ESP32 Big filament clock

Inspired by one of Big Clive's YouTube videos (see here), I fancied doing something with the LED "filaments" you can get from AliExpress. 

I fancy making a big 7-segment display clock. 

Some where duly ordered up, and arrived some time later.

Now they need around 47V to light... and (allegedly) 30mA , but experiments show they are more than bright enough at 3.3mA for our purposes...

So, let's make a clock display!

Problem no 1: The power supply. 

Unlike Clive, I'm not going to run the clock off rectified mains ... I did contemplate a capacitive dropper, but, again, it's not safe. 

A quick dig about in the junque box did provide a couple of suitable transformers... one with 18-0-18 and one with 15-0-15, with 220V primaries. As we are normally somewhat above 220V here, and the secondary will very lightly loaded, these would probably do, and some tests are run. Getting an low voltage output to run the micro will mean another transformer ...

The I though about using one of the ebay buck-boost modules... except for they top out at 35V .... or do they?

Now the module is based on the LM2577_ADJ. Shown here in it's boost configuration. R1 on the module is a 10K multi-turn pot, and R2 is a fixed 360 ohm resistor.  

The output voltage is controlled by the feedback pin, so that the feedback pin is always at 1.2V. The module differs from the schematic shown here, as it has two inductors to allow for buck and boost operation, but the voltage feedback section remains the same.

 A quick calculation shows that making that resistor 250 ohms would give us an output voltage of 50V. The existing R1 is removed... 

... and replaced with a 220 ohm resistor soldered to the rear. 220 ohms will allow us some adjustment. 

The output capacitor is removed, as it's only got a 35V rating.

... it's replaced with a 100uF 63V part.

The module is powered up and adjusted for 50V

And a filament is tested. The unit is being powered from 5V, useful, as I can power the whole thing from a 5V wall wart. 

To keep power consumption down a bit, I'm going to multiplex the display. It also reduces the wiring somewhat. Each of the 4 digits of the clock will have it's own anode supply, with only one digit ever lit at a time. This allows us to connect each of the segment's cathodes together. 

The anodes will require a high-side drive, that will happily hold off our 50V. After looking around in the transistor box, it looks like a 2N7000 will do. It's a 60V FET. We'll need two per drive. I've got a load of them, and they're cheap. 

The micro's 3.3V high signal comes in via the 10K resistor, and switches on the lower 2N7000. This removes the gate signal from the upper 2N7000, and removes the anode's supply. The two resistors (100K and 47K) are there as a potential divider, so we do not exceed the maximum gate-to-source voltage of 20V. We'll need to remember the logic is inversed, input of high turns the output low, but we can sort this in software. 

It's built up and tested on a breadboard. 

For the cathode drive, we can use our old friend the ULN2003 (just, it's rated to a maximum of 50V... we could have used individual transistors again, say the 2N7000... anything rated to 50V would have done) 

The micro's inputs come into pins 1-7 and the output to each cathode via it's own 1K current limiting resistor. 

Excellent.

We'll use the ESP32, as we can utilise it's Wifi to get the current time via NTP, and it has it's own RTC. It's proved excellent in the ESP32 analogue clock. 

The end schematic is rather simple. I've added switches for wifi reset (more about that later) and 12/24 hours, and an output to drive the colon separator between the hours and minutes.

I did contemplate a PCB... but it's quite simple, so it's implemented on a piece of perfboard.

So, how to mount the filaments? 

Fusion 360 is fired up, and a design created. 

It's just a bit too large for my 3D printer, but it should be OK... 

I did have to scale it slightly to fit on my Anycubic Mega S, a friend of mine has a Pursia, and he rekons it'll fit on that if this works out too small. Anyway, it's printed, and sanded back to get rid of any lines, and painted with a little high-build primer. 

and finally finished in a bit of satin black. 

The LEDs are tested on the bench supply, some are intermittently faulty, or just dead.

Each filament gets a couple of wires soldered to each end, and threaded into the back... 

Much tea later ... 

The filaments are quite fragile, and I lost a few along the way... 

A couple of warm white LEDs are inserted to form the colon, and everything sealed into place with a dab of RTV. 

The whole thing is wired up ... each digit's anodes are connected together, and each segments cathodes are connected together..

What a rats nest !

The software owes a lot to the ESP32 analogue clock from earlier in the year. It still uses NTP to acquire the time, and uses the ESP's in built RTC. The new parts are a software multiplexer for the display... it goes like this....

1. All anodes off.

2. Set cathodes.

3. Switch on one anode.

... and repeat until the display is complete. It's identical in principle to the display driver for the STD305D turntable The ESP32 is more than capable of doing this faster than the eye can make out (in fact, it's capable of doing it FASTER than circuitry can switch the LEDs on, so there's a delay in the programme to slow it down!), so it just looks like the whole display is lit. 

The other nice bit about the software is the use of the WiFiManager library, which gives us an easy way to connect to wifi without all that tedious hard coding. That's also why I've provided a reset button, which erases the WiFi credentials, and resets the unit, so you can add a new SSID, if yours changes or you move the clock to another location with a different SSID.

Power consumption of the running clock is a bit over 250mA 

The software, as usual can be found on my git, along with the 3D model for the back panel. 

https://github.com/andydoswell/filamentclock

WiFi Manager in use ...

Just for fun, I did run one of the filaments at 30mA just to see what would happen... 

It was incredibly bright ... and lasted for about 30 mins before starting to flicker, eventually going open circuit after another 20 minutes. 

The ESP32 Analog (Analogue!) Clock

As promised, here's the new stupid analogue clock. 

It consists of two 270 degree analogue meters, thoughtfully calibrated 0-70V and 0-300A, kindly donated by Andy from North Hill Audio in Malvern, and an ESP32 with it's nice in-built RTC, wifi etc etc...

The idea is almost identical to the previously stupid analogue clock, found here. This time, we will use the ESP32's built in wifi to connect to the internet, and pick up the time from an NTP server, and set the ESP32's inbuilt RTC. There's an example in the ESP32's arduino library to do this, and utilises the standard time.h library to maintain the clock. I'll set a random time to refresh the RTC randomly between 12 and 24 hours, I don't want to keep bothering the NTP server. I can then output hours and minutes using PWM to drive the meter movements. 

First off we're going to need to do something about the scaling on our meter movements, as I can't easily drive 0-70 volts, nor 0-300 amps! 

Examining the meters shows the 0-300A meter is actually a 75mV full scale deflection meter (FSD), designed to be used with an external shunt of 0.025 ohms (or does it ... read on), this is good, as we can easily drive that.

The 0-70 volt meter reads directly. Disassembling it shows it has a "multiplier" resistor (actually 3 resistors in series, one adjustable) with the earthy end of the movement. 

Shorting out the multiplier gives us a meter with about 200mV FSD, perfect. The meters are dated 1972.

Both meters pull a fair amount of current ... 

A schematic is conjured up. There's not much too it.

As ESP32 has no native support for analogWrite, but we can use the ledc function. I don't care it's not driving an LED, and neither should you 😉. Each output pin, feeds the base of BC547 NPN transistor which is driving the meter. There's a 1K pot in the emitter to allow FSD for the meters to be set accurately. The venerable 7805 regulator supplies our +5V to the ESP32 and meters. 

There's an output on IO18 for a PM indicator LED, and an output on IO15, which I'm going to use to drive a relay, to give the clock a ticking sound ...

The script has some differences to the original ATMEGA328P script, as there's no 409Hz interrupt timing required, as the RTC is doing that job for us, and the code to receive the 433MHz timing signal has now gone, as that's now taken over by NTP. There's a section to extract the individual hours, minutes and seconds from the RTC, and load them into separate variables that we can manipulate. There's also no "smooth" function, as both hours and minutes are smooth, and there's no 400Hz interrupt timer to give us 1/409th of a second to drive a "smooth" seconds meter, and there isn't one anyway! The clock is also permanently set to 12 hour mode, with a PM indicator on pin XX. I've retained the fabulous "IsBST" routine. It just works. If you're outside of the UK, you'll need to modify the code to suit your timezone.

The software can be found on my github page at https://github.com/andydoswell/ESP_32_analogue_clock

The electronics is mounted up on perf board.

So that's the electronics and software sorted. 

I ordered a wooden craft box from eBay to mount it all in.

A couple of stains and finishing oil are tried out on the inside... 

.. and while they dry, I use libre cad to make a cutting plan for the front. 

The template is printed out , and taped to the box.

The two holes are cut out for the meters

The box given many coats of finishing oil to bring out the grain, and give it some shine.

I had to get a bit creative when  mounting the meters, as I didn't have the original mounting hardware..

I added some redundant 19" rack handles as a stand. I'm pleased with the results.

There, and goodnight, it's 250 amps past 51 volts!

I notice there's a kickstarter, where you can now buy something very similar. 

Pyers read this last night, and has pointed out the value of the shunt is wrong. I simply glanced at the back of the 300A meter, and mis-read it... I had wrongly assumed 0.025ohms was the value of the shunt... But 75mV across 0.025 ohms only gives us 3A... not 300.

Reading the label correctly, it simply states "Use with external shunt" , and the 0.025ohms is "lead resistance". The shunt should be 0.00025 ohms for a full scale deflection of 300A. 

Arduino TV clock with vintage clock source (or another stupid Arduino clock, or "The Ping-Pong Clock")

A while back, the boss presented me with a gift ...

"Found this, I thought you'd like it"

It's a pair of 460KHz Crystals in a standard B7G glass valve envelope. Pretty.












It's sat on my desk for months ... I keep looking at it, and wondering what to do with it...

About the same time, my mate Alan gave me a small CRT video monitor.













I think it's come from a reversing camera from a lorry...










After working out the power pins on the rear, it springs into life, and a stupid plan is hatched....

Take the crystal, and build a CD 4060 oscillator/ripple counter, divide the output down to produce ~28Hz (460000/16384). Feed the 28Hz into an Arduino interrupt pin, and get it to run a clock. Run the TV out library as well, and produce an image on the screen. Brilliant. Stupid. All in one go.

Now, with a crystal oscillator, we need to know the crystal's "load" capacitance. This is formed by the two capacitors off each crystal leg to ground. Every crystal needs these. Get it wrong, and it's unlikely to reliably oscillate (if it does at all). For example, the usual Arduino crystal of 16MHz needs about 22pF on each leg to reliably start up. We'll also need to bias the 4060 so it always starts. Regular readers of this blog will know I used to include a 1 megohm resistor (unnecessarily) across the arduino crystal in my stand-alone designs. We'll need this in our oscillator. (We didn't in the ATMega328 because the bias is supplied by the IC). We'll also need a load resistor. The load resistor is there to make sure there's a voltage on the crystal with which to start the oscillation.

The issue we have is we have no idea of the load capacitance required, or it's required load resistor. There's a rule of thumb about load resistors. 1mm thickness of crystal = 1K ... our crystal is a little over 1mm thick, let's opt for 2.2K. Capacitance? No such rule of thumb. I was going to put 2 adjustable 100pF capacitors in each leg, and twiddle until it would reliably start. Sadly, I could only find one 100pF capacitor... so that went in one leg, and a 22pF fixed capacitor on the other.

Result? Nothing. No amount of twiddling of the 100pF capacitor helped. OK, add another 22pF capacitor to the circuit to give 44pF ... and it's oscillating, but it frequency isn't stable. Remove the two 22pF capacitors and replace with 100pF ....


... and bingo! ... 28.061 Hz pops out of the Q14 pin. It's reliable and steady as a rock. It's a shade off frequency, but crystals do tend to drfit with age, and I doubt my load capacitance quite meets the original spec. It should be 28.076171875Hz.. (460000/16384)








It's all built "fugly" style on a bit of perf board.



























Here's the circuit

The pulses are fed into an Arduino Uno which allows a bit of development to go on, and a discovery is made (should have read the read.me!) .... you can't use an interrupt when using the TVout library, as the interrupts are being used to generate the timing need for video ... no problem, we'll use another arduino to generate the video and pass the data to it over a serial interface..

Now Arduino no1 is just doing the final bit of dividing down and outputting seconds over it's hardware serial interface, it's given it's own 16MHz crystal and mounted on the perfboard along with a 7805 to provide the 5V. The Arduino Uno is now used to develop the video software.







Before long, we have a rather nice clock display running.... (The photo doesn't do it justice, this is the tiny colour monitor I use in the workshop, and the photograph has artifacts..) . There's still plenty of memory left though ....








How about an animated "Pong" (Copyright Atari, the dawn of time) clock? Oh this is getting silly... Yeah, OK...

There appears a number of Arduino "pong" clones on the web, that will (allegedly) sit happily with TVout. After trying, I can't find one that works properly. Shoddy ball/bat collision (it's at best hit and miss 🤣), the ball flying right through the bat half the time.. we don't want that, so I re-engineered the code to suit.

Next thing... We don't want the monitor on all the time, consuming power and wearing the CRT cathode out. So some means of switching on the monitor when someone maybe looking at it. I did contemplate using a passive infrared sensor (PIR), but that's overkill. A simple sound detector will do. We can implement this on Arduino 1. 

A small electret microphone is amplified using an op-amp, and the output fed to Arduino 1's A0 pin.
This input is measured a few times, averaged, and compared to a value. There's an interesting bit of code here, whereby the audio is "rectified" in software, this means any negative audio is "flipped" over,  and a negative peak has the same value as a positive one (I've got a cunning plan for this bit of code, watch this space...). Anyway if this level increases above a certain point, there's some noise about, and we can switch our monitor on.

The output pin feeds a BC547 transistor, which in turn switches on a P-channel FET as a high-side driver, which supplies 12V to the monitor. A minimum on time is specified, as the monitor takes 8 seconds for the CRT cathode to warm. Even with the monitor off, Arduino 2 is still doing it's thing and creating the video waveforms and dealing with time.  

A temperature sensor is added to display the temperature as well... 

This is connected to A0 of Arduino 2. It's just a cheap 10K NTC thermistor. 

Meanwhile, we need a method for setting the clock. 3 push buttons are added, one for hours, one for mins and one for set, and connected to Arduino 2. When Arduino 2 starts up, it automatically enters this screen, and the hours and minutes can be set with the buttons. Once the set button is pressed, a reset pulse is sent to Arduino 1, and it's restarted, setting the seconds back to zero.

I also added a bit of code to enable easy calibration of the clock. Grounding pin D6 (pin 12 on the actual microcontroller) puts Arduino 1 into a calibration mode, whereby it outputs minutes elapsed since start-up and seconds over the serial interface. Disconnect Arduino 2, and connect Arduino 1 to an FTDI converter and monitor the serial output. Open the arduino monitor , and enable time stamp. Write down the time the sketch started, and the difference between that and the current elapsed time. Leave it for hours, then calculate how far it's drifted and use this value to alter the calibration value in Arduino 1 (old cal factor * ( number of seconds expected / number of seconds counted) . Excellent, it now keeps good time.

Excellent ... what about a chime?

I don't want just a beep or something from the micro... I'd like a proper chime.

(It's at this point, it begins to dawn on me I may have taken leave of my senses...)

A small solenoid is purchased from eBay... 

...and a bicycle bell...

.. all conjured up into a chiming assembly! 

... and a driver circuit created to be driven from Arduino 2.

So the final thing looks like this ...

The vintage crystal is top left, with the 14 bit ripple counter to it's right, and the 7805 voltage regaultor to the right of that.  The horizontal ATMEGA328P is our seconds generator, and controls the power to the monitor, the microphone and amplifier can be seen just below it. The high side FET is there too, and runs very cool, as the monitor draws about ~520mA when running. The vertical ATMEGA is creating the video, and also drives the chime, via the low side FET under the stripey ribbon cable, which leads off to the three push buttons, hours, minutes and set. Video is connected to the monitor via the small coax just above the ATMEGA.

The final schematic.

The clock display itself has three modes.


Clock ...












... pong ...












 ... and 3D cube ...

The software can be found, as usual, on my github page https://github.com/andydoswell/stupid-video-clock , and I've included the TV out library as well, as I've modified it so it compiles without issue, has a "degrees" symbol and tweaked the timing a shade. 

Now, I'd better start thinking about a case for it... some sort of perspex thing?

My colleague , Alan, has given it a good name... "Well, it goes "ping", and plays "pong" ... it's the ping-pong clock" 😁