29 December 2015

Striped print on Epson inkjet printer

We own an Epson XP-215 inkjet printer/scanner/copier. We print/copy very seldomly and very few pages each time. One of these days I had to copy a drawing and it came out with wide horizontal white areas. A second copy of the same document was even worse!

In order to better understand the problem I created a simple print with increasing numbers, one per line, and you see the result on the right.

Initially I suspected something had gotten into the printer cylinders and periodically moved paper away from the print head, but further prints were even worse. With the help of an Internet search engine I found and watched some tutorials that showed how to clean nozzles that had dried ink all around.

All you need is a piece of kitchen paper, some liquid for cleaning glass (windows) and thin gloves (else you loose sensitivity in your fingers). The quest for these tutorials is left as an exercise to the reader.

Around the black ink nozzle there was something that looked like dry ink. I removed it with the method mentioned above.

Once I re-installed ink cartridges the first test print came out like this:

Almost completely readable. Almost. Then I did a print head check printout (from the Epson tool on Windows). Then after 10 minutes I printed again and everything looked normal.

Mission completed. 60€ printer saved from dumping within one year from purchase. Happy family members.

24 December 2015

biNixie clock: completed.

Less than six months, from first purchase to a completed device. This is a personal record!

In May 2015 I bought some Burroughs B5092 Nixie tubes. Since I couldn't build something meaningful out of three working tubes I opted for a dual-digit clock, and a third B5092 stays as spare.

Inside a decoupage light-wooden box I fit the two sockets, a perfboard with K155DI BCD decoder/drivers, an Arduino clone board with DS1307 RTC and a lumos.sk high voltage PSU.

This clock has been built with components of the Cold War era, from both sides of the wall, mixed with modern microcontrollers and switching power supply.

The firmware shows HH, MM and SS. Every hour all symbols are looped once in a basic cathode de-poisoning routine. I have not implemented any other fancy feature like fading, DST auto.adjust, night mode and so on: it is a simple addictive clock.

Yes, I caught people staring at the box for few cycles before asking what are those numbers.

Edit March 2016: the schematic diagram is now available.

20 December 2015

TIL305 on driver's back

Fast forward to the arrival of ATtiny4313 through few design considerations about controlling TIL305 displays "remotely", and a couple of these dot-matrix displays have jumped on the back of their drivers according to the drawing published earlier on this blog.

Meanwhile I have also completed the onboard font with 0-9 numbers and 7 possible symbols (check the ASCII table and you will see there are seven "holes" between number 9 and capital letter A).

For this first release I have left out current limiting resistors on anodes, so I rely on the average current since in the worst case each pixel is ON only 1/7th of the time. The resulting "component" is compact and TIL305 displays can sit next to each other without short circuits.

With proper fuse settings the ATtiny4313 works between 1.8 V and 6 V, so intensity can be controlled through either "pixel ON time" or DC power supply (but not through PWM dimming).

14 December 2015

CCFL laptop inverter board for Nixie/valve testing - 3 - keeping it on

As it can be seen in the video of a previous post, the CCFL driver enters a shutdown mode if there is no lamp connected. According to the chip datasheet it measures both output voltage and current.

I found a couple of papers describing how the impedance of a CCFL changes at startup, and that Z profile cannot be reproduced with passive components.

Nevermind, how about trying to satisfy the current feedback check with a beefy resistor? With a little bit of Ohm's law, considering that a CCFL lamp should work with 5 mA:

R = V / I = 600 (guesstimate) / 0.005 =  120'000 ohm

I cannot measure the actual voltage, but I am sure I am in the right region. Note that the 120 kohm resistor will have to dissipate 3 W: that's why I wrote "beefy".

In the junkbox I located a 100 kohm 2 W resistor. Put it across output terminals and the frequency counter probe thereabout: powered it up and, as almost expected, the HVAC driver kept running after the first second. The resistor did not get too hot either, so the voltage is probably lower than the estimated 600 V.

Now I need to scrounge a resistor with adequate power rating and find a proper housing for the Nixie tester.

13 December 2015

ATmega48 controls TIL305 with custom font

Before ordering a bunch of ATtiny2313/4313 I wanted to be sure I could control the TIL305 dot matrix display. After few trial and error I chose to power each column for 4 milliseconds and multiplex per-column rather than each individual pixel. This way flickering is acceptable, even if I will probably reduce it to 3 ms as well as the current limiting resistor so to get a tad more brightness than shown in the video below. Right now I picked 1 kohm and 5 V power supply (to the uC) at the video shows the result.

Well, the overall idea and firmware does work, so I can order ATtiny4313. At least 6 of them, if I want to build a clock.

04 December 2015

INPUT_PULLUP and dot matrix display

I tried to use the Atmel uC pullup resistor on input lines to limit the current through an LED (the TIL305 dot matrix display in my case). The declared value on the datasheet is 20 kohm minimum: that means 160 uA flowing through a red LED.

The idea does work, so if you need to source that little current through a pin, you can do it.

But 160 microAmps are too little for the old LEDs inside a TIL305 and they can be seen only in real darkness, and even then they are very dim. Because of this result I have not tested how fast the internal pullup resistor can be switched in and out.

Now I will use column drivers as outputs with proper external current limiting resistors. I will probably use small 1/8th Watt resistors to keep everything compact.

02 December 2015

Dot matrix display programmable interface - for TIL305

Thinking how to use those TIL305 matrix displays, I came up with a way to interface them with an ATtiny2313 saving the correct pins for I2C/TWI communication. The idea is taken from the "tinymatrix" project, but the coupling TIL305/ATtiny2313 is the one shown on my diagram. The decimal point LED has to be sacrificed if you want to keep TWI communication ability, but it is not a big loss for an alphanumeric display.

TIL305 on top of ATtiny2313.

The firmware will scan each dot, one by one, and turn ON those needed to form a predefined pattern. The scanning algorithm will be: column 1 / row 1, column 1 / row 2, ... column 1 / row 7, column 2 / row 1, and so on.

As with every LED a current limiting resistor is needed. They will be installed on the five anode lines ("columns") rather than on the 7 cathodes ("rows"). But how about using the pull-up resistor embedded in most Atmel microcontrollers? According to the datasheet it is 20 kohm typical, or more. At 5 V supply that means very little current through a red LED, but worth a try. How to implement it? When a column is active the corresponding pin will be configured as INPUT_PULLUP (in Arduino terms): this will allow some current to flow through the pin to the dot activated by pulling low the corresponding row. All other four columns will be set as high impedance INPUT. The Atmel datasheet does not mention how fast the pullup resistor can be switched in and out, so this is another thing to learn from this project.

01 December 2015

A bag of matrix display from the past: TIL305

3 types of TIL305.
At the last edition of the local HAM flea market I got a bag of TIL305 5x7 dot matrix displays. [I got quite a lot of them, more than I can possibly use, so if you need them to repair a piece of equiment and landed on this page because of that, write me.]

A new challenge is up! There are few possibilities for driving these displays and most of them rely on equally osbolete hardware or expensive parts (like those from Maxim IC). On hackaday I located an interesting project called "tinymatrix" that uses a similar display with an interesting approach.

The display is soldered on top of a ATtiny2313/ATtiny4313 microcontroller, which handles pixel multiplexing. While the original idea has hardcoded animations, with my TIL305 I can go one step further and use the microcontroller as a matrix driver connected on a I2C/TWI bus. The "driver" can then carry onboard some symbols or receive temporary ones from the master controller.

Time to get some paper and do some preliminary work.

30 November 2015

Simply dangerous sample - NOT

The self-luminescent keychain mentioned in a previous post has arrived today. It is weaker than advertised, but still visible in semi-darkness. With a bit of disappointment I found out that it does not emit radiation types and levels that can be detected on my Samsung smartphone with the Radioactivity Counter app.

22 November 2015

CCFL laptop inverter board for Nixie/valve testing - 2 - video

Want to see how a Nixie tube lights up when it is close to an HVAC source? Here you go:

The board is from an Acer laptop. Power supply is about 11 Vdc plus 3 Vdc for the control circuitry. Since there is no CCF Lamp connected the circuit shuts down after a couple of seconds, but that is enough to see the gas inside the Nixie so show its presence. Re-apply the 3 V control and HVAC appears again.

Remember: no light means broken Nixie.

21 November 2015

CCFL laptop inverter board for Nixie/valve testing - 1

A WORD OF CAUTION. This experiment involves dangerous high voltage AC. I take no responsibility for whatsoever damage may result in reproducing what is described below: if you don't know how to deal with HV, just read and learn.

Surfing around and reading about Nixie testing I came across a text that describes a way to check these tubes in a wireless fashion. With such a device, that is battery operated, it is possible to test Nixies found at flea markets so that you don't buy broken items.

The device produces high voltage AC through an antenna and when put close to the Nixie glass the gas inside lights up. There are commercial products starting from 12€, but a DIY solution is possible too.

The laptop CCFL (screen backlight) driver circuit fits this purpose. You need an older laptop with CCFL and not LED backlight. So all those laptops pre-2008 should do (do your homework first and check its specs before taking it apart).

The circuit usually sits in the lower part of the screen. How to get there probably will be covered in a future post. Hint: you don't need to take apart the whole laptop, just the screen. Otherwise you unscrew a lot for nothing.

Possibly do this job on a laptop you can power up, so that you can reverse engineer the connections. The driver has a least 4 inputs: broad range Vcc (like 8 to 20V), GND, enable and luminosity. The latter two signals usually take up to 3V, some are analog other digital, but in general if you pull both "high" you'll turn on the booster at max strength, that is what we need.

Here is the circuit powered with my own bench PSU and an extra battery for 3V lighting up the original CCFL:

The frequency counter can pick up the 56.1 kHz without any physical connection to the booster: radiated energy is enough! The yellow crocodile brings 3V DC to both Enable and Luminosity pins.

What happens if the CCFL is disconnected? This booster controller chip has a cut-off feature that stops HV if there is no current flow after few seconds. Anyway even without a wire going to the CCFL the frequency counter picks up the signal (at 80 kHz) and a good Nixie lights up too!

Next ... a video showing the effect on a B5092 Nixie.

17 November 2015

Simply dangerous sample?

I have not yet closed the Android radiation monitor case. Besides building a stand-alone circuit with parts that I have at home - sooner or later - I need to locate a source that does not require walking through security checks. One of the Chinese big online sellers has something that might come handy in the form of a self-luminescent keychain. Today I ordered one for about 12€.

Results in a couple of weeks, when the sample will arrive. Testing will be straightforward.

06 November 2015

biNixie clock - firmware complete

I have just tested the latest (and probably the last) firmware for my biNixie clock. It lacked a way to set hours/minutes, that has been implemented with two push buttons and the useful INPUT_PULLUP feature of AVR microcontrollers (the pin is set as input and an internal pull-up resistor is toggled in, so you save external components).

I am now able to understand how precise the RTC module is, so I will decide whether those buttons can be hidden into the case or they need an easier access.

Going happie (or nixpy?) to sleep.

19 October 2015

Real world test of Radioactivity Counter app

As described in a previous post, a German team has developed an Android and iPhone app that detects ionizing radiation ("radioactivity") through the obscured camera sensor. But, how to test that your device actually works if - thanks God - you lack a radioactive source?

Airport security checks and alike is the answer.

I recently flew to Berlin, so for the sake of science and personal learning I gave it a try. I obscured the smartphone camera as described in the App manual. Please note that you need two layers of black cardboard to be effective even in full sunlight. I fired up the App, slipped the smartphone in a pocket  and waited for my turn. The jacket went into the usual box and through security scanning. I tried to stay calm when the operator at the screen asked for a second pass of the box containing my jacket. Got caught? No, I had simply piled up stuff and they couldn't see clearly everything. I grabbed my stuff with a nice smile since my counter disguised as a smartphone passed twice into the radioactive area!

This is the screenshot of both passes. The App log keeps track of hits per minute and probably the second pass was just across the minute change:

The two bars measure about 25k and 10k CPM.

Let us move forward. Upwards, actually, to the sky. When airborne we are exposed to higher doses of radioactivity because there is less air filter between us and the outer space. So, probably, the App could detect something during the flight too. Note 1: flying is perfectly safe with regards to radioactivity!! Note 2: the app ran for 21 minutes on each flight, but I kept the phone screen hidden under a paper or in my pocket.

First flight was before sunrise:

And the second flight after sunrise:

Not a big difference between the two airborne situations, but I can tell that I get fewer CPMs when on Earth.

Last but not least, I dared the security check experiment on the way back, thinking that it would have been fun to explain German security personnel what I was measuring and why. Well, I had plenty of time before boarding. I have no screenshot to show since during my stay in Berlin I needed my smartphone camera and I had no means to cover it back. So I simply fired up the app and threw it in the internal pocket of my Winter jacket: darkness in there is enough not to produce false hits.
The X-ray scan produced 17300 clicks, which is interesting.

Incidentally, the sum of both scans at my home airport is around 34k. Since the phone had gotten Xrayed twice, that makes an average of 17k CPM over those two minutes, which is very very similar to the third scan on the way back.

So. The Radioactivity Counter app does work. I do not claim it can produce calibrated uS/h readings, but it can detect ionizing radiation above normal/natural levels. Just let it self-calibrate, provide a good black screen (try it aiming the camera towards the Sun while the app is running, and look for 0 hits) and try it in a probably radiation free environment (your home, your basement) so that you know what is normal for your specific App/smartphone combination.

Good luck in not finding radioactive sources in your neighborhood!

15 October 2015

A simple way to detect ionizing radiation (Geiger counter and alike)

One object I could not find last June in Friedrichshafen Ham Radio fair is a Geiger-Muller counter. What for? Mainly for curiosity of measuring if my home is radiation free given the amount of surplus around. And a Geiger counter can be re-sold easily afterwards. Failing my purchase, I started an online quest for an alternative.

First I came across a small dongle that plugs into a smartphone earphone/mike hole. It is made in Korea and costs about 30 EUR/USD, with an accuracy of ~30%. Few days of thinking later, I remembered an article on Hack A Day blog about an Android and iPhone app that acts as a Radioactivity Counter (that's the app name). It works by counting how many pixels of an obscured camera sensor turn white because of a high energy particle passing through (beta and gamma).

That's cool! Just need to cover the camera sensor area with thick black paper and let the app run. Fine. Almost no readings around home. It is a good sign, but I need a proof that it works when there is a radiation going on.

So, I have been looking for something that you would never want to find in your life: a beta/gamma radiation source. It is a paradox: you look for something and you hope to never find it! Granite, dangerous energetic jewels, ... all something that I would have to dispose properly afterwards. Then I came up with a different approach.

The most easy way to expose a smartphone to ionizing radiation is to get it X-rayed at some security checkpoint. I realized it when entering the Milan Expo 2015, but the lens was not covered (this "source" is even written on HaD post, right on top!). Too late. Next chance would be an airplane trip to Berlin.

Been there, done it.

Results in the next post!

19 September 2015

Transforming 7-segment LED clock into IV-6 VFD clock

Take a cheap 4-digit 7-segment LED clock kit (like Bangood SKU142210, about 6.5 USD [red, without case]), design an adapter board and replace the display with four IV-6 VFD Russian tubes. That's what I have been working on during the last three weeks.

According to the schematic, the original clock uses common anode displays and multiplexes all four digits: IV-6 VFD satisfy this requisite, even though I have no idea how fast the multiplex is and whether tubes can react that fast.

IV-6 on veroboard, note the 3rd tube leads.
These VFDs require a grid and anode drive of 12-30 V or more, while LED displays and the clock run at 5 V. After few tests I opted to keep the 5 V input voltage (ubiquitous USB...) and insert a step-up module to obtain 12 to 35 V, which also controls luminosity.

The clock microprocessor outputs a low logic level to turn on segments, while VFD requires a high "logic" level, so the adapter board must both adapt voltage levels and invert the signal. Well, the ULN2003 darlington transistor array is fit for this purpose. Since I need to control 7 (segments) + 1 (digital point) + 4 (tubes) I need 12 lines, two ULN2003 chips (total of 7 + 7 = 14 transistors).

For the sake of simplicity I opted for wiring in series the four VFD filaments that operate at 1 Vmax and add a voltage-drop/current-limting resistor on the cold end: this ensures that the anode voltage is below the grid potential, so the segment turns completely off (otherwise it could still be visible in complete darkness).

Wiring up the boards requires a lot of concentration.

In order to limit the current through ULN2003 transistors when they are "ON", meaning a segment is "OFF", I needed to choose a suitable pull-up resistor value: too high and the current will not be enough to switch off, too low and the overall current consumption increases as well as unnecessary heat dissipation. 3k3 ohm is fine but pretty low, 9 mA apparently not worry much, but they mean 270 mW if I run anodes and grids at 30 V, on a single resistor. 51 kohm with their 51 mW (@ 30V) are too much and Darlingtons don't turn off properly. At least at a first test, but I want to add decoupling capacitors on the high-voltage side because moving wires around seem to fix the problem.

31 August 2015

QYT KT8900 - digital noise on audio

One of the most noticeable defects of QYT KT8900 is a digital buzz that can be heard when the stock microphone is plugged in. This problem has been tracked down to the digital logic on the microphone communicating with the main radio board.

The picture shows how the loudspeaker signal looks like on an oscilloscope:


Spikes are about 25 mV peak, one every 3.5 ms or so. I couldn't get a stable trigger on the signal and the picture shows my best capture (100 MHz analog Tek, 100 MHz probe at 1x).

Any idea for a simple fix?

30 August 2015

AA to C cell adapter

[In Italian: "Adattatore da stilo a mezza torcia"]

So, your kid just received a new toys that requires batteries. Not plain simple AA cells, but C size. "Batteries not included" the box says. After the fist moments of unspoken words, this is a good chance to prove your audience your homebrew ability with a macgyverism!

Grab two ubiquitous AA cells and wrap them with cardboard until their diameter fits comfortably into the battery compartment. Fix the wrapping with a bit of adhesive tape. That's it!

AA on the left was partially extracted from the adapter.

How is it possible? AA and C cells share the same height. A C cell is an AA put on steroids. Theoretically it holds 2-3 times the capacity, but some people reported fake C batteries that were wrapped up AA's.

Have fun with the new toy!

26 August 2015

QYT KT8900 current consumption

Is it me, or the QYT KT8900 manual does not specify the current consumption? I measured it. The unit I got my hands on, performs as follows:

RX: 0.27 A (backlight on, stand-by)


Readings taken at 12.5 V out from the PSU.

I have no idea what the RF power output is. But I noticed that it will not transmit below 10 Vdc, while the current drain is constant between 10 and 13 Vdc supply.

24 August 2015

QYT KT8900 arrived

The parcel containing a QYT KT8900 has arrived today. It took less than 20 days and it was shipped through Germany to Italy. I haven't opened it yet...

22 August 2015

QYT KT8900 transceiver

Online reviews do not rate it at the top of the category, but QYT KT8900 VHF/UHF mobile transceiver has worthy features for less than 100 USD. And it is small, very small.

I have ordered one, that will come once it completes its long journey from CN/SG/HK. I am looking for two info that current reviews do not cover:
  • spectral purity of the transmitter
  • microphone sensitivity

My own tests will tell me how it performs. The challenge will be to build a probe that turns 25W RF into something acceptable by the spectrum analyzer...

14 August 2015

Insulated element support for ultralight VHF/UHF Yagi-Uda beams

Once upon a time, 10 years ago, I used to build VHF beams for high mountain contesting that could be carried in the backpack. My best result was a 5 element Yagi of DK7ZB design whose boom was a telescopic fiberglass fishing pole, elements attached with clothes pegs.

The system could not grow, though: longer poles have larger diameter, while there are no large clothes pegs.

I had built a 4 metre long aluminium square boom cut in two sections, that is OK for backpack transportation where there is no vegetation around. I broke it before the first contest.

Fast forward 6-7 years and I still haven't found a solution to the larger clothes pegs. But while hanging around "special offers" from Chinese online retailers I came across this product: Adhesive cable ties in packs of 6. I couldn't pass on them at the incredible price of 0.8€/pack, including shipping. Their original purpose is to keep cables (like USB) from falling off (behind) furniture.

But their base size matches my aluminium boom (6 or 8 mm per side) and the cable hole is perpendicular to the longest side: do you see a better, lightweight and cheap alternative to large clothes pegs?

The clip sitting on my old boom.

8 mm element into the clip.
The hole diameter is about 7.5 mm. An 8 mm dia. element fits perfectly with the right friction. A 6 mm dia. element needs to be embossed at the junction point with some rubber pipe.

While I could stick them to the fishing rod once fully extended, I will rebuild the Yagi beam with a square boom and element supports already in place.

Side view of the element inserted into the support.
Too bad I got only one color!

09 August 2015

biNixie Clock current consumption

The biNixie Clock deserves a box with an antique look. But the technology inside remains modern, so why not use a rechargeable battery supply?

Without any optimization, at 12 Vdc the biNixie Clock drains 82 mA with two digits ON and 65 mA in stand-by. This means a normal battery pack or 2000 mAh would last between 20 and 30 hours. I have to identify how much current each module drains so that I can plan the power supply and the box size accordingly.

I would like to keep it running off a 3S LiPO pack!

05 August 2015

Vacuum Flourescent Display from a Toshiba VCR

On the wave of my recently born passion for neon/vacuum displays I remembered I had a VFD (Vacuum Fluorescent Display) sitting in the "might come handy" box.

The board was removed from a Toshiba VCR and it was the front panel circuitry. Under the display there is a Toshiba TMP47C1670 chip. It is a pretty complex device, probably programmable like a recent microcontroller, plus it can drive VFD natively.

How to light it up? Two things helped tracing down connections: the TMP47C1670 (partial) datasheet and a DVM.

With DVM in continuity test mode I looked for the filament. Incidentally they are the outermost pins. The datasheet provided hints on which pins drove grids (output pin called Gn) and which were for anodes. With the help of DVM I mapped IC pins with display terminals.

That was it. I fed the filament with about 4Vdc from a battery. Then my 12V PSU ground went to one side of the filament and +V to grid and anodes. Moving around grid and anode touch points I could lit all segments:

This experiment was fun and rewarding. I learned a bit more about VFD technology. Unfortunately the display has too many pre-shaped symbols that make it unattractive for reuse. It is offered free + shipping to a good home (with a documented project).

31 July 2015

biNixie clock - preview video

I have finally produced a working firmware for my own implementation of a two-digit clock using two Burroughs B5092 Nixie tubes from late 1960's. The biNixie Clock. It uses
  • an Arduino clone,
  • a lumos.sk high-voltage PSU,
  • K155DI BCD decoders and
  • a DS1307 RTC for timekeeping.
It works from 12Vdc. Each tube anode has an independent variable resistor to set luminosity.

I need to complete the circuit and firmware with a couple of (hidden) buttons to set the time and a proper wooden housing.

I will release the Arduino firmware after tiding up the code.

28 July 2015

Friedrichshafen 2015 - What I bought

This is mainly a log for my future self of what I bought at Friedrichshafen HAM fair 2015:

  • WW2 headset with PTT (1€).
  • Bencher BY-2 paddle.
  • A pair of Hama gamepad USB controllers.
  • Diamond HF-20FX 20m mobile antenna.
  • Burroughs Nixie sockets (priced between 50c and 3€ each).
  • Russian VFD 7-segment IV-3 (1€ each).
  • Russian VFD 7-segment IV-6 (1€ each).
  • Russian VFD IV-1 (1€ each).
  • Nixie driver chips K155DI (equivalent of 74141, 1€ each).
  • Blank spiral mic cord (6 wires, 1€).

17 July 2015

B5094 Nixie now fully operational

The B5094 Nixie was tickling me: why only two symbols light up? Why others don't? I asked around and the hint was to increase the voltage supply (as well as the current limiting resistor).

Worth a try. My PSU module reaches 186 V: not enough to light up other symbols. Then I added in series a 12 V battery pack for a total of 198 V and ... all 3+3 symbols said "Hello!".

The series resistor was increased to 47k that is probably still too low. I need to measure the working current of that specific Nixie and make sure I do not exceed the 3.5 mA stated in the datasheet.

The need for an increased voltage is a sign that the tube is "tired". The 4th B5092 doesn't fire with 198 V. I may try to go higher but that's not too easy to achieve with batteries, or keep it as a souvenir.

09 July 2015

Friedrichshafen 2015 - What I have not bought

These are my regrets. Objects I have not bought at "Ham Radio 2015" in Friedrichshafen but I should have.

A Fluke 8050 bench multimeter (~50 €).

A Heathkit G-1 signal generator. Found under a table. Looked nice.

A strange 1960's telephone loudspeaker (1 €).

A metal box with three flanged N sockets (probably 1 €!).

An empty metal box with handle (probably 1 €1).

There were other interesting things too, mainly stuff older than me. Either too expensive for a home ornament or too big to transport in a already fully packed car.

29 June 2015

1000 km round-trip

Walk for seven hours.
Smell of bakelite.
Dusty hands.
Smell of dusty bakelite.
Curious objects.
An horizontal Tower of Babel.
Refrain from compulsive shopping.
Smell of not-so-healthy food.
Look for old-time friends.
Regrets (I should have bought that, that and that!).
One thousand kilometres round-trip.

That was my first visit at the largest European HAM fair in Friedrichshafen 2015.

I joined a local group of HAMs which included the person that about 30 years ago introduced me into the world of CB.

19 June 2015

IoT thermometer online again

The newer firmware I compiled a couple of night ago did not work. The old firmware, compiled with the latest Arduino-ESP8266 combo, did not work either!

So, the problem lies either in the 1.6.4 combo or the OneWire.h library. Or both.

I can't speak for 1.6.4 vs 1.6.1, except that the resulting binary firmware is larger with the latest GUI version. 46% vs 36%, that's a lot, but I should check the actual value in bytes.

Then OneWire.h library was modified to use the internal pull-up resistor in the ESP8266 microcontroller. I think that is quite wrong since the 1wire specification calls for a 4k7 pull-up resistor while nobody really knows the internal R value, that can vary in the future too. Moreover the new pinMode setting breaks existing implementations - like mine - that diligently used the external resistor.

I reverter to the 1.6.1 compiler and the original 1wire library, with the result that my newer firmware compiles, it is small, and it works.

I added a second loop counter to distinguish between counts from power-up and counts since the last TCP o WiFi failure.

17 June 2015

IoT thermometer offiline

I did few changes to my IoT thermometer firmware (ESP8266 + Arduino IDE) and something broke: it does not read the DS18B20 anymore. I tried two 18B20: their address is read, which confirms they work and are connected properly, but temperature data is never returned.

Throughout the month since I changed the firmware I updated the Arduino IDE, the ESP8266 libraries and did a Windows update too. Too much to debug in a single night.

As a result my ThingSpeak thermometer is offline.

16 June 2015

10k online samples - reset

All was OK, and overnight my DS18B20+ESP8266 online thermometer recorded the 10'000th uninterrupted sample.

But this morning at 11h35 after 10'435 cycles the counter variable was reset. Another reset occurred at 13h50. Either AC power, WiFi or xDSL line were disconnected, or several of them together.

The firmware needs new features that help keeping track remotely of failure reasons. Back to coding, then!

15 June 2015

10k online samples

If all goes well, my ESP8266-based online thermometer will reach ten thousand uninterrupted samples in the next hours. That's a bit more than 6 days. Actually it survived a xDSL outage, and the counter was reset (memo: keep two counters). Nothing to be really proud of, but it somehow proves that the code produced by the ESP-Arduino IDE is stable enough for a long-term project.

09 June 2015

DIY socket idea for Burroughs Nixie B5092 (13 pin) - part 1

I am gathering parts for my Nixies at a rather constant pace. Unfortunately proper sockets are not easy to source these days (at a decent price). The person that sold me my Nixies had no sockets for them. So I looked up online pictures for DIY Nixie sockets and a new idea sprang up.

Several sources point out that the old serial and parallel DB9/DB25 female connectors can donate sockets whose inner diameter matches the one on B5092 pins. Glad that I kept them aside for you-never-know-when! Taking these apart in a clean way is an art itself, but I got there too.

Now I needed a way to mount all those pins (11 for each Nixie).

First idea was to get a socket 3D printed or laser cut, and insert pins in there. I have not found online a model to use, so I would have to design my own and perhaps do few reprints: more expensive than the rare surplus sockets.

On a second thought I opted for etching my own socket. Here's how.

Basically I used the way I built adapters to solder inside MiniDIN8 plugs and my Solder Paddle morse key. Take a piece of unetched copper clad board: that is our conductive sheet on which we will draw insulation segments. Sit the Nixie on the copper and mark the position of each pin with a permanent marker. Burroughs B5092 have 13 + 1 missing = 14 equidistant pin locations.

Now draw lines that join couples of opposite pins inside the rectangle that will be the Nixie base.

Time to etch. Use a heavy cutter and a helping ruler to cut along previously drawn lines. Remember PCB material releases volatile dust: use skin, eyes and breath safety tools! Also take precautions not to get hurt or damage anything around you in the process (like your workshop table).

With an ohmmeter check all 14 slices are insulated. If they are not just pass the cutter few more times.

There you are a 14-slices base for your socket.

To be continued.

05 June 2015

My first Nixie light

With the Lumos PSU kit I was able to test my Nixies in less than an hour (the time needed to build the voltage upconverter). Here she is:

Out of 5 Nixies purchased, three B5092 work. One B5092 is dead and the B5094 only lights up "u" ("micro") and "A" symbols.

Still, I am impressed they do work 46 years after the manufacture date!

03 June 2015

Voltage supply for Nxie tubes

Now that I own four B5092 and one B5094 Nixies I have to turn them on. Datasheet (a scan of) can be found on Internet, together with a wealth of information on these displays. Anyway, as I understand it, it is pretty simple: 170 Vdc and a series resistor (10k in case of B5092) on the anode. The current is less than 3 mA per filament switched on.

Either use a stack of batteries or a DC-DC upconverter. Plenty of projects online for DC-DC step-up converters. But ... I have also found a kit from Slovakia for less than 12€, designed just for Nixies and alike.

In four days it came to my door, even with a real stamp on the envelope! Yes, I used to collect stamps. Top quality PCB and components. Clear instructions in real English language. Assembly should take less than an hour to everyone.

So far I am a satisfied customer of www.lumos.sk kit.

01 June 2015

From the past: nixie tubes!

Some blogs I follow sometimes write about Nixie tubes. Those authors are enthusiastic enough to spark some interest in me. I checked usual online sources and prices are way too high for my curiosity.

Then at the local HAM flea market in May 2015 my friend Angelo spotted a box with few Nixies on a table full of valves. I explained my interest in trying out the technology and few minutes later I walked away with four Burroughs B5092 and one B5094 (we thought they were all 5092's). I wanted just one but the price was just too good to be true (two cappuccino's).

No warranty whatsoever, of course. Nevertheless they look cute, don't they?

When I told this story on the local repeater on Monday morning I learned that Nixie tubes in Italian language are referred to as a female noun. So now I've got to light up five little ladies.

29 May 2015

MacBook Pro battery - not a happy end

Once charged, the 3S(eries)2P(arallel) Li-po pack from a MacBook Pro, was not too keen to keep the top voltage. Each cell parallel slowly decreased voltage from 4.20 V to 4.05 V or so in a couple of days at rest.

So I plugged my usual discharge load: a 12 V / 5 W automotive lamp. It drained about 350 mA. Then I left it there unattended for a couple of hours when the pack voltage had reached ~11 V.

When I came back I found a very faint lamp and this:

Six heavily inflated Li-po cells!! No heat. No smell (these square cells have a sweet smell when they loose gas - and it's toxic of course!).

Each parallel was at 2V thereabout. I took the pack outside and vented each cell to prevent explosion from excessive pressure. Not a big loss since it would not retain enough energy. I have always admired the battery duration on Apple laptops and I think they did a good optimization job in their energy management processes, both software and hardware. A depleted Apple battery is probably really exhaused and not good for reuse.

Morale. Do use a discharge protection circuit if you care of Li-xy cells!

26 May 2015

MacBook laptop battery teardown and recycle

A MacBook laptop battery pack came into my hands. Pretty old stuff from 2006, but Li-po, rated at 60 Wh. Maybe I am about to get a new battery pack. Lacking the laptop for proper charging, I had to teardown the object. Pictures below show the process. It is almost impossible to open it in a way that it can be built back, so I used some brute force here and there to get past some resistance.

The MacBook Pro battery from 2006.

The aluminium lid is glued.

Cells are under the black plastic sheet.

Configuration is 3S2P.

Electronics on-board. Promptly discarded.

And finally charging it.

18 May 2015

It's IoT time! Wireless online thermometer with ESP8266 and DS18B20

IoT, Internet of Things, is not rocket science anymore. Grab an ESP8266 WiFi module, a USB-to-serial adapter and few wires later you are programming it through the familiar Arduino IDE. The most annoying part is providing 3.3 V at 300+ mA as power supply. That's about 10 USD if you already have a computer for programming.

For this initial test I am uploading my shack room temperature to thingspeak.com in the "ZYW Home" channel.

The circuit reads temperature every 60 seconds (almost 62 seconds, actually) and uploads it to the cloud. The temperature sensor is a 1-wire DS18B20. The dashboard shows the average temperature in buckets of 10 minutes and how many measurements have been made since power-on.

Incidentally the latter number provides two more information:
  1. power outage at home if "cycles" restart from 1
  2. xDSL outage or problems if there is a gap or no data upload
 I want to add another temperature sensor for outside and, when ESP-201 modules will come, a solar panel for reading solar illumination. Since ESP-201 allows deep-sleep state, the whole thing could even be battery powered and solar recharged.

10 May 2015

PowerBook G4 ambient light

The Apple PowerBook G4 was a powerful laptop ... in year ~2002. I was given a non-working 12" unit in case I could salvage some parts for my projects. Well, besides the battery pack that still holds the charge and lots of small screws, not much in there can be reused. But I was intrigued to find out how the typical Apple cover pulsing light works. So I started carefully dismantling the laptop until I was left with the metallic cover only.
While I have not really understood where the pulsing light is, I had a bright idea: use the metallic cover as heatsink for 1W LEDs and create a diffused light lamp.

After four LEDs and a little wiring the object finally shone:

The lamp held at about 10cm from the wall
In the beginning I connected the 4 LEDs in series, requiring about 12V 300 mA, but that voltage was a problem for a battery powered object that had to keep the low and sleek profile of the original source. So I rewired LEDs in parallel, and now I can use a 3.7 V LiPo cell that can sustain 1.5 A of current. That's much easier to recharge and doesn't need a booster circuit either.

The lamp with other room illumination.

Battery and (mesy) wiring.

I am really satisfied with the result. In a dark 20 sq.m room the lamp produces enough light for reading. I need to find a simple way to hang it 10 cm or so from the wall and a convenient power switch (now it is a jumper, the blue one in the third picture).

Possible improvements: PWM the current so that intensity and current drain can be reduced; use a microcontroller to control each LED and put up a light show, ...

29 April 2015

Pixie kit - I give up

I have checked everything 10 times.
I have removed key components.


The 40m Pixie kit oscillator does not start, unless the varicap diode is shorted meaning there is neither frequency agility nor TX/RX shift.

I give up. In the time spent debugging I would have built the same circuit dead-bug style. Maybe I will. One day.

28 April 2015

Pixie kit - does not oscillate

The 40m Chinese Pixie replica does not work. The oscillator does not start. Can't get simpler than this.

Besides some capacitor values very different from most common designs, I cannot get the oscillator to start if the RIT/VXO arrangment is in place.

When the varicap diode (reverse polarized 1N400x) is shorted the oscillator runs smootly. Otherwise nothing happens. The supplied diode is OK, because I replaced it with no improvement.

The coupling capacitor to the PA stage isn't guilty either, since I removed it and still no oscillation.

There must be something wrong in the cap value at the RIT/XIT side. It is one of those capacitors where the documented value doesn't match.

Further investigation is needed. With a Manhattan/dead-bug layout would have been easier and safer.

24 April 2015

Battery upgrade for electronic keyboard

As most families with children, we happen to inherit toys and alike. We got an electronic keyboard (musical instrument) called "Miles Electronic Keyboard 3738". To my surprise it was built around year 2000, not earlier as the electronics inside would suggest. It works either on 220V AC or 9 V DC provided by six 1.5 V D cells. My older daughter is currently playing with it around home and I don't like her constantly looking for an AC outlet. But no way I buy six heavy D cells that are probably made of AAA on steroids.

Time for a battery upgrade!

First of all, the PSU board is very simple, with a bridge rectifier and a 78L09 9.8V zener: the required current must be low. In fact the current drawn peaks to 40 mA when emitting sounds, and 16 mA in stand-by. Why D cells?! Anyway, I decided to fit a rechargable battery in there, so I chose a recycled 18650 3.7 V Li-ion (originally 2000 mAh) followed by a boost step-up cicuit (I had bought a stock of them). The current requirement does not call for two cells in series.

So I rearranged wiring in a way that the on-off switch acts between the battery positive pole and the boost converter. The AC PSU was already always on when plugged in, only audio electronics were switched off. This fact gives me the chance to repurpose the PSU as a constant voltage Li-ion 1-cell charger. It is now voltage conditioned and fed through a diode (protection against reverse discharge) right to the battery.

The first circuit uses two LEDs (red and white in series) to step down from 9 V to 4.2 V max, but the current flow is too small to recharge in a decent amount of time. It's about 20 mA, as much as the stand-by device current consumption. We've got plenty of non-use time, but I don't want to have the keyboard around, keeping an AC outlet busy just for charging.

Second thought was to fit a Zener diode as voltage conditioner (Resistor - Zener to ground - Diode to battery). Still not optimal.

Third thought: replace the 78L09 with a 78(L)05 and drop voltage to 4.2 V with a couple of diodes in series. Make sure that the recharge current lies within the voltage regulator specs and let them play! Nope! There is no 78L09 in there, just a Zener. So, nevermind, I will leave it at 9V, keep the LEDs and the slow charge.

The messy area inside the keyboard.
The battery and the voltage boost are the red heatshrinked block, that has been superglued to the keyboard top cover.

Pixie kit - so cheap because ...

While buiding the cheap Pixie kit I came to a dead spot when placing few capacitors. So I did few investigations.

First of all, the 10uF felt too small for the marked 25V rating. I searched my junk box for an older 10 uF 16V cap and they are physically the same. See the picture:

Something else puzzled me: the value of C2, C4, C8, C11. It is specified 100 nF but the marking to look for is 103, 10*10^3, 10'000 pF = 10 nF. Which is the correct one? I will search online for other 40m Pixie diagrams and place the correct value.

Then I moved my study to resistors. I had felt something wrong when bending leads: too thin. The comparison with "old" 1/4 W and 1/8 W volunteer samples from the junk box confirms the anomaly. Even the 1/8 W resistor has thicker leads! (click the picture to get a larger version).

Same conclusion for the supplied "dummy load" (blueish resistor): much thinner leads than an old 0.5 W resistor (I have no idea when I bought it in the last 20 years).

Summing it up. I think the Pixie kit is so incredibly cheap most probably because the supplied components are not top quality. Electrolitic caps with wrong voltage rating. Resistors with wrong leads, at least from 3 dB ("one stop" in photography terms, or half) lower power rating. The print error on parts list would be even easier to correct. I wonder what's wrong with the supplied BNC socket (lossy? wrong impedance?) and transistors.

Nevermind. I have already soldered a socket for transistors so that I can try different, more trusted, parts. At 7 MHz the extra stray capacitance should not matter. The price paid is good for the PCB, if it is OK as well...

22 April 2015

Pixie kit - progress

I couldn't resist and started building the cheap Chinese Pixie 40m RTX kit the same day I received it. I looked for a modular route but there is no real block separation, so I decided to solder all passive components before the smoke test.

Building notes:
  • I got far more capacitors than needed, which can be confusing for the inexperienced builder,
  • components have very thin leads, it almost feels they fall apart when bent,
  • components themselves (mainly capacitors) seem too small for their value,
  • the PCB quality is OK but will probably not withstand solder heat applied too long

I am half-way soldering it. In the next building round I will probably reach the smoke test moment.

21 April 2015

Pixie kit, arrived

The Pixie QRP CW RTX has arrived. The PCB is really tiny, 5x5 cm. I am unsure if I should build it AS-IS or apply few mods to improve user experience, like:
  • tuning pot in place of the small resistive trimmer
  • improve the VXO
  • do not use onboard CW and earphone connectors, to ease installation in a repurposed aluminium candy box
I will certainly add a socket for the LM386 IC, for the XTAL and the two transistors. If I am able to swap transistors I can possibly improve TX power or RX sensitivity, or both. And a removable XTAL adds some extra frequency agility.

11 April 2015

Modding RTC DS1307 for non-rechargable battery

I wanted a stand-alone clock for my solar irradiance monitor. So I bought a batch of 5x DS1307 RTC modules for Arduino. The item description and pictures claimed their backup battery was a rechargeable R2032 cell, but they all arrived with a standard CR2032 battery.

Click to zoom
Not a bad purchase, considering the overall cost and how much I would pay for a single CR2032 battery here in Italy (the same price of the whole RTC module), or the stand-by current consumption of the RTC, in the order of uA (microAmps!) that equals the self-discharge current of most battery chemistry.

The problem is that the RTC board includes charging circuitry and when supplied with 5V it tries to recharge a non-rechargable battery. This process damages the backup battery and can cause all sort of troubles (failure to retain clock time, leak, fire). How to fix it?

This is the relevant portion of the backup circuitry:

You may apply two different mods, both valid.

First of all, the recharge can be disabled by removing the (voltage drop) diode. This is enough, but purists may wish to go further.

A fully charged CR2032 cell exhibits a voltage of 3-3.2 Volts, while a rechargable Lithium battery reaches 4.2 Volts. Therefore the module designer added a resistive voltage divider to drop 4.2 V to 3.2 V or so. That's what R4 and R6 do.

So, now you may operate in two different ways, both shown in the picture above. Either remove R4 that goes to ground or also remove&short R6 to feed the battery voltage directly to the DS1307 RTC chip. In the first case the current drawn is so small that the voltage drop across R6 is irrilevant. The second choice is obviously the best fix.

Please note that the procedure described above needs to be reversed if you want to use a LIR2032 in the future.

31 March 2015

Pixie kit, for fun

I read it on another blog. I checked the eBay item and I simply couldn't pass on this one ... for less than 5€@2015 shipping included. Most of all, I was interested to find out if the kit arrives with the ceramic capacitor sticking out of the BNC socket, since it is there in all the pictures showing the product.

Most likely I will have a simple QRPp rig to play with: add VXO and switchable XTAL, improve the RF bandpass filter, increase the output power.

23 March 2015

Breadboard power supply module destructive failure

If you have been shopping for DIY electronics lately, you have probably met the small board pictured here. It fits on the solderless breadboard and provides both 5V and 3.3V regulated output from a 6-12 Vdc input. My two specimens are marked "YwRobot Power MB V2 545043" and probably came from two online sources given that silkscreens look different.

/* If you are in a hurry, you can fast-forward to my conclusions right after the second picture. */

I used this one in my Arduino solar energy monitor and I noticed that AMS1117 regulators were getting pretty hot to touch. A Nano clone, an idle SD card adapter and an HD44780 display shouldn't drain that much current, and if they did they would be warm to touch too. [edit: the whole circuit draws about 50 mA except, I think, for those short moments it writes to the SD card once a minute.]

So I wanted to measure the input current, since that's the easiest point to insert an ammeter. I disconnected the positive lead from the 12 V PSU and inserted there my DVM ... the power adapter switched on but Arduino was behaving erratically. I did a power cycle and nothing happened anymore, no reassuring LEDs saying hello, not even on the YwRobot board.

The DVM internal fuse could have blown, so I restored the original 12 V positive lead and powered everything up.

Arduino LEDs were blinking awkwardly, or not blinking at all. Another wiring check confirmed that nothing had moved or shorted.

Just to be sure I measured if 5 V were still there and ... SURPRISE! I measured 12 V where I was expecting 5 V!

Since I was feeding 5 V directly to the Arduino board, it was fried (the FT232 USB-to-serial adapter survived, though!). The SD card adapter uses its own 3v3 regulator, so it was not damaged either.

I removed the power supply module and verified:
  • 5V output is now few hundred mV below input voltage, that's 12V or so
  • 3.3V output is not present anymore

What the heck is going on here? I decided to sacrifice my second power adapter module, that was equally heating up in my final application circuit.

This time I played it safer and removed the YwRobot adapter board from the solderless breadboard.

First of all, I wanted to know the idle current of the power adapter board alone. I removed it from the breadboard, added my DVM on the positive lead, fed 13V and the current drawn settled to ~20mA, which is fine according to AMS1117 datasheet (10 mA minimum load to keep regulating, and the board mounts two in series). After 5 seconds the board emitted the typical crackling sound of pine wood in the fireplace. And the unpleasant smell of burned electronics.

This time the DVM fuse didn't blow (10 A ;) ) and no other electronics were harmed.

The two broken adapters. Notice silkscreen differences... that don't make a difference!

Post-mortem analysis revealed that:
  • the AMS1117-5 (5V regulator) shows a resistance of 14 ohm between input and output pins
  • it passes through the input voltage
  • the first board's AMS1117-3.3 blew up, the second is still fine
I suspect that those YwRobot MB V2 boards go crazy if input lead lengths differ (I added about 100 cm of wire on the +V): both of them failed this way. What worries me is that failure breaks regulation and passes input voltage to the output pin instead of opening the circuit.

In both cases 220V AC to 12V DC was provided with a solid-state 5A PSU that mounts two 78M12 in parallel.

Curiously enough I find no mention elsewhere of this kind of failure. I do not own other YwRobot adapters and I do not plan to buy any more. Please let me know your thoughts in the comments.