Nixie watch as a lockdown project - help with voltage clamping

[Max DN]

Hi all,

I'm been following this group for a while and finally decided to go back to a hoppy that I had 15-20yrs ago or so, to brush up my old knowledge and to build my personal nixie watch starting from what David at Cathode Corner kindly shared, but looking to improve its efficiency with new chips and to add functionalities similarly to Kopriso watch or else. I like the capacitive buttons etc.

This is all very steep for me but getting there. I have managed to learn to use a CAD software to design the PCBs, I still haven't worked out how to properly solder an ADXL335 by hand (it's an LFCSP_LQ) while I can solder a QFN40 with no issues, I find the LFCSP challenging even with a microscope but what I have done so far seems fine for a prototype project.

I have put it all together and it works! Almost: only issue I have is the voltage clamping, which I don't understand well. The ADXL335 seems unreliable, probably due to poor soldering but I'll work on that later.

I couldn't find any B5870 Nixies, so for now while testing I'm using some cheap Q27 Nixies.

I do get a stable 158V from the power supply. I believe I need 180V as per descriptions on Cathode Corner's webpage (the Q27 Nixie should take 120-170V or higher). Looking at the resistor divider, I should be getting 163V, so the HV circuit seems to work, I think. Not sure why David says it should give 180V. So I may be doing something wrong or the values of the resistor divider on the schematics are incorrect (I very much doubt so!). I calculated 1.22V*[1+(510k+510k)/(7.5k+150)]=163.8V as per datasheet of the LT 1308B.

Thing is, the whole thing works, keeps the time etc but if I connect the VCLAMP (to clamp the 083A darlingtons to 51V - see link to schematic below), then nothing seems to happen, only one of the Nixies goes on, but it's very fainted (and it keeps the time).

What am I missing?

This is what I am looking at:

http://www.cathodecorner.com/nixiewatch/theory/theory.html

http://www.cathodecorner.com/nwl/NWLD5schem.pdf

Obviously I could experiment with different values of the resistor divider but I don't understand why I should do that as I'm sure Dave's design is proven by many years and many watches he sold. So, I'm guessing I'm doing something wrong here. Or it's as simple as the Q27 tube needs a higher maintaining voltage than the B570? I cannot find a datasheet for the Q27. But then again, why Dave talks of 180V out of his HVPS? I'm going in a loop...

Any help would be much appreciated.

Thanks,

Max

[Max DN]

Just a quick update. I have put a resistor of 2.2k between R1 / D5 cathode and pin10 of COM of the Nixie driver and it seems to clamp the voltage to a stable 25-40V (average, just measured with a simple multimeter). I chose 2.2k just because I had it on the table, nothing scientific but I'm thinking that the 20-30cm cable of the prototype maybe providing some resistance which this is somehow offsetting. Not sure if any of this makes sense?

[Kevin A.]

In order to assess what is happening with your circuit, we need more information. Schematic, bill of materials, and pictures of your pcb artwork (showing copper and component placement) and the actual test prototype. All of that information is critical if you want a serious response about potential problems and how to move forward. 

On Mon, Aug 31, 2020, 6:55 AM Massimo Di Noi <flata...@gmail.com> wrote:

Hi all,

I'm been following this group for a while and finally decided to go back to a hoppy that I had 15-20yrs ago or so, to brush up my old knowledge and to build my personal nixie watch starting from what David at Cathode Corner kindly shared, but looking to improve its efficiency with new chips and to add functionalities similarly to Kopriso watch or else. I like the capacitive buttons etc.

This is all very steep for me. I have however managed to learn to use a CAD software to design the PCBs (still haven't worked out to properly solder an ADXL335 by hand, I find that challenging even with a microscope but what I have done so far seems fine for a prototype project.

I have put it all together and it works! Almost: only issue I have is the voltage clamping, which I don't understand well.

I couldn't find B5870 Nixies, so for now while testing I'm using chip Q27 Nixies.

I do get a stable 158V from the power supply. I believe I need 180V as per descriptions on Cathode Corner's webpage. Looking at the resistor divider, I should be getting 163V, so the HV circuit seems to work, I think. Not sure why David says it should give 180V. So I may be doing something wrong or the values of the resistor divider on the schematics are incorrect (I very much doubt so!). I calculated 1.22V*[1+(510k+510k)/(7.5k+150)]=163.8V as per datasheet of the LT 1308B.

Thing is, the whole things works, keeps the time etc but if I connect the VCLAMP (to clamp the 083A darlingtons to 51V - see link to schematic below), then nothing happens, only one of the Nixies goes on, but it's very fainted.

What am I missing?

This is what I am looking at:

http://www.cathodecorner.com/nixiewatch/theory/theory.html

http://www.cathodecorner.com/nwl/NWLD5schem.pdf

Obviously I could experiment with different values of the resistor divider but I don't understand why I should do that as I'm sure Dave's design is proven by many years and many watches he sold. So, I'm guessing I'm doing something wrong here. Or it's as simple as the Q27 tube needs a higher maintaining voltage than the B570? I cannot find a datasheet for the Q27. But then again, why Dave talks of 180V out of his HVPS? I'm going in a loop...

Any help would be much appreciated.

Thanks,

Max

--
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/d0451e2f-a7ec-470a-a268-eecedd2ffc3en%40googlegroups.com.

[Max DN]

Thanks Kevin. You are absolutely right. I'm actually thinking that you guys are so far ahead with these projects that it's not a good use of your time. Anyway, since you are familiar with this type of work, hopefully it's obvious what I am doing wrong if anyone is willing to help without spending too much time on it. Or I'm thinking it's simply that the cables for the prototype are just too long at 20-30 mins each or the QS27 Nixie I'm using to test.

Anyway, I've attached a picture to show the cabling, doesn't add much to what you may need. But other then that the actual prototype, everything is 100% as per this schematic in the link below. Checked probably over 10 times before ordering the PCBs and checked 3 times over after assembling the components. http://www.cathodecorner.com/nwl/NWLD5schem.pdf

I even managed to source the flyback transformer 31105, which seemed difficult to find these days. So, everything is 100% as per that schematic, list of components etc.

[Kevin A.]

Sure thing. Assuming that your circuit is 1:1 schematically and component wise, the biggest remaining variable would be your physical layout. 

Switched mode power supply performance is very layout critical, as you often have both high current and high frequency signals on power circuits, in addition to sensitive feedback circuits which supply the switching controller's logic. The layout can very well make or break the circuit, so it is important to see both where your components are placed, as well as the copper that connects them together and to the rest of your circuit. 

I went though all of this personally when I first began designing nixie power supplies. You can see some of the thought process put into words on my first power supply related posts here: https://neonkev.com/2019/08/18/project-20-watt-high-voltage-boost-converter/

Some good reading about switched mode power supply layout considerations and best practices: https://techweb.rohm.com/knowledge/dcdc/dcdc_pwm/dcdc_pwm03/2734

Let's start with seeing the layout of your design and the physical test setup. A picture is worth a thousand words. 

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/1474844f-bc92-41e6-80ce-f242a09179f6n%40googlegroups.com.

[Max DN]

Indeed I came across your work yesterday. It's really well written and clear. Thanks for sharing the other link too, very useful. I have already started to redesign the PCB as it may be that.

I have attached a picture of my pcb for the power supply (pardon my soldering skills, first time ever I solder 0201 components since I used to etch my pcbs over 15 yrs ago for through hole components only!). I've also added the copper mask etc.

I have indeed realised that I made at least a couple of mistakes with the layout: 1/ the resistive divider (red circle) is too far from the LT1308B (red square) as the diodes (green square); 2/ I mistakenly chose 0201 resistors which are 50V each rates (it seems to work but not ideal) and are a pain to hand solder... Oops.

With all of that, I'm surprised that it works at all. 

On pin 2 of the LT1308B (feedback pin) I get 1.17V vs a reference voltage of 1.22, so something is not right, but that means that it should put out an even higher voltage if I'm not mistaken (I'm getting 158V, Dave/Cathod Corner says it should be 180V). I may need to redesign the layout of the power supply, but I'd like to understand why the voltage clamping works better (not perfect) if I add a (randomly chosen) 2.2k resistor in series along VCLAMP as per schematic shared before). It has to do with either the feedback voltage or the cheap QS27 Nixie I'm using or too many cables being used. I should add that I do get a purple dot in one of the tubes, so I suppose clamping is not perfect...

I'll read agin what you sent and I'll keep on thinking. No point reprinting a new layout if I don't understand well what's happening first.

[David Forbes]

Massimo,

Hi. I see that you are trying to make the Cathode Corner Nixie watch
circuit work, using your own circuit board design. it's not an easy
circuit to build, because the design was optimized for size and
efficiency, given the parts I had available.

The power supply is designed to make 180V with no load. It is designed
to make about 10mA of current with one Nixie tube lit, where the Nixie
tube regulates the voltage to about 150V. There is a third voltage
level that is selected by the TiltP signal, which reduces the HV to
about 90V by changing the feedback divider current, through that 33K
resistor. This is done to provide blanking for the tubes while the
power supply is active, which makes the multiplexing cleaner.

You can see that all of these parts interact, and the feedback
resistor values were tuned after building a few watches, to get the
display to look the best. I wouldn't have made it this way if I was
only building one watch!

The clamp voltage is not critical, since the clamp Zener diode is only
conducting when high voltage is on yet no tube cathode is enabled.

The circuit board layout is critical. I do not think that it will work
properly if you have any length of cable on the Nixie tubes. This is
because the capacitance of that cable, combined with the high votlage
switching of mutliplexing, will generate noise and affect the timing
of the cathode turn-on and turn-off.

Here's a picture of the PC board design for the round watch power
supply. You can see that I made all the high current 3V power and
ground traces very short and wide, to prevent problems with the
switching currents radiating into the feedback circuit. In fact, my
first prototype PC board had the feedback line running across the
power supply, and it failed, so I had to add a wire to move that
sensitive line away from the noisy switching circuit.

I hope that this description helps you to understand where to focus
your attentions.

[Max DN]

Now this is an honour! David Forbes himself! I'm a big fan.

David, thanks for taking the time to reply in detail. I have been studying your circuit for 2-3 months now (I know, for most people here it may take a day or so, but my knowledge of electronics is very basic however it's always been a passion of mine). My day job is in a totally different field (banking) and since Covid lockdown started in the UK I have dedicated some of my spare time to this project, just as a way to switch off, so to speak :-)

My wife insists I should just buy your watch but I like a good challenge (more so I'd like to try to make a HVPS at least as efficient as the Kopriso but that's some time away for my basic knowledge). I may still get your Cathode Corner watch as a courtesy for sharing your entire firmware and schematics. Beside, I like it as well and perhaps start from there and try to update the software to add functionalities (such as date etc). I'd also like to improve the efficiency of the high voltage circuit and have worked on 2-3 projects in parallel, but I'm leaving those in standby for now given my limited spare time.

I do think my issues are the amount of cables currently running to the Nixies as well as poor layout. And perhaps many more. In fact, I'm still surprised that it works at all, although not perfect (the accelerometer is not working at all or not working well, I'll look into that later, probably due to poor manual soldering skills there). 

I have noticed that if I even try to just measure the voltage on the feedback pin (I only get 1.17V or so) the brightness of the Nixies changes. 

The multiplexing doesn't work well with my current set up and you can clearly tell it's there. I have taken a video if you are curious at all. I'm loving this project and would like to keep the watch size of 38mmx38mmx16.5mm which I think is achievable.

Needless to say, David, your PC board design is compact and neat, my first mistake was to have the feedback line running across the entire board as well as the switching pin so far from the transformer and with its copper wire running under the 1308B! Oops.

Last, it's a puzzle for me that a (random) 2.2k resistor in series with the VCLAMP line seems to improve stability. And works better with 3.2k. It's probably affecting the resistive divider somehow.

I think you are spot on with all of your suggestions (of course). I'll focus on the feedback line for now and perhaps just start from scratch soon by redesigning the high voltage layout.

Thanks so much again.

Massimo

[Max DN]

David and all,

Just a quick update. I finally had some time to remove all cables and to solder everything directly, with the nixies being on sockets and I'm pleased to report that it all works great! 

Re: accelerometer ADXL335. I cannot get to solder that chip properly, quite tricky even under the microscope. I'll start again but I'm not too concerned about it, maybe I'll replace it with a chip in a different package, I find the LFCSP_LQ too tricky to hand solder even when following the recommended soldering profile.

Re: HV 1308B - Turns out that the PCB layout that I posted here - while not ideal - works ok (ok-ish?) or at least it's not a disaster.

I've added some tiny LEDs along the anode lines and they work great but they want a 1k resistor in series otherwise the tubes glow a bit blue, so I need to work around that.

Next I'll look to add a photodiode to control brightness through PWM I suppose, the intention being to save power more than to dim the tubes. But I have some PIC assembly to chew first for PWM and perhaps I'll add touch buttons too...

Nothing new for this group, but a whole new world for me. Feeling like I've been missing out!

Thanks again for the suggestions given,

Max

[gregebert]

Have you tried using solder paste and hot-air reflow ?

[Max DN]

Hello, indeed I have. Two chips gone (oops), I have 10, I'm sure I'll learn before I get to the last one :)

It amazed me that I managed to hand solder with hot air flow and soldering liquid paste a QFN40 first time with no issues, the LFCSP_LQ is obviously a different story. I tried to put less solder on the big center pad and all pins seem to be touching well, perhaps I burnt it but I was careful with the heating profile on the way up to the peak as well as down and how many seconds I spent on each phase.

 But hey! Who doesn't like a good challenge on it after a long day at work! I'll get there.

[Max DN]

gregebert, I finally managed to find some time to hand solder an LFCSP_LQ package. Achievement :)

What worked for me, it is easy and reliable: clean pins of chip with alcohol, apply no-clean flux, apply normal soldering wire on the tip of the iron (I think melting point 215-220 celcius), slide the tip of the iron on all pins and center pads, clean PCB, apply flux, apply soldering paste (eutectic point at 217 Celcius) on the PCB, align the pins of the chip to the footprint on the PCB, hot air (max 230C), being careful not to heat the chip too fast) and done. It takes just about 2 minutes to do it all.

Hot air reflow alone wouldn't work as it's not guaranteed that the soldering paste would stick to the tiny pins (no matter how clean) and there is close to no chance to be able to inspect the bonds properly.

I thought I'd share. 

Best,

Max

[Max DN]

Hi all, I hope you are well and healthy.

This little project is progressing really well, although not as quick as I'd like to have the time to... I got it all working and it's nice and smooth. However I do have a piece of code which I'm confused about and I was hoping that better trained eyes could help.

The code that I'm using (verbatim from here for now: http://www.cathodecorner.com/nwl/nwld1.asm.txt, that is helping me to learn PIC assembly so to write my own code later on) is doing analog conversion of outputs of the accelerometer ADXL335 (datasheet here https://www.analog.com/media/en/technical-documentation/data-sheets/ADXL335.pdf) to sense motion / tilt of the wrist. It works beautifully and it's very reliable, coded by DF of Cathode Corner on the X-Y axis (he's also done it on Z-X on another project, so Z should work well on this sensor). My issue is that I have moved the chip such that I need to use Z-X instead of X-Y as in the code. Now that should be an easy code edit... Problem is that I can see the correct mV changes on X-Y-Z on the chip and all pins are individually well connected to the PIC but somehow the PIC ADC isn't picking up Z, so I'm thinking I'm doing something wrong with the channel selection.

Schematic here, except pin 8 Z-ADXL which I have actually connected to pin 21 RA4 of the PIC): http://www.cathodecorner.com/nwl/NWLD5schem.pdf

X-ADXL connected to AN1 -> So here I should select ADCON0 CH0=1 CH2=0 

Y-ADXL connected to AN0 -> So here should be ADCON0 CH=0 CH2=0 (X and Y seem switched in the code but given some logic later on, it's fine, no issues so far).

Z-ADXL connected to AN4 -> So here I should select ADCON0 CH0=0 CH2=1 (but when I do that... it doesn't work, I get 0 voltage sensed, odd) 

Voltage monitor (resistive divider) connected to AN5 -> now this is odd as I would have expected to select ADCON CH0=1 and CH2=1 but it works only with ADCON0 CH0=0 and CH2=1 (so this is confusing to me! I don't understand why this works only on channel 4 when it's actually connected to channel 5 and portA is set as '00110011' (2 and 3 being outputs) with '00110011' (AN0, AN1, AN4, AN5 being analog inputs).

Am I missing the obvious elephant in the room? Relevant code below (full code in the link above). I'd appreciate if anyone has 5 mins to spare to look into this please.

Thank you,

Max

movlw   B'00110011'     ; PORTA bits 2,3 outputs [pin 19 RA2 and pin 20 RA3]

        movwf   TRISA

bsf     BSR,BSR1        ; point to high bank

        movlw   B'00110011'     ; AN0 of pin 17 connected to Y-ADXL, AN1 of pin18 connected to X-ADXL, AN4 of pin21 connected to Z-ADXL, AN5 are analog [I have connected to  pin 19 and pin 20 are digital outputs

        movwf   ANSELA

bsf     BSR,BSR0        ; point to high bank

        bsf     ADCON0,ADON     ; turn on the A/D converter module

        bcf     ADCON0,CHS0     ; select A/D channel 0 = X

        bcf     ADCON0,CHS1

bcf     ADCON0,CHS2

        nop                     ; wait one millisecond

        nop

        nop

        nop

        nop

        nop

        nop

        bsf     ADCON0,GO       ; Start first A/D conversion at 2 milliseconds

        nop                     ; Wait for X conversion to complete

        bsf     ADCON0,CHS0     ; select A/D channel 1 = Y

bcf     ADCON0,CHS1

bcf     ADCON0,CHS2

        movfw   ADRESH          ; get X tilt from A/D

; lo RAM

        bcf     BSR,BSR0        ; point to low bank

        movwf   XTilt           ; save X tilt

; hi RAM

        bsf     BSR,BSR0        ; point to high bank

        bsf     ADCON0,GO       ; Start second A/D conversion

        nop                     ; Wait for Y conversion to complete

bsf     ADCON0,CHS0

bsf     ADCON0,CHS1

        bcf     ADCON0,CHS2     ; select A/D channel 3 = Z

        movfw   ADRESH          ; get Y tilt from A/D

; lo RAM

        bcf     BSR,BSR0        ; point to low bank

        movwf   YTilt           ; save Y tilt

; hi RAM

        bsf     BSR,BSR0        ; point to high bank

bsf     ADCON0,ADON

; can remove the line above

        bsf     ADCON0,GO       ; Start third A/D conversion

        nop                     ; Wait for Z conversion to complete

nop

nop

nop

btfsc ADCON0, GO

GOTO $-1

bcf     ADCON0,CHS0

        bcf     ADCON0,CHS1

bsf     ADCON0,CHS2     ; select A/D channel 4 = Volts

        movfw   ADRESH          ; get Z tilt from A/D

nop

nop

nop

; lo RAM

        bcf     BSR,BSR0        ; point to low bank

        movwf   ZTilt           ; save Z tilt

; hi RAM

        bsf     BSR,BSR0        ; point to high bank

        bsf     ADCON0,GO       ; Start fourth A/D conversion

        nop                     ; Wait for V conversion to complete

        nop                     ; Wait for V conversion to complete

        movfw   ADRESH          ; get volts from A/D

        bcf     ADCON0,ADON     ; turn off the A/D converter module

[David Forbes]

Max,

Hi. I wrote all that code.

The NWR watch that I designed uses all the capabilities of the
PIC16LF722 chip. It was carefully engineered to do exactly what it
needs to do and nothing more. Every pin does useful work, some of them
doing two or three jobs at once. It has absolutely zero room for extra
features.

If you need to read all three axes of the accelerometer (I don't know
why you would), then you need to use a PIC chip that has more pins. I
used such a chip to make the NWL watch. It uses an additional analog
input to read the battery voltage.

I'd suggest starting with that deign if you want to add capabilities.

[Max DN]

Thanks David.

Indeed that is your code as I had mentioned, it's really helping me picking up pic assembly for this purpose. And the chip I'm working on is PIC16LF1519 as you use it in your NWLD5 (http://www.cathodecorner.com/nwl/NWLD5schem.pdf). 

What I had found confusing was that the battery voltage input is on AN5 and so I thought you'd select channel 5, instead it only works on channel 4. I can see now it's probably because there is no ADC on bit 4 of ANSELA. I had connected Z of the ADXL to AN4 (pin21) and I couldn't record an ADC conversion. I can see now why! Oops..

movlw   B'01110010'     ; internal RC oscillator, AN3 Vref

movwf   ADCON1

movlw   B'00111011'     ; AN0, AN1, AN3, AN4, AN5 are analog

movwf   ANSELA

You are right, you don't need X-Y-Z for the motion detection and it would use more power for no reason but I wanted to try and set the time using the accelerometer instead of the buttons. It may not be a good idea, it will still be a good learning exercise for me. 

Best regards,

Max

[Nick]

Elektor claim that they have a nixie watch project starting next issue...