1
\$\begingroup\$

I have designed a circuit that drives a +/- 20 to 30 peak to peak external signal received as an input to 16 different outputs. Another requirement is that I have to be able to switch the state of these outputs in less than 100us (ideally 50us).

To meet these requirements, I am currently using a AQS221N2S low capacitance optocoupler array from Panasonic, precisely 4 of them. The actual circuit has been produced and tested and is functional for the specs.

However, I am now required to scale up the circuit to a 64 outputs, which brings me to reconsider my part choice for two reasons:

  1. AQS221N2S are relatively expensive.
  2. The AQS221N2S's turn on time meets the specs in the typical value (0.03ms) but not on the maximum value (0.2ms).

I cannot use relays or solid state relays as they are often slower and more expensive than the actual part.

The alternative I was considering was using an AC switch, such as the MAX14756, which is both faster and cheaper than the optocoupler solution I am using currently. Unfortunately, using an AC switch requires to add a bit of extra hardware since I will need to rectify the input AC signal to generate VDD and VSS necessary to the IC, which has the drawback of adding complexity to the design and more development time on my part.

Before I start redesigning the whole system, I was wondering if anyone had a suggestion of a part that would have the benefits of using an AC switch without the drawbacks of needing to rectify the input AC signal?

Edit: Here are some additional information about the circuit:

  • The circuit is currently powered from a single Micro-USB 5V supply, then a 3.3V linear regulator is used to supply 3.3V to the circuit.
  • The AC signal that I am driving ranges between 55 to 65 KHz
  • The AC value given above, 20 to 30V, is a peak to peak value, it was a mistake from my part to identify it as VAC.
  • Each input of the AQS221N2S are in series with a 390 Ohm resistor, giving a 5.4uA current flowing through and a switching speed below 50us according to the datasheet.
\$\endgroup\$
14
  • 1
    \$\begingroup\$ Hi! Interesting problem. Which supply / ground voltages are already present in your system? \$\endgroup\$
    – mmmm
    Aug 18, 2021 at 17:08
  • \$\begingroup\$ Why would you need to rectify the AC signal. That part is unclear. Ditto the previous comment. \$\endgroup\$
    – Andy aka
    Aug 18, 2021 at 17:14
  • 1
    \$\begingroup\$ The MAX14756 probably won't work. AC is usually specified in terms of RMS. For a 30 VAC signal, this suggests a pk-pk value of +/- 43 volts, and the MAX part is only rated to +/- 35. \$\endgroup\$ Aug 18, 2021 at 17:23
  • 1
    \$\begingroup\$ MAX1480x would handle the voltage \$\endgroup\$
    – bobflux
    Aug 18, 2021 at 17:42
  • \$\begingroup\$ @Andyaka The reason why I would need to rectify the VAC is because, in my understanding, AC switches typically need the AC signal to be between the VSS and VDD of the part. Therefore in order to generate those VSS and VDD it seems my best option is to rectify that signal to supply the chosen AC switch. \$\endgroup\$
    – jaun_dough
    Aug 19, 2021 at 20:17

3 Answers 3

1
\$\begingroup\$

You've run into a common problem. For a large number of switched input-outputs, the common industry solution is analog switching/multiplexing. The Maxim part you picked is from that category. These devices offer no isolation, and you don't seem to need it. They offer some of the fastest t(on) t(off) achievable (some under 100nS), and are reasonably priced compared to the opto-isolated solid-state relays. They will need the Vdd and Vcc supplies that encompass your AC signal with 0.3V overhead.. If your Vcc and Vdd is +15 and -15, your AC signal cannot exceed +14.7 and -14.7 V respectively. The power requirements are usually quite low, in the fractions of uA. You will also find that the 8xSPST devices often need some type of serial communication interface, while direct control devices are limited to 4x SPST. For a system with 64 controlled outputs, a serial control might be a good idea, even if it's just using serial-in parallel-out shifters. If your outputs don't have to be on at the same time, there are 1x32 multiplexer devices as well.

I think for your application, developing the proper supply rails would be well-worth the benefit of using the analog multiplexers. They are designed for this exact purpose.

Thank you @tobalt for bringing it up, I have forgotten about the "Beyond the Rails"(TM) line from Maxim that has integrated charge pumps to operate from 3.3 or 5V while internally still achieving those higher voltages necessary to pass through your AC signal amplitude.

Edit2: in the comments after some additional discussion, the solution that was accepted as the answer was: MAX335, octal SPST, tie together all the COM(X) lines, and it has a built-in shift register for control. Just get it the correct rails, which can be literally one part such as RB-0515D.

\$\endgroup\$
6
  • 1
    \$\begingroup\$ at least Maxim also has a family of switches with integrated charge pump, that might be able to use only the 5V supply. I think they use beyond the rails as name for those. \$\endgroup\$
    – tobalt
    Aug 20, 2021 at 7:45
  • \$\begingroup\$ @tobalt You are right, I have forgotten about the "Beyond the Rails"(TM) line from Maxim. There are trade-offs that would have to be looked at closely, such as turn on and off times, as well as total system efficiency. \$\endgroup\$
    – DSI
    Aug 20, 2021 at 13:26
  • 1
    \$\begingroup\$ Great point, I didn't mention it at first but I am controlling the opto-isolator using 2 shift registers, which would scale to 8 shift registers for the 64 output design. Multiplexing would not be an option since multiple outputs can be activated at once. I need to double check my options, as I am hesitating between developing rails or using "Beyond the Rails". \$\endgroup\$
    – jaun_dough
    Aug 28, 2021 at 20:08
  • \$\begingroup\$ @jaun_dough MAX335, octal SPST, tie together all the COM(X) lines, and it has a built-in shift register for control. Just get it the correct rails, which can be literally one part such as RB-0515D. Just set up decoupling caps on the max335. \$\endgroup\$
    – DSI
    Aug 30, 2021 at 3:25
  • 1
    \$\begingroup\$ @DSI Thank you very much for the part suggestion! I was going to wait for digikey's filtering interface to work again (seems to be working today) but with the suggested parts and the beyond the rails as an alternative (still ends up 50% cheaper than my original design) I can clearly see the DC-DC converter + AC switch is the way to go. Thanks again, I will mark this as the accepted answer. \$\endgroup\$
    – jaun_dough
    Aug 31, 2021 at 13:32
0
\$\begingroup\$

The AQS221N2S are rated at low diode current of 5mA and perform far better at 20mA. It seems from the Ton vs If curve is inversely related so you can improve the speed at 20mA. The nominal Ton/If = 50us/mA so at 20mA the Ton= 2.5us nominal for the conditions specified. Thus you have 5% of your goal nominal.

You can adjust \$I_f\$ to suit the ambient range and device tolerance margin you need. But the speed is adequate with more drive If current.

https://www3.panasonic.biz/ac/e_download/control/relay/photomos/catalog/semi_eng_rf4a_aqs22_s_cr10.pdf enter image description here

\$\endgroup\$
4
  • \$\begingroup\$ Thank you for your reply, the resistors used in series with the photodiodes of the AQS221N2S are 390Ohm resistors, which paired with the 3.3V supply allows 8,46uA to flow through. The speed obtained by this configuration is currently considered fast enough for the application, thus changing the part is mostly a matter of reducing the cost for now. \$\endgroup\$
    – jaun_dough
    Aug 19, 2021 at 20:25
  • \$\begingroup\$ You dont compute current from 3.3/390 , you use( 3.3-1.2)/390= 5.4 mA . As long as you don’t use that for high yield in production ok, otherwise, you may get fallout in volumes. So to switch on at the same time , you can run some in series with a higher V. Like 10mA \$\endgroup\$ Aug 19, 2021 at 20:31
  • \$\begingroup\$ also the MAX14756 is not Opto isolated. do you need this? \$\endgroup\$ Aug 19, 2021 at 20:47
  • \$\begingroup\$ @ Tony Stewart EE75, That is a great point about the current, I forgot to consider the voltage drop in the opto isolator. That being said, the current version of the circuit works as intended. The circuit being opto isolated is actually not a criteria of the design, as I said, I need to scale up this design to go from 16 output to 64 outputs, therefore going from using 4 AQS221N2S to 16. I am looking for AC switches as an alternative since they are both faster and cheaper, but since they require modifications to the initial design, I was wondering if other alternatives existed. \$\endgroup\$
    – jaun_dough
    Aug 28, 2021 at 19:41
0
\$\begingroup\$

This is not a full answer but might solve part of your problem with optoswitches, and is too long for a comment.

A neat little trick to speed up optoswitches is: Place a capacitor in parallel to the series resistor.

  • Dimension the resistor so the resulting current keeps the switch barely closed. That way the turn off time will be fastest.
  • Then add a cap in parallel to the R to allow the full short circuit current of the driver during turn-on which will speed up turn on.
\$\endgroup\$
2
  • \$\begingroup\$ Thank you for the answer, I had not considered this way of speeding of the circuit. A concern that I have with this method is that the current drawn by the opt switches might be to much for the previous IC to supply in case that all 16 switches need to be switched on at the same time. \$\endgroup\$
    – jaun_dough
    Aug 19, 2021 at 20:30
  • \$\begingroup\$ Indeed, this might be too much for the IC (such as a shift register) if it can occur that many of the LEDs are switched on simultaneously. It is a way to push the envelope a bit for optoswitches but certainly slower than directly driven CMOS switches \$\endgroup\$
    – tobalt
    Aug 19, 2021 at 21:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.