8
\$\begingroup\$

I've noticed that most of the opto-isolators (optocouplers, opto-triacs) have a white package. Why is that?

\$\endgroup\$
1
  • 5
    \$\begingroup\$ I've only seen black ones \$\endgroup\$
    – endolith
    Aug 14, 2011 at 2:19

4 Answers 4

12
\$\begingroup\$

Summary:

  • In optocoupler devices which are essentially identical apart from being either white or black in color, manufacturers'data sheets show differences in switching speed and thermal performance, white being superior in each case.

  • The most notable physical parameter of the actual devices is a much lower capacitance for the white package. It seems likely that the lower capacitance is caused by a different dielectric constant in the white material and that the lower capacitance allows faster switching.

  • Numerous detailed "white versus black on the same graph" comparison curves are available in the cited datasheet.

  • The thermal performance of the white material is within the range that would be expected from superior radiation characteristics (and possibly also enhanced thermal conductivity.)

  • Note that many early IC's were white due to the use of ceramic packages - quite different than the material being considered here.

DATA-SHEET BASED DIFFERENCES

  • Some significant to very significant differences are evident between otherwise apparently identical or nearly identical parts where the datasheet provides comparative white and black pkg data. However, some manufacturers do not offer white and black alternatives for the same industry std part numbers where others provide a choice (eg Fairchild offer white & black for 4N25).

Where a colour choice is offered, the most notable differences are

  • A 1.5x to 3x improvement in switching times for white packages compared to black.

  • Somewhat better thermal behaviour from white packages.

    The tightness of correlation is somewhat spoiled by the manufacturers making minuscule mechanical differences which are almost certainly not relevant but which leave a very small amount of uncertainty.

Differences which can be seen, based on real world data sheets include:

  • White packages have better thermal characteristics.

    Thermal resistance is lower and

    Parts can be derated at a smaller amount per degree increase in ambient temperature.

    Maximum allowed dissipation can be higher.

  • Input to Ouput capacitance is lower for the white package - presumably due to a difference in dielectric constant.

  • Switching speed is faster for the white package. Varies with load resistor. Toff affected more than Ton but both significantly difference. Ton 2x to 3x faster in White !!!

Examples of all the above can be seen in the Datasheet for Faichild 4N28 optocoupler

This version of the 4N28 can be obtained in white (with "-M" suffix) or in black. Published data sheet differences include:

  • Total power dissipation. 250 mW at 25C in each case but derated per degree C at

    Black - 3.3 mW
    White - 2.94 mW.

  • DC average forward input current. Note that this seems to go against the trend but is not obviously directly thermally related.

    Black - 100 mA
    White - 60 mA

  • LED - power dissipation and derating per degree C. Again, a "mixed message".

    Black - 150 / 2 White - 120 / 1.41.

  • Detector power dissipation. 150 mW at 25C in each case but derated per degree C at

    Black - 2.0 mW
    White - 1.76 mW.

  • Input - Output isolation voltage. A bizarre result but they do seem to be differentiating. Note that 5300 VAC RMS = 7500 VAC_peak for a sine wave. While the reason for this "different but the same" spec can be debated, it's bizarre and misleading to specify it this way. For a pure sine wave these specifications are identical but one is for 1 minutes and the other for 1 second.

    Black - 5300 VAC RMS, 60 Hz, 1 minute

    White - 7500 VAC peak, 60 Hz, 1 second

  • Isolation Capacitance. This seems to be significant for some applications BUT they spec each slightly differently , which prevents certain comparison. Note that while the White value is only 40% of the Black value, which seems likely to be highly significant, the White max is 1000% o the Whitetypical but the Black typical is not stated. Very sloppy.

    Black - 0.5 pF typical

    White - 0.2 pF typical, 2 pF max.

  • Package dimensions. Agh! Idiots.

    The black and white versions have their own package specs and there are various minor dimensional differences in many of the dimensions in each of through hole, SMD and 0.4" spaced versions. :-(.

  • Absolute Current Transfer Ratio - CTR.

    No differences between white and black in numeric data.

    This seems wrong based on inferences which may be drawn from relative CTR data.

  • Normalised Current Transfer Ratio - CTR. Idiots again, it seems.

    Graph axis are different scales (very poor practice)
    Normalisation to CTR = 1 relative at 10 mA prevents full comparisons.
    Black peaks higher wrt 10 mA and at lower mA than white.

  • Switching Speeds. Ton, Toff, Trise, Tfall. Graphs page 8. These vary with load resistance, especially Tr and Toff which are dependant on time constant of pullup resistor and device capacitance.

  • At typical load resistor values (1k to 10k) Ton is about 2x to 3x faster for the white package !!!

    At 1k load Toff black is about 1.5x Toff white

    At 10k load Toff black is about 2.2x Toff white

    At 100k load (unusually high) Toff black is about 3x Toff white


Note: The sample size on which the above is based is impressively small. However, the differences do seem to be real and significant.

\$\endgroup\$
5
  • 2
    \$\begingroup\$ You have to be careful about pseudo-correlations. There's no physical reason why white should be faster, so that's probably a pseudo-correlation. It's like "People with small feet are bad at high-density matter physics". It's true: children don't understand HDM physics and they have small feet. Just don't put a "therefore" or "because" like that between apparently correlating statements. \$\endgroup\$
    – stevenvh
    Aug 14, 2011 at 16:31
  • 1
    \$\begingroup\$ You have to be careful about absolute statements, I think :-). And you have to check to see if a possible physical linkage has been suggested in the (aka my) answer. As it has been in this case. Correlation does not prove causation, as they always say - BUT if causality exists its useful to find it. viz it seems likely that the dielectric constant of the encapsulant used signifcantly affects the inter and intra electrode and/or interstructure capacitances and that this affects time constants. A look at the switching time graphs in the document cited give good credence to this theory.. \$\endgroup\$
    – Russell McMahon
    Aug 14, 2011 at 16:42
  • \$\begingroup\$ Touché. People will take things for granted all the time, like that there's no physical reason why white should be faster, or that minuscule mechanical differences are almost certainly not relevant ;-) \$\endgroup\$
    – stevenvh
    Aug 15, 2011 at 7:26
  • \$\begingroup\$ @stevenh - You will, I'm sure, appreciate that the comment re minute physical differences is, of course [tm], a considered engineering judgement based on the available facts (plus the waving of hands and a few dead fish) and was made only after carefully [tm] comparing the dimensional diagrams for the 3 pkg types x 2 colors and looking at whether eg differences between White_1 and Blk_1 were matched in W2/B2 & W3/B3. They weren't consistent. Colors but not pkgs are graphed in detail and as the pkg differences do not intercorrelate well with spec changes I think the statement is "safish". \$\endgroup\$
    – Russell McMahon
    Aug 15, 2011 at 8:14
  • 1
    \$\begingroup\$ The above is in fact true (apart from the fish). If I'm going to say things like " ... minuscule mechanical differences which are almost certainly not relevant ..." I (almost always) try to ensure that such an assertion is a considered one. It may be based on extensive research or checking, or extensive experience or, more usually, some mix. Sometimes I "just say things" but hopefully vanishingly seldom. Even my original answer, which I now consider to be incorrect (aka 'there is no reason') was based on experience and some data sheet checking. Not enough of the latter as it turned out. \$\endgroup\$
    – Russell McMahon
    Aug 15, 2011 at 8:21
4
\$\begingroup\$

They are not always. In fact, I see very few white optoisolators. Below is a power supply removed from a TV; there are two prominent black optos crossing the high voltage boundary.

Optoisolators on a power supply

The reason may be for safety or to assist identification. For example, why are most class Y capacitors blue? On an assembly line or during repair, it would prevent mixing the capacitors up with other devices, which are not rated for the same application. There could be other reasons too; perhaps the particular material (maybe ceramic instead of plastic) that the white optoisolators are made from is more rugged with regard to external damage or can sustain a higher voltage. Maybe the engineer which designed them just liked white chips ;).

\$\endgroup\$
2
  • \$\begingroup\$ I didn't say they are all white, but most of the ones I bought were white. The safety theory makes sense. \$\endgroup\$
    – m.Alin
    Aug 14, 2011 at 0:55
  • 1
    \$\begingroup\$ Interesting notion, especially since most chip substitutions would result in a product which would be non-functional, but not patently unsafe. Putting an IC where an optoisolator was supposed to go could create a hazardous situation with galvanic connection bridging the isolation barrier. Depending upon what the opto was supposed to be doing, it would even be conceivable that a device might seem to mostly work, despite the dangerous condition. \$\endgroup\$
    – supercat
    Aug 14, 2011 at 15:46
3
\$\begingroup\$

Toshiba's Photocouplers and Photorelays catalog (PDF, 3 MB) says:

Internal Structure of Toshiba's Photocouplers

\$\endgroup\$
2
\$\begingroup\$

Thomas's safety argument doesn't sound bad, but then you would expect them all to be white, and he gives himself a counterexample.
So not all optocouplers are white, though I've seen only few which aren't, but also they're not the only white ICs:

White IC

This is the package of a resistor network, which isn't a classic IC (simplistically: meaning lots of transistors), just like the optocoupler (having a barrier between it's both parts). So there are white and black optocouplers, and white ICs may or not may be optocouplers. I don't think there's a specific technical reason.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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