Relay Coil Suppression with single Zener Diode across BJT 4

[JDW0]

In terms of relay coil voltage spike suppression, the following thread focuses on using a 1N4007 diode in series with a Zener across the relay coil:

thread248-291225

But there are times when you need to suppress coil spikes but cannot put components across the relay coil. In this case, a single Zener diode across the Collector and Emitter of a common-emitter BJT (which drives the low side of the relay coil), should be adequate to suppress the coil's voltage spike to the Zener voltage. And using a Vz = 2 x Vcc (with Vcc being the coil's positive voltage) would ensure the presence of the Zener does not slow the switching mechanism in any significant way.

My question is how to properly calculate the wattage / power rating of the Zener. While we could just use a multi-Watt Zener "for safety," that is a rather "shot-in-the-dark" method that ignores size and cost. A 0.5W 1N5252B 24V Zener costs much less than a 5W 1N5359B 24V Zener, and there is a large size difference as well. As such, it is advantageous to know if the smaller and lower cost 0.5W Zener would be a long-term reliable choice to suppress relay coil spikes.

Consider this circuit:

The OMRON G8NW-2 relay shown is a single device that houses 2 coils and 2 switches. Here is the datasheet:

https://components.omron.com/compon...5AE814485257201007DD4FE/$file/G8NW-2_0607.pdf

In the above circuit we are using an NPN BJT to switch the low side of both coils, with their high side connected together at an automotive +12V. (That 12V is filtered, so don't mind about Load Dumps.) The Zener voltage was chosen to be twice Vcc, which is 24V. The resistance of one relay coil is 225Ω at 20°C, but the datasheet shows a worst case of 180Ω at -40°C. Low resistance means more current flow. At 225Ω we have 0.64W and 53mA current, but at 180Ω we have 0.8W and 64mA current; and again, that is for only 1 of the 2 coils.

I built a test circuit that matches follows the above schematic, first without any suppression and then with the Zener across the collector and emitter. On a scope I measured relay coil spikes at the low side of the relay coil (at the Collector of the BJT) to be from 100V up to about 143V:

The transistor used has a Vcc=50Vmax specification. Vce (voltage across the Collector and Emitter) will obviously be lower than Vcc with the relay in the circuit, low side connected to the Collector. But even if Vce was spec'd at 50V, that's still only half of the measured voltage spikes coming off the relay coils. As such, we need to protect the transistor, and that's where our Zener across the Collector and Emitter comes in.

Assuming a safe maximum Vce of 40V, it takes between 150us and 175us for the spike to decay to that level (varies by Vcc voltage level):

And with the 0.5W 24V Zener connected across the transistor, we see this on the scope:

But what is the minimum reliable wattage rating we should choose for the 24V Zener? That's my question. These are short-term voltage spikes, not continuous high voltage. And again, there are times when you need to reduce both cost and size, making a random selection of a high-wattage Zener impractical. I used a 1N5252B 24V 0.5W Zener across the transistor with the OMRON relay shown in the schematic for my measurements above. I've even used that same 0.5W Zener in a bench-top test circuit with 2 larger automotive relays which draw 122mA each. I only tested for a matter of minutes, not hours or days or weeks. But the 0.5W 1N5252B did not blow. (And actually, even without any suppression at all, for my short-term bench testing, the high voltage did not fry the transistor.) So I am curious if a 0.5W rated 24V Zener across the transistor would be a long-term reliable choice for relay coil spike suppression in this particular application.

I would appreciate hearing your thoughts.

 

[JDW0]

SparWeb,

I'd actually seen that PDF (or something similar) before. Interestingly, your PDF is date stamped Nov. 1998. But despite that good advice from 19 years ago, many engineers continue to use a 1N400x across the relay coils, unconcerned about possible tack-welding implications. I believe that it's one part ignorance (which also means the engineer isn't fully reading the relay datasheet), one part cost-cutting (a single 1N4007 is dirt cheap), one part a lack of concern for long-term product durability, and one part "doing it like everyone else." Honestly, most people suppress the coil with a 1N400x across the coil. But that doesn't make it a proper suppression solution.

For now, a Zener rated at 2*Vcc (or higher, so long as Vce is not exceeded) placed across the transistor seems to be the best overall solution, even for circuits where the relay is on the same PCB. It's good, cost-effective suppression.

But if anyone wishes to add something new to this discussion, I am truly all ears.

Best wishes to all who have kindly contributed to the discuss thus far.

P.S. I just saw your new post with additional simulations. Thank you! If you scroll back up this thread and review my scope measurements, you won't see any ringing. Note my scope images showing the circuit with the Zener being used and note the scope is set to 1ms/division. Also note in my scope measurements that when 14.4V is used the Zener suppressed spike is just under 2.0ms, and at 12V it's closer to 1.5ms.

 

[Skogsgurra]

JDWD, when you say that "Other engineers have been chatting about this for a while now" I understand that you are rather new to electronics design. Kick back reduction in relays and contactors has been used ever since relays first appeared back in thirties-fourties and many different methods have been used. Selenium stacks were popular in the fifties and there were also varistors (MOV) that are still being used. A varistor may be what you are looking for, BTW. There are also snubbers, usually RC combinations, gas discharge devices and still some.

A subject like this is not something that one shall "discuss". Like all engineering, it is about setting up criteria that need to be filled and then either use your own knowledge or the collected knowledge of manufacturers and other engineers. Even text-books can be a valuable source of information. "Discussions" and chats where the level of the participants is unknown usually leads nowhere. This thread is partly showing that.

If the delay as such can be tolerated, then use a simple diode. Delay will typically go from less than 5 ms to between 10 and 20 ms.

If the delay cannot be tolerated, then add a resistor in series with the diode. As was done in my
18 Jun 17 19:11 answer, second picture. Delay is then reduced from (in this case) 12 to 6 ms. Doubling the resistor doesn't help much (a reduction of delay with another millisecond).

If the bounce and/or contact dynamics is a problem, then you will need to use another relay. One that doesn't bounce. Diode or not - bounce will be there - and not very much dependent on delay.

Zeners, transient voltage suppressor diodes and other "expensive" components bring a reliability problem with them. If they short out, you can get other, more difficult problems. A simple diode + resistor doesn't short out and that can be a good thing in itself.

So, I seldom use anything but a parallel diode (1N4148 or 1N400x) or the diode inherent in driver circuits like ULN2803 and similar ones. If there is a delay problem, I add a resistor that has roughly the same value as the (combined) coil resistance so that peak voltage gets up to around twice the nominal coil voltage.

If that isn't good enough, I have to search for other means of doing the job. SSR, opto-coupler with high CTR or any other component. Or I may need to rethink the design goals - do I really need that speed? Or will I ever switch those 10 A? and so on.



Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[JDW0]

Skogsgurra, I appreciate your input. Thank you for performing a practical test and posting the results.

Firstly, I should explain that my citing of that one thread on StackExchange was not a matter of me saying "this thread is when all the chatter started." I primarily linked to it because it presents an easy-to-comprehend graphic showing several suppression methods, one of which is the Zener across the transistor which I've been speaking about in this thread. And if someone reads this thread and learns something about how to avoid tack-welding, then this "discussion" has indeed led "somewhere." But it is up to them to practically "test" the "theory" they are reading about in this thread, just as I have by building a circuit and taking measurements over time. I think the term "discuss" should refer to a valid exchange of engineering ideas that have practical use within the scope of the said discussion.

As to the "level" of all participants, I think that is only partly relevant. I've met a lot of PhD's in EE who had no practical experience in electronic design. They could talk theory but they lacked experience getting their hands dirty. What matters is if the solutions proposed can be tested and properly measured data is exchanged in an understandable way. That is why I connected an actual circuit on the bench and provided scope waveforms. SparWeb supplemented my measurements with SPICE simulations. I think this is all useful information, despite it being a "discussion," for any EE to contemplate, along side their textbooks. I believe your earlier post with measured data contributes positively to this "discussion."

Speaking of your earlier post with the series resistor, such that I could put your data into perspective, what is the coil resistance of the relay you used? And if you don't mind revealing it, what is the brand and model of relay that you used?

Regarding varistor use for relay coil spike suppression, I am inclined to believe that a Zener would yield longer life when it comes to suppression of relay coil spikes. MOVs have their place -- for example, in the power supply of a circuit used in an automobile to handle Load Dumps. You won't typically have a load dump daily or even weekly. But a relay might be switched ON/OFF 20 times or more per day.

As to the "delay problem," it is not so much that 2ms or 8ms or 10ms is "too long" in human terms, but rather is a matter of what may happen over time to the relay contacts. Tack-welding is a serious problem the basically renders the relay useless, and that is why, as has been mentioned in this thread, that data sheets of major relay brands warn against use of a 1N4xxx diode across the relay coils. For example, OMRON datasheets will often say this:

Note: External coil suppression will cause a measurable increase in release times and may cause the relay’s characteristics to fall out of the specifications given here.

SONG CHUAN relays often come with this warning:

The use of a single diode in parallel with the coil for transient suppression causes longer contact release time. On power relays, longer release time may reduce relay life.

But so long as that kick-back voltage is below Vce for a transistor driving the ground side of the said relay, and so long as the kick-back is suppressed not much lower than 2*Vcc (Vcc being the coil voltage), the above datasheet warnings become moot in that there is no risk of tack-welding the relay contacts due to slow operation.

In my experience, an opto-coupler is usually not a good choice when you are trying to control costs. Optos are expensive, even in high volume. And as has been discussed in this thread, cost is a major concern when speaking about the best approach to suppression of relay coil spikes. Furthermore, optocouplers draw a noticeable amount of current, which in and of itself may exclude them from many low-current designs.

As has been discussed in this thread, the design premise presented in this thread is that some circuits may need to use an open-collector transistor to drive an external relay which may or may not have suppression. We cannot necessarily expect the end user to add coil suppression external to the circuit, and therefore we should assume no external suppression and consider how best to protect the transistor. That is precisely why a Zener across the Collector and Emitter makes good design sense, and a Zener is a cost effective solution as well.

As to your 0.25W Zener (1N4118) suppression solution with in-series resistor, would you mind posting a schematic so it is more clear how you have everything connected? (A quick hand-drawn schematic will be fine if creating an electronic one is troublesome for you.)

An SSR (Solid State Relay) is much more expensive than a normal relay, and again, this thread is about suppression while also keeping costs as low as possible.

As to what happens when a Zener fails, that is an important consideration seeing many of the Zener with wire leads fail shorted (which would keep the relay locked in its switching position -- powered). SMD Zeners could fail open (not affecting the circuit but eliminating the protection for the transistor controlling the relay). But to keep this in perspective, you mentioned use of MOVs which can be argued as having a shorter usable life than Zeners when it comes to absorbing repeated kick back energy. Cumulative degradation from repeated surge absorption is a legitimate concern when considering varistors, and while thermally Protected MOVs may yield longer life and perform more safely, they are more expensive as well. Again, low-cost is a key consideration in this discussion.

 

[itsmoked]

I agree with Skogs on this. For decades I've used the FWD across the relay and have never ever seen a relay weld because of it. Maybe it's because I never see how close I can get a relay's rating to the actual load. I read all of the data sheets typically and can't recall ever seeing a FWD prohibition though I've seen notes that said the relay will open more slowly.

The 5 most commonly used relays do not mention not using FWDs.

That said, I totally understand your issue of user supplied relays possibly not having FWDs.


Spar; I think a henry is way too big for a typical relay.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

Now, this is turning into a valuable exchange of points-of-views.

First, the MOV doesn't degrade if you don't overtax it. The degradation when subject to surge currents does not happen in relay coil applications. At least not at current levels seen in normal control and signal relay circuits where coil currents above one or two hundreds mA are very rare. I have, in my 50+ years of field experience, never seen a MOV fail due to repeated surge currents. Yes, they fail in other applications like being subject to high input voltage with "unlimited" current or repeated switching of highly inductive loads like lifting magnets, magnetic couplings or DC machine excitation. But never in a relay application.

Second, the fact that there CAN be a problem with material migration does not mean that it will happen. That, again, is something that I haven't seen. Never. But I have seen it in circuits where long cables (capacitance), low resistance filaments (incandescent lamps) or transformer inrush can destroy contacts after only a few hundred cycles. I may have been lucky, but I have the feeling that those warnings are a little like "Never dry your Poodle in the micro-wave oven" or "Do not wash your baby in the Whirlpool". It has been put there because it doesn't cost anything and - as the saying goes - Better Safe than Sorry.

Third, how do you measure contact speed? What I have seen so far is the duration of the kick-back voltage. That is not delay or contact speed. You need to measure time from de-energization to opening of contact if you want to measure delay. And then the difference between opening of the NO and closing of the NC contact if you want to compare contact speed. I think that you have inspired me to do so. I will test a few different relay models there.

If you wonder why I do this, I must mention that it is part of a book that I have been working with for several years. It is named "The Automation Engineer and the Reality" and contains lots of things just like this. So I also get something out of the subject.

I'll be back with more results soon.

The resistor is put here:

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[JDW0]

Skogsgurra, thank you for the clarifications and for the schematic. Curious as to why you prefer the resistor there instead of a Zener (with cathode facing the transistor's Collector)?

Regardless, you are putting the suppression across the relay coil. As I've been saying, there are designs where one must merely provide an open-collector output that drives an external relay, regarding which you know nothing (if suppression is used externally or not). As such, to protect that transistor, a Zener placed from the Collector to Ground, with Zener voltage 2*Vcc should be an adequate suppressor, if the wattage spec of the Zener is properly chosen (which was a big reason I opened this thread). I demonstrated the Zener's effectiveness in the scope measurements I posted earlier in this thread. So when you reply back with "more results" it would be appreciated if you would comment on this particular case, where you have an open-collector transistor driving an external relay that may or may not be suppressed.

 

[sreid]

The 1.5KE was just an example. How about a Vishay SA22A transzorb, 500W peak Wattage.
$0.14@1000 ea. The Zener is spec'd at 190 ma Once.

 

[JDW0]

190mA surge "once forever more" on the 1N4749A 24V Zener? At this stage, we need to define datasheet terminology and review multiple datasheets.

Consider this Vishay datasheet which says nothing other than 190mA for Tp=10ms:

http://www.vishay.com/docs/85816/1n4728a.pdf

Based on that datasheet alone, one would conclude "so long as I don't exceed 190mA for more than 10ms, this part will suffice."

Now consider this Vishay datasheet, which says it's 190mA "non-repetitive and in accordance with IEC 60-1, section 8" -- a 1980's standard that surprising one cannot download for free to see what it has to say:

http://www.vishay.com/docs/85153/bzd27series.pdf

At least there's a graph that shows iRSM in a percentage, which indicates that the climb to the peak of 190mA needs to be 10us or less and the peak must fall to 50% within T2 - T1 = 0.9ms. Even so, how does one define "non-repetitive"? Is it really "once forever more"? Or rather is it that the peak must never be exceeded and if reached the current must decline in accordance with the graphic, followed by a cooling off period (which is perhaps defined in that IEC 60-1 standard), after which it can in fact repeat?

This Vishay datasheet says something different still:

Surge current is a non-repetitive, 8.3 ms pulse width square wave or equivalent sine-wave superimposed on IZT per JEDEC method


But at least it references "the JEDEC method" which can be downloaded for free here:

https://www.jedec.org/sites/default/files/docs/jesd282b01.pdf

And it would seem they calculate that 8.3ms using Figures 4.8 & 4.9 of the JEDEC method, 1/60s = 16.666ms / 2 = 8.3ms.

Reading section "4.2.2.4 Nonrepetitive peak reverse voltage test" we see this:

This is a nonrepetitive rating and it represents the maximum value of reverse voltage which may be applied nonrepetitively to a diode without causing permanent damage.

Reading that alone seems to indicate "once and forevermore," but it still is not crystal clear, so let's read further down that page:

5) The time between voltage surges shall be long enough to permit the device virtual junction temperature to return to its original thermal equilibrium value.

Wait... "between voltage surges"? I thought it was "one surge forevermore"? Now the waters have become even more muddy.

Also in that same JEDEC pdf we read this:

nonrepetitive peak reverse voltage (VRSM): The peak reverse voltage including all nonrepetitive transient voltages but excluding all repetitive voltages.

and

repetitive peak reverse current (IRRM): The peak reverse current including all repetitive transient currents but excluding all nonrepetitive transient currents.

Figure 1.6 shows a graph of a sine wave with voltage multiple (repeated) voltage spikes on it, one being a higher voltage than the others (i.e., the "peak"). The presence of other lower voltage peaks indicates that voltage spikes can be repetitive so long as they are lower than the peak. But that is still vague because the question then becomes "how much lower"?

Let's say one argues based on a given datasheet that an iZSM of 190mA (Non-repetitive Peak Reverse Current) is one and for all. In other words, if you surge at 190mA again, the part shall be destroyed in all likelihood. Fine. What if the current "peaks" at 180mA rather than 190mA?

You see, it is rather vague. Specifics are lacking, even in that one datasheet which provides a graph.

How do you gentlemen interpret iZSM, and based on what documentation do you base that conclusion?

Thanks.

 

[Sparweb]

If you wonder why I do this, I must mention that it is part of a book that I have been working with for several years...

And I'm trying to get an illustration credit!

STF

 

[JDW0]

sreid,

Furthermore...

Earlier in this thread I posted (14 Jun 17 02:22) a scope measurement and the following current-peak calculations:

At that CH2=26.0V, the other probe (CH1) consistently measured 25.2V. So the voltage drop across Rtest is:

26.0V - 25.2V = 0.8V

Therefore:

i ("current pulse") = e/r = 0.8V/9.865 = 81.1mA

The 1N4749A 24V 1W Zener datasheets mentioned in my previous post all mention the peak current is 190mA. But I have calculated the peak current in the actual bench-tested circuit to be 81.1mA peak. Even if we assume that current level to remain constant (which it actually doesn't -- it declines a bit), the entire duration of that 81.1mA would be about 1.2ms @Vcc=12V or 1.6ms @Vcc=14.4V.

That again is why we must clarify the meaning of that "non-repetitive peak reverse current" specification in the datasheet. I would once again refer everyone to my previous post about that.

One more consideration...

If one argues that a US$0.14 (@1000pcs) Vishay SA22A transzorb would be best to suppress the relay coil spike, I could counter-argue that lower-cost (overall) Vce=350V 0.5A BST39 NPN transistor would allow one to forgo coil suppression entirely. For truly, the primary reason to even suppress the coil spike is to protect the transistor that switches the coil ON/OFF. With a transistor that can handle 350V (more than double my measured coil spike of 143V), coil spikes are no longer an issue for the circuit. Another very minor benefit of using a high-voltage NPN is that you can reduce the parts count (no suppressor diode required) and thereby gain more PCB space as a result.

Gentlemen, I look forward to hearing your thoughts.

 

[IRstuff]

There may be other reasons to suppress the BEMF, such as EMI control.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[JDW0]

Understood, but when reviewing my two previous posts, please ignore EMI considerations and let me know your thoughts regarding (a) your interpretation of "non-repetitive peak reverse current" on zener diode datasheets and (b) whether a 1N4749A zener would suffice in light of the current-pulse being only 81.1mA for a duration of 1.6ms or less (datasheet says the "non-repetitive peak reverse current" = 190mA).

Thanks.

P.S. SparWeb, I am unable to download your LTSPICE file for some reason (it results in a blank web page):

http://files.engineering.com/getfil...=212727627.11.1497931559518&__hsfp=2626819137

And my PartSim simulation (Transient Response) is way off with the spike being in the kV range:

https://www.partsim.com/simulator/#82014

 

[IRstuff]

The non-repetitive limit is 5x the working current, while this one is about 2.1x. I think you are simply risking premature failure from either junction spiking or metal migration. Death by a thousand cuts.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[JDW0]

If the non-repetitive limit is indeed 5x the working current, and if one argues that 2.1x the working current is "death by a thousand cuts," then how many cuts will result in death for say 1.5x the working current? What about 1.2x? Does even a teensy but above 1x working current spell death?

You see, there's much vagueness here.

 

[IRstuff]

Ok, I'll be willing to get you an answer for a measly $150k, for time and materials.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[zeusfaber]

Does even a teensy but above 1x working current spell death?

Death is one of life's two certainties.

Might be worth having a look at Mil-Hdbk-217 - not necessarily with a view to doing the calculations (in fact, the only firm recommendation I'd make is that you don't get yourself buried in the arithmetic) - but just to see the way that, for most component and stress types, reliability/mean life is already on the slide well before you reach the rated values.

A.

 

[JDW0]

zeusfaber, thank you for the tip (and for not charging me $150k). More specifically, you are referring to this document from Dec. 1991:

http://www.sre.org/pubs/Mil-Hdbk-217F.pdf

Since we are specifically speaking about Zener diode reliability, the only mention of a Zener in that entire PDF is on page 6-2. The base failure rate is Lb = 0.0020 for a Zener, according to that page. Using the max junction temperature of a 1N4749A of 200°C, we can calculate the Temp. Factor to be 10.5. Taking all other Factors to be 1.0, we then have a Lp = 0.0020 x 10.5 = 0.0211. And Lp = Failures/1E6 hours = Failures/114 years. Hmmm...

 

[MacGyverS2000]

I have a solution that's (nearly) free...

Put a note in your user's manual that tells the end-user not to use the specific relay setup tat will cause you issues.

Barring that, put in a reasonable solution (i.e., relatively inexpensive), put a disclaimer in the manual that usage beyond a certain limit may cause damage, and be done with it.

Otherwise, it appears you're looking for a unicorn. You want guarantees for everything under the sun and you want it for a couple of pennies. As one engineer in here is fond of saying, TANSTAAFL (there ain't no such thing as a free lunch).

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

The factors in MIL-STD-217 are for constant conditions. Otherwise, you'd have to have a weighted value. Moreover, the single, nonrepetitive pulse case kicks the acceleration factor substantially higher. That's free to you. ;-)

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[LionelHutz]

Most circuit board that house an open collector output typically also have a 12V or 24V terminal making it possible to still put protection between 12V or 24V and the collector even with the relay coil being externally mounted. Just saying.

IF turn-off was critical then I'd use a ~10V, 0.8W zener diode in series with a ultrafast diode connected between the 12V and the collector. This arrangement allows the collector side of the coil to rise to a similar 24V while halving the power in the zener. Otherwise, I'd just put a flyback diode in the circuit.

We have a circuit board with a 15V, 1W zener and 1A, 1000V diode in series wired across a transformer coil. The circuit is 24V and driven by a NPN transistor at about 25kHz. I'm fairly certain the peak current in the zener must exceed the 1W rating. Knowing the circuit, I expect the current in the transformer coil is around 200-300mA when the transistor turns off. Regardless, the zeners have never been an issue, even on boards that have been in the field for 10+ years.

So, I wouldn't be to concerned about putting the 10V, 0.8W zener in a circuit such as yours to protect a relay coil that is sees a low duty cycle.