Suitable driver for latching relay array from MCU

[sijmen]

Hi,

I have a project where I need to drive 30 Omron G6KU-2F-Y 3VDC latching relays.  These are single coil latching relays.

My plan was to use a microcontroller connected to a serial to parallel register (several of them) and then somehow drive each coil through 2 pins from the output of the register.  In this way, I could control the direction of the current through the coil.... I think.

My problem is, from the following datasheet, I can see that the 3 volt relays require 33mA to drive the coil, which is much higher than what any serial to parallel register can handle.
https://au.mouser.com/datasheet/2/307/en-g6k-348800.pdf

Is there any recommendations for an component I can use between the Serial to Parallel Register and the relay coil with sufficient power output, other than resorting to using discrete transistor and the like, which would make my board design too complex.  I was thinking something like an Opto Isolate array with darling transistors might work.  I wonder if an opto isolator would even work in this situation.


Appreciate any information.

Regards,
Sijmen.

[themadhippy]

I was thinking something like an Opto Isolate array with darling transistors might work

Does  it need to be opto isolated?if not  uln2803 or uln2001-2004 maybe suitable

[MasterT]

TPIC6B595, TPIC6B596, TPIC6C595, TPIC6C596

[Ian.M]

None of those chips are suitable for driving single coil latching relays, which require opposite coil polarity for their lathing and unlatching pulses.

Basically, to drive a single coil you need a small H-bridge.  If the supply is only 3.3V with 3V relays, you'd need a low-voltage MOSFET H-bridge.   
If you can get away with only changing one relay at a time, you can get away with arranging the N coils in a chain and only use N+1 half H-bridges.

Due to the low current and shortish pulse time, you may be able to get away with using MOSFET drivers in place of H-bridges.

If you need to be able to arrange the coils in a matrix to reduce the number of drivers, you basically need four steering diodes per coil (which had better be Schottky due to the low voltage drop requirement), and a double matrix with separate row lines for drive high and drive low, and the same for the column lines.   Its *NOT* worth it unless you've got ludicrous numbers of single coil latching relays to drive!

[Calvin]

Hi,

I had a similar prob and haven´t found any part specialized on this functionality.
There had been special ICs from SDS in the 90s iIrc ...but those are long obsolete.
Even asked here in the EEVBlog and no one knew anything.
So I put together a discrete circuit.

(rem: the part around Q1 and Q2 is just for LED-Indication).
C18 stores the energy for the resetting pulse.
It´s value needs to be designed to match the Relay´s (K10) requirements.

regards
Calvin

[Gyro]

Have you purchased the relays yet? I notice there are 4.5V and 5V coil variants too, with correspondingly lower coil currents.

A bit of an off-the-wall solution would be to use pairs of 74ACT540 inverting and ACT541 non inverting buffers. Each of the data inputs would be be connected together, also the output enable pins. The relay coils would be then connected between respective pins of the two buffers.

This would provide 8 differential coil drives per pair of buffer packages, with output tri-states to allow pulse driving of the coils. If you check the 74ACT54x datasheets, they are rated for 24mA per output, with Absolute maximum rating of 50mA (both output and VCC/GND current per pin). 33mA is pushing the specs a bit, but things get better for 4.5 and 5V coil variants. The relays have a 'must operate' voltage of 80% of rated voltage.

I've used a similar approach for switching sensitive relays in an audio attenuator. It's probably not something you would want to put into production, but for a one-off project it should be fine. It would of course require 8 packages (4 of each) to drive 30 relays.

Remember that 74ACT are fast devices and with the accumulated current of 8 relay coils will be quite high, so you need to provide good local decoupling. On the up side, relays are slow devices which are insensitive to a bit of output noise. You should also avoid output enabling the buffers for longer than is needed to flip the relays.


P.S. You could also significantly reduce the control signal requirements by commoning the inputs of all the buffer pairs and selecting which block of 8 relays is to be toggled using the output enables. This would reduce the control signal requirements to 8 data bits + 4 output enables, at the cost of not being able to toggle all 30 relays at the same instant (but very closely).

[voltsandjolts]

I prefer dual coil latching relays as they just need a couple of transistors to drive them, single coil is a bit more faff.
There are quite a few options. You could use BJT common emitter followers. Or a dual mosfet gate driver, like IX4426 or whatever.

[dietert1]

In the thread about low thermal EMF scanners, you can look here: https://www.eevblog.com/forum/projects/diy-low-thermal-emf-switchscanner-for-comparisons-of-voltage-and-resistor-stand/msg610758/#msg610758. He showed how to drive the coil with a capacitor and how to parallel multiple gates to get the necessary drive current.
In my own build i am using 74ACT138 parts to address each coil https://www.eevblog.com/forum/metrology/scannermultiplexers-for-voltage-references/msg3569019/#msg3569019.

Regards, Dieter

[Zero999]

Have you purchased the relays yet? I notice there are 4.5V and 5V coil variants too, with correspondingly lower coil currents.

A bit of an off-the-wall solution would be to use pairs of 74ACT540 inverting and ACT541 non inverting buffers. Each of the data inputs would be be connected together, also the output enable pins. The relay coils would be then connected between respective pins of the two buffers.

This would provide 8 differential coil drives per pair of buffer packages, with output tri-states to allow pulse driving of the coils. If you check the 74ACT54x datasheets, they are rated for 24mA per output, with Absolute maximum rating of 50mA (both output and VCC/GND current per pin). 33mA is pushing the specs a bit, but things get better for 4.5 and 5V coil variants. The relays have a 'must operate' voltage of 80% of rated voltage.

I wouldn't worry about exceeding the maximum current rating of the buffers. The latching relays only need to be switched on for 3ms and the ICs should be able to withstand a much higher current, for that length of time. The buffers might not be able to drive 5V relays, off a 5V supply, due to the voltage drop, of the relatively high resistance outputs, but stand more of a chance with 3V relays.

[MasterT]

Onsemi makes PCA9655E port extender. Serial I2C, 16-bits, 50mA.
https://www.onsemi.com/products/standard-logic/i-o-expanders/pca9655e
Or

PCA9535E, identical to the PCA9655E but with the internal
I/O pull−up resistors removed, has greatly reduced power
consumption when the I/Os are held LOW.

https://www.onsemi.com/products/standard-logic/i-o-expanders

[dietert1]

Datasheet, Table 4, Note 6: Each output must be limited to 25 mA output current. Roughly the same as ACT logic. There is a difference between destructive limits and operation limits.

Regards, Dieter

[sijmen]

Thank you everyone for all the feedback, but I think I might have found a suitable driver in the MAX4821, after having come across the following article:

IC drives up to four single coil latching relays - https://www.eetimes.com/ic-drives-up-to-four-single-coil-latching-relays/

For anyone interested, the datasheet is here: https://datasheets.maximintegrated.com/en/ds/MAX4820-MAX4821.pdf

In my case I think the MAX4820 is better suited for my needs because I can daisy chain multiple together and send the data in serially, to then control all 30 relay in one pulse.  The MAX4821 is quite a bit more expensive than I hoped, but having found it, I now have a starting point of what I'm looking for, (I think) and can look for cheaper alternatives.

Regards,
Sijmen.

[Ian.M]

Hmm...  Your datasheet link is borked!

Have you *READ* the MAX4821 datasheet?  The MAX4821 is open drain, hence the pullup resistors in the EDN design.  If you chose the 3V relays because your supply is 3.3V you are S.O.L. because the pullups + relays need 5V to get 3V across the relay coil.

OTOH if you've *got* a 5V rail with enough current capability (and remember, the pullup will draw an extra 100mA when that side of the coil is pulled down). it should work adequately - *IF* you can source the parts!

sci.electronics.design saying: "Friends dont let friends buy Maxim!"
https://www.eevblog.com/forum/chat/is-designing-with-maxim-worth-the-effort/
https://www.eevblog.com/forum/projects/maxim-love-_em-or-hate-_em/

[dietert1]

Yes, the eetimes proposal is bad. When driven low the 50 Ohm pullup takes 100 mA from 5 V, so it already exceeds the 70 mA output spec of the driver. Nothing left for the coil.
You could replace each pullup resistor by an active pullup, i.e. a small p-channel mosfet driven by the same pin of the driver. Plus weak pullups. Four extra parts per relay instead of two.

Regards, Dieter

[Kleinstein]

It is slight overkill, but mosfet gate drivers like TC4428, MIX4427, MCP14A... and similar are an option. However most types want at least 4.5 V supply.
The coil would be between 2 such drivers. More relais could be used as a chain - to N+1 drivers for N relais.

With some limitation to the initial phase one could also use 1 driver and a series cap (some 100 µF).

With a 5 V supply something like 2 parallel channels of a ACT540 (or similar) each could also work.

The method with the extra resistors to one side would need a higher rail (e.g. 2 x coil voltage) and more powerfull drivers, like ULN200x.

[Ian.M]

@Dieter:  Whoops, I didn't spot that the driver is only guaranteed to sink 70mA!

[Gyro]

Have you purchased the relays yet? I notice there are 4.5V and 5V coil variants too, with correspondingly lower coil currents.

A bit of an off-the-wall solution would be to use pairs of 74ACT540 inverting and ACT541 non inverting buffers. Each of the data inputs would be be connected together, also the output enable pins. The relay coils would be then connected between respective pins of the two buffers.

This would provide 8 differential coil drives per pair of buffer packages, with output tri-states to allow pulse driving of the coils. If you check the 74ACT54x datasheets, they are rated for 24mA per output, with Absolute maximum rating of 50mA (both output and VCC/GND current per pin). 33mA is pushing the specs a bit, but things get better for 4.5 and 5V coil variants. The relays have a 'must operate' voltage of 80% of rated voltage.

I wouldn't worry about exceeding the maximum current rating of the buffers. The latching relays only need to be switched on for 3ms and the ICs should be able to withstand a much higher current, for that length of time. The buffers might not be able to drive 5V relays, off a 5V supply, due to the voltage drop, of the relatively high resistance outputs, but stand more of a chance with 3V relays.

Datasheet, Table 4, Note 6: Each output must be limited to 25 mA output current. Roughly the same as ACT logic. There is a difference between destructive limits and operation limits.

Regards, Dieter

Yes, I didn't claim that it was a perfect design, but I think it would work for a one-off.

The 4.5V relay pulls 23mA, and the 5V pulls 21mA, both at rated voltage of course, before the voltage drops of the ACT outputs. The bistable relay has a 'must operate' voltage of 75%.

It would take a little bit more fiddling with the specs and a calculator to double check but I think it's workable with 4.5V relays. With 5V supply, the guaranteed ACT tristate output voltage drops are ~0.75V (high) and ~0.44 (low) @24mA so about 1.2V total (typical at 25'C is more like 1V total). With the 4.5V relay you need a minimum of 3.4V to guarantee switching. leaving 400mV of slack for the full temperature range ACT spec (600mV typical at room temperature).

A little slim, I agree, but achievable. Bumping the supply up to 5.5V would make things more comfortable (approx nominal coil voltage). The ACT logic would be operating within spec, and at half absolute maximum rating.

The 3V relay would work too, the 33mA current would probably increase the voltage drop of the ACT outputs to bring the coil voltage to nearer 3V (the ACT spec sheet stops at 24mA). It is definitely pushing the outputs harder though (outside nominal spec) - depending on how frequently the relays are going to get switched.

Switching the relays in blocks of 8 (even if closely timed) would drop the peak supply current to something more acceptable - 30 relays switching simultaneously at 33mA is almost an amp. 8 relays at 23mA is only 184mA.

[sijmen]

Hmm...  Your datasheet link is borked!

Have you *READ* the MAX4821 datasheet?  The MAX4821 is open drain, hence the pullup resistors in the EDN design.  If you chose the 3V relays because your supply is 3.3V you are S.O.L. because the pullups + relays need 5V to get 3V across the relay coil.

OTOH if you've *got* a 5V rail with enough current capability (and remember, the pullup will draw an extra 100mA when that side of the coil is pulled down). it should work adequately - *IF* you can source the parts!

Thanks Ian.M, I've now fixed the links.

Yes, I did read the datasheet, but being honest, some if it didn't make sense.  However, its from the helpful information from people like yourself which helps me focus in on those areas which I need to understand.  I really do appreciate the feedback.

Originally I was thinking I could run the whole circuit from a 3.3V source, but I've come to the realisation I'll need it to be 5v.

Regards,
Sijmen.

[BrianHG]

Have you considered low-side gate driver ICs.  They can drive a good 1 amp out, do a logic level conversion from 3.3v to their VCC & you can run them with 12v relays if you like since they can usually be supplied with anything from 5v to 18v dc.

Your only problem is that you get 2 per device.

Something like a IX4340UETR from IXYS may work.

What about combining the Toshiba's 'TBD62789APG' for high side with a normal 8 transistor array for the pull down?

How about the MC33880?

You can always use dual low voltage logic level N/P channel mosfets.
Like the diodes incorporated 'ZXMHC3A01T8'

I used to use one similar to these as a common and doubled my MCU outputs in parallel to drive 5v latching telecom relays in an array.

[Zero999]

If you have a 5V supply. How about this. Horribly inefficient, cheap and nasty, but will work. The relay only needs a 3ms pulse to activate, so perhaps the inefficiency is tolerable.

By the way, it doesn't need any diodes, because the resistors dissipate the back-EMF quite nicely.

[Kleinstein]

The lenght of the requited pulse depends on the relais type. With 3 ms the relais may just reactor, but for a reliable operation it may need a slightly longer pulse. Some recommend 20 ms. Still this is a relatively short pulse and can be OK.

[Zero999]

The lenght of the requited pulse depends on the relais type. With 3 ms the relais may just reactor, but for a reliable operation it may need a slightly longer pulse. Some recommend 20 ms. Still this is a relatively short pulse and can be OK.

You're right. I misread the data sheet. The minimum set/reset pulse width is 10ms but it's a good idea to use longer pulses.

https://au.mouser.com/datasheet/2/307/en-g6k-348800.pdf

[Terry Bites]

Once youve solved the drive curent issue, you can use the "C" circuit to latch and unlatch. https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwjEp4rDrojyAhVSi1wKHWOBAuIQFjAEegQIDBAD&url=https%3A%2F%2Fwww.te.com%2Fcommerce%2FDocumentDelivery%2FDDEController%3FAction%3Dshowdoc%26DocId%3DData%2BSheet%257F1-1773931-5%257F0717%257Fpdf%257FEnglish%257FENG_DS_1-1773931-5_0717.pdf%257F1-1773931-5&usg=AOvVaw2b0tm96w2TU_od0mCjaxLv

[voltsandjolts]

Attaching that linked PDF here for posterity...

C-circuit refers to the capacitor in series with the relay coil option (similar to the LTspice circuit shown above).

[Zero999]

The downside with the capacitor is the latching relay must always be operated in one direction, when the power is first applied. I suppose two capacitors could be used, but it would act as a voltage divider.

Here are a couple of more ideas. Both are more complicated than the previous, but have the advantage of only requiring one MCU pin, per relay. Setting it to the H-Z state will turn the relay off.

The first one uses all common emitter transistors. It has a high quiescent current and low saturation voltage, so the relay & MCU can be powered off 3V. The resistors on the left hand side of the bridge are a slightly lower value, to compensate for the output resistance of the MCU.


The second one uses an emitter follower for the one side of the bridge so can't really work well off 3V, but will be fine from 5V and 3.3V might be adequate.


It's important to select resistor values, to keep the base voltage low enough, to prevent the transistors from turning on in the quiescent state.

D1 to D4 can be bridge rectifier modules, or dual diodes, to keep the part count low.