EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: cprobertson1 on March 15, 2015, 07:48:30 pm
-
Yet another of those pesky nixie clocks.
Is mine different? Not really - I could hijack somebody else's post - but tbh then I can't show of sweet pics of my project...
"Sweet pics"? :palm: - okay I won't ever say that again...
This is the counterpart, the contra medium of my "Nixie Clock... Without the nixies!" (http://contra medium) - which is effectively the same project, but using DIY nixie tube replacements because it seemed like a fun idea to make something easy-to-use.
Well, to make thing short - I have 8x IN-14 nixie tubes to drive from an AVR ATMega168-20PU (which is totally overkill, but it's what I have around and I can't be bothered picking up something else - it also means I have all the IO I could ever wish for)
My plan is to drive the nixies directly, using off-the-shelf parts: I want everything to be easily replaceable and readily available.
I'm thinking:
uC > 8x 16-Bit SIPO Shift Registers[1] > Optocoupler[2] > Transistor[3] > NIXIE
[1]: each consisting of 2 serial 8-bit shift registers - all driven from the same clock/reset lines - but each group with it's own data line (10 I/O pins in total)
[2]: Which type of coupler do I want here? Will an optotransistor allow me to avoid having to use a high voltage transistor to drive the nixie cathode?
[3]: BJT or MOSFET with a high voltage rating - all tied to Vcc/GND - acting as simple switches to power the cathode.
NB - a separate Optocoupler > Transistor will be used to open the Nixie's anode while we're shifting data into the registers (i.e between digit-changes) (8 I/O pins in total)
This brings us to 18 I/O Pins in use, which leaves plenty for our I2C bus (for the RTC, GPS clock and whatever else I decide to add to it in the future) (ATMega168 has 23 I/O - including the reset pin, so 22 I/O just to be safe)
-
You can get Russian made 74141 chips which are binary to decimal decoders with built-in high voltage drivers. Then you can drive them with SIPO shift registers with output registers so that the outputs don't see the shifting data - it just gets broadside loaded after the shift in has finished. Also, no real need for an opto-coupler.
-
Oops - sorry! Was editing my original rambling post when you replied!
I'm trying to use off-the-shelf parts if I can help it =/ Was trying to shy away from the vintage chips (which is a pity, because it'd make it so much easier!)
Is there a modern equivalent to the 74141 chips or have they all gone low-voltage nowadays :P?
--EDIT--
Optocouplers aren't particularly cheap when you need a lot of them - even when you get arrayed versions xD Don't suppose driving High Voltage Transistor (BJT/MOSFET) directly via a series diode is an option?
-
Once you get up above about 70V there aren't many integrated solutions still commonly available. I'd look at doing something like this x the number of electrodes you need to control with discretes in SMD.
o-----------------o-----o-----o
+HT | |
.-. |
| | |
| | |
'-' |
| |/ Q2 HV NPN
o---|
| |>
| | To Tube
| o-----o
| |
| V D1
| -
| |
Logic o-----'
Signal |
___ |/ Q1 HV NPN
o--|___|---o----|
| |>
| |
.-. |
| | |
| | |
'-' |
0V | |
o----------o------o-----------o
(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de (http://www.tech-chat.de))
Although you could use a MOSFET for Q1 and omit the base resistors, if something goes wrong and Q1 fails, a bipolar is far less likely to kill your drive logic. If you are paranoid, add clamping at the logic input!
The pullup resistor should be capable of handling the dissipation with Q1 on all the time. It should also be sized to provide enough base current to Q2 without excessive drop assuming worst case Hfe and max load current. The transistors and diode should have a voltage rating at least 50% greater than +HT
It shouldn't be too tedious to assemble if you get a PCB + stencil made and have reflow capability.
-
Take a look at this range of chips http://www.microchip.com/wwwproducts/Devices.aspx?product=HV6810 (http://www.microchip.com/wwwproducts/Devices.aspx?product=HV6810)
These will do your shifting, latching and high voltage driving all in one go. It doesn't look like a they do a very high voltage, but they should work due to the volts drop across the nixie.
-
I used an HV5122 32 channel open drain driver in my Nixie clock, worked fine.
http://ww1.microchip.com/downloads/en/DeviceDoc/HV5122%20B072213.pdf (http://ww1.microchip.com/downloads/en/DeviceDoc/HV5122%20B072213.pdf)
32 channels will make a 4 digit clock, with the caveat you only get numbers 1 and 2 on the left most digit. You can easily cascade a similar chip on for another N channels though, or add another discrete driver channel to control the missing leading zero. Two '5122s cascaded will drive a 6 digit clock with a few channels left over.
The '5122 does require a >10V low voltage supply with similar logic levels, so a level shifter will be needed between your digital stuff and it. I just used the simple two-resistors-and-a-mosfet approach; it's only 4 signal lines so no big deal.
-
No need for so many components, a nixie only needs an active pull-down (BJT or MOSFET), and maybe a high ohm pull-up to counter leakage. The switching device doesn't need to handle all 170V+, but only the ~50V change between "on" and "off" states. The rest is taken up by the current limiting resistor.
You could arrange common-(emitter/source) or common-(base/gate) switching (in the latter case: B/G to VCC via resistor, MCU pin to E/S with optional current limiting resistor).
Whichever permutation you use, I wouldn't consider it a risk to the MCU. The HV supply should be current limited, both by brute force (the resistor), and by design (you only need a few mA, so why make a supply capable of piles of current?). MCU pins are more than capable of sinking this kind of current, in either state, accidentally or otherwise.
Tim
-
I used an HV5122 32 channel open drain driver in my Nixie clock
I hadn't actually realised that family of chips were still in production: I think I'll invest in a few for simplicity's sake. I'll need to etch or buy DIP converter boards from somewhere (I'll try out an etching tonight to see how it goes - toner transfer method doesn't favour fine pitch components I'm afraid!) If that works, score, saves me £1.50 per board ;)
@T3sl4co1l: Thanks! I'll have a play around at some point once I get some parts in. I need to get a high voltage source sorted out as well: I'm thinking a MAX1771 (http://www.farnell.com/datasheets/74526.pdf)using the application notes for the schematic.
I'll probably look around for a DIP version of that chip as well for the sake of simplicity
Thanks for all your help!
-
If you want to go seriously retro (steampunk?), the cathode of a vacuum tube, such as 1/2 12AT7, can be driven from TTL levels if the grid is grounded. However, the plate will not swing as far negative as a semiconductor. With a high-mu triode, the TTL level swing should be sufficient to cut off the tube, and the driver must sink the cathode current in the on state.
-
One advantage that you'd have with suitably sized triodes is, the plate resistance naturally limits current. So you could drive the nixie anode from fixed voltage and use an array of 12AX7, or AT or AU, to drive the cathodes with the required current. Just a small simplification, but hey, saving resistors and all that...
FWIW, driving the grid positive (say, to VCC) allows pretty reasonable [saturation and current draw] performance from triodes; though you won't get such good cutoff in such a case! :)
Tim
-
Depending on the current you want, use either 12AX7 (very low current) or 12AT7 (medium current). Their high mu value requires a smaller cutoff voltage. In the circuit shown, the cutoff bias would be about 5 V, if the driver swings to the +5V rail. To first order, the cutoff voltage is Vco = Vpk / mu, and the mu of a 12AU7 is only about 20, compared to 60 for the 12AT7. All three tubes are pin-compatible.
-
Righty-ho!
Getting near the finalisation of my parts-order: just need to sort out the power supply
I've been hum-ing and ho-ing and to-ing and fro-ing over what to do for the power supply:
-Should I buy one? (Nah! Making one is more fun!)
-Should I step down from Mains AC? (Overkill: plus I was hoping to have it battery powered!)
That leaves DC:DC step-up controllers; stepping up 9V or even 5V USB to the HV required for the Nixies - current draw should be low enough for batteries. 8x Nixies running at max allowable currents is only 24mA (and that's with me being generous and assuming 20% extra per tube on top of the max allowed in the datasheet)
I've seen both the MAX1771ESA (http://uk.farnell.com/maxim-integrated-products/max1771esa/controller-step-up-dc-dc-12v-1771/dp/1379871) and the MC34063AP (http://uk.farnell.com/texas-instruments/mc34063ap/ic-dc-dc-controller-34063-dip8/dp/1053579) used in various PSUs.
The MC34063AP has the benefit of being a DIP package which saves me etching my own board (which I wouldn't trust at high voltages as I'm stuck with the toner transfer method!) it's also cheaper than the MAX1771. The datasheet (I believe) says its rated to 40V output: but a number of people appear to have used it so I may be misinterpreting the datasheets!
The MAX1771ESA comes in an SO-8 package - which means I'd need an adapter (which is easy enough to do - I have adaptors for the HV5122 packages anyway so it's not a huge deal: just means I have a few daughter boards kicking about the innards) - these are considerably more expensive than the MC34063, but has a higher voltage rating (I think I saw the max output was 100V? but again, people have driven these at 170 and 180VDC so I'm again unsure how to interpret the datasheets!
Can anybody advise me further on these matters?
Many thanks!
-
Personally, I'd stay away from direct mains sourced voltages due to the inherent danger and the pain of wiring for AC.
For chip based power supplies- The reality is, if you're driving your inductor with a MOSFET,like most of the designs out there, you can have any terminal output voltage (within reason) as the chip is not exposed to the o/p voltage directly. Do a search on either chip with the term 'Nixie Power Supply' added and have a look at the large number of designs out there. The MAX1771 is slightly more efficient apparently.
I have a MC34063 design producing 180V quite happily, although this is as far as I've got on my nixie clock. :)
I'll be building mine on stripboard to give it a rustic appearance, point to point wiring would be a little too laborious :-DD
Good luck!
-
if you buy the russian or ukrainian tubes, often the seller will also have the 74141 style driver chips. I bought some from a US seller but I could have gotton them overseas, too. they are equivalent to the US old style chips (TI?) but for the very expensive tubes (I forget the name, but they are the super huge nixie that costs more than $200 for a set of 8) the US chips are supposed to drive them better (something about a blue spot appearing on those nice expensive tubes when using russian driver chips).
if you use the more affordable and smaller nixies, the russian chips are fine. I have them in my nixie clock and for the past year that its been alive, the chips have not failed me yet ;)
they are still buyable. not sure for how much longer. probably a few years, at least (guessing).
-
I'd stay away from direct mains sourced voltages due to the inherent danger and the pain of wiring for AC.
Agreed!
Mains AC scares me: a) because it can kill me fairly easily and b) because it's AC which I rarely play with.
I have a MC34063 design producing 180V quite happily
I've settled for the MC34063 as the DC/DC controller: still to work out the support circuitry for it - that side of things can wait until later though ;)
they are still buyable. not sure for how much longer. probably a few years, at least (guessing).
Ah, and there's the rub! I'm wanting to use (as much as possible) off-the-shelf parts that are still in production or are easily replaceable :p Which is a pity, because some of these old driver chips would have done the business! Though some of them are a more than alittle costly anyway (at least to buy in the UK (at least as far as I've seen!))
And that brings us to today...
I tried to order the bulk of my parts only to discover that the HV5622 I had originally settled on as the nixie driver have a minimum purchase order of 96 from Microchip Direct: and I don't have a spare £400; unavailable from uk.farnell; and I couldn't be bothered transferringmy entire order into digikey. Actually, importing it probably wouldn't be too hard, but I can't be bothered figuring it out.
Great! Now I need another driver solution (I thought I had nailed it with the HV5622 and it's family of chips! I mean, I could get them off ebay, but most looked like they were going to take a month to arrive; and some even said they had been used (wait... what!? ew... second hand chips... yuck...))
One option is to return to the idea of driving transistors from a shift register - and by coincidence I already had a latching shift register picked out for running the LED-based fake-nixie tubes; which would almost certainly do the trick (and the latching makes things simple)
Would I be right in saying that tying the common-anode nixie tubes to the collector on the transistors (and having a resistor on the emitter which is tied to ground) work as a solution?
If so, for the sake of an extra layer of protection, would a high-voltage diode between the shift register and the transistor be effective?
Ps - I'm aware hand-soldering 96 transistors is going to be one of the most wonderful ideas I've ever came up with! But it's going to be done as a modular series of boards so I can bring the workload to a manageable level (and easily replace anything that blows up)
-
maybe helpful site: google 'arduinix'.
I bought a board from him, built it but never got around to hooking it up ;) he has onboard psu, driver transistors and uses 2 russian 74141 chips. you can drive 4 or 6 or even 8 if you mux them. he has sample code, too.
might give you ideas.
-
Ooh! Plenty of ideas kicking about xD
Right: Two questions;
Question 1: So what is the opinion on muxing these? The Datasheet (http://tubehobby.com/datasheets/in14.pdf) suggests they have a "Multiplex Mode" - but will it increase the chance of them failing? (and can anybody confirm the 1 – 1.8 kHz frequency on the PWM?). I'm aiming for longest possible lifetime on my nixies xD
Question 2: Purely out of curiosity - let's assume I screw up the driving code and accidentally open two cathodes (i.e effectively telling it to display two digits at the same time) - what happens? I would have liked to have thought the current limiting resistor on the anode would have limited the current so it doesn't power on - but I'm not actually sure!
-
The one closest to the anode will invariably win, but if a rear one is already on it will tend to keep the others from having enough voltage to strike.
-
Yeah, the cathodes only have to bob up/down by the difference between lowest glow voltage and highest breakdown voltage. Something like 30 or 50V, rather than all 140+.
And distance will yield a higher drop, so there's that.
Probably, releasing one cathode while pulling down another (within a few microseconds) will push the plasma around, so that the current flow, and anode voltage, remain fairly constant rather than drooping out as would happen with two completely separate lamps. That is to say, the presence of a glow on any electrode will encourage any other electrode to discharge, so it doesn't have to switch the full voltage, from fully off (un-ionized) to on. Which means the change in voltage could be even less.
I wouldn't think multiplex will really gain you much, and you don't have much current capacity and therefore brightness to work with. They're not like LEDs, which you can pulse well over the static ratings. Which stinks, because you get the biggest mux savings by using large division ratios.
Tim
-
@SeanB - so it won't wreck them then; that's good news ;)
Well this afternoon has been spent trying to lay out a PCB to make the soldering easier.
Yuck - layout design problems; my favourite! I wonder if a transistor array would make life easier... except... they don't tend to come in high voltage-tolerant versions do they...?
I'll post up the schematic diagrams once I get one that doesn't look horrific!
-
Well, a 2N3904 style array will already get you more than enough (60Vceo, breakdown around 100V). If you demand full supply range, there's probably something in the MPSA42 range. Although, when you're talking 200V on an array, you can't really miniaturize below SOIC level, and even that's pushing it.
Or among ULNxxxx types, there's probably something over 30V too.
Tim
-
Oops sorry, I meant an integrated array - as opposed to discrete transistors (which is what I was using in the PCB: just a line of them for each Nixie. I was thinking of using the PHD13003C - and that's where my layout problems were coming from: spec-wise they're overkill for a nixie but fairly cheap and I was already using them for a different project so it wasn't a huge problem to tack an extra... 96 of them onto my order. I am NOT looking forward to soldering that...)
Reckon these (http://www.farnell.com/datasheets/1831140.pdf) might do the trick? (and am I right in saying that it's a simple case of applying current to the base (1B through 7B) and a current will come out of the collector (1C to 7C) based on the Emitter/Common voltage?)
Wait... why does it have 7 transistors and not an even number? Weird xD Might try to find one with 8 transistors to make thing simpler... anyway - will that style of array work if I can get it in a 100V Collector-Emitter Rating?
Out of curiosity - what sort of inductance do nixie tubes have? Would you consider them to be an inductive load (obviously not on the same sort of level as a motor; but worth taking into account?) I couldn't find much data on the inductance of them I'm afraid (though I didn't look for particularly long: too much stuff to do :()
-
The inductance should be negligible: they are just a glorified neon bulb or VR tube. You can see the short leads inside the glass envelope between the base pins and the electrodes. The main weird thing about their behavior is hysteresis between the "strike" voltage and the "sustain" voltage, but that is a non-linear, non-reactive effect.
-
Aye, some of their properties are a bit weird ;) My favourite being their negative resistance once struck xD
They're such fascinating little devices!
Now to get back to the drawing board yet again and find yet another driver solution ;) I should also probably get back to work at some point - I'm totally not in the office at the moment.
--EDIT--
That 7-transistor-IC is causing problems... could really do with an 8-transistor version of it! Which... it turns out (at least from Farnell) only come in 50V variants.
Take it a 50V Transistor is too low (bearing in mind a c.170V strike voltage: how much does it drop after it's struck? I thought it was only something like 20-30V drop?)
Ps - I'll do some proper research tonight on the subject - I'm just throwing my thoughts about just now!
-
No need for 500mA darlingtons, probably undesirable in fact, as you have to contend with max 0.5mA collector leakage!
The heptet package is simply what fits in a DIP16. DIP18? What's that? :P
Tim
-
I think I'm going to just go for my original plan and just do it all through-hole on matrix board with individual transistors: 4x driver boards each with 24x transistors and feeding two nixie daughter boards each: with bus lines for data and clock etc going across the boards and going to the power and control circuitry.
Why not put it on one board? Well, I could, but I'm wanting a modular design xD
@T3sl4co1l: ST do 8-transistor in DIP-18 packages? I'm confused? Are they rare or something? I can't say I know much about packages outside of the little bubble of components I regularly use.
--Edit--
Just to confirm (and stop me from :palm: again) collector leakage is just the (usually small) current across emitter>collector when base isn't powered? Right?
In the case of the nixie tubes, will there still be a small current coming from the unlit cathodes or are they effectively open-circuit/high impedance when not struck?
-
--Edit--
Just to confirm (and stop me from :palm: again) collector leakage is just the (usually small) current across emitter>collector when base isn't powered? Right?
In the case of the nixie tubes, will there still be a small current coming from the unlit cathodes or are they effectively open-circuit/high impedance when not struck?
One should assume a very small leakage current flowing when the cathode when not struck.
"Collector leakage" can be specified in different ways on the data sheet:
Icbo is collector-base current with emitter open circuited. This is the basic reverse-biased diode leakage current.
Ices is collector-emitter current with base shorted to emitter. Essentially the same as Icbo.
Iceo is collector-emitter current with base open circuited. The collector-base leakage flows into the base junction and is amplified by the current gain {beta}. This is the most stringent spec, relevant to open-circuit drive to the base in off state.
-
@T3sl4co1l: ST do 8-transistor in DIP-18 packages? I'm confused? Are they rare or something? I can't say I know much about packages outside of the little bubble of components I regularly use.
Ah... then, standard DIPs are 4, 6, 8, 14 (there may be custom sizes inbetween, but rarely encountered), 16, 20, 24, 28, etc. They often skip a few pins at a time, though the 14-16 pairing seems peculiar.
Example: LM319 has four NC's because it requires ten pins. So, DIP14.
SMT packages are far more varied, so you'll find MSOP8 and 10, and all matter of alphabet soup from DFN/QFNxx to chip-scale BGAs/LGAs.
--Edit--
Just to confirm (and stop me from :palm: again) collector leakage is just the (usually small) current across emitter>collector when base isn't powered? Right?
In the case of the nixie tubes, will there still be a small current coming from the unlit cathodes or are they effectively open-circuit/high impedance when not struck?
Yes and yes; actually, the most recent Mailbag episode illustrates this. The indicator tube passes fractional mA (and little or no light) below the glow threshold.
The trouble is, you have leakage fighting leakage, so whether you end up with a trace of visible glow (annoying but not likely a problem) or not is a dice roll. It will be worse at higher temperatures, where the transistors leak more.
The easy solution is a pull-up resistor, sufficient to sink the worst-case transistor leakage at the required voltage (say, 50V or more -- so from a 200V supply, a (200-50)V / (0.5mA) = 300kohms resistor). Of course, every single component you add, you have to add N*M times over, so even a simple pullup starts to look bothersome.
Good old MPSA42 is hard to beat... :-DD
Tim
-
I think I'm going to just go for my original plan and just do it all through-hole on matrix board with individual transistors: 4x driver boards each with 24x transistors and feeding two nixie daughter boards each: with bus lines for data and clock etc going across the boards and going to the power and control circuitry.
I've been using TLP627 high voltage optocouplers in my nixie projects. They're easy to use, take up little board space, isolate the HV and are cheap:
http://www.mouser.com/ProductDetail/Toshiba/TLP627-4F/?qs=f0GatGUIxV0H1uxeyGCvdQ%3D%3D (http://www.mouser.com/ProductDetail/Toshiba/TLP627-4F/?qs=f0GatGUIxV0H1uxeyGCvdQ%3D%3D)
-
@kwass: might become a bit expensive for what I'm doing I'm afraid - I think I considered them at some point actually. I remember getting a similar price when calculating something!
£32+VAT (on farnell - cheaper on mouser) for the optocouplers vs £5.61+VAT for the transistors.
I'll add it to the list of things to consider though :P might try firing my order through mouser at some point just to compare prices
@TimFox: Thanks! That's cleared things up a bit :)
@T3sl4co1l: Ah-ha! - thanks very much xD As I said, I know of the packages (for instance, I could tell you what they all are) - just wasn't sure on the sort of distribution of those packages - that's cleared things up a lot! [takes note for future reference]
It's been a good day for learning! I even learned how to reset one of the big CNC machining centers when the operator breaks it :)
Good old MPSA42 is hard to beat... :-DD
Aye, I reckon if I'm going back to using individual transistors they're the best bet: cheap too - my favourite combination!
I'll have one (or maybe even two) more whack(s) at a PCB layout and then decide whether I'm going to run with it or do it ye aulde way with the matrix board. We shall see!
Thanks for all your help folks! It's been very much appreciated - I can't thank you all enough ;)
-
taking an unfinished older project and bringing it closer to something like completion ;)
(https://farm8.staticflickr.com/7622/16845741629_1744282f57_b.jpg)
there is no case yet. the board was thrown together on a moment's notice and I added extra stuff to it, recently (the dc/dc converter is that hot-melt glue mess in the middle, lol). the RTC module is a $10 very common ebay module (ds3231, more than good enough for local time keeping). cpu is arduino nano v3, also a $10 common ebay module. chips are russian style 74141 and philips pcf8574 i2c port expanders, that give the bcd outputs to the nixie driver chips.
ugly as hell, but once I throw it into some kind of box, it will clean up well enough ;)
I still need to integrate an xbee receiver module in there, somewhere. somewhere....
-
Nice :-+
I'll look forward to seeing it all cased up ;) One of those extruded aluminium cases with a clear/tinted front may look pretty nice on it. What was it you were planning on doing? If you're any good at woodworking a nice hand-carved case would be the shiz
As for where to put your XBEE module.......em... might need more of that hot-snot ;)
I've had a bit of an interesting problem today: because of my projects modular design (which is very reliant on 2.54mm headers) - the cost of headers and sockets now accounts for 40% of the project cost... nearly £35 of connectors...
Well, I can tell you right away...... that... I... oh would you look at that... I miscalculated, forgot they were all double-row headers and doubled the quantity of everything... feck......... Well that's embarrassing...
You know what? Sod - it I'm going to the pub O0
-
here's what I came up with, over the weekend. hope you don't mind a flood of pics, but its a sequence, a tear-down of my own poor clock, if you will ;)
(https://c4.staticflickr.com/8/7630/17048609862_465d279201_c.jpg)
(https://c4.staticflickr.com/8/7690/17024062706_f3e52d478e_c.jpg)
(https://c4.staticflickr.com/8/7667/16429908103_49ed720778_c.jpg)
(https://c4.staticflickr.com/8/7599/16863834599_66f301f636_c.jpg)
(https://c1.staticflickr.com/9/8801/17049227351_c8d5430ec2_c.jpg)
mostly a reshape and refit of the previous clock case style I used. still not quite right, but good for a first pass. the blue plastic was all they had in stock; I would have chosen black instead. not sure I'll go with the orange top, either; might redo it in clear. the plan is to have a single rgb led that you can view and see as a non-confusing status light. you choose what you want the colors to be used for, but you get one colorful led and any 255/255/255 combo you want, but only one at a time. keeping it simple might should make its use simple but still useful (I hope).
I spent some time with the arduino Time.h and TimeZone.h libraries and also pulled in SoftwareSerial for the xbee. for almost no original code, I was able to put the xbee into listen mode, have it look for that simple demo time string (the bash 'date %T' that you send to the usb-xbee on the far end) and when it sees it, it sets its local millis() software clock to that time. there is NO more RTC in hardware, on my board! its a clock-less clock! lol. it gets its time from a remote time-master that is on the other end (the send-only) end of the xbee. that's my time-cloud (haha). so, its a clockless clock that gets its time from a local cloud.
for fun and grins, I have it set so that it blinks 12:00 (haha) when it powers up and until it receives the first time-master broadcast. if you start this thing up it blinks twelve at you. you then run over to your raspberry pi, start the bash script that echos out to /dev/ttyUSB0 and that finally sends out a time string over the xbee rf channel. the clock is in always-listen mode and when it gets its first string, it stops the stupid VCR-dummy mode and shows the real time. from then on, as long as you don't power it down, it keeps software-based time that is steered by frequent (or not) time-master xbee broadcasts.
its really fun playing around with all this stuff. its mostly glue-work. lots of payback for just a bit of gluing things together.
anyway, my IoT clock, of sorts. lol. lots more to be done, but this is a nice little test-bed for me to hack around with.
oh, and the sending code? its this simple:
#!/bin/bash
time_int="${1}"
while [ 1 ]; do
date +T%s > /dev/ttyUSB0
date
sleep $time_int
done
my rasp pi is a statum-1 time server (via a $15 ebay gps module and some more glue code). it always has 'good time'. that thing also has a usb-xbee on it. so far, for POC, I just run the same format of xbee 'message' as the sample arduino code expects, the Tblah-blah-blah stuff, in ascii. as a starting point, it works fine. and it didn't cost me any real effort to this all this working ;)
-
Nice xD I approve xD
The orange and blue work quite well actually - being complementary colours (opposite each other on the colour wheel) they contrast quite well.
I had forgotten you had access to the laser cutter - what I'd give for one of those - when I looked around a year ago I discovered the cheapest ones were cheap-o Chinese models that didn't have proper cooling and some didn't even have 'door open' switches to stop the laser when you opened them up...
And they weren't all that cheap either - £1000-ish for something that would need modified upon receiving it (to bring it up to some parable of safety conformity)
I might build (or buy) a CNC mill at some point - It can do a similar function, though to a lower accuracy than the laser (BUT it can do it in 3D and can mill PCBs directly! You can use the laser to etch PCBs but you need to paint them first, etch the negative into the paint and then etch the exposed copper with any chemical etchant)
Anyway! I'm rambling! What are you planning on doing with the clock next?
-
Sorry, quick question that just occurred to me...
See if I have a 180Vdc and a 5Vdc rail... How do I handle the ground lines for them? Just occurred to me I've never worked with multiple voltage lines at the same time before!
Two questions!
Obviously they both must eventually hit the ground line at some point:
Should I a) keep both ground rails separate and join them as close to the power supply as possible (or can I just connect high and low voltage components to any old spot on the ground rail)? And b) do the lower- voltage components need to go to ground via a diode to prevent the higher potential difference from the high voltage lines driving a reverse current through them (or does the ground sink that current before it can happen?)
PS - sorry for asking something so basic - as I said, I've never worker with mixed voltage rails before!
Pps - the number of nooby questions will decrease as I plod though the art of electronics 3rd ed. Which arrived on Saturday ;) Great price for it as well! It was half the price of much the smaller textbooks for immunology and organic chemistry that I bought at uni!
-
I did nothing special for grounds. in fact, the dc/dc that I used (ebay module, US designed and maybe made here, not sure) just has 3 terms that matter: 5v in, gnd and high dc out. you can't help but share grounds that way.
it might be a good idea to have an island or star that is 5v gnd and everything ties there for 5v levels. the dc/dc will be off on a corner of the board (not too close to other things; when it was close to the RTC module, I was getting i2c random errors). and it will have a trace from its module to the center star 5v gnd. but that any logic stuff that would flow into the star gnd and then back over to the dc/dc gnd would not want to - its a longer 'round trip' and so there's no 'motivation' for the current to flow that way.
in reality, I used a rat's nest of p-p wiring and didn't really do anything neatly on this perf board. biggest thing is to move the dc/dc to a corner of the board and keep things away from it, at least 1cm or more.
as for what I plan to do with this module, I have been seeing a lot of talk about 'IoT' (internet of things) these days and for a job that I had just recently (that ended), I got a chance to do some real IoT work. it got me thinking about what I didn't like about their approach and how I'd do things differently. so, this is a nice little test platform for that area of research. this box can be a 'message receiver'. usually its 'messages' are time-of-day broadcasts from some timemaster on the same xbee 'lan'. but if some other numeric value should be displayed, the xbee master (the only transmitter on that group of nodes) will announce some other event or thing to be displayed and the clock will receive that and change its display to show the new value in the new correct way (if its an egg timer, its a count-down. if its an alarm that fired, its a blinking timer value and maybe a sound, as well).
for proof of concept, I also want to send other messages to this, just to show what can be done. indoor/outdoor temperature, new unread email count; maybe some code to flash if you left your garage door open (lol), or if the front doorbell is ringing. or, in the more unusual (but fun) case, if I change the volume on my DIY audio preamp, I would like to see an update flash on this clock display to show the dB value. since I wrote the preamp code (also arduino based) I can have it send updates to the xbee 'lan' and this guy will have the ability to subscribe to those and show the volume-change events when the user changes the audio level.
to really show the concept, I want to have several different builds with different displays and abilities but all running the same IoT slave code and all listening to some single IoT master xbee who broadcasts updates for any thing in the house who 'wants' to listen ;)
so, much much more than a clock. once you have a data receiver module (for me, I just love xbees) you are now part of a distributed mesh of nodes. being a networking person by trade, this interests me quite a lot ;)
-
I had forgotten you had access to the laser cutter - what I'd give for one of those - when I looked around a year ago I discovered the cheapest ones were cheap-o Chinese models that didn't have proper cooling and some didn't even have 'door open' switches to stop the laser when you opened them up...
you have to do a lot of work on the cheap ones and I still would not waste money on those, personally. maintenace is high, good software support is essential. these are not toys and never will be. they are serious industrial tools. you can buy toy 3d printers and that's fine, but laser cutting is serious stuff. I don't think we'll ever see on-the-cheap ones that are worth using.
I am lucky being in the bay area, we have 3 tech shops (techshop.ws) near us; san francisco, san jose and midway between in redwood city. each store has 3 or more cutters and you simply pay a monthly fee for access ($125, roughly) and then you book 2 hours at a time, up to 3x a week. but that's per site; if you need more time than 6 hours a week, you can drive to the other sites and get another set of 3x2hrs there. the one near me (RC) is also open 7x24 (recently) and if no one is using the machine, you are allowed to hop on informally for an hour at a time. at 3am, though, I can't see it being in huge demand (lol) and so in reality, you can use them as much as you need to, as long as the machines are not in use at the time.
they have 45w, 60w and a new 120w unit that I have not used yet. you take a 2hr class on safety and use, pay a one time class fee (usually less than $100) and then you are 'certified' to use that machine. that's how it works for all their big gear (plasma cutters, cnc, sandblaster/powdercoaters, lathes, milling machines, etc). I can't even guess how much money that place has in heavy industrial machinery. that's why its over $100/mo for access. once you take a class on the machine you want to use, you're free to use it as long as you can reserve time on the calendar (ie, no one else booked it for that slot). I still have yet to take the metal tormach cnc class but I did take a shop-bot cnc wood cutter class. such fun!
they have a small stock of plastic at good prices there, but when I went, they were out of black and clear and so I had to resort to using that cobalt blue stuff ;) for $14 a sheet (18"x24" is the max bed size) its not too expensive and I can afford to waste a sheet to verify the design and that it fits and holds together. it always takes at least 2 tries anyway to get something complex fully debugged, so a 'blue run' is not a big problem as long as its not my final run, lol.
-
a lot of work on the cheap ones
- aye, that was the conclusion from my last search for them ;)
Aye, at work we had a plasma cutter and a small laser-cutter (both of which have moved off-site to a separate fabrication plant when we expanded - so that our current shop is just big CNC mills, turning and machining centers)
I occasionally use some of the CNC mills - but everything except the Bridgeport is complete and utter overkill for milling a circuit board (though I have done it once or twice on some of the smaller machines just to see if it'd work)
And by "overkill" I mean "these machines are usually used for precision CNC-ing 10-tonne pieces of steel" and also "you can quite easily step inside one of these machines" ;)
(http://www.mullermachines.ch/MachineImages/Large/18732_1.jpg)
On occasion you can borrow one of these machines, under supervision of a trained operator, for home-jobs: but only when the machine has no actual jobs due to run on it (and of course, we try to minimize such downtime because when it costs £80 an hour (what's that? About $120, maybe $130, these days?) you can see why!).
The CNC Bridgeport (which of course you can run manually too) is less accurate, but a much better size for small jobs (plus we hardly ever use it except for fixing bits of the larger machines!)
(http://www.strandindustrialmachinery.com/photos/15200.jpg)
- and I've had decent success with milling PCBs on it (though it was time consuming: you obviously need to start with the 2D picture of the tracks, which you then need to put into G-Code (which can take a while unless you're a genius on the software! (which to be fair, is designed for doing complex 3D shapes, not 2D shapes ;))) - and then the actual milling takes a fair while as well for even small boards (biggest I can do is 160x100mm due to EagleCAD limitations: unless you just make two and back them up together ;))
Anyway my ultimate future hobby-wise is to just buy or build a desktop CNC unit. Ultimately the construction is the same as a 3D printer... but with a really fast scary rotating bit instead of a really hot scary hot bit. Oh, and some vibration dampening ;)... maybe a tool-changer...
Sure, using the stuff at work will get me by for small one-off projects, and for big things that I can't fit in my bedroom (no seriously, some of these machines fit bits of metal 20ft long, 10ft wide and weighing 30 tonnes - and then spin them several revs per second on a turntable... it's actually terrifying to see such power in action!)
Anyway, back to a proper reply (I do love rambling!)
Where I live the nearest tech-shops are in Glasgow (25 miles away - except that's a two-hour journey to get in, and then another hour of pottering about with local transport to get to one of the a) hackerspace b) toolshops. Luckily I work at a CNC machining shop with about 30 big machines, but that's no good for small things (or rather, overkill for small things!) - which are fine for the odd job here and there (time and scheduling permitting) but I could do with something a little closer to home!
I'm jealous of all the US tool/machine shops and hackerspaces kicking around ;)
Ps - thanks for the stuff on voltage rails - I've still to have a proper read at it (I actually didn't realise it was even there until I scrolled down and said "ooh what's this block of text here!?")
-
simple little box for my usb xbee (USBee, lol).
with the little antennas - and if they are placed on an upper shelf, it can reach from one end of the house to the other.
(https://c1.staticflickr.com/9/8780/16882735758_7f0c9ae6db_c.jpg)
I set them at 9600 baud just to make software-serial easier, on the arduino.
-
simple little box for my usb xbee (USBee, lol).
Nice xD I love laser-cut acrylic!
I tried milling it once and got a decent finish, but the edges just aren't as nice as what you get with the laser (imagine sanding it to shape instead of lasing it xD)
Would you mind doing me a wee favour? Could you get me a ballpark figure on the value of the inductor on your power supply?
My calculations are saying it should be 20H. Yup, 20 henries which seems a little.... high... [in fact I looped one up on Farnell and mouser and they're huge!] - and running through the same calculation for the example in the datasheet gives me a 70 henry inductor! (whereas they use a 170uH inductor - which is a much more reasonable size!)
My only assumption can be that
a) I've done something stupid
b) I've mixed up units somewhere (i.e a 20mH or even 2mH would seem to be much more in the ballpark: even though they;re both pretty massive, 200uH seems reasonable, but 20uH seems a little small!)
c) my design is way off-spec (which I don't think it is; I'm basically following the application notes=/)
Any ideas?
Ps - if anybody wants I'll post up my working - figured I'd poke around before doing anything else (and double check everything before purchasing anything!)
Pps - power supply components are the last on my to-buy list and then I can get to work! Woo!
--EDIT-- I just rememeber your supply is gunked up with hot-snot... whoops xD
-
I can post a link to the guy who makes those psu's. I have no idea what is inside it. that was the point; its done, working, he did all the thinking and testing and hopefully its also safety tested and safety designed. better than I could do, especially for the price.
when I do a board for my own clock, I'll use his dc/dc module since I see no reason to reinvent it. and when I do a board, it won't be .1" locked so I -can- mount his horizontally and not have to bodge it with glue. I still might have a safety cage around it, with plastic, just for extra measure. its very likely that this system will be opened and I don't want anyone to have to worry about touching high voltage. I'd use the same 'tab box' style as a cover over this module, once its on the pcb. have to figure out some clever way to secure it, though.
-
Apparently, at least according to the datasheet, the PSUs are fairly simple to set up: with a voltage divider used to set the output voltage: the capacitors and conductors are used to set the frequency and ripple on the output. I might just need to buy a range of inductors and play around with them. I'll probably need low ESR capacitors as well just to keep things stable - we'll see!
Any idea what sort of frequency I want to set the PSU to? Is a higher or lower frequency more desirable for Nixies? (these ICs can manage 10-100KHz if I recall)? I seem to recall a lower frequency was more efficient (or was it a higher frequency?) - the aim here is to get the highest efficiency - is that the only thing I need to worry about in this case? (I'm going to try and filter the output with an inductor/capacitor in addition to what's regularly coming out the PSU)
-
AHA! That 20H figure was using seconds in the equation! Forgot to scale down to microseconds, so that means that 20H is six orders of magnitude too large... guess that's 20uH then!
Seems a little small though? Hmmm... wonder what the predicted ripple is for that...*
I've used various parameters and got between 20uH and 20mH: with 500uH cropping up repeatedly: I'll probably just buy a range and experiment with what gives the best result.
*actually that's a good point - how will I measure the ripple on that? My scope's only rated to 50V if I recall...? What would happen if I put a dropper resistor in series with the probe (i.e solderered a resistor onto the board to use as a test point)? Except I think that turns the rest of the circuit into one of the legs of a voltage divider... - how would one measure voltage above an oscilloscopes rating?
-
Sorry to - woah - triple-post!
I just realised that the comma's run on a lower max current than the digits :scared:
Digits run at 2-3.5mA max: commas at 0.3mA
I was *originally* going to just put a dropper resistor on the Common Anode of the tube (I think I worked it out as a 15K or 20K resistor?) - with each cathode going straight to the switching transistor and then to ground.
Can I limit the current on the commas with a second resistor between the comma's cathode and the transistor (or will that act as a voltage divider?)
OR would it be a better idea to put an individual resistor on each cathode (instead of just one on the anode)?
-
I think you can do both. one R on the anode and one more on the comma or dot/period cathode. leave the rest of the cathodes alone.
-
Aye, I had figured on either dropping every cathode individually (and having nothing on the anode) OR additional dropper on the commas' cathodes (but not the other ones ;))
Will be excellent if I can just drop an extra resistor on the commas though! Will save a small fortune on resistors ;)
Parts should arrive tomorrow, will try and prototype the PSU over the weekend and will hopefully get the uC sorted before next week: should have innards sorted within 3 weeks and should have it in it's final form within 5 weeks!
I'll post pictures whenever I do anything interesting ;)
Thanks again :) :D
-
Almost... half way there.
Breadboarded up the power supply today - and managed to get a maximum of 75V out of the MC34063
Wonder what I'm doing wrong... in fact I wonder if that breadboard is coupling all sorts of parasitic into it*... and the resistsance on those jumpers will add up too... hmmm...
Reckon it'd be worth soldering up on some matrix board and see if that gets me anywhere?
*translation: there are more parasitics than you can shake a tick** at!
**this is a bad joke, not a typo.
-
For starters...
MC34063...
Really?
Reeeally?
Second, that yellow-white core is a #26 powdered iron, widely renowned for its performance as a cheap resistor moreso than as any kind of inductance. Well, almost. The Q factor is typically 10 at most frequencies, 20 tops. And way worse under any kind of transient condition, like a flyback spike that's intended to be really, really thin.
You should have a tapped winding or step-up transformer to begin with, but with a core as bad as this, it is a requirement!
(How are you supposed to *know* that an inductor is bad? Well, experience more than anything I'm afraid. The Micrometals website (http://www.micrometals.com) does have info on powdered iron toroids: ratings, color scheme and performance.)
This is a pretty reasonable circuit, uses far fewer transistors than the MC34063 (semantically speaking), and works better:
(http://seventransistorlabs.com/Images/Deadbug_Sch.png)
Tim
-
Ha! That's me told ;)
The PSU was the really dodgiest part of the project tbh ;)
I think I might just buy a module from somewhere - I had originally planned on making one as a challenge, but if I've went and bought crappy inductors then there probably won't be much to salvage unless I go out and buy more components (which I can't be bothered doing xD) - ironically I had nearly bought some power inductors but went with these because they were cheaper - would they have worked any better?
The problem is lack of experience, I usually only work with straight digital circuits, none of these crazy inductors and voltage magic tricks ;)
At least I can still use the MC34063 in other projects - oh well!
Thanks very much mate - much appreciated, and it's saved me messing around with this circuit trying to debug/fix it!
-
You can still find inductors of that type, commercially, but most of them are lower losses. If they specify ferrite, you can be pretty sure the Q factor is good. "Powder" tends to mean different things these days (molded powder composite cores, which work surprisingly well), instead of that archaic but temptingly cheap stuff.
A step-up transformer like that isn't too common, but some likely candidates might be a "capacitor charging" type, such as by Coilcraft and others; or a very small converter transformer (say in the 120V to 5V, 1A range, wired backwards).
Tim
-
archaic but temptingly cheap stuff.
- aye, damn my tightfistedness! - just checked out a few datasheets for these inductors - q factor isn't even listed, must be a good sign ;)
Let us assume we tried to salvage the existing design (I'll get a schematic up later - though it's pretty much the step-up application circuit with this extra part (http://www.changpuak.ch/electronics/High_Voltage_Power_Supply_MC34063.php) to it to bump up the duty cycle and a little bit of extra filtering on the output)
Would a 2200R series inductor (http://www.spiratronics.com/data/5670.pdf) or similar be of any use for the power supply?
For starters...
MC34063...
Really?
Reeeally?
Oh aye! Meant to ask, what's wrong with the MC34063 out of curiosity :P? (might as well turn this into a learning experience!!!)
I was originally going to use a MAX1771; but once again was tempted by the lower price and the DIP package of the MC34063. I was also (perhaps naively) reassured by the number of nixie clocks using the MC34063 in it's PSU.
I don't think I cut corners anywhere else though; so with any luck nothing else will fail miserably because of a cost-saving measure!
-
Let us assume we tried to salvage the existing design (I'll get a schematic up later - though it's pretty much the step-up application circuit with this extra part (http://www.changpuak.ch/electronics/High_Voltage_Power_Supply_MC34063.php) to it to bump up the duty cycle and a little bit of extra filtering on the output)
I would suggest not, but I'll cover that in more detail... :)
Would a 2200R series inductor (http://www.spiratronics.com/data/5670.pdf) or similar be of any use for the power supply?
Umm, probably. Looks to be a ferrite bobbin type. Winding losses are dominant on those, so make sure you get a low enough DCR to handle it (minus skin effect and all that).
What you really want is a step-up autoformer or transformer, like in my circuit. Then you don't need to worry about ridiculous duty cycles, fast pulses and huge losses. Examples are also in the '34063 datasheet.
Oh aye! Meant to ask, what's wrong with the MC34063 out of curiosity :P? (might as well turn this into a learning experience!!!)
- It's old. OOoooollld. Not necessarily a bad thing, but that brings inevitable process and design limitations that modern parts don't have.
- The operating frequency is low.
- The control method sucks.
- Actually, they never even say what the control method *is*. There's that "current sense" block, but it's not peak current mode, it doesn't turn off a latch. It just kind of... pokes at the oscillator duty cycle. WTF?
- Regulation is typically (always?) hysteretic, so the output ripple is incessant, low frequency, chaotic and hard to filter.
- Because the frequency is low, you spend extra on caps and inductor(s). Doubly so if you need to filter down the ripple.
I was originally going to use a MAX1771; but once again was tempted by the lower price and the DIP package of the MC34063. I was also (perhaps naively) reassured by the number of nixie clocks using the MC34063 in it's PSU.
This lesson repeats time and time again: ;)
Just because everyone is doing it, doesn't mean it's a smart thing to do.
Weekend projects are very frequently that sort of thing. Slap together a nixie clock, or Arduino-whatever, or a Tesla coil, or... Almost none of them have real consideration involved, they're just following what someone else did. But usually making "this looks close enough" compromises that degrade performance in poorly defined ways. It's still useful learning or "doing" experience, but don't mistake it for understanding.
As for modern/ish suggestions:
UC384x -- also bipolar, but somewhat faster (<500kHz), and designed to drive MOSFETs (high efficiency!); peak current mode control; Micrel makes a BiCMOS version MIC38C4x.
UCC3808 -- CMOS, made for push-pull applications, acts like a 3844 with both output phases brought out to driver pins.
Any of a number of integrated and external-switch controllers, from LT, ADI, TI and others. TPS54xxx series are generally good.
I tend not to remember a wide selection of "newer" parts unfortunately, but sticking to the brand names and following appnotes will generally go well.
All of these are more expensive, but inductors and capacitors are expensive, too. You save a lot of space by going to a higher frequency, and not having to dissipate half your converted power as heat!
Tim
-
I just shouted "SOD IT" (exceptionally loudly as well) and bought a cheap PSU (based round the MAX1771 I believe) from some nixie enthusiast.
That will tide us over and allow me to get the clock up and running - and then I can start the building of a proper power supply with better efficiency ;)
Thanks for all your input! It's been very useful - and with any luck will teach me a thing or two when I get round to building my own supply! I'll need to have a good study of power supply design at some point.
If only I went to uni to study all this... I probably would be sick of it and never do it for fun again... so actually starting as a hobbyist is probably a better idea (for me at least!)
-
this is a good example of 'make vs. buy'. for me, it was clear : 'buy'.
its so cheap, its already worked out, why re-invent what is commodity?
figure where your value-add is. for me, the psu is NOT a differentiating factor. the software and features are.
-
its so cheap, its already worked out, why re-invent what is commodity?
Ah, you see my problem is I like a challenge too much - figured I'd take something I don't know a whole lot about and make myself learn a whole lot about it ;) It's much more fun than just reading a book!
Called it wrong on this one though; should have bought a module in the first place and learned about stepping up DC afterwards. Auch well, nothing ventured; nothing gained!
While I'm waiting for the PSU to arrive I'll take another crack at a fake-nixie. I can use the spare MC34063s for the LED drivers. Soldering these 0803 leds is going to be great fun though. Once I finally get a bigger place with dedicated lab space I'll get an oven and try my hand at reflow soldering - just now everything is done by hand. Which works fine 90% of the time, but it does generally limit me to through-hole parts.
I suppose I should also start developing the driver boards for the nixie clock at some point. Might leave that to the weekend though!
-
Did you get the photos and circuit I sent you CP?
-
Did you get the photos and circuit I sent you CP?
I certainly did! :D Helped quite a bit actually! I've tried it and another couple of variants and got between 40 and 76.4V - highest voltage design I found was pretty much the application note with an additional NPN on it before the main switching transistor. As T3sl4co1l mentioned, my choice of ferrite is probably a big part of the problem (amongst other things of course)
Efficiency wasn't all that great on the circuit I had either, I reckoned it was less than 60% (based on power in vs power out (when supplying a dummy load via the step-up converter)) - tested at 40.0V, 60.0V and 74.6V (+/- error in my meter ) - efficiency varied depending on the exact test circuit - I was actually surprised I managed to get it as a high as 60% - I was expecting even less for the first couple of runs due to contact resistance in the breadboard.
I got a cheap...ish board for £10, which seemed to be the going price for many of them on ebay (bought this one - claims to be "professionally designed" - from some nixie enthusiast website that seemed about as reliable as you can get for this sort of thing. Time will tell if my investment has paid off!)
-
PURELY out of curiosity - could a Cockroft-Walton generator with a pulsed DC input be used to provide the necessary voltage for the nixies? (as I said, I've bought a module - was just curious xD)
-
Yes, but: not from a very low supply. You probably need a transformer or inductor to get enough voltage worth starting with. And if you're doing more than a few digits, the current consumption gets a bit unwieldy too.
Reason being, if you need say +170V from +5V, you can generate 5Vpp with something like a MOSFET driver chip (which pulls fully rail to rail with low resistance). You need effectively 170Vpp at the end, so you need a 34 stage multiplier. But each diode in that multiplier drops at least 0.3V (assuming schottky), so you need at least 10.2V to supply all those diodes alone! Instead, it'll peter out somewhere around 50-60V DC.
If you start with 12V, the ratio is much lower, and the total forward drop smaller as well. Tedious, but possible. You still need a substantial current supply, which needs big bypass capacitors to handle the ripple drawn by the thing. Which you'd need anyway for the inductive case, but the inductor will save some size and cost over the stack.
If you use both positive and negative multipliers, you get better performance: you only need +/-85V. Still not likely to work at 5V, but pretty easy at 12V. You do need a level shifter to communicate the digital signals to the drivers down at -85V -- or a pair of big coupling caps to bypass the drive point to the driver, grounding the negative end of the chain. (This might not be too bad of a compromise; electrolytics of that rating aren't a terrible problem, at least if you only need a pair.)
There are other configurations that trade DC voltage across the capacitors for diode drops. Instead of a straight ladder stack, you can use the "half wave" style configuration, where capacitors are used to couple AC voltage to higher and higher doubler stages; because the AC signal isn't being transmitted through the entire diode chain, diode drops subtract linearly (per stage) rather than incrementally. But the capacitors get much larger.
It's noteworthy that you could use an H-bridge driver. But now the ground reference of the multiplier chain is bouncing up and down. A balanced, symmetrical chain might be possible, or if using level translators for the negative end, simply having devices with enough CMRR would suffice (who cares if the negative end is bouncing between -90 and -80V, as long as the signals still get there in order?).
Tim
-
Another very informative reply!
Thanks again - you're a goldmine of infoT3sl4co1l!
-
Auch, you were right about those powder-core inductors - 40 ohm impedance and 1.6kohm resistance at 10khz test frequency. Youch!
Tried a different inductor with a solid core and managed to get 150V out, but the efficiency was in the region of 35%
Anyway! I'll get some pics up soon - I've decided to see if I can mill some PCBs on one of the CNC centres (overkill much?), so I'll get pics of that once they're on the go.
I've also had a test run of the nixie tubes and they're looking good!
Anyway, pics to follow!