You are cutting it close IMHO

You are correct in saying that when you connect those capacitors in series the voltage rating for the resulting stack is 5 x 2.7v = 13.5 v,

You say that since you are using a 12v DCC system that this means the capacitors don't need any additional over voltage protection.  I don't think this is a good idea.

Several points: In practice, designers leave plenty of safety margin here.  A common practice is to use capacitors that are rated to withstand twice the maximum voltage that you expect your design to produce.  Plus, although your supplier says that the capacitors have a working voltage of 2.7v, what happens if a marginal batch gets shipped to you that is 5% below spec, with a working voltage of 2.565?  A stack made of those has an actual working voltage of 12.825v-- getting pretty close.  Your supplier doesn't state the tolerance for its working voltage spec, but if I were betting, it would say it is no better than 5%, and might be 10%.  And don't forget, some low cost suppliers ship product that doesn't fully meet their specs.

In addition, your 12v DCC system is producing a signal with a peak amplitude that is "nominally" 12v or less.  However, depending on the design of your DCC system, it could go higher than 12v during a surge in the AC supply voltage or as a result of other conditions.  In addition, depending on your DCC wiring, there could be ringing on your DCC signal lines that could produce spikes that exceed 12v at the rising or falling edge of the signal.

If I were doing this, I would design for a capacitor stack that has a working voltage rating of 25v, or else add over voltage protection circuitry to your 13.5v stack.  Better to be safe than sorry.

Dennis

Allied examples, over-volt protection

Dear MRHers,

There are a number of related articles here on the MRH Forum (and have been published in the MRH magazine)
which may be useful reference material, esp in investigating Pyg's circuit above

https://model-railroad-hobbyist.com/magazine/running-extra/2018-12/stay-alive-n

https://model-railroad-hobbyist.com/magazine/mrh2019-06/electrical-impulses

https://model-railroad-hobbyist.com/node/31852

https://model-railroad-hobbyist.com/magazine/mrh2020-06/electrical-impulses

https://model-railroad-hobbyist.com/magazine/mrh2019-05/electrical-impulses

I would also note that adding a suitably-spec'd Zener Diode accross Pyg's "series capacitor chain" would act as effective Over-volt protection, with almost no appreciable increase in physical size of the overall package...

(How-to DIY guide C/O Scale Sound Systems.com)
https://f5479436-5972-4fe2-a17a-b4fbd2ea046e.filesusr.com/ugd/7e84c8_35ab6ae3df2843d680250332307fd953.pdf

Happy Modelling,
Aim to Improve,
Prof Klyzlr

On overvoltage protection

Thank you very much for the comments. Let me start with a disclaimer:

I am not recommending not using a Zener diode, I am just describing my setup. If you use a Zener diode, it will add another 2mm to the stay-alive.

Now let me continue with my reasoning: 2mm does not seem like much, but for some reason the width of virtually every Mehanotehnika train cabin is 30mm. Perfect empty space for stay-alives like the one above.

Also, I'd like to say something about low cost suppliers on AliExpress. I have been buying electronics there for six years - admittedly nothing fancy like chips for $100, but elements for up to $5 each. I have only had a problem once. The chip just would not work right. Long story short: After buying chips at Mouser and troubleshooting for two months, I found that the US chip manufacturer's manual was flawed (doh!) and the low cost supplier's chip worked just fine...

Of course, there's a chance you are getting an overvoltage from the power supply - a more unlikely cause is a faulty power supply and much more likely is a lightning strike. However, if I ever decide to play with my locomotives during a thunderstorm and a lightning induced voltage surge occurs in my home, I am pretty sure broken $4 stay-alives will be the least of my worries.

Finally, most stay-alive circuits on the internet contain Zener diodes. If the stay-alives are created with five 2.7 V supercapacitors, the Zener diode is usually 13 V. My impression is that five supercapacitors in series can handle up to 13V, and that the Zener diodes are more of a protection in case someone accidentally runs their locos on a 16V DCC system. In that case, no Zener diode stay-alive definitely means ex-stay-alive.

It seems that the biggest problem with these supercapacitor configurations is the uneven distribution of voltage between all the capacitors. I think that in my situation, putting a large resistor in parallel with each capacitor for voltage balancing would be much more beneficial than using a Zener diode. But maybe I am wrong.

Best regards,

Use an SMT zener.

I use a 13v zener diode in a SMT case.  I have yet to NOT be able to shoehorn my KA in any model including a BLI Trackmobile.

KA in Trackmobile

@Nelson......your quote caught my attention as I have a couple of trackmobiles I hope to get converted to DCC operation plus lighting effects ,...   https://forum.mrhmag.com/post/adding-lights-to-a-bli-trackmobile-12200739

But I don't need, nor want, sound in them,...how does that effect the room for KA in them? (non-electrical person asking)

No sound

No sound in mine Brian.  I would see James Regier's (hope I spelled it correct) as to how he did his.  There are some components that are all SMT that can be used.  But I will tell you that it will run for quite a long ways with out track power. 

Long Run Time?

I don't think I need a long run time,...rather just good running over Peco insulafrog turnouts.

Of course as someone has pointed out I must consider how closely together these potential insulated frogs might be in reference to how long it takes the KA's to recharge to make it across multiple ones?

Voltage

I have to agree to the over voltage protection. Similar situation .... my DCC track voltage is set to 13.8 volts. I used a single 16V capacitor in a passenger car, assuming I had enough headroom. After about ten minutes of running the capacitor exploded.

After considerable time cleaning up the mess it creating inside the car, I replaced the capacitor with an identical value but with a 25V rating. Still working as it should many years later.

Mark. 

Zener protection against permanent and pulse overvoltage

I decided to do some calculations and determine the protection that a 13 V Zener diode provides for 5 x 2.7 V capacitor packs.

Permanent overvoltage

Causes: Putting the locomotive to the DCC system with UU
Effects: IMHO the biggest problem is the variance in capacitance in the capacitor pack. All capacitors in the pack have the same charge, but the capacitor with the smallest capacitance has the largest voltage. Let us calculate the largest possible deviation protected by a 13V Zener diode. The system can be decomposed into a capacitor with the smallest capacitance b * C (b < 1) and the remaining capacitors with an average capacitance of C, where C is the nominal capacitance (statistically, the average tends to the nominal value for many elements). We can write

Q = 2.7V * b * C = (13V - 2.7V) * (C / 4)

giving b = 0.954. So a 13V Zener diode protects against a capacitance deviation of 4.5%. This is really poor protection. If you want to protect the capacitor pack against capacitance variation, I recommend a 12V Zener diode. In this case the maximum deviation is 13.9%.

Pulse overvoltage

Causes: the behaviour of other elements in the DCC system, especially motors and solenoids, lightning surges.

Effects: Supercapacitors are indeed sensitive to overvoltages. However, since they are connected in series with a current limiting resistor (typically 100Ohm), we have a RC circuit with a huge time constant t = R * C. It will take a very long time for the voltage to rise from, say, 12V to 13V. The formula is

t = R * C ln(U₁/U₂)

Where U₁ and U₂ are the voltage differences at the start and the end. For a 14 V overvoltage pulse, U₁ = 2 V and U₂ = 1 V, which gives t = 4.9 s for R = 100 Ohms and C = 70 mF. For 15V overvoltage, U₁ = 3 V and U₂ = 2 V, we get t = 2.8 s and so on.

Normally these overvoltage pulses are only a few ms long, much too short to have any serious effects on the capacitors.

My conclusion: If you use a Zener diode for protection it is effective only for the permanent overvoltage, but you should at least choose a 12V Zener diode to make a real difference.

I will be happy if you can point out the errors in my calculation or logic.

Best regards,

Electronic components have tolerances

Marko-

I agree with you that transient over voltage due to possible ringing on the DCC bus will not be a problem due to the long time constant resulting from your series resistor-- I was incorrect in my original comment when I said that DCC bus ringing could be an issue.

My only comment on your calculations is that a reliable circuit design takes into account the tolerances on the values of the relevant component specifications. 

A 13v zener diode is not necessarily going to clamp at 13.0000 volts. Same goes for a 12v zener.  Some low cost zener diodes have a +/-10% tolerance on the value of the zener voltage.  You can also buy them with +/-5% tolerance.

Capacitors also have tolerances on the value of their capacitance.  The specific 0.35F 2.7v ultra caps that you referenced in your original post do not provide a reference to any data sheet.  The product listing makes no mention of what the tolerance is on the capacitance value.  If you look for 0.35F 2.7v ultra caps on various distributor's web sites (Arrow, Digikey, Mouser, etc.) you see that there is a wide range of tolerances for sale, including capacitors whose C value can be as much as 30% below the nominal value.  Yes, you can also buy capacitors whose tolerance for the value of C is -0% to +50%.  The problem that I see is that without a datasheet, there is no way to know what the tolerances are for your capacitors.

regards,

Dennis

SuperCap Characteristics and Failure Modes

Hi Marko and Dennis,

First:

transient over voltage due to possible ringing on the DCC bus will not be a problem due to the long time constant resulting from your series resistor-- I was incorrect in my original comment when I said that DCC bus ringing could be an issue
Actually...  high voltage transients are an issue, and are an issue that many modelers, even vendors regularly ignore.Supercaps, similar to electrolytic caps, do not behave the same way across the frequency spectrum. They generally loose there effective capacitance as frequency increases. The figure below:

demonstrates effective capacitance falloff as frequencies rise above 1000 Hz in large supercaps. The smaller supercaps we use in modeling exhibit similar characteristics. The RC calculation cited previously would need to account for this, yielding much smaller time constant calculations, rendering the keep-alive ineffective as a high frequency filter.

The typical failure mode of supercaps exposed to over voltage conditions does not usually yield a catastrophic failure, but is based on the high voltage exposed, the duration of exposure, and the frequency of exposure. The actual observed failure is an incremental degradation of the effective capacitance, rendering our keep-alives less "supportive" over time.

Last, I have measured actual DCC overshoot of 207% (29 Volts+) over 60% of a DCC bit time. I have also run simulations where, dependent on DCC bus electrical properties (resistance, capacitance, and inductance, and DCC transition rise and fall times), it it is even possible to generate worse overshoot, under precisely poorer conditions.

The likely result seen over years of use, will be observed "worn out" keep-alives for no apparent reason.

Your mileage will vary. Have fun! 
Best regards,
Geoff Bunza

Peanut Gallery Q : Location of KA Caps in-circuit

Dear KA circuitry luminaries,

A question from the peanut gallery: 

We've had mention of "DCC Ringing/Voltage overshoot" and "Frequency degradation",
but this reads to some as if the Capacitors are being connected directly accross the track/pickup feeds.

My understanding is that this isn't the case, a properly-connected KA is connected _behind_ or _after_ the decoder's rectifier-diode-bridge and onboard Volt regulator stages,

As such (cue "Peanut Gallery" level electronics understanding),

Surely the bridge significantly mitigates "pulsing due to polarity swing",

and the Volt Reg stage minimises the severity (amplitude over time) of "overshoot" conditions?

NB I'm not saying these conditions don't happen, far from it,
and I'm not suggesting eiter Bridge or Volt-Reg are somehow immune from Hysteresis or switching-lag,

but at the Post Bridge, Post Reg point in the circuit,
where the KA is commonly connected,
 are these conditions really as-extreme as "measuring straight accross the rails" suggests?

Happy Modelling,
Aiming to Understand Better,
Prof Klyzlr

Re: SuperCap Characteristics

Dear Geoff,

thank you for sharing your insights with us.

However, I must admit that I do not fully understand your arguments. The time constant R * C refers to DC situations. In my previous post I have modeled a situation where I have a one time voltage surge.

In the case of AC, the relevant quantity is the complex impedance

Z = 1/( 2 * pi * f * C ),

where f is the frequency. So let us modulate this situation. The AC situation could be caused by a DCC signal with a frequency of f ≈ 10 kHz. Let us assume that each voltage step causes a spike from 12 V to 29 V. This would require Fourier analysis, but it is reasonable to estimate that the fundamental has a frequency of 10 kHz and an amplitude of 17 V. Let us also assume that the capacitance is only 1/10 of the nominal capacitance, or only 7 mF. If we put that into the formula, we get an impedance of 2 mOhm! This means that the capacitor voltage will be 12 V plus a voltage fluctuation of

V' ≈ 2 mOhm/( 10 Ohm + 2 mOhm ) *17 V = 3.4 mV

around 12 V due to these voltage spikes. Simply negligible.

Too complex? Not convinced? Let us make a simpler argument. To change the voltage on a capacitor pack from 12 V to 13 V, you have to bring C * (Delta V) = 70 mF * 1 V = 70 mAs, which is a huge charge in electronics. And that should be done across a large resistor of 100 Ohms. Let us assume a current of (29 V - 12 V)/100 Ohm = 170 mA. The voltage of 29V should last for 70 mA/170 mA = 0.4 s, which is much longer than the period of the DCC signal, which is 1/10kHz = 0.0001 s.

OK, that's the theory. The practice is something else. But in practice, supercapacitors degrade even when not subjected to overvoltage. It would be interesting to have two stay alives, one with and one without a Zener diode, and compare the degradation after long use. Theoretically, you should not be able to tell any difference.

Best regards,

Re: Location of KA Caps in-circuit

Dear Prof. Klyzlr

yes, indeed, stay alive is behind the rectifier bridge. Of course, the rectifier bridge still transmits the ringing of the voltage, which as I understand occurs after any voltage reversal. So if I understand Geoff's argument correctly, the stay alive "feels" voltage of 12 V with voltage spikes of 29 V every 0.05 ms (that's half the period of the DCC signal). Yes, that's a huge spike, but it's also a very short spike, cushioned by the fact that the impedance of the 100 ohm resistor is much larger than the impedance of the battery pack, even with reduced high frequency capacitance.

Again, I am arguing only from a theoretical standpoint. In the real world, capacitors degrade even when not subjected to overvoltage. So maybe those voltage spikes can harm supercapacitors, even if simple theory can not predict it.

Best regards,

All Capacitors Degrade

While I have no argument, Marko, with your theoretical calculations, I am very much in Geoff's camp with regard to capacitor degradation particularly when subjected (however briefly) to high voltages.

Any voltage applied across an electrolytic or super capacitor causes both physical and chemical changes in the materials of which it is made throughout the capacitor's life, and the voltage rating of capacitors is determined in part by their predicted life expectancy and their operational reliability based on such changes.

In the military and aerospace areas, where I spent the majority of my professional engineering career, we are only allowed to use such capacitors at a maximum of 40% of their rated voltage, ie. to operate at 12V, the capacitor must have a voltage rating of at least 30V.

While model railroad applications are obviously not subject to such rigorous rules, in my view operating your supercapacitors rated at 13.5V with a 12V supply is pushing them to the limit, and I would not expect them to have a very long life, even without the presence of high-voltage transients.

While I do not have much real experience of building keep-alives, in designing and building other pieces of electronics for DCC applications I will always use electrolytic capacitors rated at 35V - and I would be very reluctant to drop below a 25V rating.

...which prompts the question...

Dear MRHers,

...which prompts the question, how "survivable" are the "build an N scale KA" articles published previously in MRH, using Tantalum and similar SMD caps

https://model-railroad-hobbyist.com/magazine/running-extra/2018-12/stay-alive-n

or "standard" electrolytics?

https://model-railroad-hobbyist.com/magazine/mrh2019-06/electrical-impulses

Happy Modelling,
Aiming to Improve,
Prof Klyzlr

@Prof K, Marko, and Terry re:Discussion

Hi to all,

There are a number of very real, intricate issues we are discussing, for which "enlightenment" is not a simple one sentence answer...

@Prof K:
but at the Post Bridge, Post Reg point in the circuit,
where the KA is commonly connected,
are these conditions really as-extreme as "measuring straight accross the rails" suggests?
The typical KA is connected just past the bridge rectifier, as this is the highest voltage in the decoder and it is used in its raw form, PCM switched, to drive the motors in locos. The same rectified voltage nominally feeds a voltage regulator to deliver the 5 Volts (or so) to the decoder control electronics (typically a microprocessor/microcontroller).

Now we start with the variability. Good design practice without the KA would place an electrolytic cap (35 Volt rated) after the bridge in parallel with a small cap like 0.1uF because the small cap is of different construction (like ceramic, polystyrene, film, mylar, etc.) and has much better high frequency filtering characteristics than the electrolytic. This literally means the electrolytic does not react much to high frequency signals but the smaller cap does, even though it is of much lower capacitance. This is a similar situation to what I described previously with supercaps. Last, good design practice would also place a similar capacitor combo after the voltage regulator (rated for a smaller voltage like 10 Volts) to stabilize the Voltage provided to the decoder controller.

One can find examples where some, perhaps most of what I described is omitted to lower cost or to allow for smaller board/decoder footprint. Model railroad electronics is not the only example where designers cut corners, either intentionally or due to poor design practices. If a good high frequency filter were already on a decoder where the KA was connected in parallel, it would definitely aide the DCC noise suppression for the KA.

Surely the bridge significantly mitigates "pulsing due to polarity swing",
In reality, not likely.

and the Volt Reg stage minimises the severity (amplitude over time) of "overshoot" conditions?
No, the typical monolithic voltage regulators used on decoders are slow reacting to such fast transitions. This is why they are typically spec'ed with input and output caps in their application notes.

are these conditions really as-extreme as "measuring straight accross the rails" suggests?
Typically no, but this depends heavily on how the surrounding circuit is designed. I wish I could give you a simple answer here, but this is not a simple situation. I have ripped apart about five different manufacturers decoders and have built several of my own-- never found 2 exactly the same. This does not represent a majority of those available by any stretch.

@Marko,
3.4 mV around 12 V due to these voltage spikes. Simply negligible
Likely negligible if your base assumptions were correct. They are not. The DCC noise on top the base DCC signal is similar to short "damped harmonic oscillation" provoked by the fast DCC bit transitions (which by the way are not uniform from booster to booster). As such the frequency components that need to be eliminated are 3,5,9,11 and more times the base frequency of the 10KHz frequency you cited. Further, the drop off at these frequencies of the effective capacitance of the supercap is likely approaching 1/1,000,000 not 1/10 the original value, as even at 1000Hz it has dropped to nearly that already.

@Terry
Your design criteria is spot on for high quality commercial and hi-rel design. I bet one could find a few decoders that meet the same standards (QSI might be one) but very likely not most.

@Prof K
...which prompts the question, how "survivable" are the "build an N scale KA" articles published previously in MRH, using Tantalum and similar SMD caps
The 220uF 20 Volt Tantalum capacitors primarily used in the article are a different cap, with different material internal structure. However, from memory of experience years ago, the failure mode for Tantalum capacitors is often a short circuit, sometimes causing significant circuit failure. I do not know if the caps cited are of similar construction. As Terry stated previously, a 35 Volt rating would be better design practice. I worked with several engineering groups where 50 Volt ratings were the accepted minimal standard for Tantalum caps in similar designs, based on board failures experienced by the company.

I hope this is of some value to those inquisitive modelers who actually want to know what is really going on in their DCC decoders. For those who don't just remember "ignorance is bliss!"

Have fun! 
Best regards,
Geoff Bunza

A fascinating discussion.

Thank you gents, an educational dialog. I am enjoying it and learning a lot. Still reasonably far above my level, but I am improved. Really an excellent use of the forum medium. Thank you.

An improved DIY KA circuit

Geoff:

Could you diagram what a well engineered DIY KA would look like?    I am assuming it would be a superCap, resistor, small cap, and zener diode. 

Not so much concerned about a bill of materials but the schematic of what you just described.

You contributions to the MRH DCC forums is enlightening, educational, and entertaining; and very much appreciated.

@Geoff Re: Discussion

3.4 mV around 12 V due to these voltage spikes. Simply negligible
Likely negligible if your base assumptions were correct. They are not. The DCC noise on top the base DCC signal is similar to short "damped harmonic oscillation" provoked by the fast DCC bit transitions (which by the way are not uniform from booster to booster). As such the frequency components that need to be eliminated are 3,5,9,11 and more times the base frequency of the 10KHz frequency you cited. Further, the drop off at these frequencies of the effective capacitance of the supercap is likely approaching 1/1,000,000 not 1/10 the original value, as even at 1000Hz it has dropped to nearly that already.

OK , now I see I misread the graph you shared. I thought there was kOhm at the abscissa

According to your information above, I overestimated the capacitance of the battery pack by 100,000 and underestimated fundamental frequency by 3, which means that new estimate of battery pack impedance is 60 Ohms, comparable to the resistor impedance. Now we are pretty close to palpable negative effect on the capacitor pack, so it would be useful to know the amplitude of the fundamental ringing frequency as precise as possible.  Which is however highly dependent on the quality of the decoder.

But what is the conclusion? Is the Zener diode fast enough to suppress these oscillations? I know from my measurements that the flyback diode does not completely suppress the spikes of the solenoid. Maybe it would be more beneficial to put a ceramic capacitor in parallel with the battery pack?  Or something completely different?

Best regards,

One of the best parts of MRH

One of the best parts of MRH is learning electronics from experts.

@Bob M re: Example SuperCap Circuit

Hi Bob,

The diagram below has all the essentials for a robust supercap-keep-alive:

R8 limits the charging current; D1 schottky diode allows a faster discharge; R1-R7 is for charge balancing the supercap group so that the voltage across each is relatively the same during charging, preventing a single over voltage condition, but also speeding up the self discharge of the keep-alive; C8 together with R8 acts as a high frequency cut off from the DCC bus. There are seven supercaps each rated at 2.7 Volts (in series yielding an effective 18.9 Volt rating, protected by the 16 Volt Zener diode D2. This should give modelers an idea of the issues facing a keep-alive design and what might be added. It should also give one pause to consider what is left out in quite a few postings on the web. Please note that the series configuration for the supercaps is needed due to the low voltage ratings on each supercap. The N-scale Tantalum keep-alive discussed previously, paralleled all the tantalum caps to increase the effective capacitance, since each was already rated at a higher operating voltage. Last, the tantalum caps have no need for the R1-R8 balancing resistors.

Note that the NMRA DCC spec S 9.1  Electrical Standards for Digital Command Control specifies that the max voltage powering the DCC track for HO scale is 22 Volts, and a conforming NMRA DCC HO decoder must withstand and operate with voltages up to 27 Volts. If I did the calculation correctly, it would imply that R1 would need to dissipate 0.9 Watts under worst-case operating conditions. I only point this out because only the SoundTraxx current keeper is rated at 22 Volts and most only at 15-16 Volts (see Mark Gurries useful summary here:  https://sites.google.com/site/markgurries/home/decoders/keep-alive-compatibility  ).  I do not know what protective measures have been included or omitted in most of the decoders listed. However, I suspect that not all keep-alive modules combined with respective decoders likely meet the S 9.1 standard under worst-case conditions.

'Hope this helps. Have fun! 
Best regards,
Geoff Bunza

After all has been said and done

Hi,

this really is an interesting discussion, even though the entry post said it all: "did not use a Zener diode for protection" - which means Pygmalion is acutely aware of the issue.

Personally I found the graph "supercap capacitance over frequency" very enlightening. In the graph, the capacity drops off towards zero at around a few hundred Hertz. In any DCC system there is a fixed base frequency of 10 kHz (iirc) with harmonic frequencies high above that (remember, it's a square wave form). In addition to that there is a high-frequency component caused by the PWM motor control. Unless we're talking about large-scale decoders, there is just not enough "mass" in a decoder to effectively filter frequencies before the KA circuit.

(@Prof, you would have seen the massive capacitors used in typical audio circuits just to filter 50/60Hz mains - usually paired with diminutive caps to take care of the higher frequencies).

With the keep-alive in a loco (or model car) I would think the biggest dangers are risk of fire (overheating) and risk of the caps "exploding" and causing a mess inside the model (below the model, too). Maybe even breaking the model due to expansion of the capacitor.

The circuit with a Zener diode and resistor is prone to overheating if used for "excessive" current, so the resistor has to be substantial enough.

However the danger of a capacitor exploding - I've seen the hole in a classroom ceiling caused by an older electrolytic cap "blowing its cap"... And smelled the smoke, days after. The desk it had exploded upon needed to be replaced. I wouldn't want to imagine what the model looked like after the explosion, nor the layout all around the model.

When faced with a stack of supercaps - I would want the extra protection provided by a few cents' worth of a zener diode. Looking at the picture of the pack, I'm sure there is enough space to add one without changing the overall size. There seem to be voids next to the resistor, between the caps and even on the side where the wire exits. If in doubt cut away the shrink tubing around the resistor - which would aid with heat dissipation (the other issue). It's your model.

With circuits it is like with many things - best practice is there for a reason. But with my younger colleagues I notice that many don't know and have never been taught ("teached"?).

Have fun

What protection does the Zener diode provide?

I understand the need for the drain diode D1, but what is the purpose of Zener diode D2 in Geoff's circuit?