I've added data for a new diode, 1N4001, and reworked the discussion and added new oscilloscope photos for my diode turn-on time page, viewable by clicking here or through the navigation panel at the top left of this page.

I've finished laying out the battery discharge tester's schematic and hope to start a PCB layout tomorrow.

I've also added a page to this site showing turn-on and turn-off time measurements for four power diodes:

Here's another flashlight lamp - the pre-focused type 222 lamp, used with penlights. The 222 is rated at 2.25 volts / 250 mA, lifetime 5 hours. The current scale shows a similar knee to that seen with the PR2 bulb.
 

I'm about 80% finished entering the schematic of my prototype battery discharge meter into DipTrace, the program I use for printed circuit board layout. It's taken two full days of work to enter the schematic. That seems like a long time, but I'm also making the layout footprints for quite a few parts that are not in DipTrace's stock libraries.

I was curious about the load presented by typical flashlights to batteries, so I took some data this morning using an HP 6038A programmable power supply and my GPIB control software. The first example is a standard 2-cell flashlight, using a PR2 incandescent bulb.

The data is quite different than I found when measuring a variety of instrument lamps on 21 December. There's a distinct knee around 0.4 volts and the data seems linear above and below the knee. In fact, the data is not linear, although the departure from linear is not as pronounced as with the instrument lamps. Over the range 0 to 2.7 volts, for example, the PR2's resistance increases from 0.93 ohms to 5.7 ohms. This is about half the increase in the typical instrument lamp.

I suspect the voltage versus current difference between the PR2 and instrument lamps is related to the PR2's short, stiff filament structure, optimized for low voltage / high current.  The rated life for a PR2 lamp is 15 hours. The PR2 lamp is rated at 2.38 volts / 0.5 ampere, which suggests the normal operating mode with fresh D cells is significantly over the rated voltage, with the lamp designer accepting short filament lifetime in exchange for increased light output.
 

The plot below shows the effect of the LED adapter's electronics. Useful light, by the way, starts appearing below 1.0 volts, and above 1.4 volts, the light output is essentially constant, as measured with a Gossen Pilot 2 light meter.

Looking at the data, it seems that the LED's driver begins functioning around 1.4 volts where it  operates in boost mode "1", with a clear transition to a second operation mode at about 1.8 volts. White LEDs requires about 3 to 3.5 volts across them to conduct, so I'll surmise that the drive circuitry creates a negative voltage output that, when in series with the battery voltage provides about 3 to 3.5 volts across the LED. As the supply voltage increases, there's a clear mode shift around 1.8 volts to an approximately constant power drive, mode "2." Again, this makes some sense in that as the sum of the inverted and direct voltage exceeds 3.5 volts or so, current limiting must be applied to the LED.

The circuitry allows the flashlight to use essentially all the potential of a standard or rechargeable cell, as useful light output is provided well below the normal individual cell exhaustion voltage of 0.8 volts (alkaline) or 1.0 volts (NiMh or NiCd). Additionally, as the cells discharge, the light output remains quite close to constant until cell exhaustion.
 

I've added measurements of the thermal resistance between a TO220 metal tab transistor and a heat sink to the Heat Sink page.
 

I've edited the 21 December and 22 December comments to clarify my meaning and clean up the wording.
 

I fixed several small errors in my 21 December comments.

Also, conversations with Ron, K8AQC, a retired Bell System technician, suggested that some of the WeCo slide-base lamps might have carbon filaments. Based on yesterday's measurements, I took a closer look at the slide-base type 2U and 2G lamps in my junkbox.

I made simple measurements at more-or-less room temperature and at operating temperature for two samples each of Type 2U and 2G lamps. (Voltage in volts, current in amperes and resistance in ohms in the table below.)

If the filament temperature increases over room temperature about 2000 °C for the 2U lamp and a bit more for the 2G lamp, we can compute the temperature coefficient for carbon filaments:

The two temperature changes are rough estimates as I don't have an IR thermometer suitable for filament temperatures, nor an optical pyrometer. The 2G lamp appears clearly whiter to my eye than the 2U lamp, so I've arbitrarily estimated its temperature increase as 10% higher. The 2U lamp operates at about 1.1 watts at 24V, whilst the 2G lamp is at 2.4 watts. The filament structure appears similar, so this factor also suggests the 2G lamp runs at a higher temperature than the 2U lamp.  My estimated temperature coefficient for these carbon filaments, α, is -0.0002, which is within the range -0.0005 to +0.0007 found at  Handbook of Chemistry and Physics, 44th Edition, Temperature Coefficient of Resistivity" pages 2672-2674.

Typical household tungsten incandescent lamps operate around 2800-2900 Kelvin, or approximately 2500-2600 °C. Neither the 2U nor 2G lamp filaments looked nearly as white as a standard household lamp, which leads me to place their operating temperature in the 2000-2200 °C range. Because the light emitted by a black body radiator is proportional to T⁴,  where T is the temperature in kelvins, it does not take much of a reduction in filament temperature to significantly reduce the emitted radiation. There's a second factor at work as well, in that as the filament temperature decreases, the peak emission wavelength increases (more towards the infrared spectrum in this case) thereby reducing the apparent brightness as the human eye is not sensitive to IR. These reasons suggest that my 2000-2200 °C estimate for the carbon filaments in these lamps may be on the low side.

The important point is not so much computing α, but rather than a carbon filament lamp has quite a different resistance versus temperature relationship than does a tungsten filament lamp. For one thing, there is no cold inrush current.  Second, a carbon filament lamp can't be used as a stabilizing feedback element in exactly the same fashion as a tungsten lamp, e.g., HP's 200-series Wein bridge oscillators.

If this subject is interesting, you may wish to read Basic Physics of the Incandescent Lamp(Lightbulb) by Dan MacIsaac, Gary Kanner, and Graydon Anderson, published pages 520-25, The Physics Teacher, Vol. 37, Dec. 1999. I found a copy at http://my.ece.ucsb.edu/bobsclass/134/Handouts/Phys%20Teach%20vol37%20Dec1999.pdf but if this link fails, search for the title and you will find it available at many locations on the Internet.
 

I've added a new piece of test gear to my workshop; an HP/Agilent 6038A digitally-controlled DC power supply. One benefit of a GPIB controlled power supply is that it lets you automate current versus voltage curves with a bit of BASIC code and a GPIB interface. As I've mentioned before, I recommend Prologix's USB-to-GPIB adapter as a reasonably priced flexible adapter. I've recommended it to several fellow hams and all have found it useful.

The 6038A displaying 10.6V at 10 amperes into the constant current load I'm working on.
 

The figure below shows how non-linear the current versus voltage curve is for a selection of nine small incandescent lamps. The stair-step appearance of the data results from the 6038A's output limited current resolution (1 mA steps) . (The Western Electric 2J and 2U lamps are carbon filament and have a different behavior. See the discussion below and my measurements of 22 December 2007 above.)

Tungsten's electrical resistance temperature coefficient is 0.0045/°C at 18°C (increasing to 0.0089/°C at 1000 °C) [1], so the temperature increase corresponding to a resistance increase of 9.0 is 9.0 / 0.0045 = 2000 °C over room temperature, if we use the room temperature coefficient  The filament thus operates at approximately 2300 Kelvin. This estimate is a crude one, as our simplistic computation assumes tungsten has a linear change in resistance with temperature, and the data shows otherwise. (A high temperature IR thermometer, or an optical pyrometer could be used to measure the filament temperature directly, of course.)

Carbon, used in Edison's earliest lamps and in some Western Electric slide-based lamps—including the 2J and 2U lamps plotted above—has a rather different behavior. Between 0 °C and 2500 °C, for example, carbon's resistivity, measured in micro-ohm-cm, decreases from 3500 to 900, a ratio of -3.89:1, corresponding to a temperature  coefficient of -0.0015.[2] Other data cites a resistance versus temperature coefficient α ranging from -0.0005 to +0.0007 [1] Thus, a carbon filament lamp will have a smaller change in resistance when operating compared with tungsten and the change may be higher or lower resistance, depending on the particular composition of the filament, such as impurities in the carbon. From my limited research, however, it seems that it is far more common for carbon lamp filaments to decrease in resistance with increasing temperature.)

We also see carbon's α when looking at the temperature coefficient of a carbon film resistor. Panasonic's carbon film resistors, for example, have a tempco quoted as ranging from ±350 PPM to –150 to –1000 PPM, depending on the resistance value. α is just the tempco, e.g., -1000 PPM corresponds to α = -0.001.

Perhaps the most famous use of the change in resistance of a tungsten lamp with increased applied voltage involves Bill Hewlett and the 200-series audio oscillator. See http://www.hp.com/hpinfo/abouthp/histnfacts/museum/earlyinstruments/0002/other/0002patent.pdf for a copy of his 1942 patent on an amplitude stabilized Wein-bridge audio oscillator.

[1] Handbook of Chemistry and Physics, 44th Ed., Table "Temperature Coefficient of Resistivity" pages 2672-2674.

[2] Id, Table "Resistivity," pages 2665-2671.
 

My Z100 article has made the cover of Jan/Feb 2008 QEX as the "featured article" and is available for reading here http://www.arrl.org/qex/2008/01/smith.pdf

My Z90 article in Mar/Apr 2007 QEX is available for reading here http://www.arrl.org/qex/2007/03/smith.pdf

I added a plot and photographs to my heat sink page, viewable by clicking here. The new data shows how an inexpensive surplus heat sink & fan intended for cooling a Pentium-type integrated circuit can be used for TO-220 style tab-mount transistors.
 

I've added a plot to my heat sink page, viewable by clicking here. The new data shows the improvement when I used AAVID's Thermalcote II thermal transfer paste instead of Dow Corning No. 4 silicon grease. I've had the tube of DC No. 4 for at least 35 years, so it's not surprising that better heat sink joint compounds have been developed and, I believe, DC No. 4 was intended more for electrical insulation and water displacement properties than for a heat sink compound. The results are about a 20 to 25% improvement in power dissipation for the same temperature rise.

I also have started collecting data for two fan-assisted heat sinks, but they are so good that it will take me a couple days to arrange enough power to properly stress them.

The used Z91 has been sold, and the link accordingly removed.

I've modified the Z100 Tuning Aid's firmware (release 2.2) and add it it to my Software Update page. I've also modified the Z100's Assembly and Operating manual and posted it at my Documents page. The modified manual incorporates firmware versions 2.1 and earlier, but not 2.2. Please also download Firmware 2.2 release notes from the Software Update page.

I've been working on the battery evaluator firmware for the last week and have made significant progress. I plan on a third breadboard of the discharge circuit when a few additional parts arrive next week. I will also breadboard a high rate (up to 20 A discharge for a 12 V battery) optional discharge module over the next week or two. This is an exercise in heat dissipation more than anything else.

I've also assigned a model number, the Z200, to the battery evaluator design. I'll provide preliminary specifications for it between now and the end of the year.

I ran two more discharge tests of the Radio Shack AA NiMH 2000 mAh rechargeable batteries yesterday. The first was at C/5 (400 mA) with sample no. 501, after a thorough recharge. It now performs essentially identically with sample no. 500, so the shortfall in capacity I found in earlier was likely due to not allowing the cell to fully charge. (I use a Panasonic 4 cell AA/AAA charger and must have taken it out before the end-of-charge indicator came on.)
 

As a comparison point, I measured the useful life of a fresh (March 2012 expiration) Duracell MN1500 alkaline AA cell and also re-measured one of the Radio Shack 2000 mAh NiMH AA cells after a fresh overnight charge. (I've also re-numbered the test cells from yesterday's tests.)
 

Comparing the NiMH rechargeable cells with the Duracell primary battery shows a remarkable difference in discharge voltage versus time. In fact, at 400 mA constant current discharge, the MN1500 has less than 60% of the amp hour capacity of the rechargeable NiMH cell.

Comparing the energy available from the Duracell with that from the NiMH at a constant 400 mA discharge rate shows a remarkable difference:

Duracell Alkaline: 1.632 watt hours
NiMH #500: 2.454 watt hours

The rechargeable cell provides about 50% greater available energy than the primary cell under this test condition.

We must not forget, however, that the energy extractable from a cell is a complex function of the discharge conditions, and the tests presented above are particularly bad for an alkaline cell. In this regard, http://www.duracell.com/oem/Pdf/others/ATB-5.pdf provides interesting reading.

My battery discharge evaluation design provides for constant current, constant resistance and constant power discharge cycles, along with time-varying discharge rates. I've completed a first pass at the firmware for constant current and constant resistance discharge cycles and should have the constant power firmware operating in a day or two.

Any readers interested in purchasing a kit battery discharge evaluator should drop me a note.

The figure below, reproduced from Duracell's ATB-5 reference, shows how the service life of a MN1500 AA cell varies under different discharge conditions.
 

I've re-breadboarded the PIC part of the battery tester and connected the second load to it and have been working on the controller firmware this week.

As usual, I've moved the prior month's Updates to an archive page, reachable from the table at the top of this page, or by clicking here.

I finished edits on the Z100 QEX article yesterday and faxed the changes to my editor. It should appear in the January/February 2008 QEX.

I've also been working on a revised constant current load for my battery tester project. Photos below show the first version, with a bipolar transistor and a small heatsink and the second version, with a MOSFET device and four current shunts, relay selected. The second version will dissipate at least 10 watts safely.

In a MOSFET device the source current is essentially identical with the drain current, but in a bipolar transistor, the emitter current is comprised of both the collector and base current. For a variety of reasons, its much easier to monitor the current at a ground referenced point, so a MOSFET transistor provides significant accuracy benefits in this design.