10 September 2017

Redraw Analog Meter scale with Inkscape

A dual-needle clock.
Guess what? I've had a challenging idea of a (clock) project, and it has suddenly become top priority. Because of the new challenge.

I want to make a clock using a dual needle analog meter. Ideally an SWR meter would be converted into this analog clock. Electronically speaking, with an Arduino and a couple of PWM outputs, interfacing is simple. A few lines of code and you're done.

The challenge lies in redrawing meter scales. The easy part requires to scan the original panel detached from the meter, and then it is all a learning expeerience.


Once you compute that the needle moves over an angle of 60 degrees with a radius of about 48 mm, both scales have to be drawn with proper marks: one every hour or five minutes.

Old fashioned way: draw it by hand. Fun and relaxing, but not too professional looking.

Test printout.
Do it on the computer: how? A drawing software is needed. Possibly free. And with layers so that the original scan can be used as a reference. Something in the back of my mind suggests that I need a vector graphic software and the search engine response is: Inkscape. It runs on Windows, Mac and Linux.
Not willing to install software for a one-time project on "main" home computers, I turn to the 10-year old HP Pavillon laptop on which I installed Ubuntu Linux 12.04. Once I head to the package manager and search for Inkspace I am welcomed with a warm "Already Installed" message.
The Internet is full of tutorials, but Inkscape itself is well documented too. In two hours I manage to redraw my scales the way I want and I even get the size right at the first printout. All for free.
The original scale (on plastic) vs. redrawn.
By the way, the laptop is a Core2Duo with 1.5 GB RAM and 120 GB HDD (dv6236ea Frankenstein). It runs LibreOffice just right. Firefox surfs the Internet without a problem and I can watch HD videos full screen off YouTube because the NVIDIA graphic card is fully supported. Arduino IDE works too and ... Dropbox (which does not work anymore on WinXP)!


06 August 2017

A cute VFD display

At the Marzagla rally I spotted a small box with a 7-segment symbol and "Nixie", all hand drawn/written. Inside there were three small 9-digit VFD's looking NOS. Five € for the three.

The display looked like the Russian IV-18 and IV-21, with all pins on one side, but they have no external markings and one extra digit. So they are not an IV-21. To simplify things these tubes had an insulator slipped over filament wires: that left [9 + (7+1) =] 17 more wires to idenfity (9 digits, 7 segments, 1 decimal point).

Back side. No markings!
The visual area is about 3.5 cm, and the whole tube is about 6.5 cm long. I can make a desktop clock/display since you won't be able to read at a distance.0

Since the filament was already identified, I began from it. I started at about 1Vdc and raised it in partial darkness in order to observe if the filament would start to glow (not good, sign of too much current). I settled at 1.5V and started looking for segments and digits. A good news is that it glows even at 12 V, so I will have not to worry of having and dealing with a higher voltage as most VFDs require (30 V, some even more).

Octopus!

Spreading the wires.
Luckily grid pins are on one side and segments on the other. Not willing to design a PCB for this display I took advantage of the long leads and build a simple adapter on veroboard. I will chop off the extra board under the display and let it hang out of some sort of vertical structure .... when I will be done with all the electronics involved.


A not-so-quick and dirty adapter.


18 July 2017

32600 Li-Ion batteries

Not much is going on this Summer 2017: it is just too hot! Meanwhile a friend gave me a couple of battery packs from a burglar alarm system. They get replaced yearly during periodic service. The label states Li-ion, 2x 3.6 V 13 Ah. Pretty little beast: it can almost direct-drive a bunch of while LEDs and make a lot of light, or last very long. But all contacts measured 0 V, or open circuit. Time to... pry it open!


I was expecting to see some 1SxP configuration and instead I found two 32600 cells. They are huge, like D batteries. Also they are not connected in parallel.

I recall that some Li-Ion batteries have an embedded protection circuit from over-(dis)charging, that disconnects the battery (= open circuit, or 0 V) when it is too low. In fact once I fed some 4V to each cell it sprang to life and started charging.
32600 vs AA

Unfortunately there do not seem to be cheap flashlights that can host this cell size (32600 or the more common and slightly longer 32650), so I will have to build mine. If this cell has a capacity close to 6000 mAh, I can run a 1W while LED for about 15 hours, or a normal 5 mm LED for 300 to 600 hours! Or a booster to 5V makes it a bulky powerbank.

Well, I spoke too soon. Since it already happened to me that a Li-Ion element started self-discharging and self-heating, I was periodically checking voltage under charge and no-load as well as temperature (to the touch). It went like this. Both elements stabilised at 3.96V. When I removed the charging source the voltage dropped steadily down to 3.60V,  about -1 mV/s. At 3.6 V thereabout they heated up, warm to the touch. Voltage stable. After one hour they were ambient temperature and reading 0 V again.

I will not bother with the second pack and bring them both to the recycling center.

Next!

15 June 2017

My take on CloudAtCost surprise maintenance fee


Note: this post goes beyond the interest of the main audience of this blog. But it is the only way I have to share my thoughts. If you have no business with CloudAtCost you can skip this post. But please don't do business with them.

Few years ago I bought a virtual server from a company called CloudAtCost. It was sold with a one-time risible fee that left me wonder what kind of business model was behind it. Now, on June 15th 2017, chickens are coming home to roost. I received an email from them titled "24 Hour Server Suspension Warning" stating that there is an unpaid invoice. So I logged into their user panel to discover that they changed their Terms of Service, and now they include a recurring annual maintenance fee (9 USD). The invoice was issued one month ago, and no email warning had been sent. Also, as I recall, no email warning was issued when ToS were modified.

Over these years their service has been quite lousy, and got worse month after month: I could live without my overseas basic server.

But but but.

I have an invoice of 9 USD expiring in a couple of days. I looked everywhere on their user portal for an "account termination" button, but I couldn't find any. I took the time to read their current ToS and found these worringly interesting points:

9.9 All bills and receipts will be sent to Customer electronically at the current email address provided by Customer in the Customer Account. Customer is responsible to keep such email address up to date with CloudatCost. [NOTE: I got no bill for the maintenance fee, but first pay then complain - in 30 days - ]

9.11 Interest will accrue on any amount not paid for thirty (30) days following the billing date, as and from the billing date at a rate of 2% per month (26.82% per annum) or the maximum legal rate, if less.

9.18 Customers with a onetime payment service is subject to an annual maintenance fee of $9 which will be invoiced 12 months after using our service. This does not apply to users that have a monthly paid service. This Maintenance fee will ensure proper hardware upgrades and maintenance to reduce degradation of onetime payment services. [this generated the current mess]


And, last but not least:

20.1 The terms of this Agreement, including fees, charges, features, content or any other aspects of a Service, may change at any time and without prior notice. The Customer is responsible for frequently reviewing this Agreement posted on CloudatCost' web site to obtain timely notice of any such changes.

Paragraph 20.1 struck me: they don't care to notify their Customers of ToS changes! And they can change the fee of §9.18 anytime, without giving any notice. I am speechless.

Given their Customer-unfriendly manners, I have a strong feeling that all those unpaid maintenance fee invoices generated by "forgotten" accounts will end up at a debt collection firm(s), together with the 26.82% interest rate: what a great way to financially support their "business". A note to European users: their debt would be transferred to a EU-based firm and then subject to EU laws. Improbable? Maybe. But in 5 years such a debt could turn into a request of 100€/$.

My strategy is to pay the invoice and begin the account termination procedure in due time. UNLESS, they REALLY improve their service.

12 June 2017

5870 vs 5750

Nixies 5870/ZM1332 (and alike) and B-5750 share the same pinout, the same digit dimensions and shape, but not the height. Here is a side-by-side comparison of the two tubes during a test in my clock-to-be.

Left are ZM1332/5870's and the rightmost is a 5750:

Left to right: ZM1332, ZM1332, ZM1332, B-5750.

26 May 2017

Staging the Nixie line-up on PCB

First things first: do ZM1332/5870S fit and align on my fresh PCB's? Do they look right? In order to get a first impression I picked four of those Nixies and inserted them into one PCB.

Probably the holes in the footprint I made in KiCad could have been a tad larger. Nixie pins need to be 100% straight to enter, but the black plastic base can be pulled down to keep them aligned. On the other hand, the little beauties stand up perfecly: good when I will solder them to the PCB.

Front view.

Top view.

Back view.

23 May 2017

My firstpcb.com's first PCB's have arrived!

Finally! After a slooooow journey across the world, my first PCB's have arrived! Ten pieces of a board holding 4x ZM1332/5870S Nixies, the driver IC and transistors. The tubes are supposed to be multiplexed offboard.

There is a lot of free board space, I know. I could have packed many more components, but I really wanted something as general purpose as possible that would simply hold the tubes in place at a proper distance. And, a PCB that would work right away (test still pending).

These PCBs were designed in KiCad and fabbed by www.firstpcb.com. I haven't found a problem but, as I said, these are simple two-layer boards.

In case you want to give a try at the whole PCB production process, you may want to register through my referral link https://www.firstPCB.com/mi_odrCr8 which gives both of us a 10 USD bonus (new customers get it nevertheless).

The boards packed in vacuum. Cool.



19 May 2017

My review of "57° Mercatino di Marzaglia" - May 2017

Top quality from the past!
For the first time I attended the famous Mercatino di Marzaglia near Modena (57th edition, May 2017). I believe it is the largest Italian radio-electronics flea market event, only for private sellers and used items. No cheapo Chinese modern stuff, no drones (not even in the air!), not much computers either. It takes place in a large camping site twice a year (mid May and mid September, on Saturday): bring sun lotion, baseball hat, water, ... Free entrance and parking for visitors, small contribution for exhibitors.

What was there? A lot of vintage valve radios in excellent conditions. A lot of 80's and 90's audio equipment. The usual assortment of valves ("NIB" conditions) and maybe overall 15% tables with pure HAM radio stuff (RTX, antennas, accessories), components. Few instruments, very little retrocomputing, some militaria.

I bought (memo to my future self): a 200Vdc Wester Electric voltmeter (without resistor!), 3x IV-21 VFD tiny displays, some HV electrolytic caps for Nixie PSU's. I haven't found other display technologies as spares (Nixies, VFD, early LEDs). The list of things I have not brought home is too long, but I do have good memories.
Parking lot - 7 A.M.!

I arrived at 7:00 am and most stands were already displaying their stuff. I think you shouldn't arrive later. Overall I walked 3h30' and I think I haven't missed a stand, but I have skipped over those sellers that were not calling for my attention. With reference to Friedrichshafen 2015 (1,75 halls) and 2016 (1.5 halls) flea markets, it seems to me that Marzaglia has the same size if not larger. Exhibitors in Marzaglia are mostly Italian. I spotted two Hungarians, two Germans and a French, not counting those that were selling stuff I was not interested in and I didn't even get close.

It has been a positive experience, combined with enogastronomic and cultural tourism in the area (Modena, Reggio Emilia, Parma). I encourage interested visitors to get there early, so sleep over in the area. You better get there by car or stay at the campsite if you're that kind of traveler (you need to bring your own tent/caravan/camper). I have no comments about eating on site because I left early, but it probably gets really crowded at lunch time, so bring your own food. Also, the bar/restaurant building is at the campsite entrance. Plan a stay of at least 5 hours if you want to do some serious business and see everything. Add chat time!

Personally I will attempt a second visit to Marzaglia and perhaps skip FN2017: I fear too few exhibitors will go there in mid-July!
Need some valves?
They were playing "Stayin' Alive" and dancing :)
Six Dekatrons in a row. Rare instrument I dare to say.

Keep on walking. I took this picture and I
missed interesing stuff on the grass!
More, more, more sellers!


18 May 2017

HB-100 300 metres "DX" and opaque windows

When the first "long distance" 10 GHz reception test failed (about 300 m as the crow flies), I knew something was utterly wrong. Based on the indoor NLOS SNR I recalculated the link budget and determined I had enough signal to cover a much longer distance.

The setup was as follows. TX in a plastic box outside at the far away site. RX indoor at home, with RTLSDR and laptop. Just bare HB100 radar modules. Full optical visibility.

No reception at all. I suspected broken wires and cold joints, so I re-checked everything and successfully re-tested at home.

For the second attempt the RX station was on the balcony and I replaced the RTLSDR with the spectrum analyzer: ta-da! My transmitter' signal was there, at 74.8 MHz I.F. as expected. It was fun to notice how millimeter waves bounce everywhere if you're surrounded by concrete buildings and combined with the broad radiation pattern of HB100 antennas I could not get a null.

Then a doubt arose: could the RTLSDR dongle be too insensitive at that frequency? So I powered up the laptop on the kitchen table, the dongle and the receiving module. This time the door to the balcony was open, and I was standing between it and the nearby closed window. Since I was holding the RX in my hand, I moved it around looking for a peak on the spectrogram and noticed it would appear and disappear, loosing 20 dB down into the noise. What was going on? Just five minutes before I couldn't get a null and now the signal was gone still pointing the antenna at the right direction?!

Slow down Paolo. Repeating my movements I could see that the signal was lost when the receiver was behind the window, and present when looking out of the open door. What?! Wow!

The window is made of double-glazing and I know it contains some form of Sun shield: my home windows are also a very effective 10 GHz filter!!

Now I can look for a longer LOS path!

14 May 2017

10 GHz TX ready for the first mission

Now that I have completed, debugged and tested the WFM modulator for HB100 radar modules, it is time to try an outdoor transmission. Since I will place the transmitter at my parent's place I needed an enclosure.

Peter PA1SDB (probably him) measured the attenuation of various materials, and plastic was found to introduce negligible attenuation at 10 GHz. So I bought a plastic box for food at the Chinese shop around the corner (1.19€) and this is the first result:



I love working with plastic because most of the time it can be drilled by hand. So far I have secured the HB100 module (TX only), while the breadboard is a tight fit in there.

The modulator is based on DJ7OO circuit and was built with parts that were in the junk/parts box. Main difference is that the tone generator is built around a NE555 and potentiometer P2 on the voice chain has been connected properly. The NE555 might be replaced with a microcontroller so that more functions can be executed, like bitonal beacon and MCW ID.

What's missing before the test in the outside world is a DC socket and a power switch. Then a beacon/voice toggle and a microphone input too.
The test transmission is scheduled for Tuesday.

10 May 2017

A ionizing experience

In 1987 Italians voted against the use of nuclear power plants to produce electricity on the Italian land. This meant an immediate shut-down of all four nuclear plants and the beginning of a long decommissioning phase. Few years later, real decommissioning works could begin (around 1994) and since few years ago the former nuclear sites are safe enough to be visited by the general public.

SOGIN, the company owning the plants and handling the whole decommissioning process, opened their gates last weekend (6-7 May, 2017) with two possible guided tours: unrestricted and controlled zones. I managed to get a seat in the remainings of the controlled area tour.

Taking pictures was strictly forbidden, so this will be a blind blog post.

The plant in Trino used a PWR reactor build between 1960 and 1964. We dressed up covering shoes and wearing gloves, helmet and white coat. Some participants were given a dosimeter, while the staff wore their own. We climbed up four floors to see the pools from above (still full of water), then moved to the area that was above the reactor (temperature during operation: 60-70 °C). The top position offered a clear view of the vessel from above. Then we went outside to a temporary storage of nuclear waste, full of drums, some already categorised, some not. Probably this was the most active area visited even if the highly active material (reactor fuel) has been moved to France and UK for reprocessing. Lastly, before leaving the controlled zone, we removed all the protective clothing and got two full-body scans for possible contamination. Everyone was clean.

It was a really unique experience. It showed how complex and costly is to decommission a nuclear power plant, regardless it happens after its useful life or before. It was clear that the technology itself is safe, as long as a human error, human stupidity mixed with bravery (Chernobyl disaster) or natural event (Fukushima) do not interfere with the operation of the plant. In my humble opinion the largest problem of this technology is the kind of waste that is produced, which requires ages to become safe. Don't forget that the same kind of waste is produced by nuclear medicine and research centres, even if in smaller quantities.

I asked the tour leader if, in two years from now - the possible next Open Gates event - there will be something left to see in the uncontrolled zone. She nodded. Dismantling the Trino plant will end somewhere near year 2030.


09 May 2017

First WFM on 10 GHz (HB100 radar)

The transmitter of this test.
After fiddling for a while with the modulator circuit by DJ7OO, I managed to transmit audio using Wide-FM through a HB-100 radar module. The single-tone beacon was already working (with a NE555), even though I should try to widen its resulting RF bandwidth (which means increasing the amplitude).

The ability to increase the receiver's bandwidth on an SDR automatically compensates for the short-term frequency instability of the HB-100. At least for "operations" within indoor conditions, that are pretty much static.

The test setup was as follows. RX: RTL-SDR dongle receiving the IF output of a stock HB-100 module. TX: retuned HB-100 driven by my own version of DJ7OO modulator (LM317). No external antenna aids, no real line-of-sight but a loose zig-zag in the house for a total of about 10 metres. There was also a slight polarisation mismatch.

My WFM signal at about 10.452 GHz, as received on another HB-100 radar sensor.
In the conditions described above the signal is about 30 dB peak above noise. I think the peak could be higher if LOS could be established over the same distance. But now it is really time to test it over a longer distance!


25 April 2017

More on HB100 10 GHz frequency agility

In the previous post I wrote about measuring how much a stock HB100 radar module can be moved in frequency. I did another test to answer "where does the frequency go if the metallic shield is removed?"

This test is purely academic since the exposed DRO is much more sensitive to hand/object proximity and temperature changes, therefore unsuitable for most - if any - HAM radio applications. I got a delta of about 360 MHz, so the naked circuit was transmitting somewhere around 10'165 MHz.


21 April 2017

HB100 10 GHz module frequency agility

Before I can get my hands on two Sat-TV LNB's and work on a 10 GHz WFM transceiver, I wanted to measure the frequency agility of a stock HB100 radar module.

The plan was to use one module as a fixed receiver on the factory preset (presumably 10.525 GHz) and tune a second module. Remember that HB100 is both a TX and a RX at the same time, and it emits a continous carrier. If the two TX frequencies differ, their difference will be visible at the ultra-broadband (unfiltered) I.F. output.

At first I connected the IF output of a receiver module to the frequency counter (100 MHz), and I could read up to 10 MHz delta.

Then I routed the IF output to the oscilloscope (100 MHz bandwidth, 10 MHz probe) and, as long as the signal was above 1 mVpk the sine wave period indicated a signal of about 20 MHz.

Time to turn on the spectrum analyzer, that starts at 45 MHz. Bingo! The maximum delta I could get between an untouched HB100 (let's assume 10.525 GHz) and a fully unscrewed tuning harness was 232 MHz, so it was operating roughly on 10.293 GHz.

Also, the closer the tuning screw is to the DRO, the higher the frequency. Screw in to go up in frequency, unscrew to decrease frequency.

Last but not least, I confirm that the receiving mixer works better if it is terminated on some medium impedance. I will try to visualize the signal on the oscilloscope in parallel to the S.A.

14 April 2017

Measuring frequencies around 10 GHz

While changing the frequency of HB-100 10 GHz radar modules is pretty simple, it is not so easy to understand where exactly it has been set. Unless you have got access to an expensive frequency counter, of course.

A common workaround seems to be using a satellite TV LNB so that the whole 10 GHz band is downconverted below 2 GHz, which is more affordable to be measured even with an RTLSDR dongle! As long as everything sits in the same room, there should be enough signal to do meaningful measurements.

I need to dig out one of those LNB's and do some cut and solder.

10 April 2017

Changing DRO frequency: add dielectric (hypothesis)

In order to bring HB-100 radar modules into the 10 GHz HAM allocation, their frequency has to be reduced. Reports say that the stock screw is enough, better if replaced with one with finer thread.

Then I investigated how Dielectric Resonator Oscillators work (wikipedia, no more no less) and tried few simulatons with the provided formula: the more dielectric material, the lower the frequency.

So, assuming that we have some "extra" dielectric laying around, like from a similar (dead) module, it could be worth trying to add it on top and see the resulting frequency. Besides providing a sort-of fixed frequency, it would reduce the number of factors that are influended by temperature and cause frequency instability.

For first experiments I will stick to the screw method. Reports of success/failure are welcome.

02 April 2017

Just ordered my first PCBs

After fiddling with KiCad for few month, reviewing the same circuit over and over again, I decided to stick with a relatively general purpose PCB for four ZM1332/NL5870S nixies. It is a multiplexed design that includes the decoder IC on-board (7441 or equivalent). Anodes must be multiplexed on the logic board since everyone has his preferred way of doing it (pnp, optoisolator, pmos).

This is the 3D render (by KiCad) of the boards I have ordered on firstpcb.com:
Since I was not satisfied with the result of the embedded autorouter I did it myself. I have never designed a 2-layer PCB, so I used this extra degree of freedom only when I was stuck. I ended up with only two vias, and other transitions were handled at pads when needed.
ZM1332 cold cathode displays are small and not too tall, so I put all components on the back side of the board, leaving only tubes on top. There are no overlapping components, so it won't matter which side I start soldering.
In order to simplify routing I have remapped the outputs of driver IC to actual digits, so this will have to be taken into account in the firmware. The two headers mantain a 0.1" spacing even if they are far apart.

I am really curious to see the resulting boards, and to build them of course! By the way, the size is 10x5 cm.

If all goes as planned, I will publish what is needed to reproduce this project. Fabricator emailed me they should ship on April 7th, 2017. No idea how long it will take to get here!



30 March 2017

Challenges of the HB100 10 GHz module

As listed in the previous post, the HB100 10 GHz sensor module poses some challenges when repurposed as an RTX. Let's list them, and possible workarounds.

1) The RF power is in the order of 10 mW (10-12 dBm).
On 10 GHz it is easy to assemble and use a high gain antenna. Some people report that an IKEA lamp has the perfect shape. Just remember that higher gain means narrower beamwidth.

2) Frequency stability was not a design goal for the original destination of use.
Frequency instability can be tamed with proper thermal insulation of the module. The more the merrier. In any case all narrow band modes are out of question. WFM is the way to go.

3) Receiver is direct conversion.
This one, combined with #2, is a bit harder to tackle. You can't do CW or SSB. You can't do FSK. The solution proposed by a U.S. HAM is to work full-duplex. Transmitters are on different frequencies. The received signal is then at an Intermediate Frequency equal to TXQRG difference. If the difference is about 88-108 MHz, you know which WideFM receiver can be used! Actually an RTLSDR dongle receiver allows more frequency agility and flexibility, thus allowing to operate I.F. outside the crowded FMBC band.

There are other reports that HB100 is sensitive to microphonics (mechanical vibrations are picked up and turned into electrical/RF signals). Not hard to keep under control, either.

29 March 2017

Easy way to 10 GHz

I read it on hackaday, then again on G3XBM's blog: there is an easy way to play with 10 GHz. And (very) cheap too!! The idea is to repurpose something originally meant to be used as something else. Like the RTLSDR TV dongles, the 74HC240 buffer and many more in this wonderful hobby.

The HB100 is a microwave sensor module designed to be used as motion and speed (doppler) detector. It operates on 10.525 GHz and can be retuned below 10.500 GHz into the 3 cm HAM band (Italian bandplan) with a screwdriver. It can be frequency modulated through the power supply (I guess you get some AM too). It features both the transmitter and the receiver, (patch) antennas included.

How much? Less than 3 EUROs including shipping. That's three espresso coffees standing in an Italian bar. Or three of the cheapest burgers in the "M" restaurant (their own definition, not mine).

Drawbacks (A.K.A. "challenges"):
1) The RF power is in the order of 10 mW (10-12 dBm).
2) Frequency stability was not a design goal for the original destination of use
3) Receiver is direct conversion

I have ordered 2 pairs and a spare one. I am curious how far the unmodified version will go.

By the way, I have spotted a similar radar device operating on 5.8 GHz and others on 24 GHz (InnoSent IPM165). Maybe ... ?

27 March 2017

UDN6118A VFD driver IC

I haven't tried it myself yet, but the (obsolete, discontinued) UDN6118A IC is an 8-line driver specifically designed for driving vacuum fluorescent displays. With two chips of these, up to 8x "7-segment + decimal-point" displays can be controlled through multiplexing.

Apparently UDN6128A and XO-951 are suitable replacements.

These IC's don't seem to be cheap either, but they do simplify wiring.

18 February 2017

Raytheon 2051 Thyratron vacuum test

The WikiPedia page about Thyratrons mentions that they contain some gas. Either neon, argon, mercury or xenon. This means that the tube can be tested in the same way I do with Nixies, using the high-voltage AC generator.

Now that I built a HV-AC tester out of a CCFL driver crcuit the test is fast. And the picture shows the result:


My Thyratron lights up, so it contains some gas. Which one, according to the color?!

Only the lower half of the glass lights up this way, which confirms something I've read: the gas tends to leave deposits on the glass. I am now holding the tube upside down to see if something changes.



15 February 2017

Tektronix 7A13 plugin ... fixed!

Contary with what I stated in the previous post, I tried to fix the 7A13 plugin. I couldn't leave a burned capacitor in there.

Messages on the TekScopes list suggested a simple test: measure with an ohmmeter the impedance across the tantalum capacitor: if it is zero, then there is a problem. And it measured zero.

So I cut off the original C158 capacitor and promptly replaced it with a modern 100 uF 25V electrolytic. Why cutting instead of desoldering? Because pads might not sustain several re-heating and, in any case, that component will not be reused. Even if the board is densely populated, I could operate without disassembling it, that could have been a nightmare.

With the new capacitor in place, the meter reading was still zero. So I located another couple of tantalums and one of them was dead short even if it looked brand new. Off it came, and a shiny 10uF 16V electrolytic got in the scene.

No more 0 readings across those capacitors. Time to a test in the 7603 chassis ... GO! The baby lives, again.

Left: C158. Right: C165.
This 7A13 has at least one more tantalum that I should replace before I forget where it is located. Then I should do the same replacement in my other plugins, just in case!

Out of curiosity, the burned capacitor measures open circuit and does not smell: I suspect it went up in flames long ago. The smaller one measures short circuit.


11 February 2017

Small change in HP 5082-7300 clock firmware

The picture doesn't show "3" is weaker.
I always feel awkward when I flash a new firmware in a clock. That was the case again with my 4x HP 5082-7300 bedside clock.

After more than six months of uninterrupted operation, the tens of minutes display started loosing brightness and the decimal point burned out. In order to prolong its life I changed the code to switch off the display between 9.00 and 21.00 unless a large change in the incoming light is detected. In that case the time is shown for few seconds, and then off again.

Moreover during night hours the display is PWM'ed to reduce brightness, and heat dissipation is decreased as well.

The updated code is on github.

06 February 2017

Tektronix 7A13 module on fire (almost)

I spotted it in a drawer, so I tried to plug a 7A13 input module into my Tek oscilloscope chassis. In less than 10 seconds on the CRT a bright spot appeared and then all lights went faint. I shouldn't admit that I replugged the unit and tried a second time, and the experiment lasted just two seconds since all lights were faint since I pulled the switch (on this device you pull to power up and push to switch off). No smoke/smell was released.

Next I did two things:
  1. visual inspection of the 7A13 but nothing was obviously burnt (but I met many nice looking components)
  2. searched on TekScopes Yahoo group for similar failures
The search revealed that tantalum capacitors do fail abruptly, and they were used to decouple power lines inside the instrument.

Back to the book then! The user and service manual is a work of art itself, so it was a pleasant experience to go through it. Looking at parts list and their position I finally spotted a polarised capacitor (C518) hidden under wires and behind the module structure:

C518 is the burnt blob in the centre.


I really doubt it originally looked "burnt brown", with a shade of orange towards its pins. So, this is the starting point to fix this instrument.

I will not go much further, since I don't really need this module and it already had a broken relay when I acquired it (debugging and broken relay posts from 10 years ago). Needless to say, this piece of history is the result of 1960's and 1970's engineering and I am glad to have had the chance to see and touch it. Oh, is has a digital display too!


28 January 2017

Arduino Nixie voltage booster: 12 V to 200 V

Nixie PSU kits are cheap and robust, but sometimes they can be "too much" for a project: too big, too costly, too powerful . It is the case of single tube circuits, where the current requirement is pretty low and board/case space too.

The Net is full of projects that use the 555 in place of specialised IC's: a PWM signal controlling a MOSFET it's (almost) all that you need. Since my circuits are usually based on Arduino, why not use its embedded PWM generator? Again, there is at least one fully working Arduino code and diagram on the Net, and that's where I started (thanks to Ian of nixieclock.biz).

While I could get clean HVDC, it was too high: even more than 350 V! But but but it would drop to about 200V when current was drawn. It behaved like a far-from-ideal voltage generator with a high internal resistance.

All of this could be fixed with careful run-time trimming of PWM parameters in software after a thorough (software) calibration, but still the firmware would lack the real-time response provided by specialised circuits. So I went a different way.

I used a Zener diode to keep the voltage at 180 V. When the Nixie lights up it "pulls" the current its way, effectively cutting out the Zener. The trick here is to calculate the drop resistor Rz before the Zener in a way that it will load the power supply slightly less then the Nixie. When displaying a digit this Rz becomes part of the anode current limiting resistor you need to add anyway. Since the Nixie sustain voltage is less than the striking/Zener voltage, you need Ranode > Rz with Ranode = Rx + Rz. Just do the math in such a way that the anode current is a few hundreds of microA higher than the "stand-by" current.

Yes, I know I am wasting power into the Zener, but this way I keep control over the maximum voltage if something fails upwards in the PWM chain. Moreover, since I fully control the firmware, I can save power by turning totally off the voltage booster when nothing is to be displayed. Last but not least, as in the original circuit, a resistive voltage divider lets me read the output value though one of Arduino ADCs.

I will share the diagram in a second post on this topic.

12 January 2017

From Eagle to KiCad

For Xmas 2016 I gave myself a little time to test KiCad as a replacement of Eagle.

Even if I used it for very few hours, I could not get the right feeling with Eagle interface and most of the boards in my wishlist would not fit the free version limit.

So I jumped in the Hack-A-Day tutorial series on "Creating a PCB in Everything" and tried KiCad. I love their old-school keyboard shortcuts. I love the way the interface responds when drawing a schematic diagram. I could easily (cough... I had to) design my own symbols for Nixie and VFD and their respective footprint.

So far it has been a positive and productive experience with ZM1332/5870S and IV-6 drawn in KiCad, both symbol and footprint.

I need to define a couple more tube display components and then I'll work on my first PCB to send to a fab house.

Meanwhile I keep up-to-date my own KiCad libraries on github.

09 January 2017

Nixie Tube Ciapapuer - 3 - final firmware

Lately I have pushed to github a new version of the Nixie Tube Ciapapuer firmware.

The main change lies in the way the random number generator is initialised, that now relies on an external library called Entropy. For simplicity I added the library in the github area too. It relies on small speed differences of oscillators within the Arduino microcontroller.

A second change is that the depoison routine (or slot-machine effect) can go either backwards or forward, and it's all chosen randomly.

That's it.

03 January 2017

Ham calls in a stationery store

I was walking in the neighborhood this afternoon and at the stationer's shop window I saw these bags:


They show supposedly U.S. vanity car plates with HAM callsigns. W6HQF seems to be expired, but NCA is not.

Looking up other callsigns is left as an exercise to the reader.