Using a Schmitt trigger IC vs making one from op-amps or transistors

Question

In a professional setting, what is the benefit of using a Schmitt trigger IC as opposed to making one from op-amps or transistors? In a high volume application, wouldn't an op-amp implementation be easier for the supply chain because it wouldn't be using a custom IC that will be harder to replace?

Answer 1

It really depends on the requirements first. If there are multiple ways of meeting the requirements, then things such as using cheaper parts, more available parts (perhaps multiple-sourced) etc. will play a part in the design decisions.

Things like ST-input gates are easily available, cheap and quite fast, however the guaranteed trigger thresholds are very loosely specified. They also work from logic supplies- although 4000 series CMOS parts are still widely available, especially simple gates.

If you want more tightly specified thresholds you might choose a comparator chip.

If you need high precision and speed is of little concern then some op-amps with precision resistors and perhaps a precision voltage reference will do a fine job in those specific types of applications. Or maybe you have a free leftover op-amp in a multiple package and can press that into service.

Note: Some op-amps effectively have diodes across the inputs (usually with some series resistance) so they're unsuitable for most comparator applications where differential input voltage can exceed a few hundred mV, so care is required (datasheets are often not explicit in this matter- and if there are internal series resistors the differential input voltage rating may still be something like 30V but the inputs will draw significant current). Again, care is required. And recovery from saturation is seldom specified on op-amp datasheets and it can be extremely long (like tens of microseconds in a MHz GBP op-amp).

Or maybe you use an ADC and implement the hysteresis digitally with an MCU. That would almost certainly be the choice for an HVAC controller.

Answer 2

In a professional setting, all of the below can be considered to have some value to the supply chain or manufacturing. The various values can be traded off one against the other in an infinite number of ways. Anything that scores points in all categories is a winner. Most things improve one category and make others worse.

1. total BOM cost (cheaper is better)
2. total number of lines on BOM (minimizing unique parts count)
3. total number of parts placed on PCB (minimize pick and place run time)
4. multiple sources for each part (single source parts cause problems)
5. expectation of long term support (last time buys are expensive)
6. minimization of board area (this can feed back to cost since smaller boards are cheaper)

Design engineers have the reputation of being callous to the difficulties of supply chain and manufacturing. It is often thought that designers simply design however they want, then throw the design over the wall to manufacturing and let them deal with it, no matter how painful. While the operations people are pulling out their hair and begging for changes, the design engineers have moved on to the next shiny object and have no time or interest in the design that is already (in their minds) done.

But in a well-run organization, manufacturing and supply chain will have input into the design process from the very beginning to make sure the design can be manufactured and tested and supported, etc. There will be an opportunity for them to educate the designers about the pain they are feeling during production. So if there are two ways to do something, and you think one is better for the supply chain team, just ask them. In this case, they probably will not care whether you use a comparator or a schmitt trigger buffer.

Answer 3

I can think of a few use-cases where you might.

As mentioned, Schmitt triggers are off-the-shelf parts and standard in a lot of logic (even things you might not realize without reading the datasheet closely, like microcontrollers or SDRAM). So, other than those.

1 - Tighter Thresholds

For example, a logic analyzer has a variable threshold or reference, so it can be used for most any logic family: from the narrow voltage margins of ECL or LVDS, to the wide swings of CMOS (or deprecated families like HTL, or bespoke discrete logic). This might be done with a DAC driving the reference, and an array of comparators to receive the input signals. Note that a comparator is also a 1-bit ADC, so this effectively converts the potentially continuously-varying input signal to a digital signal as such. To be even more precise, a signal is never not analog, but it's digital when we can apply the assumptions we use to design digital circuits -- short risetime, well defined pulse widths, some propagation delay, etc. Put another way, the comparator converts the potentially messy outside world into the well-defined signals internal to the analyzer.

There are some precooked (fixed) examples, typically receivers like RS-232 (high voltage tolerant, 0-3V threshold), RS-422 (differential), 74HC7014 (60% V_DD), etc. The latter is interesting, being a 74HC family part but with accurate thresholds (most CMOS thresholds fall a gross 30-70% range).

2 - Not Worthwhile

If we only need a few channels worth, maybe it's not worth putting in a logic hex-gate version like 74HC14 or CD40106B -- we might use a comparator or even op-amp (if the slow speed is tolerable) that's already used elsewhere in the design. Mind, with the availability of TinyLogic and friends (74HC1G14, etc.), this is hard to justify, but does happen sometimes. Probably the most common case is at high voltages, as CD4000 family logic isn't available in "tiny" versions.

Note that we need about three resistors to turn a comparator/op-amp into a Schmitt trigger. There are knock-on costs here. Also keep in mind, the full (quad/hex/etc.) gates don't cost much more than the smaller versions, so there's not much wrong with tossing one in and tying off a load of unused sections. Mainly they're annoying when you're trying to save space, and even a TSSOP version has a large footprint for what you're doing.

The most likely use-case is when you only used one comparator of a dual, or three of a quad, and you need one more signal done, and well, there you are. Harder to make the case if you need two signals and so would add at least a dual to the design (or change dual to quad, but then routing is complicated -- the quad footprints are usually somewhat annoying to use).

3 - To Impress?

When an economic case doesn't need to be made, you might just go and build a whole bunch of the circuit in discrete i.e. plain transistors and resistors with very few or no ICs. Maybe as an exercise, maybe just because you can. The design is certain to be large and inefficient, needing many parts and a lot of supply current; and it will be expensive to assemble (a typical ballpark figure is $0.10/part to assemble PCBAs, but this varies wildly with who/where the assembly is being done, how many parts and types there are, etc.). See MOnSter 6502 or Discrete Pong for a couple exceptional examples. (It's hard to make a case for "professional" examples, though. Possibly the former counts, as an educational or academic product; certainly it is not practical in terms of raw functionality alone!)

Answer 4

The transistor version of the Schmitt trigger is the most inconvenient to use. A significant disadvantage is that the output voltage cannot reach zero because there is a voltage drop across the common emitter resistor RE.

(from Wikipedia Schmitt trigger page)