

The HP-35 production started in California end of 1971, in the Advanced Product Division building, 10900 Wolfe Road, Cupertino (now home of the Apple Park), for an official launch on January 4, 1972. The initial planned volume mentioned in an internal HP board meeting held on October 5, 1971 was 12000 units for the HP '72 fiscal year ending October 31, 1972.

Here is an extract of the HP Measure January 1972 issue, highlighting the newly launched HP-35:



The order rate was much higher than anticipated and the HP-35 went quickly on backorder. Six month after introduction it was estimated that the fiscal 1972 sales would be three times the initial expectations despite a 12-week delivery. See the HP Measure June 1972 article "A superstar is born":



In the second half of 1972, the HP Singapore plant was set-up for calculator production for markets outside the U.S. and started to build the HP-35 followed by the newly launched HP-80. This is mentioned by Bill Hewlett in the HP Measure October 1973 issue:



The HP-35 was produced from 1972 until 1975, when it was discontinued and removed from the HP Price List on July 1, 1975, being replaced by the cheaper HP-21, according to the HP Measure August 1975 issue:



For the launch of the HP-80 one year after the HP-35, the HP Measure magazine published a detailed article about the teams and people behind the HP-35 and the HP-80, from Corporate R&D, component divisions, calculator engineering groups and also manufacturing in Cupertino and Singapore. See the HP Measure February 1973 "Critical Mass" article.

Serial Numbers

The serial numbers of HP Calculators manufactured before 1975 are organized by ranges which are specific to each model (35, 45, 65, 80).

Numbers are in the format YYWWC##### where :
YY Is the number of years since 1960.
WW Is the week (roughly) in which the first unit of the range was made.
C Is the country code: A = America (USA), S = Singapore
##### The serial number within the specified range.

The fact that each YYWWC date code is unique to a given model give us a view on the number of units of the model that where build.

Beginning of 1975, the date code format moved from YYWWC to YYMMC, MM being the month where the unit was built, and this prefix was shared across all models built during this month. So we loose the visibility on the production volume by model.

Here are the different ranges that appear in my HP-35 virtual collection:

• Built in America (1292 units):
  • 1143A
  • 1230A
  • 1249A
  • 1302A
  • 1346A
  • 1503A
    
• Built in Singapore (357 units)
  • 1143S
  • 1302S
  • 1402S
  • 1503S
    

Strangely the first units built in Singapore have a 1143 date code, same as the first units built in America one year before.

Excluding the 1503A/1503S date codes from 1975 (a few units) the first thing we can do with all these serial numbers is to look at the difference between two consecutive numbers to see how they are spread.

Starting with units built in America this gives the following chart:



They are quite evenly spread, with 4 major gaps in the numbering.

The first one occurred quite early in the first months of the HP-35 production: between the serial numbers 1143A 05426 and 1143A 25550, which represents ~20000 units missing from the 1143A range.
The second gap is between the serial numbers 1143A 59208 and 1230A 70094, which represents 10886 units
The third one is between the serial numbers 1230A 94870 and 1249A 00151, with 5281 units.
And the fourth gap is between the serial numbers 1249A 13498 and 1302A 20214, with 6716 units.

The first gap is difficult to explain, I don’t know the reason why HP skipped ~20000 serial numbers in the early production months.

The three other gaps occurred when the date code was changed which was generally a way to track a significant internal design change. And at these occasions the unit number was reset to (or close to) the next multiple of 10 thousands.

Looking now at units built in Singapore we have the following chart:



I have proportionally two times less units from Singapore than from USA so they are less evenly spread. There are several differences just below or above 2000 units and one bigger gap at 3278 units between serial numbers 1402S 19110 and 1402S 22388. I don't think this is statistically significant enough to indicate a real gap in the product numbers, rather than a gap due to the lower sampling.

Production Volume

An estimation of the total number of units built can be provided by the different ranges (the numbers below are likely underestimated by a few hundreds/a few thousands as the first and the last serial number of each range should be missing from my virtual collection).

My estimation is that there has been at least 346133 HP-35 units produced overall, split as following:

For units built in America there is a total of 223277 units with 0.58% of these units in my virtual collection or 1 out of every 173 units produced:


I don't know the reasons for the transition from 1143A to 1230A and from 1230A to 1249A.
For the transition to 1302A, according to an internal memo from February 1, 1973 published in Remembering the HP-35A by Richard Nelson, it is indicated that serial prefix 1302A and higher are bug free. This means that the transition to 1302A was driven by the ROM update to fix the '2.02 ln e^x' bug. This is confirmed by the highest serial number in my virtual collection for a unit with the bug, which is 1249A 11780. So we can estimate that all units with prefix 1143A, 1230A, 1249A were produced with the bug, which represents 77194 units at least. I have units with the bug spread across these three date codes.
The next transition, from 1302A to 1346A, occurred when the factory ran out of numbers in the 1302A range.

For units built in Singapore there is a total of 122856 units with 0.29% of these units in my virtual collection or 1 out of every 350 units produced:


The transition from 1143S to 1302S seems to have occurred to mirror the transition to 1302A in the USA to identify bug free units, however in my virtual collection the highest serial number from Singapore with the bug is 1143S 3965, so I can't confirm that all 1143S units were shipped with the bug.
The transition from 1302S to 1402S occurred when the factory ran out of numbers in the 1302S range.

If you have any evidence of an HP-35 unit with a serial number outside of the ranges listed above, let me know and I will gladly update my charts.

Production Dates

Another interesting information is that some of the units in my virtual collection can be linked to a date close to the production date, either from an invoice document, a warranty form or a stamp on the shipping box. With these dates associated to the serial numbers we can plot the evolution of the number of units produced over time.

HP-35 units built in America over time



It starts with a very slow ramp in the first half of 1972 due to the supply constraints, then in the second half of 1972 there is a significant ramp up to cover the backlog of orders. The slope decreases gradually until end of 1974 then the production was further reduced until the end of life mid 1975. I have collected two units with a 1503A date code which would indicate that the last HP-35 units from the USA were built in March 1975.

Looking at the first serial numbers with a date in my virtual collection we can see the slow ramp up as well as the 20k serial numbers skipped in May 1972:

1143A 02546 : HP original invoice dated 11 February 1972
1143A 04969 : Warranty form dated May 1, 1972
1143A 26391 : Warranty form dated May 19, 1972
1143A 27228 : HP original invoice dated as ordered 16 March 72 and shipped 25 May 72.

HP-35 units built in Singapore over time



In Singapore the production ramp is much more linear than in USA, until end of 1974. According to the 1503S date code on 6 units in my virtual collection, the last HP-35 units from Singapore were also built in March 1975.

If you have any evidence of an HP-35 unit with a serial number associated to a date, please let me know.

[TomC]

That's incredible! Thank you for your work and sharing the results. I've tried to do similar type research with both the 35 and the 48 series, but gave up - it is a most tedious process (as I'm sure you know!).

Regards,
TomC

[rprosperi]

Holy smokes Didier, this is amazing! I've only just scanned over it briefly but will read it thoroughly later today when I can read slowly and take it all in.

Words fail to convey sufficient thanks, but A BIG THANK YOU for the huge amount of time and attention you've put into, a product of both many years effort and much passion.

[Massimo Gnerucci]

Thank you Didier for this scrupulous analysis. It surely took many hours, days, months to collect all this info.

Where's the like button?

[AndiGer]

Hi Didier,

Very interesting post. As I read on the 20,000 numbers gap immedíately this came into my mind - from the description on the HP-35 on this site:

"Based on a marketing study, it was believed that they might sell 50,000 units. It turned out that the marketing study was wrong by an order of magnitude. Within the first few months they received orders exceeding their guess as to the total market size. General Electric alone placed an order for 20,000 units."

Maybe they made a reservation of those numbers? But unlikely none of those machines escaped to the wild. Maybe never produced/delivered?

Andi

[Didier Lachieze]

(01-16-2022 10:57 AM)AndiGer Wrote:  Maybe they made a reservation of those numbers? But unlikely none of those machines escaped to the wild. Maybe never produced/delivered?

Initially I had the same thoughts but I highly doubt that if these 20000 units were produced none escaped to the wild unless they were sent to another planet.

Furthermore between unit 1143A 04969 with a warranty form dated May 1, 1972 and unit 1143A 26391 with a warranty form dated May 19, 1972 there is less than three weeks, no way they could have produced more than 20000 units in such a short period.

So these numbers may have been reserved, but they were not produced.

[teenix]

After flipping through the "Remembering the HP-35" document, which is very interesting, on page 6 it mentions the bug was caused by a "6 bit programming error".

I don't know how the 6 bits figure into the changes in the .0002 ARC TAN error, but in the later HP-35's, the way one of the division routines was entered was changed after doing an arc tan math routine. This changed the A and P register values.

At the end of the division in the early HP-35, the result in the C register was 1, whereas in the later HP-35's the result in the C register was 2.

If at this point the C register is changed to 2, then after all the remaining calculations in the early HP-35 were completed, the result came out ok.

I haven't checked the other errors, as they may have other compensation mods in the code.

cheers

Tony

[KimH]

This was a great read, thanks for the hard work in getting this together and especially sharing it here.

What it suggest - maybe stating the obvious here, sorry - is that initially the serial number beginning with YYWW was indeed showing what revision we were dealing with, it did not stay that way.

I can see that on my many Titans, they change YYWW weekly.

If this assumption is correct, it would be fun to get a POW on which models where the last one(s) with the same approach as the Classic 35.

[Didier Lachieze]

(01-17-2022 03:21 PM)KimH Wrote:  What it suggest - maybe stating the obvious here, sorry - is that initially the serial number beginning with YYWW was indeed showing what revision we were dealing with, it did not stay that way.

I can see that on my many Titans, they change YYWW weekly.

Yes the serial number format changed, initially the week WW in the serial number prefix was the first week of the range. This format was used in 1972, 1973 and 1974.

Then in 1975, as I described here, the prefix moved to YYMM, with MM being the Month Manufactured.

It changed again in 1978, I think (but the exact date would need to be confirmed), to YYWW, but this time WW was the week where the unit was manufactured. This format applies to the 41, the Voyagers, Titan and so on until end of the 90's.

[Ren]

Thanks Didier!
Others have said Thanks more eloquently, and more sincerely,
but, Thanks!