(Re-)Print labels on back side of classics

[Sadsilence]

(09-03-2015 04:36 PM)Didier Lachieze Wrote:  

(09-03-2015 04:09 PM)Sadsilence Wrote:  HP-45 with it's "hidden" timer feature is my personal favorite. Inserting a tiny PCB with an adjustable oscillator as replacement for installed inaccurate LC curcuit or unavailable 784 kHz quartz you can get timer to an accuracy of about 0,01% or better instead of 10%...

That's interesting ! Can you share more details about this tiny PCB ?

It was a spontaneous idea after reading this forum arcticle as I realized, that there are simply no 784 kHz crystal units available any more.

Idea is based on LTC6900 (LTC1799 is same with different resistor values). Frequency of those oscillators is set by a simple resistor value. Indeed I use a series or parallel curcuit of two resistors to be able setting frequency correctly. Together with a voltage filter capacitor and an output resistor for voltage level adjustment to HP 45's LED curcuit I put all components on a SOT to DIP PCB, using DIP pads as pads for 0603 smd resisitors and 0603 smd capacitor. What makes those chips so appealing are their little current consumption compared to normal VCOs and the easy setup. No programming, no highly precise voltage etc.

Linear states, that precision is only 2%, which would not pay off the effort. But that is only half the truth. Exactly it means, when using same resistor value for 100x LTC6900 output frequency of all would vary between +/- 2%. One LTC6900 in itself is very stable, you just have to get right resistor value for your unit.

You do not need to have great measurement equipment or high precision resistors. Later one is even bad as you cannot compensate LTC6900 variation. Just use your standard DMM, select resistors (24k9 + 560R from 1% default line in series works great for me) and let HP-45 do the measurement, comparing timer mode after 10000 seconds (2h:46m:40s) with your radio controlled watch. When HP 45 timer is to slow, reduce resistor value, using one out of your selected pool, that shows a little less value on your DMM and repeat test. 10 Ohm change normally affects oscillator frequency at around 0.003%. When timer is too fast increase resistor value. One or two iterations are normally enough to get a difference less than 1s for 10000s run time. To do more is not senseful as LTC6900 frequency is voltage supply depending (battery voltage decreases during runtime = NiCD/NiMH characteristic) and temperature depending. It is not really bad but in 0.01% variation range themselves.

During optimization phase I use long wires and let HP-45 open, supply it with an external 3.75V power suppy. 2 supply wires are soldered to battery supply line of HP-45 (please use a HP-45 pad after on switch for + line. Otherwise battery has to deliver around 1 mA, even when calculator is switch off. Frequency output pin of LTC6900 has to be connected to 330 uH inductor pin like mentioned in crystal article. Do not remove inductor! It is important not directly use output pin, as voltage levels are different. A decoupling resistor with any value between 1k and 10k does a good job.

When being happy with the result, PCB is glued with doublesided tape to inside of GOWANDA chip and wires are truncated and resoldered. As wires I use wirewrap ones' or small insulated copper wires.

Of course there is plenty of room for improvement, e. g. a stabilized Vcc for LTC6900 (easiest way: Z-diode) or a better output curcuit, allowing to remove inductor on HP-45. Nevertheless it is a 1000 times improvement for less than 10 Euro material value with an additional load of only about 1mA.

If there is interest I can produce some already optimized "ready to solder" units together with an "installation guide".

Best regards,

Werner