Capacitors. blah blah blah

[peteski]

Probably the key to my success is that I rarely reverse the locomotive.

That sounds like a logical explanation.

[jdcolombo]

John, similar super-caps are used in the keep-alive modules from various DCC manufacturers.  I bough one of the TCS modules to evaluate its construction. It does have a bank of 5.5V rated super-caps connected in series to increase their working voltage to 15V (at the cost of much smaller total capacitance).  There is also a simple diode/resistor circuit which limits the charging current by passing it through the resistor while allowing much higher discharge current (through the diode, bypassing the resistor), to allow enough current to power the motor during power drops.

Ah.  So this is what TCS uses in their KA's, eh?  Hmmm.  The interesting part is that you presumably wouldn't need 15v for an N-scale decoder.  I wonder what the minimum voltage is to keep, say, an ESU Loksound running?  The motor itself clearly doesn't need 15v; heck, I suspect that most of the time I'm running my locos at 6v or less to keep the speeds to something prototypical.   If I remember correctly, I set my LokSounds in my LL Berks to a maximum value of 100 in CV5, which is about 40% of the max.   So the real question would be what is the minimum voltage that will keep the digital circuitry in the decoder running.  I know 11.5 works, because that's the track voltage I use.  So if you could get by with two of these in series for 11v, that would make a much smaller package than a TCS KA-1 or KA-2.

In fact, I was poking around Digikey and found rectangular 3.3v supercaps made by Seiko, .011F (11000uf) that are only 3mm long by 2.5mm wide by 1mm high.
http://www.digikey.com/product-search/en?mpart=CPH3225A&v=728&v=729 

If you "stacked" three of these, the package would be 3mm x 2.5mm by 3mm, and you'd have a third of the total capacitance (say about 3800uf) at 9.9v.  That's a package that would easily fit in most tenders.  Or you could put three of them side-by-side for a package that was 3mm x 7.5mm x 1mm.  Or any of a bunch of other combos - heck you could stash for of them in the corners of the tender wired by 30-gauge flex.

But how to do I keep them from blowing up when charging from an 11.5v track voltage?  Is this possible?  Am I crazy for thinking this way?

John C.

 

[VonRyan]

Probably the key to my success is that I rarely reverse the locomotive.

This.


With small Nn3 layout not built for switching maneuvers, reversing is pretty unimportant.
I only installed a direction toggle on the throttle for Killashandra so that I can run trains facing the opposite direction for filming purposes.

[peteski]

Ah.  So this is what TCS uses in their KA's, eh?  Hmmm.  The interesting part is that you presumably wouldn't need 15v for an N-scale decoder.  I wonder what the minimum voltage is to keep, say, an ESU Loksound running?  The motor itself clearly doesn't need 15v; heck, I suspect that most of the time I'm running my locos at 6v or less to keep the speeds to something prototypical.   If I remember correctly, I set my LokSounds in my LL Berks to a maximum value of 100 in CV5, which is about 40% of the max.   So the real question would be what is the minimum voltage that will keep the digital circuitry in the decoder running.  I know 11.5 works, because that's the track voltage I use.  So if you could get by with two of these in series for 11v, that would make a much smaller package than a TCS KA-1 or KA-2.

I'm not sure if I understand what you are trying to accomplish. Are you willing to lower your overall track DCC voltage to under 11V?  Various sound decoders seem to need around 6-10V to start functioning.  That is in DC, but I assume similar requirement in DCC.

I would really not recommend using anything lower than 15V rating (even if you know your track voltage is lower). What happens if you forget and run the loco on someone else's layout? Or if you sell the loco and again, forget about your reduced-voltage caps installed in it. 

[jdcolombo]

OK.  Let's see if I can clarify my thinking (maybe not  .

Let's suppose that a DCC sound decoder running the sound circuits and motor needs a minimum of 9v at .3 amps to run.  When this decoder gets 12v from the track, it runs it through a bridge rectifier. After the bridge rectifier, that 12v becomes, say, 10.5v (I'm assuming a 1.5v drop from the bridge, but I realize it is likely to be less).  The 10.5v is then split into different paths; the full 10.5 is available to run the motor; the other path is the digital circuitry and sound amplifier - I'm assuming that there is a voltage regulator circuit in there somewhere to provide a steady 3.3v or whatever to the digital circuitry and sound amp.

When you connect a keep alive to a DCC decoder (any DCC decoder, including just a motor decoder) you wire it up just after the bridge rectifier, but before any other circuits.  So under my hypo above, there's say 10.5v to charge the keep-alive.  When power is interrupted, the keep alive discharges.  If the keep alive has been charged with 10.5v, then it has a certain amount of electrical energy available.  How long it can keep the decoder running depends on how much storage (capacitance) is available and the voltage and amperage required by the decoder.  Like a battery, as the cap discharges, voltage drops, until at some point it can't supply the voltage needed to run the decoder.

Now because I'm not an EE, I don't know how to calculate how long a capacitor can supply a given voltage at a given amperage with a particular capacitance and charge.   For example, suppose you have a 1 farad cap charged to 10.5v.  How long can that cap supply 9v at .3 amps?  However long it can do so is how long you have keep-alive under my initial assumptions (that the decoder needs 9v at .3 amps to keep the motor and sound circuits running in a typical application).

OK.  Now the working voltage of a capacitor (or bank of series caps) is the outside limit of the voltage you can apply to charge the cap.  If you have a 16v cap and try to charge it with 20v, it blows up.  But for safety reasons, you never want to charge a cap with its full working voltage.  Maybe you want a 20% safety factor, so instead of charging the 16v cap with 16v, you charge it with 12.5v, which is about a 20% safety factor.   Now notice, this works out pretty well for most model rr applications.  Say the track voltage is 14.4v, and the voltage out the bridge rectifier is 13v.  If you're charging a 16v cap with 13v, you have some safety range (as a practical matter, you might want to use a 20v cap for even MORE safety range, but bear with me on this - I'm just hypothesizing).

Now when that cap discharges, it can supply 13v for some (tiny) period of time depending on it capacitance and the current draw.  It presumably can supply 9v for a slightly longer time.  Once it can no longer supply 9v, the decoder quits working, the sound drops and the motor stops.

With that background, let me get to my main point.  If you need 9v to run the decoder, there are all sorts of ways to get there.  You can have a cap that's charged to 13v or 20v or whatever, but you don't actually need 13v.  What you get with the 13v charge is a longer time the cap can supply 9v at a given amperage draw.  So suppose instead that you charged a cap to 10v.  That cap can also supply 9v for some time (not as long as an equivalent capacitance cap charged to 13v).  But you can expand the time available by upping the capacitance.  Again, because I'm not an EE, I can't calculate what the exact relationship is, but there must be such a relationship, because the cap is just storing energy; the larger the capacitance, the more energy it can store, just like the larger the charge voltage, the more energy it is storing.  This is why higher voltage caps are physically bigger, and higher capacitance caps are also physically bigger  (I think - again, because I'm not an EE, I might be thinking about this completely wrong).

So let's say you had a bank of supercaps that had a 13.2 working voltage (four 3.3v caps in series) and a total capacitance of about 12000uf (1/4 of 47000uf).  You charge that bank of caps with 10.5v coming from the bridge rectifier.  Presumably, that bank of caps can supply the 9v at .3 amps for some period of time.   My first question here would be "how long" (there must be formulas for figuring this out).

But here's an even more important question.  I want to make sure that the cap bank is charged at no more than 10.5v.  If I set my track voltage at 12v, and I'm sure that there's not more than 10.5v coming from the decoder's bridge rectifier in these circumstances, I'm golden.  But I'm worried.  Maybe I accidentally select the HO setting on my Chief, which ups the track voltage to 14.5v, and now there's 13v coming from the rectifier.  Or maybe I take my engine to someone else's layout and they have a track voltage of 15v.  Not good.

So the question I was trying to ask earlier is whether there is a simple way to make sure that I don't charge the cap bank with more than 10.5v?  If I insert a diode on the charge line from the bridge to the cap bank, then I get a .6 or something voltage drop.  That helps.  Two diodes in series gets me 1.2v   Of course, the problem is that I also don't want to charge the cap with much less than 10v.  So maybe I've answered my own question - there isn't any easy way to do this short of using some kind of voltage regulator circuit, and that would take up too much space; I'd be better off in these circumstances just using a capacitor bank that has a 15v working voltage.  Which is probably why the TCS KA's are designed the way they are - a 15v working voltage, with a bunch of supercaps in series that gives you a high total capacitance even after taking into account the series capacitance loss.

What I was hoping was that I could use, say, 4 of these 3.3v supercaps in series to give me a working voltage of 13.2v with something like 10,000uf total capacitance, and use about the same space as 4 220uf, 16v tanatalum caps.  I'd presumably get a lot more run time from this arrangement.  This would probably work fine as long as my track voltage is set to 11.5, where it is now.  But you're right that I'm playing with fire if I put this engine on a layout running track voltage at 14.5 (or even higher in some cases).  And there doesn't appear to be any easy way to guard against this other than adding a fifth cap in series, which (1) takes up more space and (2) reduces the total capacitance.

Hmmmm.   Will have to think more about this. I probably need to take a "Electronics for Dummies" course . . .

John C.

Late Edit:  According to this page: http://www.members.optusnet.com.au/mainnorth/alive.htm a TCS KA-1 has 36,667uf of capacitance at a working voltage of 15v, while the KA-2 has 200,000uf at 13.5v  If TCS thinks 13.5v working voltage is fine, then it seems to me that 4 series supercaps for a working voltage of 13.2 would be OK, too.  4 of the Seiko smt's wired in series would give me 2750uf of total capacitance, which isn't huge, but it is a significant improvement over the 880uf I get with 4 16v tantalums wired in parallel.  And the cost is about the same ($2 per cap).  I might try this with one of my steamers and see what happens . . . I'll wear goggles 

   

[SandyEggoJake]

This wouldn't work for a regular N locomotive because every time you tried to stop at a certain spot the loco would want to keep going a little further.

Chris, I've been considering the spotting issue you've hinted at.  It is an issue I've been thinking about some as I'm building a scratch 0-6-0 fireless (thus without a tender) in N scale, and wondering how to best stuffing it so that it can work somewhat flawlessly over a complex yard environment. 

Obviously, accurate spotting would be required should you wish to add operations (filling and/or emptying of carts) to your action.  Given it seems in the video yours is a dedicated loop with a battery powered loco plus the cap, if the battery was also located in the train, you might be able to use a second independent circuit (power via the track) to manage stopping via a reply that governs the current between the cap+battery to one lead of the motor.  And if you really wish to get fancy, a relay (or replays) could be selected that govern an array of resistors between these leads, affording you stepped speed control as well.  Toss in some simple logic for timing of these replays and you would open the door to having automation on this track that takes it beyond a simple roundy round.

[SandyEggoJake]

John C,

Avoiding your question on an equation for time to capacitance discharge (I'm not an EE either), and focusing on your assumption that "there doesn't appear to be any easy way to guard against (over charging of capacitors)" you may wish to explore the various ways to use cheap & super small voltage regulator available from use of a zener diode in reverse bias. 

-Jake

[peteski]

John,
looks like you have this super-cap keep-alive thing pretty much figured out.  Now that it has been mentioned, I think that the TCS unit I opened up did have a Zener diode giving some over-voltage protection to the caps.  The diode works together with the charging current limiting resistor to protect the caps.

You should be able to build a circuit similar to the TCS unit (when I have some some time I can draw a diagram) while using your choice of caps.  As far as trying to figure out the length of time the circuit will be able to supply high enough voltage to keep the decoder and motor alive through the dead spots, I thing it might not be worth trying to calculate that.  Just going by the drastic increase in the available capacitance, you can tell that it should supply sufficient power to keep the model running through the dead spots (even if your overall DCC track voltage is lower than the standard 13V.  The website you pointed to does have extensive info on where to attach the super-cap based unit on various sound decoders.  That makes things easier too.

I don't think you need to wear goggles.    While super-caps are sensitive to over-voltage and reverse-bias, they aren't nowhere as dangerous as tantalum caps. In most cases they will simply leak the electrolyte (unless you were to really abuse them).  Plus, if you add the voltage and current limiters then even if the caps were overstressed, the result would not be explosive.

When tantalum cap goes into a failure mode, that causes a chain reaction - hence the explosive results.  Super cap constructed more like a conventional electrolytic cap. When it fails, the electrolyte leaks or evaporates and its internal resistance will increase (which will reduce the failure mode current). So the cap will fizzle out rather than burning up.

[jdcolombo]

John,
looks like you have this super-cap keep-alive thing pretty much figured out.  Now that it has been mentioned, I think that the TCS unit I opened up did have a Zener diode giving some over-voltage protection to the caps.  The diode works together with the charging current limiting resistor to protect the caps.

You should be able to build a circuit similar to the TCS unit (when I have some some time I can draw a diagram) while using your choice of caps.  As far as trying to figure out the length of time the circuit will be able to supply high enough voltage to keep the decoder and motor alive through the dead spots, I thing it might not be worth trying to calculate that.  Just going by the drastic increase in the available capacitance, you can tell that it should supply sufficient power to keep the model running through the dead spots (even if your overall DCC track voltage is lower than the standard 13V.  The website you pointed to does have extensive info on where to attach the super-cap based unit on various sound decoders.  That makes things easier too.

I don't think you need to wear goggles.    While super-caps are sensitive to over-voltage and reverse-bias, they aren't nowhere as dangerous as tantalum caps. In most cases they will simply leak the electrolyte (unless you were to really abuse them).  Plus, if you add the voltage and current limiters then even if the caps were overstressed, the result would not be explosive.

When tantalum cap goes into a failure mode, that causes a chain reaction - hence the explosive results.  Super cap constructed more like a conventional electrolytic cap. When it fails, the electrolyte leaks or evaporates and its internal resistance will increase (which will reduce the failure mode current). So the cap will fizzle out rather than burning up.

OK - I'll leave the goggles off.    A wiring diagram would be really helpful; I'm not sure I understand the zener diode thing.  But I'm definitely going to try the supercaps in my MRC/Model Power Mikado, which is still a tiny bit pickup-challenged despite having 880uf of tantalum caps inside.   The Seiko supercaps are the same size, so I should be able to fit them easily inside the tender.   Off to NYC for a conference, but this is on next week's agenda.

John C.

[SandyEggoJake]

I'm not sure I understand the zener diode thing. 

Me either!  Involves electron quantum tunnelling which hurts my brain and leaves me worried (and at the same time not) about Schrodinger's Cat!

But it's a cool phenomena that results in a very useful and cheap component with a very small footprint.   

The Bachmann Spectrum 4-6-0 uses one in reverse bias in its headlight PCB01 (N811X) in parallel with the LED.  While not used here for capacitor overcharge protection (and I'm not certain if/or how such might work in such an application) but rather simple voltage protection of an LED.  In that application, the backwards zener does nothing till the circuit sees too much voltage for the LED ... but then above the Zener's so called breakdown voltage, it "opens" and creates a shunt that allows an alternative path to the dump excess above that breakdown voltage.  Not sure if it's an accurate analogy, but I kind of think when a zener is used in a reverse bias, it can be used in a way similar to a "pressure release valve". 

You may wish to check out tutorials on the subject such as here http://www.evilmadscientist.com/2012/basics-introduction-to-zener-diodes/ and here  http://www.electronics-tutorials.ws/diode/diode_7.html

[VonRyan]

What if I just want to change the capacitors on the sound decoder of an Atlas S2 to ones that are actually effective, what should I use?

Even with clean wheels and clean track, it likes to find every atom of dirt and sputter, just to spite me.

I've been thinking about upgrading the speaker in mine, so I figured I should also upgrade the capacitors to effective ones while I have it opened up.

[peteski]

What if I just want to change the capacitors on the sound decoder of an Atlas S2 to ones that are actually effective, what should I use?

Even with clean wheels and clean track, it likes to find every atom of dirt and sputter, just to spite me.

Cody, I have one of these locos but I haven't had a chance to open it up yet, but yes, you will most likely be able to add additional capacitor to the decoder.   Unfortunately, larger value caps are also physically larger and (with a sound decoder and a speaker already in there) there isn't much free space to install the cap.  So I'm not sure if this will be a viable solution to your problem. This loco shouldn't have *THAT* much problem with electrical dropouts. Maybe there is something wrong with it - something that needs to be repaired, cleaned, or serviced?  I know it is a new loco, but we all know that this doesn't guarantee that it is perfect or in optimal running condition. 

[peteski]

John,
here is a sample diagram of a keep-alive circuit.



The circuit I evaluated seems to use diodes with a very low Vf (around 0.2-0.3V).  Most likely Schottky diodes.  The Zener diode is either 12 or 13V rated diode.  Cx was a bank of five 1F 2.7V connected in series.

When charging, the current travels through D1 and R1.  R1 limits the inrush current, but its resistance is low enough to charge the Cx within several seconds.  When the circuit is discharging, the current is traveling in a reverse direction, through D2. These diodes should be able to withstand a total current of all the devices which will be consumed by the decider, motor and active function outputs.  For N scale I should say use at least 250mA rated diodes.  Or maybe even 0.5A

If the charging voltage is too high then the Zener Diode will start conducting (clamping the voltage down to its rated voltage). The extra voltage will be dissipated as heat across R1.  A 1W rating for the Zener diode should work (judging by the size of the diode in the circuit I examined, I think they used a 0.5W diode).

[SandyEggoJake]

Peteski's diagram looks good to me - but I repeat - I'm no EE. 

The key will be to select a zener that has a zener voltage (aka breakdown voltage, which thus gates the voltage applied to the cap array) less than the working voltage of the cap array.  If so, as long as the zener if fine (and in reverse bias) you cannot overcharge the caps.  The array will only charge up to the level of this applied voltage.     I think.   

[peteski]

The diagram is done by me by analyzing a keep-alive module being sold by one of the DCC companies.  It better work!    But I didn't unsolder the diodes to precisely check their electrical attributes (I just took some measurements in-circuit).