Is there any glaring mistake in this Neon Bargraph Driver PCB?

[zlaurin]

Hello,

I am a hobbyist this is my second time designing a PCB and my first time designing a high voltage PCB (It has both 5V and 150V DC) and I was wondering if anyone could take a quick look at what I've got and tell me if I made any mistake? It's a driver circuit for the IN-13 Neon Bargraph Nixie tube. It requires 150V and lights up proportionally to the current on it's indicating cathode pin in a range of 0 to ~4.5ma. I'm getting the 150V from a step up DC-DC converter (A Nixie tube power supply I got on eBay).


Here is the schematic. The right part is a very common circuit to drive it I found online. The resistors I'm using are 1206 resistors and the potentiometer is a ST-4. Both are rated for 200V and 0.25W. The left part is a DAC to generate an analog voltage (and it's I2C with 8 addresses so I can have 8 PCBs total to make a spectrum analyzer or something like that) and an op amp.





Here is the PCB. I'm not really sure what to do for voltage creepage / clearance. All driver components are on the bottom and all 5V components on the top. All 150V and 5V traces on the same layer are at least 1mm apart however some are superimposed on top of one another across layers. I tried putting as much of the 5V components and traces on top of the resistors but the SDA, SCL and transistor base traces overlap with the transistor emitter trace. There's also a 0R jumper resistor because I couldn't figure out how to route it without it. All traces are either 0.25mm or 0.4mm.






Here are some renders of it with the component models. Also I forgot to mention it but the ground plane is only in the corner of the board where the components with a pin connected to ground actually are and the clearance between the ground plane and the traces is 0.7mm.





So yeah my question is does anyone see any glaring issues with the schematic or PCB? What I'm most worried about is voltage clearance / creepage. Is it acceptable in this case to have HV traces on one side and LV traces on the other side overlapping? That's a 1.6mm spacing no? I'm also unsure about the op amp I picked. I chose that one (LM321) because it's the single version of the LM324, which in turns should be equivalent to the LM358, which I've seen used in other drivers for the same tube online here: https://www.tubeclockdb.com/media/kunena/attachments/589/in13_driver.pdf

Thanks in advance for the help,
Zachary

[jwet]

Overall, the board looks really nice.  I like the keepouts and conservative rules.  I'm also amazed at good Kicad has gotten!  All very good.

Some constructive criticism:

I don't see a connection between the ground terminal of the input terminals and the other grounds- is there a plane or other not shown?

The 150v trace is fine and running this on FR-4 with opposite signals is ok.  This is ok to about 500v depending.

The pot doesn't make much sense.  You've got an 8 bit dac, you don't need a pot do you. (see below)

The REF pin on the DAC is a no connect as I assume you're using Vcc internally.  This would be a better place to have an adjustment if you need one- a pot, etc. but I suspect you can calibrate in software only.

The Op-Amp isn't a great pick- its not rail to rail in or out and there are lots of options that would fit better- maybe a MCP6001 since you've already got a Microchip supplier.  It might work but isn't optimal.

[T3sl4co1l]

Heck, I'd be fine without the op-amp. How precise do you really need that current?  It's a neon tube, it won't mind 10% here or there, and you probably need to calibrate to its scale or intensity anyway.  A 0-5V DAC range, offset by ~0.7V, will get you plenty of linearity for the remainder of that range.  A digital calibration would be pretty straightforward to eke out just a little more precision (down in the exponential range 0-0.8V say, bonus points if a temperature sensor is equipped, to cancel out Vbe tempco -- at that point, the op-amp is preferable though).

An R+C snubber on the open collector might be desirable in case of long wires potentially causing oscillation, both in terms of the neon lamp (a negative resistance device) and the transistor (making some manner of oscillator with its capacitances -- unlikely for merely a MJD340, but easily addressed regardless).  Ballpark values might be, oh, 100pF + 1k?

Voltage and temperature limits might be desirable, but can also be solved externally with suitable caution (e.g. series limiting resistor plus tube; fault monitoring, maybe check that Vout is in nominal range?).

Tim

[schmitt trigger]

Many years ago I built a spectrum analyzer based on IN-13 tubes. I can tell you that every one who sees it in operation is awed at the beauty of these devices.

Now my comments. Back then, before the actual RU-UA conflict, there was plenty of information on the web.  I was able to procure original Soviet data sheets. On one of them, they emphasized that the pilot cathode MUST BE lit up prior to current being applied to the main cathode. As such, there is an additional low voltage transistor and a resistor VT2 and R2 to ensure this happens.
I enclose image of schematic.

Other than that, I also suggest that you employ a R-R opamp, as you are only using 5v. I used 12v, and thus was able to employ a LM324.

[zlaurin]

Hi, thanks a lot for your answer

I don't see a connection between the ground terminal of the input terminals and the other grounds- is there a plane or other not shown?

Yeah there's a ground plane in the corner of the board where the components with grounds are. Forgot to unhide it but it's visible on the 3D render.

The pot doesn't make much sense.  You've got an 8 bit dac, you don't need a pot do you. (see below)

The REF pin on the DAC is a no connect as I assume you're using Vcc internally.  This would be a better place to have an adjustment if you need one- a pot, etc. but I suspect you can calibrate in software only.

The potentiometer is there to set the max current on the indicating cathode pin. All IN-13s are a bit different and it can be useful to set the top of the light bar, as well as increase the max current it as its lifetime decreases. Here it is simulated: https://tinyurl.com/27zqtrhs

The Op-Amp isn't a great pick- its not rail to rail in or out and there are lots of options that would fit better- maybe a MCP6001 since you've already got a Microchip supplier.  It might work but isn't optimal.

Thank you for the suggestion. I'm not 100% sure how op amps work, I just know I need one but not what specs it needs to be. I'll switch to the MCP6001.

[zlaurin]

An R+C snubber on the open collector might be desirable in case of long wires potentially causing oscillation, both in terms of the neon lamp (a negative resistance device) and the transistor (making some manner of oscillator with its capacitances -- unlikely for merely a MJD340, but easily addressed regardless).  Ballpark values might be, oh, 100pF + 1k?

Voltage and temperature limits might be desirable, but can also be solved externally with suitable caution (e.g. series limiting resistor plus tube; fault monitoring, maybe check that Vout is in nominal range?).

Thank you for the suggestions I will look into that.

[zlaurin]

Many years ago I built a spectrum analyzer based on IN-13 tubes. I can tell you that every one who sees it in operation is awed at the beauty of these devices.

Yeah they're absolutely beautiful. The DAC I'm using has 8 possible I2C addresses and so I intend to make a spectrum analyzer with 8 sockets controlled by an Arduino.

I was able to procure original Soviet data sheets. On one of them, they emphasized that the pilot cathode MUST BE lit up prior to current being applied to the main cathode. As such, there is an additional low voltage transistor and a resistor VT2 and R2 to ensure this happens.

Yeah I've seen this datasheet online and I initially intended on having another transistor for that but the DC-DC boost converter I'm going to use, the NCH6300HV (https://omnixie.com/products/nch6300hv-nixie-hv-power-module), has an active high enable pin so in the software I will make sure that the DAC is outputting 0V on startup before powering the tube with 150V.

[schmitt trigger]

Apparently you are mis-understanding the information I gave you.

You don’t delay the +150v. You immediately apply +150v to the anode, and then the pilot cathode which is always grounded thru a resistor, should strike and light up on its own. Only after the pilot cathode has become active, you should apply current to the indicating cathode, which is driven by the transistor.

This is necessary because some times the pilot cathode takes a while to actually strike. I have personally seen it happen.
TR2 actually senses that the pilot cathode is conducting, before allowing a current path to ground for the indicating cathode.

[TimFox]

I don’t know the reason, but when I tried an IN9 on the bench, it worked fine from a series connection of voltage source, ammeter, and resistor, but gave weird results (broken line) when driven from a lab constant current source (with high compliance voltage), so I set it aside.  Have you seen anything similar?

[T3sl4co1l]

I don’t know the reason, but when I tried an IN9 on the bench, it worked fine from a series connection of voltage source, ammeter, and resistor, but gave weird results (broken line) when driven from a lab constant current source (with high compliance voltage), so I set it aside.  Have you seen anything similar?

CCS with output capacitors?  Put another way: CCS only at DC, or up to some frequency as well?

Tim

[TimFox]

I don’t know the reason, but when I tried an IN9 on the bench, it worked fine from a series connection of voltage source, ammeter, and resistor, but gave weird results (broken line) when driven from a lab constant current source (with high compliance voltage), so I set it aside.  Have you seen anything similar?

CCS with output capacitors?  Put another way: CCS only at DC, or up to some frequency as well?

Tim

It was a Fluke DC calibrator in CC mode, so there was an output capacitor, which I presume was before the current sense resistor.  It was a while ago, and I don’t remember looking for relaxation oscillations.