Why not use digital isolators in hobby microcontroller and USB designs?

[Nominal Animal]

I like to use digital isolators in my own hobby designs related to microcontrollers and USB.  Sometimes the isolation itself is the purpose, sometimes it is only a tool to achieve what I want – for example, power domain separation, or safety against experimental circuit failure.  One particularly good not-for-isolation-per-se example is I²C voltage level translation, with their supply requirements and low level threshold differences: I prefer to just grab a TI ISO1640 to avoid all those issues, even if I end up tying/shorting the two grounds together.

Why u no use digital isolators?

I mean, I see people complain about digital supply noise and ground loop issues all the time, but it is extremely rare to see e.g. the I²C DAC and amplifier circuits to be digitally isolated, even though it'd completely avoid the problems ahead of time.  For quick one-time tools, isolating a SPI ADC frontend lets you reference it to the target 0V rail even when running, without having to rely on AC coupling via a capacitor or some such.  To me, the digital isolator chips seem like a vastly underused tool that would solve a lot of the daily problems many are encountering, if used in the design in the first place.

Just like trying to add security on top of an existing system, isolation tends to be expensive and ineffective if added on top afterwards.  So, why not design it in, from the get go?

Edited to add: Let's exclude stuff like RF and high-frequency analog stuff that requires linear supplies anyway.  I'm talking about simple hobby stuff like your own USB audio projects, interfacing your computer to your own USB or microcontroller projects, and similar that will anyway be using switchmode supplies and the noisy USB +5V.

[coppice]

If someone could get the cost really low the volumes would soar. If someone could get the cost low for devices with integrated isolated power the volumes would skyrocket.

[Nominal Animal]

Is it really the cost?

(A honest question: no doubts or subtext intended.)

Mouser sells ISO774x four-channel isolators (capable of voltage translation and very high data rates) for about 2€ apiece in singles.  ISO1640BDR is 3€ in singles.

Isolated DC-DC 5V to 5V unregulated 2-watt converter modules cost between 4€ and 8€ in singles, these being sufficient after filtering with a low-drop regulator like MaxLinear SPX3819M5-L-3-3 (0.53€ in singles) to get a low noise 3.3V rail.

For bipolar ±3.3V to ±4V rails (max. 200mA each), one can use something like Traco TBA 2-0521 (4.40€ in singles at Mouser) to get unregulated ±5V rails.

For a low noise ±9V or ±10V bipolar rails (80mA max), one can use something like Traco TBA 2-0522 (4.40€ in singles at Mouser) to get unregulated ±12V rails, and filter and regulate down.  Or TBS 2-0523 (±15V 65mA max).

My preferred design is such that the microcontroller is always powered from the host USB 5V, but the isolated side is powered using an isolated DC-DC converter, preferably hardware-switchable (DPDT) between host 5V and a separate power input.  Notably, when using a separate power input, it should not be tied to the USB ground, and should be completely separate from the USB host.

Getting a low-noise isolated 3.3V rail is easy; +5V rail is much harder.  It is easier to use an external AC-DC supply, really.  However, Mouser now carries Murata MGN1S0508MC-R7 for 3.62€ in singles, which can provide 125mA at 8V from 5V 310mA, which one can easily filter and linearly regulate down to a low noise +5V rail.  I haven't used this one myself yet, but it and the MGN1D050603MC-R7 for 3.90€, +6V and -3V at 111mA each definitely look like a good match for my needs.  (+6V filtered and regulated to +5V using SPX3819M5-L-5-0, to -2V or -1.8V using TPS72301 for example.)

I understand that the cost is the driving factor behind commercial designs, but for hobby designs, I'd expect to see isolation used much more, considering how many problems one can exclude or mitigate with them, for relatively little increased cost.

[coppice]

Is it really the cost?

Mouser sells ISO774x four-channel isolators (capable of voltage translation and very high data rates) for about 2€ apiece in singles.  ISO1640BDR is 3€ in singles.

Yes, its the cost. i haven't seen real prices for these for several years, but I doubt they have changed that much. If Mouser sells you one IC for 2€ you might expect to pay a few cents when you buy a decent volume. That is not the case with isolators. They still cost tens of cents, which is a big hit on a high volume product BOM.

[nctnico]

I like to use digital isolators in my own hobby designs.  Sometimes the isolation itself is the purpose, sometimes it is only a tool to achieve what I want – for example, power domain separation, or safety against experimental circuit failure.  One particularly good not-for-isolation-per-se example is I²C voltage level translation, with their supply requirements and low level threshold differences: I prefer to just grab a TI ISO1640 to avoid all those issues, even if I end up tying/shorting the two grounds together.

Why u no use digital isolators?

Because they are pushing RF currents around and can cause other problems. The isolator chips need really carefull layout to create enough capacitive coupling between the side so the RF current isn't going to seek other paths.

Same for DC-DC converter modules; these have significant RF currents between the primary/secundary side which needs to be diverted (=shorted between in & output) and filtered. All in all galvanic isolation causes more problems than it solves. Especially if you tie the grounds together. In that case it makes no sense to isolate.

The worst example I've seen is a open-collector comparator followed by an opto-coupler which also creates an open collector output. Everything tied to the same ground 

For I2C level translation, TI has a whole bunch of bi-direction I2C level translators which work really well.

For power supply separation, you can use filtering between supply sections. In RF equipment where filtering is critical you often see an inductor or PI network followed by a local, low noise regulator. A simple RC filter before a regulator can also help but needs line regulation analysis.

To avoid digital noise from entering a DAC or ADC, you can filter the digital signals to & from in order weaken the edges (=reduce high frequency components).

Oh, and do not split grounds!

[coppice]

Why u no use digital isolators?

Because they are pushing RF currents around and can cause other problems. The isolator chips need really carefull layout to create enough capacitive coupling between the side so the RF current isn't going to seek other paths.

If you are doing something like digitising an analogue signal specifically to be able to put it through an isolator into a different DC domain what you say is true. If you have high speed digital data you already have those noise issues. If you put an isolator in the path, to create two separate DC domains, you really haven't made things any worse.

[Nominal Animal]

For I2C level translation, TI has a whole bunch of bi-direction I2C level translators which work really well.

You haven't used them in a situation where the other side logic level varies from lower than to higher than the microcontroller side?  Nor been bitten by the asymmetric low-level thresholds and limited pull-down abilities of them?

The most common one is PCA9306, I believe.  VREF2 > VREF1 is required, and bias current from VREF2 to VREF1 can play havoc with VCC1 side regulators.  So, to use it right, you need to use a VREF1 pulldown and a current-limiting resistor on VREF2.  If you don't know beforehand what logic level voltage the device uses –– for example, you want to use it for a couple of different modules you can switch between during experimentations –– good luck getting that right.

I like to use digital isolators in my own hobby designs.  Sometimes the isolation itself is the purpose, sometimes it is only a tool to achieve what I want – for example, power domain separation, or safety against experimental circuit failure.  One particularly good not-for-isolation-per-se example is I²C voltage level translation, with their supply requirements and low level threshold differences: I prefer to just grab a TI ISO1640 to avoid all those issues, even if I end up tying/shorting the two grounds together.

Why u no use digital isolators?

Because they are pushing RF currents around and can cause other problems. The isolator chips need really carefull layout to create enough capacitive coupling between the side so the RF current isn't going to seek other paths.

Same for DC-DC converter modules; these have significant RF currents between the primary/secundary side which needs to be diverted (=shorted between in & output) and filtered.

Most of hobby projects where digital isolation would help already already use switchmode supplies.  USB in particular is quite noisy in that sense, both from computers and from wall warts.

Let me amend the initial post to reflect what I actually meant: hobby projects using USB, not RF stuff that require linear supplies.

On typical machines, when you have your microcontroller and a mouse connected to the same hub, you can usually tell from the supply noise when the mouse moves (look at glitches occurring at very steady 1/2/4/8/16 ms intervals).  On many USB sound cards, this translates to audible noise.

Another case where I'd prefer to see more digital isolation is in stepper controllers.  There is no need for the very electrically noisy motor supply and the digital control circuitry to share the ground.  If you isolate the step-and-direction signals, you can use beefier AC-DC supplies for the motors, and a separate AC-DC low-voltage supply for the logic, or even allow safely connecting a computer via USB to the logic.

[coppice]

Another case where I'd prefer to see more digital isolation is in stepper controllers.  There is no need for the very electrically noisy motor supply and the digital control circuitry to share the ground.  If you isolate the step-and-direction signals, you can use beefier AC-DC supplies for the motors, and a separate AC-DC low-voltage supply for the logic, or even allow safely connecting a computer via USB to the logic.

From what I've seen motor control is a big part of the isolator market. That's why you see a lot of isolated sigma-delta modulators now.

[Nominal Animal]

Another case where I'd prefer to see more digital isolation is in stepper controllers.  There is no need for the very electrically noisy motor supply and the digital control circuitry to share the ground.  If you isolate the step-and-direction signals, you can use beefier AC-DC supplies for the motors, and a separate AC-DC low-voltage supply for the logic, or even allow safely connecting a computer via USB to the logic.

From what I've seen motor control is a big part of the isolator market. That's why you see a lot of isolated sigma-delta modulators now.

High-end, yes.  Hobby projects, absolutely not.  This is exactly what I am wondering about.  You don't see isolation used in OSHWy projects like Smoothieboard or RAMPS.  For four drivers (enable, step, dir) you only need two TI ISO7760's, which Mouser sells for under 2.50€ in singles, 1.80€ in lots of one hundred.

Most of the higher-current driver modules on heatsinks have optoisolated inputs, but for hobby projects –– including small servo robotty things –– digital isolators just are not used.  Is it the same kind of "they bring more problems than they solve" assumption/attitude that we can see in nctnico's response above?  For me, digital isolators are very useful components, and I honestly haven't been bitten by any issues yet.

[ajb]

Another case where I'd prefer to see more digital isolation is in stepper controllers.  There is no need for the very electrically noisy motor supply and the digital control circuitry to share the ground.  If you isolate the step-and-direction signals, you can use beefier AC-DC supplies for the motors, and a separate AC-DC low-voltage supply for the logic, or even allow safely connecting a computer via USB to the logic.

From what I've seen motor control is a big part of the isolator market. That's why you see a lot of isolated sigma-delta modulators now.

High-end, yes.  Hobby projects, absolutely not.  This is exactly what I am wondering about.  You don't see isolation used in OSHWy projects like Smoothieboard or RAMPS.  For four drivers (enable, step, dir) you only need two TI ISO7760's, which Mouser sells for under 2.50€ in singles, 1.80€ in lots of one hundred.

Most of the higher-current driver modules on heatsinks have optoisolated inputs, but for hobby projects –– including small servo robotty things –– digital isolators just are not used.  Is it the same kind of "they bring more problems than they solve" assumption/attitude that we can see in nctnico's response above?  For me, digital isolators are very useful components, and I honestly haven't been bitten by any issues yet.

A high power motor driver module is exactly the kind of hobbyist/dev-board sort of thing that SHOULD have digital isolation at the IO port, but more for controlling motor return current paths than for noise.  Many an arduino (et al) have met their end when a ground connection to a motor driver came loose and motor return currents found their way back to the battery via the signal wiring harness of some hobbyist/prototype project.  So isolating digital signals where they cross from a 'safe' domain into a high power/high risk of dramatic failure domain makes a lot of sense, since it eliminates a path for escalating failure. 

If you're putting everything on one PCB, though, like on the smoothieboard, you don't have to worry about that fault mode.  The best thing to do in terms of noise is to put everything on one solid ground plane across the whole PCB, so there's minimal impedance between everything, and carefully place everything so that large return currents and noisy signals are kept well clear of anything sensitive.  You do not split the ground plane, to be clear, but you do split the layout between digital/analog and/or high/low power so that noise and return currents stay in their respective zones.  In some rare cases you might slot the ground plane to help control DC return currents, but you *never* route tracks across that slot.  You can of course add filtering between the power domains, or on signals that cross from one domain to the other wherever needed.  You may need to protect signal lines from backfeeding into logic supplies in a partial-power state (like if the motor controller logic supply goes out with the motor supply fuse but the MCU keeps going), but that can be done with level translators or just a transistor + resistor -- you only need to block the signal connection, there's no need to break its return connection. 

Even if you want to use separate supplies for the motors and the logic, you can still do that perfectly well by bonding the two supply grounds at the PCB in most cases.  If you have a larger, possibly distributed system of many PCBs and a big harness between all of those things, then you're back to the first scenario where maybe isolation does make sense.

As far as optocouplers instead of digital isolators, an optocoupler input has the advantage of flexibility -- you can make it work with a sourcing or sinking output, and with active high or active low logic quite easily.  (Of course, it's not so flexible if the cathodes are just connected to the local ground, which I've also seen on those cheap relay boards!)

As far as why you don't see digital isolators in hobby projects specifically, even presupposing that they are the right solution in a given scenario....well there are a lot of hobby projects that get away with far worse!

[exe]

I'd say cost and board space are significant for isolation.
If we add a digital isolator, DC-DC converter, output LDO to filter the noise, it's already more complex than most of my projects) and take quite some real estate. Also, the isolators mentioned in this thread are just simple digital isolators, while I'd prefer  to use a usb isolator. Those, unfortunately, even more expensive (or please point me out to a suitable device).

Though I do like the idea of isolation from PC. I just wish there was a simple and cheap USB module to achieve just that.

Just the other day I connected my oscilloscope to a circuit powered by my laptop. I had horrible waveforms, and my hdmi monitor went black. I spent quite some time troubleshooting. The moment I disconnected scope from the charger (it's battery powered) all problems were gone. I'm glad I didn't get any damage, especially to the laptop and monitor.

[tszaboo]

Is it really the cost?

(A honest question: no doubts or subtext intended.)

Mouser sells ISO774x four-channel isolators (capable of voltage translation and very high data rates) for about 2€ apiece in singles.  ISO1640BDR is 3€ in singles.

Isolated DC-DC 5V to 5V unregulated 2-watt converter modules cost between 4€ and 8€ in singles, these being sufficient after filtering with a low-drop regulator like MaxLinear SPX3819M5-L-3-3 (0.53€ in singles) to get a low noise 3.3V rail.

For bipolar ±3.3V to ±4V rails (max. 200mA each), one can use something like Traco TBA 2-0521 (4.40€ in singles at Mouser) to get unregulated ±5V rails.

For a low noise ±9V or ±10V bipolar rails (80mA max), one can use something like Traco TBA 2-0522 (4.40€ in singles at Mouser) to get unregulated ±12V rails, and filter and regulate down.  Or TBS 2-0523 (±15V 65mA max).

My preferred design is such that the microcontroller is always powered from the host USB 5V, but the isolated side is powered using an isolated DC-DC converter, preferably hardware-switchable (DPDT) between host 5V and a separate power input.  Notably, when using a separate power input, it should not be tied to the USB ground, and should be completely separate from the USB host.

Getting a low-noise isolated 3.3V rail is easy; +5V rail is much harder.  It is easier to use an external AC-DC supply, really.  However, Mouser now carries Murata MGN1S0508MC-R7 for 3.62€ in singles, which can provide 125mA at 8V from 5V 310mA, which one can easily filter and linearly regulate down to a low noise +5V rail.  I haven't used this one myself yet, but it and the MGN1D050603MC-R7 for 3.90€, +6V and -3V at 111mA each definitely look like a good match for my needs.  (+6V filtered and regulated to +5V using SPX3819M5-L-5-0, to -2V or -1.8V using TPS72301 for example.)

I understand that the cost is the driving factor behind commercial designs, but for hobby designs, I'd expect to see isolation used much more, considering how many problems one can exclude or mitigate with them, for relatively little increased cost.

What you are describing competes with the total BOM of many designs.
I understand why you would think that this cost is insignificant compared to your hourly wage, hunting down issues.
But most hobbyist will buy this from their already taxed salary, and they might work in different jobs as you.
Plus, you can buy these as separate blocks, and place it at the end of any USB cable.

[Ice-Tea]

Why u no use digital isolators?

Because it violates rule #1 of..., well, everything. KISS.

[EPAIII]

Not the price? I recently purchased a combo package of two power strips. They had three AC outlets, three USB outlets, surge protection, and even a power indication light. The TWO together PLUS FREE SHIPPING cost $10 US. That's $5 each. And they were made in China so they had a lovely ocean voyage to get here.

When I opened one of them up I found three PC boards in addition to all the power sockets. Three separate PC boards! I can't imagine what each part on those boards must have cost. FIVE DOLLARS each! That is the world we live in.

Now tell me that a ten cent part is a negligible addition to a product. You and I can add $5, $10, even $50 to the price of a personal project and think nothing of it. But anyone designing a product today will be pricing each part down to the 1/1000 of a cent, if not better.

[Nominal Animal]

I understand that the cost is the driving factor behind commercial designs, but for hobby designs, I'd expect to see isolation used much more, considering how many problems one can exclude or mitigate with them, for relatively little increased cost.

What you are describing competes with the total BOM of many designs.
I understand why you would think that this cost is insignificant compared to your hourly wage, hunting down issues.
But most hobbyist will buy this from their already taxed salary, and they might work in different jobs as you.

I am poor, and have less than 50€ each month for my projects.  As I've described in other threads, I suffer from repeated burnout and depression, and have been unable to work for years now.  I am only a hobbyist on the electronics side too, with some formal education on it, but not that much practical experience.

For me, it is not so much a cost calculation, but a simple observation on what I have had to do to get my projects working.

One previous similar step for me a few years back was level translators.  Other hobbyists struggle with all kinds of transistor circuits.  I checked the datasheets, and realized I could use 74LVC1T45 –– my Odroid HC1 had a 1.8V UART –– and avoid all that: in SOT23-6, it's even dead-buggable directly to wires.  Later on, I discovered the TI TXU0n0m, and shifted to those for my UART and SPI shifting needs.

Why not bidirectional ones?  They typically have a low-level "glitch" when "erroneously" shifting directions.  If the MCU or device has a sufficiently high logic low threshold, that glitch does no harm, and everything works fine.  But, when you do have low logic low thresholds, or you really don't want the voltage level translator to try and drive your MCU output pins, using unidirectional level shifters like the ones I mentioned above avoids all the problems beforehand.

Thus, you could describe my own design and implementation strategy as the opposite of throwing components on a breadboard and seeing what seems to work: I want to avoid the problems I can foresee.  It is why I like to use NX138AKR MOSFETs in my 12V PWM fan controller –– they're like BSS138, but slightly faster ––, why my small display module LED backlight driving scheme needs the possibility of individually adjusting the LED intensities in case they're not well matched, and so on.  Because I can easily and cheaply avoid the problems I can foresee at design time, I like to take the time and do so, instead of having to later redesign (or worse, bodge a workaround) because those problems become a showstopper.

Plus, you can buy these as separate blocks, and place it at the end of any USB cable.

The cheap ADuM3160/ADuM4160-based full-speed (12 Mbit/s) isolators suffer from poor DC-DC converter choices.  They are typically this, and their output voltage without external load can shoot well over 6V and break things.  Wagiminator has a much better design, with a 5.1V Zener limiting the output voltage to about 5.3V even when not loaded.  I do find its 1W (5V 200mA) DC-DC converter choice limiting, though.  A good 2W output one (5V 400mA) at 80% efficiency should only draw 500mA from the host.

ADuM3165/3166/4165/4166 and TI ISOUSB211 -based high-speed (480 Mbit/s) isolators are much more expensive, with the cheap ones off eBay suffering from the exact same DC-DC converter choices.  What we really need, is a design where the DC-DC section can be omitted completely for self-powered designs, and the DC-DC input can be isolated – say with a pair of jumpers – from the USB host.

I have thus far not seen a single one with a boosting DC-DC converter to allow for filtering the switchmode noise and linear regulation to a fixed voltage.

Thus, while I personally do have some cheap ADuM3160-based isolators and use them quite often, they're not exactly a panacea; and the high-speed isolators are either quite expensive or suffer unnecessarily from similar limitations.  DIY solutions are still better, exactly due to commercial solutions having to be the cheapest possible, and the niches having users who don't know exactly their own requirements.  (This is best exemplified by the Audio segment ones, which don't even describe the technical properties of their USB isolators, only their effects to the audio...)

[tszaboo]

Plus, you can buy these as separate blocks, and place it at the end of any USB cable.

The cheap ADuM3160/ADuM4160-based full-speed (12 Mbit/s) isolators suffer from poor DC-DC converter choices.  They are typically this, and their output voltage without external load can shoot well over 6V and break things.  Wagiminator has a much better design, with a 5.1V Zener limiting the output voltage to about 5.3V even when not loaded.  I do find its 1W (5V 200mA) DC-DC converter choice limiting, though.  A good 2W output one (5V 400mA) at 80% efficiency should only draw 500mA from the host.

ADuM3165/3166/4165/4166 and TI ISOUSB211 -based high-speed (480 Mbit/s) isolators are much more expensive, with the cheap ones off eBay suffering from the exact same DC-DC converter choices.  What we really need, is a design where the DC-DC section can be omitted completely for self-powered designs, and the DC-DC input can be isolated – say with a pair of jumpers – from the USB host.

I have thus far not seen a single one with a boosting DC-DC converter to allow for filtering the switchmode noise and linear regulation to a fixed voltage.

Thus, while I personally do have some cheap ADuM3160-based isolators and use them quite often, they're not exactly a panacea; and the high-speed isolators are either quite expensive or suffer unnecessarily from similar limitations.  DIY solutions are still better, exactly due to commercial solutions having to be the cheapest possible, and the niches having users who don't know exactly their own requirements.  (This is best exemplified by the Audio segment ones, which don't even describe the technical properties of their USB isolators, only their effects to the audio...)

So take a soldering iron, and replace those converters. Or add a through hole Zener to the USB connector.
It's not like you would need more than 2-3 of these isolators unless you work with everything connected all at the same time.

[joeqsmith]

Why u no use digital isolators?

No need and don't design by sprinkling in parts. 

[Nominal Animal]

So take a soldering iron, and replace those converters. Or add a through hole Zener to the USB connector.

It is still cheaper to DIY one.  For example, JLCPCB sells ADUM3160BRWZ for assembly for USD $4.45 apiece right now.

It's not like you would need more than 2-3 of these isolators unless you work with everything connected all at the same time.

At most, two or three full-speed ones and two or three high-speed ones.

Probably the most common scenario for me is to connect my computer to the console (UART) of one of my SBCs or routers.  Their logic level varies; I have seen 1.8V, 2.5V, 3.3V, and 5V ones.  If I connect it to my laptop, I can just run it on battery power, and tie the laptop and SBC/router grounds together.  Or, I can use a full-speed (12 Mbit/s) USB isolator.  Or, I can use a TI ISO6721/ISO7721 for RX+TX or TI ISO6742/7742 for RX+TX+RTS+CTS, for both isolation and level translation.  Mouser sells ISO6721 for 1.39€ apiece in singles, 11.80€ for a set of ten.

Considering comparable voltage translators (TI TXU0202 is 6.93€ for a set of ten), why would one stick a voltage level translator at the end of their USB-UART cable, and not a TI ISO6721/7721/6742/7742 isolator?  I only need at most two of these, though.  As I run Linux, instead of one of the dedicated USB-serial converter chips, I can just use a cheap Cortex-M0 or RISC-V MCU with a native USB interface, so that if I need some kind of translation or software handshaking, I can implement it on the MCU instead of on the host.  Back in the days when ATmega32u4 "pro micro" clones were cheap, I used those.  Later, I used Teensy LCs.

I'm thinking of making a new slim-line one with WCH CH552G, as JLCPCB has both TI ISO6721 and CH552G in stock for assembly ($1.19 and $0.58, respectively; with AP2112K-3.3 for the 3.3V regulator at $0.08, plus some supply bypass capacitors, maybe a button and a resistor for boot mode/flashing).  I'll solder a cable directly to through holes, scavenging a cheap but known-okay USB A-B cable I no longer need.  It will be dirt cheap, and be much better than any ready-made USB-to-UART cable you can buy.

[shabaz]

Just to add an (unrelated) idea for your USB-UART board, perhaps it could be something of interest: break detection. I don't know how useful it will be on a USB-UART board, but it could occasionally be a good problem-solver.

The attached circuit (untested) shows what I was thinking about; the break detection output is made available on the output connector, along with 3.3V and 5V break-interruptable outputs, so that optionally circuits-under-test can be easily reset from the serial console connection. As mentioned, it is probably not an extremely useful feature, but it is possibly worthwhile on a DIY USB-UART for development purposes. 

I've been using a DIY USB-UART board for ages too. Mine happens to be based on CY7C65213 (no real reason, just that it supported reasonable speeds, and had battery-charge current negotiation capability which I wanted to experiment with but never got around to). I'm hoping to add that break detect circuit next time I modify it.

[T3sl4co1l]

This is actually a perfect application for this meme...



Those who don't know about them, might only know where they've seen them (not often) and assume they aren't very useful generally, or only used in special places (when ya gotta, ya gotta).

Those who know everything about them, know they're useful, but also know the hazards that can be created by them (ground slots, EMI), and are also aware of mitigation to avoid them (dealing with ground loop voltage/current, differential design, etc.), even if it costs extra time and effort to plan around (hey, we're into that kind of stuff~).

Those who know a bit, can use them, at added cost of course (and really, it's not much, not for few-offs anyways), and maybe run afoul of EMI issues, but often EMI isn't a functional requirement for little projects, and yeah, that can ease a lot of potential annoyances with hooking things up.

Tim

[Nominal Animal]

This is actually a perfect application for this meme...

You meanie!

I'm definitely in the "isolators should be used everywhere it makes things easier" camp, but even I don't use them everywhere; only where useful.  Tool for the job, really.  It's just that I don't see them used in hobby designs at all.  For example, consider the isolated USB-to-UART example above, and try and find a similar hobby project.

Furthermore, I do not agree that digital isolation is closely related to EMI issues; only isolated DC-DC converters are (and only to the same degree as any other switchmode DC-DC converter).  There just isn't enough energy involved/lost in digital signal isolators to generate/cause much EMI.

[coppice]

Furthermore, I do not agree that digital isolation is closely related to EMI issues; only isolated DC-DC converters are (and only to the same degree as any other switchmode DC-DC converter).  There just isn't enough energy involved/lost in digital signal isolators to generate/cause much EMI.

Some isolators inductively couple. Some capacitively couple. I seem to remember people having issues with the inductive types, but its a few years ago.

[Karel]

One reason: EMI.

I designed a medical device (ECG, cardiac floating) that uses a multi-channel digital isolator and
because of the patient leakage current constraints you can't use capacitors between the different
sides to suppress EMI. It was very difficult to get the device within the limits.

Using a variant that provides also power to the secundary side was a nogo.
It was impossible to get it within the limits without using capacitors or overlapping groundplanes
that caused the leakage current going through the roof.
So, for power I used a HF transformer.

Ofcourse, for non medical devices that are connected to a USB port of pc, there's no requirement
for isolation so you can apply capacitors between both sides to suppres EMI but then, how useful
will be that isolation. I guess it helps against ground loops and short circuits but any HF will leak
through.

[T3sl4co1l]

This is actually a perfect application for this meme...

You meanie!

I'm definitely in the "isolators should be used everywhere it makes things easier" camp, but even I don't use them everywhere; only where useful.  Tool for the job, really.  It's just that I don't see them used in hobby designs at all.  For example, consider the isolated USB-to-UART example above, and try and find a similar hobby project.

Furthermore, I do not agree that digital isolation is closely related to EMI issues; only isolated DC-DC converters are (and only to the same degree as any other switchmode DC-DC converter).  There just isn't enough energy involved/lost in digital signal isolators to generate/cause much EMI.

Yeah, to be clear, the meme is written with some hyperbole.   I am, after all, in the right-hand camp writing about the middle and left. I am, right?...

Usually EMI is low, but some can misbehave.  Keep in mind for example the ADI family are monolithic transformers, pulsing in the 100s of MHz; the isolators I think are generally well-behaved while the DC-DC ones can potentially be problematic, though I haven't used and tested any of the latter (and mostly used the former in industrial applications where I wouldn't notice EMI anyway).  The TI ones are monolithic capacitors, differential, and pulsed or otherwise coded at similar frequencies (I haven't actually seen what frequency and coding they use, would like to know), which should generally do well but in the event it ends up off balance, that would be a cause.

The thing is more that, isolating a cable or set of cables, makes a high impedance gap where voltage can build up, and where CM currents can't flow.  Except when the barrier capacitance is dominant, which -- depends.  In short, it's more reactive components in play, more considerations, and unusual conditions.  Maybe one is accustomed to dealing with common-ground EMI situations well enough, but doesn't know as well how to deal with this, and maybe it turns out alright, maybe it mucks up.  (Most people aren't familiar with either, anyway; EMC is such an obtuse topic.  They might well be helped by using isolation more often? dunno.)

For example, I'm fond of putting a R+C across most isolation barriers, just to add some loss to the isolation capacitance.  Maybe this is justified (de-Qs cable resonances, makes ferrite beads more effective), maybe it's worse (exacerbates CM-DM mode conversion, increases CM current in high-CMRR or dV/dt applications), but in any case it depends on application.

And yeah, the nearly-inevitable DC-DCs are a big concern, but I mean independent of them, which is why I didn't mention it.

(I actually built one of those applications the other day... no DC-DC, that is.  Just a little test equipment project with a logic level trigger input, figure why not make it a 5-30V universal input ala SSR, simple enough to add an opto with current limiter.  If something can provide power to the opto LED, that's all that matters.  And as a bonus, I provide local 12V on nearby pins so I can trivially plug in a pushbutton for manual activation when I don't need it sequenced electronically.)

Tim

[joeqsmith]

Edited to add: Let's exclude stuff like RF and high-frequency analog stuff that requires linear supplies anyway.  I'm talking about simple hobby stuff like your own USB audio projects, interfacing your computer to your own USB or microcontroller projects, and similar that will anyway be using switchmode supplies and the noisy USB +5V.