dc coil surge supression 11

[electricpete]

I was asked to provide a “quick” (i.e. not my real job…helping someone else out) general recommendation for dc coil suppression for relay coils powered from 125vdc. . There are about 30 different coils in the cabinet, all powered by 125VDC and all drawing about 0.25A or less. Various problems have been experienced that are believed attributable to voltage spikes from coil switching. Response time is not critical in this application… 1 sec delay would not hurt anything. It is important for relays to change state reliably and even more important not to short out the dc power supply if the surge suppression device fails.

I did a quick search and it has been discussed many times on eng-tips.
Also I found:

http://relays.tycoelectronics.com/appnotes/app_pdfs/13c3264.pdf

http://relays.tycoelectronics.com/kilovac/appnotes/fig48.asp

http://www.sprecherschuh.com/library/techdocs/get/TECH_Surge_Suppression_109.pdf

I’ve read through the above and formed my own conclusions, submitted for your comments.

I think the 2 most common discussed options are:
1 – flyback diode
2 – varistor.

The first link especially seems to push the option of varistors. They highlight a concern that flyback diode can make the coil be so sluggish that it might not even operate. Also apparently when it operates slowly, it’s output contacts can be degraded. I have to admit I have not heard much about these concerns before (other than time response).

I really don’t like varistors in this application. I think every time the coil switches open there can be fairly high current at high breakdown voltage drop and these things degrade. Sure there is a rating, but they use a little life every time they cycle. If they ever short circuit, life is not good.

So I like the flyback diode. Diodes can short, but then again I have an easy solution: put two diodes in series…. Makes me feel a lot better. Either diode can fail short and not a problem. Also I’m thinking I would put a resistance about equal to the coil resistance (R = 125V/0.25A = 500ohms) in series. That should tend to minimize concerns about effect of slow field collapse upon the relay discussed in the first link. I picked 1 times coil resistance since when the full coil current switches into flyback loop the voltage is limited to the original 25VDC voltage (if I had double the resistance I could have double the voltage).

In summary I am thinking about two reverse-biased diodes (during normal operation) and a resistor 1x coil resistance... all connected directly in parallel with the relay.

What do you think? Am I grossly overlooking anything?

By the way, any tips for diode selection?

=====================================
(2B)+(2B)' ?

 

[Skogsgurra]

Pete and sibeen
The measurements were made with one objective, to see what a diode/resistor combination does to contact separating speed. So, the resistor value could have been chosen better. I have one measurement with a 470 ohm resistor (twice the coil resistance). I shall use that instead of the 4700 in a revised PM. That will reduce kickback to more reasonable values.

Pete
Most of the energy in the collapsing field is consumed during the arcing period. That's why the usual L/R relation doesn't hold.

This seems to be the start of an FAQ where AC and DC inductive loads and suitable snubber techniques will be treated. I got plenty of time (will not go to El Salvador) so this is the right time to do it. Stand by!

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[sibeen]

Gunnar, looking foward to it.

Can we expect a book?

 

[btrueblood]

Very cool stuff, Gunnar.

Yes, the little wiggles in the curve do show the opening and closing times of the armature. It's just like DC motor theory - you have a field, however transient it is, and a moving magnet. Speed of motion of the magnet drives a back-emf thru the coil, thus generating the blips in a voltage trace across the coil. In the spacecraft biz, for measuring solenoid valve opening and closing transients, I used to measure the drop across a low-ohm resistor in series with the coil to see those blips, and we would make the assumption (quite well correlated by actual pressure/flow measurements) that they corresponded to the start and end of solenoid armature motion.

Gunnar writes:

"So, the slow flux decay does actually hold the armature back a little. If that difference has any effect on contact destruction or not is an open question. Any thoughts?"

I'd have thought the answer to this would be "it's negligible"; delay in dropout yes, and measurably so, but transfer speed of the contacts should behave from first principles of force balance independently from a dissipator linked to the coil. But experience has shown me to be wrong, for certain types of contacts and high-amperage DC. And after think about it more, there could be an effect. Yes, the first motion of the armature should occur at a fixed value of the coil current (all other variables equal). But, if a dissipative device is continuing to cause the coil current (and any back-emf current from the armature motion) to fall, one could argue that this dissipation lowers magnetic forces during the armature motion, and thus elimates the hold-back force, ie. more dissipation means more acceleration of the armature, equals higher transfer speed. Somewhere in an Omron data sheet (of which there are many thousands) I think I read a description of what we are all talking about, and it had recommendations against a simple freewheeling diode (and for something with more back-emf effective resistance) to reduce relay transfer times and subsequent arcing. Will search for it...wish me luck.

 

[btrueblood]

Well...

an appnote from Tyco:

http://relays.tycoelectronics.com/appnotes/app_pdfs/13c3311.pdf

has this to say on page 1:

"Even though the use of coil suppression is becoming more significant, relays are normally designed without taking the dynamic impact of suppressors into account. The optimum switching life (for normally-open contacts) is therefore obtained with a totally unsuppressed relay and statements of rated electrical life are usually based on this premise. The successful "breaking" of a DC load requires that the relay contacts move to open with a reasonably high speed."

A typical relay will have an accelerating motion of its armature toward the unenergized rest position during drop-out. The velocity of the armature at the instant of contact opening will play a significant role in the relay's ability to avoid "tack welding" by providing adequate force to break any light welds made during the "make" of a high current resistive load (or one with a high in-rush current). It is the velocity of the armature that is most affected by coil suppression. If the suppressor provides a conducting path, thus allowing the stored energy in the relay's magnetic circuit to decay
slowly, the armature motion will be retarded and the armature may even
temporarily reverse direction. The reversing of direction and re-closing of the contacts (particularly when combined with inductive loads) often leads to random, intermittent "tack welding" of the contacts such that the relay may free itself if operated again or even jarred slightly"

a little later it says:

"The use of a reversed-biased rectifier diode in series with a zener diode will provide the best solution when the relay can be polarized. This suppression is often recommended by Siemens Electromechanical Components (SEC) for use in automotive circuits. The impact on release dynamics is minimal and poses no loss of reliability"

and

"A reversed-biased rectifier in series with a resistor may be used
successfully with some relays when maximum load switching capacity is
not required. Care must be taken to use a resistor large enough in value to quickly dissipate the relay's stored energy but yet stay within the desired peak voltage transient. The required resistor value may be approximated from the following equation:
R = Vpeak/Icoil
where;
R = resistor value in Ohms
Vpeak = peak transient voltage permitted
Icoil = steady-state relay coil current

I had to think about that statement for awhile, but am now understanding why they say it after re-reading the whole paragraph, and thinking about Gunnar's results.

They then have this to say about a diode only:

Many engineers use a rectifier diode alone to provide the transient
suppression for relay coils. While this is cost effective and fully eliminates the transient voltage, its impact on relay performance can be devastating. Problems of unexplained, random "tack welding" frequently occur in these systems. In some applications, this problem is merely a minor nuisance or inconvenience and the controller or operator will cycle the relay until the proper response is obtained. In many applications; however, the first occurrence may cause a complete system failure or even present a hazardous situation. It is important that these systems be designed with another method of relay suppression.

And I think I agree with that too.

They then give a little table to show the measured effects of various suppression methods. Note the difference in dropout times for a 24v reverse-biased zener, versus a 100 ohm resistor (both giving transient suppression of about the same voltage peak values).

 

[electricpete]

A few questions:
1 – What type of coil? Clapper type telephone coil?
2 – What is coil resistance?

I agree smaller resistor will be more representative of intended application. Ratio of peak overvoltage to nominal voltage is roughly the same as ratio of snubber resistance to coil resistance.

For all cases except for the diode-only case, we can infer the current from the voltage trace. For the diode-only case, it would be interesting to provide some trace that could be used to infer the current... like voltage accross a very small shunt resistor.

Looking at previous traces, I still say they raise some unanswered questions that we would like to understand to really know what the heck is going on in the circuit.

Regarding the difference in L/R time constant, look at the interval where voltage decays down from 60 to 20 volts. From the qualitative shape of each individual curve, my guess would have been that the arcing is already completed and the gap in the magnetic circuit has not yet opened. What we expect and what we appear to see in that region is an exponential decay following the rule exp(-t*R/L)

For slide 2 it takes about 0.002 seconds to decay from 60 to 20 volts
exp(-t*R/L) = exp(-0.002 * 4,000/L) = 20/60.
L = -0.002*4000 / ln(20/60) = 7.28H (!)

For slide 3 it takes about 0.0011 seconds to decay from 60 to 20 volts
L = -0.002*2E6 / ln(20/60) = 3640H (!)

Have I done this calc wrong? Not only does the same L act vastly different, but the inductance seems unrealistically high. Also if we consider magnetic non-linearity, the permeability is highest at normal operating point presumed slightly below saturation and monotonically decreases as we decrease from there which suggests slide 3 should have lower L (opposite of what we saw).

And again, the dropout "voltages" are the same which suggests vastly different dropout currents. That suggests that magnetic force as function of current doesn't tell the story and paints a picture of a more complicated dynamic system that I can't quite fathom. I think I can piece together a simple mental model to explain why the coil moves slower with lower snubbing resistance, but I can't get anywhere close to piecing together an explanation for these measurements.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Going to figure 3 of the very first linked Tyco article, they show the rectifier diode and zener used in place of our resistance.

The reason for the big advantage of the zener is now clear. For resistance, we are prevented from using a high resistance because it gives a voltage spike at the moment just after input switch opening. Let's say we use Rsnubber = 150% Rdc, then we get roughly peak voltage of 150%. In contrast, let's say we use zener of 150% voltage. The effective resistance of the zener just after switch opening (when we want it low) is also 150% which is in accordance with the voltage limit. But in the time after that when voltage spike is no longer a concern, the effective resistance of the zener continues to increase, drawing maximum power out of the coil for that voltage. So the zener is the best characteristic we can achieve that satisfies maximum energy removal rate at voltage limited to 150%

Sorry if this is already obvious

=====================================
(2B)+(2B)' ?

 

[Skogsgurra]

BTB

..which proves that also simple systems can be quite complex and that there are no trolls in technology, but sometimes badly understood mechanisms. I shall now put an inductive load on that relay contact and measure voltage and current and then multiply the two to get instantaneous power and also integrate the two over time to see how much energy is developped in the arc and how it depends on contact seoarating speed. A nice project for a rainy winter day.

Pete
I have added one recording with a 'better' resistor. This one is twice the coil resistance. Does it make more sense?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[btrueblood]

"So the zener is the best characteristic we can achieve that satisfies maximum energy removal rate at voltage limited to 150%"

I think that is right, and it was obvious (to me) only after thinking about it for several years. Good that those several years were several years ago, now I can sound smart...

"I shall now put an inductive load on that relay contact and measure voltage and current and then multiply the two to get instantaneous power and also integrate the two over time to see how much energy is developped in the arc and how it depends on contact seoarating speed. "

That sound like fun...or maybe the short winter days are getting to me too!

 

[electricpete]

Thanks Gunnar. I will take a look.

We contrasted 2 approaches:
1 - I liked the rectifier diode + resistor because it was robust to failure of the rectifier diode.
2 - I like the rectifier diode + zener because it seems better at quickly deenergizing the coil, which might help it switch faster, which would be easier on any output NO contacts (which is an unknown risk...probably depends on the load fed from the contacts, whether it's inductive, whether it has its own surge suression etc... better to be safe in absence of details on that).

We could probably combine the advantage of 1 and 2 above each design just by inserting an additional rectifier diode into (1), which will give us two rectifier diodes plus zener all in series... gives us advantages of zener (fastest deenergization) plus robust against single failure of a rectifier diode shorting out the power supply.

I shy away from proclaiming this is an "ideal" configuration because although I don't know much about it, the field of "snubbing" ;-) seems to be something well studied with quite a variety of approaches... perhaps motivated by subtle differences in specific application requirements that I am unaware of. FWIW, yet a few more or listed in the attached table excerpted from "Electrical Relays: Principles and Applications" by Gurevic (note some are for dc and some are for ac as indicated by circuit symbol on left of diagram).

=====================================
(2B)+(2B)' ?

 

[electricpete]

Sorry. I need a correction in bold:

"We could probably combine the advantage of 1 and 2 above each design just by inserting an additional rectifier diode into (1), which will give us two rectifier diodes plus zener all in series... gives us advantages of zener (fastest deenergization) plus robust against single failure of a rectifier diode shorting out the power supply."

should have been:

"We could probably combine the advantage of 1 and 2 above each design just by inserting an additional rectifier diode into (2), which will give us two rectifier diodes plus zener all in series... gives us advantages of zener (fastest deenergization) plus robust against single failure of a rectifier diode shorting out the power supply."


=====================================
(2B)+(2B)' ?

 

[2dye4]

250 Volt Zener in series with 1n4007 diode.

The varistor is not reliable long term and the clamping voltage is important for turn off time.

wire so that the inductive kick current flows through the zener in its breakover mode through the diode in its normal conduction mode.

 

[OperaHouse]

Besides overworking the solution, I think you guys have veered away from the actual problem. In a suppression solution there are three areas of concern: 1. Protecting the source from over voltage, either arcing of contacts or spikes to a semiconductor. 2 Limiting the speed of the device, 3. Controlling radiated emissions. You have focused on the first two but the primary problem of the customer is the last.

 

[electricpete]

Thanks OH.

All I know about the original problem was that there was mis-operation in a 24 vdc circuit during switching of a 125vdc system coil. These two systems are very independent electrically (125vdc is rectified outside the panel and supplied to the panel, whereas the 24vdc is produced from ac inside the panel). So I suspect you are right, the interference is probably not conducted, but coupled through either radiative coupling or capacitive coupling.

Aside from general EMI procedures in other parts of the circuit/wiring, how specifically should we incorporate this concern into our coil input surge suppression design? I can imagine perhaps it might be important to know how fast the zener transitions to reverse conducting?... or perhaps this is an advantage to the resistor (vs zener) ? .... or completely other design suggested?

=====================================
(2B)+(2B)' ?

 

[OperaHouse]

That doesn't sound like a very sensitive application. I am always concerned about contacts when switching 125V DC. An "open" contact continuing to conduct and causing two exclusive functions to operate at the same time overloading the line. That may have caused the 24V supply to drop out. I remember one manufacturer that used a 25A analog supply just to power one relay because of short period line dropouts. A large capacitor (20,000uF) on the 24V line might be a solution. I would be examining all the relay contacts for some more evidence of a logic malfunction. Just get the feeling your solution may not come easily. Is there any other electronics in there besides that 24V power supply? I assume this has worked for a while without a problem or is the 24V section a recent addition?

 

[OperaHouse]

Adding to the other discussion. I used to work for a company that manufactured all forms of suppression devices MOV, RC, diode/resistor, and combinations. Low tech but it was the profit end of the company vs the custom electronics which barely broke even. I got involved with a lot of exotic customer problems. Business got even better with the increase in variable speed drives. Few manufacturers cared about what noise they dumped onto the mains. Like rivers, it was thought they could handle whatever you threw in. That was true back when there were lots of incandecent light bulbs lamps to dampen the noise. Now I recommend RC networks on the power lines to absorb noise. One customer had a photo cell module that had a simple phase angle power supply. Normally that would work fine except he had 50 of them on the power line and would fire at the same point on the sine wave. A bunch of RC networks on the power line solved that. A machine tool manufacturer had noise in the cabinet from a couple of VFDs. I mention this because the line noise was so bad that we had to use real polypropylene 2uF snubber caps that were rated at 5A at 10KHZ. The polyester caps they tried before were just cooking. The only way to get rid of noise on a power line is to dissipate it in a resistor. Otherwise it will just bounce around the entire factory on the power lines.

Many years ago we built an automated inspection machine where the 24V solenoid lines had to run unshielded within the signal cable. The lowest noise resulted when a resistor is added across the load along with a diode. Remember ringing happens in both polarities. I understand the the automotive industry is now using resistors instead of diodes for high reliability. These draw power but quickly dampen ringing. A RC network gets around the constant power draw. Another advantage the RC is since often many relays are on at one time the networks are also absorbing power line spikes from other sources.

If there are no solid state devices driving the relay, a RC network is another valid solution. Most use X2 rated capacitors which are internally two capacitors in series for high withstand voltage and reduce internal corona effects. The foil is self healing and almost all failures (the weld to the foil) result in an open. If building your own, 1/2W carbon composition are suitable but getting harder to find.. These can handle repeated surges over time. Never use carbon film which will open over time even though there is no heating. Many metal film resistors have surge capability.

 

[Skogsgurra]

I do not know if this is of any help. But I made a presentation back in 2002 to describe some of the problems with relay coils to people that never really understood what was going on in their factory. I did a crude translation to English. It can be found here:

http://www.gke.org/presentationer/files/The_AC_DC_coil_arcing_EMI.pdf

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[OperaHouse]

We had one of those interference blasters. You forgot the most important part..... It has to be glued onto a big piece of foam rubber! Even then it would hop off a table.

 

[Skogsgurra]

Yes, they walked away. Not exactly silently..

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[electricpete]

I am always concerned about contacts when switching 125V DC. An "open" contact continuing to conduct and causing two exclusive functions to operate at the same time overloading the line. That may have caused the 24V supply to drop out

That possibility was examined during troubleshooting, the output contacts from the 125vdc coil were not associated with any 24vdc loads.

Interference blasters looks like an interesting toy. The whole idea of doing a survey for emi is interesting. I know Doble has a lot of fancy equipment they use for that (developed by Jim Timperly).

======

As an aside, attached is a spreadsheet that I put together as a followup to of Gunnar’s relay test graphs. Results were not earthshaking… pretty much as expected (although the parameters were SWAG’d).

Tab “model” is textbook model of a simple coil with slug, including a speed voltage term. I pretty much stuck to that model, except I omitted damping.

Tab “main” has inputs (green cells) and controls (grey buttons) if you want to run a simulation for yourself.

Tab “plotsheet” is updated to show the results each time a new simulation is run. You can adjust the appearance using green cells in plotsheet.

The simulation starts after switch has opened and we have an “initial” current that will decay.
The simulation continues until time Tstop (set to 0.05 sec). However any data after the displacement reaches top of the graph (corresponding to 0.005m) should be disregarded… I did not try to model the mechanical interaction with the open contact…so the simulation just keeps on going and eventually overshoots the anchor position of the main spring, then oscillates back.

Tab “comparison” shows comparison of results (cut/paste graphic from plotsheet into comparison) using total resistance (external plus coil) of 4000 ohm, 1500 ohm, 500 ohm. The wiggle is there, and the initially-decaying current begins to increase at the moment the magnetic contact begins to part (I presume electric contact opening is a few milliseconds later for Gunnar’s graphs). The duration of the hump following the wiggle is highly dependent on the resistor as shown in tab comparison, and it is also easy to see the acceleration is much faster for the higher resistance and the contacts.

Again there is nothing new here, just decided to try it out and share for your info.


=====================================
(2B)+(2B)' ?

 

[electricpete]

There is a factor of 10 error in the constant mu0. Should be 4*pi()*1E-7 = 1.25664E-06. You can adjust that for yourself in the green cell in the main tab. Results remain qualitatively the same.


=====================================
(2B)+(2B)' ?