New clock project with R|Z568M nixies

[newxito]

I have started today with the design of a new clock with 6 x R|Z568M. Because the controller board with the ESP32 and the HV power supply will be 1 meter away from the nixies, I tried to design an "intelligent" socket with the same size as the tube and an I2C interface. Because the socket uses the recomended anode and cathode resistors, the board has a lot of componets on one side. With these sockets I will only need a bus with 5 cables to control all the tubes. Looking forward to get and test the boards...

[martin martin]

That's a fine looking PCB!

I used these made from a fine member of our group-  I used 3 of them for a stacked design

~

mcve...@gmail.com

On Sun, Oct 4, 2020 at 3:10 PM newxito <axt...@gmail.com> wrote:

I have started today with the design of a new clock with 6 x R|Z568M. Because the controller board with the ESP32 and the HV power supply will be 1 meter away from the nixies, I tried to design an "intelligent" socket with the same size as the tube and an I2C interface. Because the socket uses the recomended anode and cathode resistors, the board has a lot of componets on one side. With these sockets I will only need a bus with 5 cables to control all the tubes. Looking forward to get and test the boards...

--
You received this message because you are subscribed to the Google Groups "neonixie-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neonixie-l+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/c410c772-c512-4a53-8883-a63b3cf0c9c4n%40googlegroups.com.

[Kevin A.]

Nice work! I had a very similar project in mind for Dalibor's tubes. Picked up a set of 6 plus 2 of his separator tubes as used in the zen and puri clocks. 

I wanted to try some funky things in addition to the modular design, like having each tube smart socket on slip rings and driven by stepper motors, so they can spin and add some flair to the overall design. Haven't gotten too far past modeling the tubes themselves and doing some mechanical outlines of possible board shapes.

Like you pointed out, minimizing the # of circuits on the control bus is ideal. I wanted to see if it could be done with 2 circuits only. They would carry low voltage DC and data would be transmitted using a 2 wire serial scheme at a slightly higher voltage. 

The socket would then have input filtering to separate DC for the on board high voltage supply and local regulation, and the data would be filtered out and level shifted as a serial input into the local smart socket micro, which would then house all of the display routines for the tube. 

On Sun, Oct 4, 2020 at 6:10 PM newxito <axt...@gmail.com> wrote:

I have started today with the design of a new clock with 6 x R|Z568M. Because the controller board with the ESP32 and the HV power supply will be 1 meter away from the nixies, I tried to design an "intelligent" socket with the same size as the tube and an I2C interface. Because the socket uses the recomended anode and cathode resistors, the board has a lot of componets on one side. With these sockets I will only need a bus with 5 cables to control all the tubes. Looking forward to get and test the boards...

--

[newxito]

Spinning nixies? That's cool, I hope that one day you will build this project!

[newxito]

I’m a little bit concerned about generating a lot of EMI in this project. I’ve added an image, so you can see what I’m trying to do. The controller will be hidden in the base of this old lamp. I will need 5 wires going all the way up into a horizontal tube. The nixies will be mounted with the sockets on top of the horizontal tube. 

The 5 wires are +170V, +5V, ground, SDA and SCL.

 I’m concerned about the long 170V wire. Will this act as an antenna? 

The wires will run inside metal tubes. Will this attenuate the EMI?   

Should I twist all the wires together? Only SDA and SCL? 

I was considering using a Cat6 cable and use one of the 4 twisted pairs for the 170V, one for 5V, one for ground and one for SDA and SCL but I’m not sure if that is a good idea.

Probably I should buy some kind of EMI meter and stop asking questions :-)

[Nick Sargeant]

You seem to have already taken care of the normal recommendations. So, first make sure the metal tube is grounded at the base. Second, make sure that it doesn't connect to the ground at the far end of the tube near the nixies - that way, you will avoid signal currents using the tube itself as a return path. Third, switching of the HV for the tubes is at confined at the top, so no emissions issues there. There is a slight risk that the switching to create the 170v will leak on to the supply rail, but given you are carrying that inside the tube, really you've done as much as you need to already. I remember when TV sets had tubes, huge voltages and currents switching, with little more than a hardboard cover. 

[gregebert]

CAT6 is a good idea; use pairs for 

• +170VDC/GND
• VCC/GND
• SCL/VCC
• SDA/VCC

If you recall, I2C uses pullup resistors, which is why you use VCC as the other signal in the pair. For SCL, put the termination resistor at the receiving end. Even though the termination resistance (typically 1K)  does not match the impedance of the twisted-pair (100 ohms), the reflections shouldn't be severe as long as the driving device doesn't have fast rise & fall times. I2C is rather slow (400kbit/sec), so rise/fall times will be much longer than the delay along your cable. If you do see reflections on a scope, you can try lowering the termination resistance, but that will put more load on the driving device. SDA is a bit trickier because it's bidirectional; I'm tempted to say start-out with 2K at each end and see what happens.

Next, be careful about ground loops. Dont connect the metal enclosure directly to your circuit ground; use a 1Meg resistor. As much as possible, follow guidelines for PCB layout to minimize ground loops, and you should be OK when you connect them together thru the cable. 

As long as you filter-out the high-frequency noise on the +170V, there wont be any radiated noise. And if you use a twisted pair, it's even lower. I would put 0.01uF caps at each end of the cable. Same for the lower voltage supply.

I have a multi-board RZ568m clock I'm building, and it's also using a serial protocol for displaying info on the tubes, reading pushbutton switches, controlling panel lamps, etc. We'll see how well it works out. Normally I dont use so many serial-connected devices but this saved me a lot of pins on the FPGA so I decided to give it a try.

[newxito]

That’s a lot of useful information, thanks Nick, thanks Greg. 

I will follow your advices in order to avoid problems with the ground. I will give CAT6 a try but first I will make some tests and measure the voltage drop, the wires are very thin and I expect a total cable length of 2m.  

On the 170V pair I expect 6W and about 1.5W on the 5V pair.

[newxito]

Forgot to mention that I will use the pairs as sugggested by Greg

[gregebert]

2 meters of cat6 cable (AGW 23) is about 0.16 ohms, plus any connectors. Since you are drawing 300mA or less, the voltage drop will be  50mV or less.

[H. Carl Ott]

Cat6 is also available shielded. Just don't pull any current through the shield. 

[newxito]

I wanted to test if the wiring through the tubes is feasible. I had to use some lube but it worked.

[gregebert]

I hope you dont have any young kids running around your house......one wrong move and it's going to be an expensive "splat".

Cant wait to see a video of the finished clock.

[newxito]

No young kids but young cats...

[newxito]

I couldn't find a cat6 cable that was thin and flexible enough for this project. So I used a 5-wire normal cable and tested it before with a 2.5m connection. The I2C bus worked perfectly so I did all the wiring for the clock. The wiring was hell because all cables are inside tubes. In the end, all 8 sockets were ready to connect. Unfortunately, I did the bus test without high voltage. As soon as I have 170V on the 5th wire, there are many transmission errors on the I2C bus. Now I must start all over again ...

[David Pye]

Could you insert two-core screened cable for the HV and HV gnd in a daisy chain setup, and reduce the amount of noise finding its' way into the I2C bus?

David

On Fri, 4 Dec 2020 at 23:36, newxito <axt...@gmail.com> wrote:

I couldn't find a cat6 cable that was thin and flexible enough for this project. So I used a 5-wire normal cable and tested it before with a 2.5m connection. The I2C bus worked perfectly so I did all the wiring for the clock. The wiring was hell because all cables are inside tubes. In the end, all 8 sockets were ready to connect. Unfortunately, I did the bus test without high voltage. As soon as I have 170V on the 5th wire, there are many transmission errors on the I2C bus. Now I must start all over again ...

--

You received this message because you are subscribed to the Google Groups "neonixie-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neonixie-l+...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/7801af98-abd6-4138-bea4-cb32aeafb153n%40googlegroups.com.

[Josh Girgis]

New to the group so I'll try my best to help.

You could consider twisted wire pairs for both i2C wires, and for the HV and HV ground. The rest of them Signal and ground, twisted wire pairs. That could help reduce the EMI.

Something to also try is the shielded cable David mentioned, just for the HV. Maybe a Triac cable?

Or maybe you could look into an I2C optical repeater if your wires are too long. That could help

https://www.digikey.com/en/products/detail/ixys-integrated-circuits-division/CPC5903GS/3182393

[Christian Riise Wagner]

What value resistors are you using for the pull-ups on the I2C bus? Also, what voltage are the microcontrollers in the smart sockets run at? If they're 5V, you'll have more EMI susceptibility on the bus, as the ESP32 runs the bus at 3.3V. A level shifter close to the ESP32 should provide a benefit here, as the HIGH level output then won't be as close to the LOW input threshold of the smart socket's microcontroller.

Also, it may be that your HV supply is overly noisy. Have you put a scope on the output?

[newxito]

Thank you all!

I think, daisy chaining is almost impossible because there is no real access to the horizontal tube. Maybe it works using a slim flexible cable and some steel U-form wires and then trying to pull out the cable trough the short vertical tubes. But if I use a separate cable for the HV, I will end up with 4 cables in each short vertical tube, I think that’s not possible.

What I did is kind of a star wiring. First, I pushed a cable all the way up inside the long vertical tube and it came out on the left side of the horizontal tube. Then I introduced a new cable into the horizontal tube from the left to the right. After that, I was still able to push 8 cables from the top of the short vertical tubes to the sides of the horizontal tube, 4 to the left, 4 to the right. Finally, I had to join all the cables on both sides. I tested the connections, they are all ok but it does not work as expected.

For the socket I’m using an atmel 328p at 5V and on the controller a ESP32 at 3.3V. The level shifter sits on the cotroller board under the base of the "lamp". I use 2 small MOSFETs. The pull-up resistor values are 3.9k on the 3.3V side and 4.7k on the 5V side. The HV supply is a flyback design with a LT3757, it works great but I don’t know how noisy this thing is. I will try using another HV source.

[Christian Riise Wagner]

A quick test you can do is lowering the values of your pull-ups. You can easily use half your current values, assuming you've only placed a single one on each of the lines. Should even be able to go lower than that. A 1k resistor will result in 5mA at the 5V side, which should be manageable by most devices. I'd probably start with 2.2k though and see if that gets you anywhere. If it doesn't work with 1k or a new HV supply, you'll probably have to take on the daunting task of rewiring the whole thing with twisted and/or shielded cables as suggested by others.

[newxito]

I found some 2.7k and 1.2k resistors. I tried with 2.7k, still got some erros but with 1.2k on the 5V side it works like a charm. I'm using the 2.5m cable, 2 sockets attached, HV enabled and sending 50 commands per second in fast mode. I'm sure that's good enough. All the heavy tasks like nixie dimming, slot machine effect and led color cycling will be done by the socket, so I don't expect performance problems on the I2C bus.

I found a TI note, they recommend 1.5k for 5V, so that's pretty close.  

Thanks again!

[newxito]

Finally, I got the time to work on this project. This is the finished clock.

Since I have neither the skills nor the tools to work with metal, I used curtain parts and an old lamp. Please notice the wi-fi antenna on the base, the controller is buried under 4mm steel :-)

[gregebert]

I hope you attach a small anchor to the wall near the top of the clock. It would be a horrible tragedy if this got tipped-over.

In our house, it would be a matter of minutes before the dogs knocked it over.

[Richard Scales]

Great work indeed - can you tell me - for the colons - are those from Dalibor? I am looking for matching metal parts to make some of my own.

 - Richard

[newxito]

You're right Greg, I really should do this, even though the glass breakage insurance would probably pay for it.

[newxito]

The colons are not from Dalibor. The wiring through all of these tubes was a nightmare. I only had 5 wires available for the whole thing so even the colons had to be smart. The cylinders for the colons are also curtain parts. Amazingly, they fit quite well :-)

https://www.bauhaus.info/gardinenzubehoer/expo-ambiente-wandlager/p/25375525

[Richard Scales]

That's amazing, thank you for the link.

[newxito]

One thing I forgot to mention: in case someone wants to design his own R|Z568M smart socket, consider using 2 LEDs for the backlight. My design has only one LED, that’s a little bit dim for these big tubes.