# Current Transfer Ratio (CTR) and opto couplers

#### robotDR

Joined Mar 17, 2020
21
Hi All,

I am trying to understand this concept.

My Optocoupler says 50% min to 600% max for CTR.

My collector current is 20mA. Does that mean I can simply solve for LED forward current by 20mA/300% if I want my CTR to be 300%?
How do I account for temperature?
What exactly does it mean to be saturated and non saturated and how does that effect me solving for LED forward current?

#### ronsimpson

Joined Oct 7, 2019
1,227
What optocoupler? Then I could look at the data to.

Saturation. Often the output transistor is used on a (on or off) mode. With a 5V supply the C-E voltage will be 5V or below 0.8V but never 2.5V. Non-Saturated the output will be in that middle land.

Some where there is information as how they want the LED driven. Too much current and the LED will dim with time.
Often I set the LED current to what works and will not age the part. Then I look at the transistor.
Say 20mA LED current is good, The with 50 to 600% CTR you know the Collector current will be as low as 10mA and as high as 120mA. Lets say 10V supply and a 1k Collector resistor. Then 50% CTR will just pull the resistor to ground. (close to ground) And a 600% CTR will also pull to ground. 10V & 1k = 10mA max.

Another thought: 10V 500 ohms = 20mA max. With 600% you can easy pull the Collector to ground. With 100% the Collector will probably make it to ground. But with 50% the Collector voltage will be at about 5 volts. (not saturated)

Hope that helps. RonS.

#### Papabravo

Joined Feb 24, 2006
16,103
No. It means you can't depend on the optocoupler to have ANY particular value of CTR, but 99% of all the devices that you can buy will have a CTR in the range. Think of CTR as a normally distributed random variable with a mean of 325% and a standard deviation of ≈ 91.67. so 6 standard deviations will cover a range of 325% ± 275%

The implications of this are that, you must design your circuits so that you don't care what the actual CTR is as long as it falls within the range. Can you do that? If not the alternative is to arrange for parts that are prescreened to a narrower range. That will raise somebody's costs dramatically.

#### crutschow

Joined Mar 14, 2008
27,184
The CTR is fixed for any particular unit and can fall between the range of 50% to 300%.
You have no choice or control in what you get unless you screen the devices.
It's similar to the variation in current-gain (Beta) of a BJT.
Typically you would worst-case design for the minimum CTR of 50%.

#### DickCappels

Joined Aug 21, 2008
7,424
I think the LEDs are better these days but one reason you should drive the LED a little harder than you would think is that the output of LEDs decreases with time and the higher the current the faster it degrades. The point is to be sure that even if you are one of the low CTR couplers, you can be sure the CRT is going to become lower over the years.

In the 1980's optocouplers became the rage for providing negative feedback to the primary side of switching power supplies from the grounded (isolated from line) side. The LEDs back then were not very good at maintaining light output at the optocoupler aged and about 2/3 of the way through the decade power supplies were dying all over the place.

Supply extra drive capability to the optoisolator to account for this kind of aging effect and temperature just to be sure, especially if you are designing something for production.

#### robotDR

Joined Mar 17, 2020
21
What optocoupler? Then I could look at the data to.

Saturation. Often the output transistor is used on a (on or off) mode. With a 5V supply the C-E voltage will be 5V or below 0.8V but never 2.5V. Non-Saturated the output will be in that middle land.

Some where there is information as how they want the LED driven. Too much current and the LED will dim with time.
Often I set the LED current to what works and will not age the part. Then I look at the transistor.
Say 20mA LED current is good, The with 50 to 600% CTR you know the Collector current will be as low as 10mA and as high as 120mA. Lets say 10V supply and a 1k Collector resistor. Then 50% CTR will just pull the resistor to ground. (close to ground) And a 600% CTR will also pull to ground. 10V & 1k = 10mA max.

Another thought: 10V 500 ohms = 20mA max. With 600% you can easy pull the Collector to ground. With 100% the Collector will probably make it to ground. But with 50% the Collector voltage will be at about 5 volts. (not saturated)

Hope that helps. RonS.
Hi Ron that is helpful. Here is a link to the data sheet (pn: tcmt4100):
https://www.vishay.com/docs/84181/tcmt4100.pdf

Sounds like I should find a better 'binned' part? 20mA LED current will quickly maximize my available current from my GPIO expander and I have several of these to drive.

A colleague of mine has been driving the LED at about 1.8 mA but that seems way too low because if we 'need' 20mA collector current, 600% won't even get us there. Might just be lucky that the device is enabled on less than 20mA.

#### robotDR

Joined Mar 17, 2020
21
I think the LEDs are better these days but one reason you should drive the LED a little harder than you would think is that the output of LEDs decreases with time and the higher the current the faster it degrades. The point is to be sure that even if you are one of the low CTR couplers, you can be sure the CRT is going to become lower over the years.

In the 1980's optocouplers became the rage for providing negative feedback to the primary side of switching power supplies from the grounded (isolated from line) side. The LEDs back then were not very good at maintaining light output at the optocoupler aged and about 2/3 of the way through the decade power supplies were dying all over the place.

Supply extra drive capability to the optoisolator to account for this kind of aging effect and temperature just to be sure, especially if you are designing something for production.
Thank you for that insight, definitely makes sense. So it doesn't really hurt to drive it higher, just makes sure the transistor turns on more or fully. Only other consideration is if I can drive that much current from other devices.

#### robotDR

Joined Mar 17, 2020
21
The CTR is fixed for any particular unit and can fall between the range of 50% to 300%.
You have no choice or control in what you get unless you screen the devices.
It's similar to the variation in current-gain (Beta) of a BJT.
Typically you would worst-case design for the minimum CTR of 50%.
I might end up getting the binned version. or as you said design closer to the 50% CTR. Thanks for your time reviewing this

#### robotDR

Joined Mar 17, 2020
21
No. It means you can't depend on the optocoupler to have ANY particular value of CTR, but 99% of all the devices that you can buy will have a CTR in the range. Think of CTR as a normally distributed random variable with a mean of 325% and a standard deviation of ≈ 91.67. so 6 standard deviations will cover a range of 325% ± 275%

The implications of this are that, you must design your circuits so that you don't care what the actual CTR is as long as it falls within the range. Can you do that? If not the alternative is to arrange for parts that are prescreened to a narrower range. That will raise somebody's costs dramatically.
Ok that makes sense. I'll check out the binned versions. I'm not so constrained on cost for this project, at least perhaps not down to dollars for a part like this. I'll look into it, thanks.

#### crutschow

Joined Mar 14, 2008
27,184
What is the load for the output transistor that you think you need so much current?

#### robotDR

Joined Mar 17, 2020
21
What is the load for the output transistor that you think you need so much current?
I measured it at 20mA. It’s the disable signal of a power supply unit from meanwell. Shorting the RC signal to ground disables the power supply.

#### crutschow

Joined Mar 14, 2008
27,184
I measured it at 20mA. It’s the disable signal of a power supply unit
You could add a BJT buffer (e.g. 2N3904) to reduce the required output current from the opto by at least a factor of 10-20.
A MOSFET buffer (e.g. 2N7000) would reduce the required opto output to just a fraction of a mA.

#### robotDR

Joined Mar 17, 2020
21
You could add a BJT buffer (e.g. 2N3904) to reduce the required output current from the opto by at least a factor of 10-20.
A MOSFET buffer (e.g. 2N7000) would reduce the required opto output to just a fraction of a mA.
So basically use a mosfet as a switch?

So like this:

#### crutschow

Joined Mar 14, 2008
27,184
Yes, but it needs a 10k (or so) resistor from the gate to gnd to turn it off.