EEVblog #1284 - How Bad Product Design Kills The Environment

[ogden]

Gentlemen, either we get back on the topic or the topic risks being locked, I am afraid.

Which exactly topic? "Design topology differences of standby supplies" is well-covered in the video already. Other topics are not supported by Dave as you can see.

[EEVblog]

Then go make your own videos.

What?! Did not expect such "argument" from you.

But if you can't at least understand my intention, even if you don't like it, then 

I said it already - you are doing business. Sometimes engineering excellence have to be traded for clicks. That's OK especially for professional youtuber. BTW existence of argument do not prove that I don't like video. Your videos are fine, they get better over time as well.

I had a point I wanted to make in that video and I think I made it, it's got nothing to do with "trading engineering excellence for clicks", if you think that then you are demonstrably wrong.
Again, you don't have to like my intention of that video

If you cared to actually watch the video:

Text overlay that looks like afterthought? - Nah. Sad that you consider critique as an attack rather than idea for your future video(s).

The problem is you that you are arguing based on something I had no intention of doing in that video.
If you don't know how I do videos, let me explain. I don't have a script, I don't plan them, I have an idea for an intention for a video and I hit record and start talking. I then join the clips together and upload and hopefully something useful comes out of it.

In this case my intention with the first video was to show that zeners based solutions waste a lot of power and here's a couple of alternative that could have possibly been used instead. It was not about redesigning that actual product and doing the engineering to select the best part to do the job. When editing that video I thought some people like yourself might pick up on the 10mA part, so I measured actual buzzer current and added that overlay saying that this would require a proper engineering solution, and that was the end of that. I wasn't going to go and shoot another hour worth of material exhaustively going through the design process of selecting the exact correct part to meet the requirements, that would have been stupid as it was not the intent of the video.

For the 2nd video my intention was the explain why I didn't have to test such a solution, and explain what "design by inspection" is. Again it was not about redesigning that actual product and doing the engineering to select the best part to do the job.

Again, you don't have to like my approach, and that's fine, but at least try and understand that I have a different intention to what you wanted, and made the content accordingly.

[kcbrown]

This is how the semiconductor industry works. If you have a part with some novelty there are small volume users who will pay a good price for it, so you set the small volume price high. However, high volume users are ALWAYS driven by the BOM above all other factors. You need to offer attractive volume pricing to win any high volume design. The markets for this device are high volume things like smoke detectors, where you can lose a design over a fraction of a cent on the BOM, and utility meters, which will be attracted to the lack of any magnetic components (they are sensitive to tampering issues), but still need to meet aggressive BOM goals.

Sure, I get that, and it makes sense.  But the manufacturer can't sell at a loss, either, especially at the highest quantities.  Their price has to be enough to (in aggregate) recover their R&D plus the manufacturing costs (and sales/advertising costs, etc.).

In this case, their competition is the wide array of SMPS controllers that are on the market, with offerings ranging from those produced by first-tier manufacturers of the same caliber as TI to the cheapest offerings from Chinese vendors/manufacturers.  Against that kind of competition, it seems to me that TI would have to sell a device like this on its novelty, not just to small volume users but even to high-volume purchasers.

Which is to say, if a high-volume purchaser can get away with using an SMPS solution then it seems to me they probably will use that in favor of using a solution built around this TI part, except perhaps under one condition: when the cost of the controllers is of the same order of magnitude as the jellybean parts, such that the difference in the cost of the jellybean parts is at least enough to compensate for the difference in the cost of the controllers, thus turning the space advantage of the TI part (due to fewer jellybean parts) into a cost advantage.  Maybe TI can sell these things cheaply enough to meet that condition and still make a profit.  No idea.

[Nominal Animal]

I am really glad Dave made this video, because it does matter.

How do I know?  Because it exactly mirrors the standby power issues a decade or two ago.  It was common for devices to waste up to tens of watts in standby power, because designers and engineers didn't think it mattered.

But it did.  I recall a time here in Finland, where those of us with devices like TVs and computers on extension cords with power switches could save on their electricity bills simply by completely unplugging the devices when not in use.  For me, dwelling in a city apartment but with lots of electronics in the late nineties, it was about a third of my electricity bill.

Some of you might have seen my Beginner thread about USB-to-3.3V, too.  That is this exact same situation in another form.  Most of the USB gadgets not running on battery power use an LDO to drop the USB 5V to 3.3V.  Sure, that keeps the BOM cost low.  But it also turns a third of the consumed power into waste heat.  In some cases, like LTE modem dongles (the USB sticks in particular), that extra heat is a real issue, and shortens the devices' lifespan and usability.

I am more than a little annoyed at so many people trying to convince others that this is a non-issue, because the BOM cost is too high to make this kind of a change.  Yet, we basically eliminated the standby power issues within a decade, as knowledge about it grew; and I'm absolutely sure the EEs at that time tried to make the same argument, that significantly lowering the standby power would never be cost effective or competitive, because of the increased cost of the designs.  Fuck that kind of statist thinking: wasting power is silly, and sooner or later people will understand its effects on their finances, and will switch.

Before any of you respond with "but Nominal, making a better but slightly more expensive product makes no business sense, so you're wrong", I'll just say that unless you have earned millions from your EE designs, your opinion has zero weight in real life.  The trajectory I am seeing has happened several times in the past (different technologies, from steam forwards), and most recently, just over a decade ago with standby power.  So I am not proposing this is anything out of the ordinary; I am saying this has happened before, will happen again, and Dave pointing out one possible (albeit small) way we could take right now along this trajectory, is a very good point.



I only wish Dave would attack USB and LiPo power supplies to 3.3V projects, because there is very little talk of this right now on the internet, we're talking about 34% of losses (5V to 3.3V using LDO is only 66% efficient; step-down converters with low enough ripple and noise are black magic to us hobbyists), and even a bumblefuck like I can find chips and datasheet designs that can reach 90%-95% for ≲ 5€ total cost per unit for a set of ten, using resources available for us hobbyists.  Yes, small Chinese step-down modules are ubiquitous, but because their ripple and noise are unknown factors (and one really needs an oscilloscope to determine them in practice), so many avoid them, and opt to use wasteful and hot LDOs instead, because at least their characteristics are easy to understand and determine, needing just a multimeter, really.

Yet, thousands of us hobbyists are doing these projects, and would love to know how to do it better/properly; and realize that just like standby power or bad product design that wastes power, this too is one way we can reduce the overall energy costs of our projects: many do not even realize how many of the practical problems of their projects are due to waste heat or too high current draw.  (The latter is more of an issue with single-board computers; desktop machines and laptops tend to have better filtering and power budgets for USB power lines.  There is a reason why Olimex has an USB module with just bulk capacitors between VUSB and ground; that kind of thing was necessary with the power-hungry 3G USB modem dongles.)  For these USB gadgets, the vastly reduced waste heat means one can use cheap 3D printed closed enclosures with much fewer issues -- which was kinda the way I stumbled on this myself.

[Ed.Kloonk]

I am really glad Dave made this video, because it does matter.

How do I know?  Because it exactly mirrors the standby power issues a decade or two ago.  It was common for devices to waste up to tens of watts in standby power, because designers and engineers didn't think it mattered.

But it did.  I recall a time here in Finland, where those of us with devices like TVs and computers on extension cords with power switches could save on their electricity bills simply by completely unplugging the devices when not in use.  For me, dwelling in a city apartment but with lots of electronics in the late nineties, it was about a third of my electricity bill.

Some of you might have seen my Beginner thread about USB-to-3.3V, too.  That is this exact same situation in another form.  Most of the USB gadgets not running on battery power use an LDO to drop the USB 5V to 3.3V.  Sure, that keeps the BOM cost low.  But it also turns a third of the consumed power into waste heat.  In some cases, like LTE modem dongles (the USB sticks in particular), that extra heat is a real issue, and shortens the devices' lifespan and usability.

I am more than a little annoyed at so many people trying to convince others that this is a non-issue, because the BOM cost is too high to make this kind of a change.  Yet, we basically eliminated the standby power issues within a decade, as knowledge about it grew; and I'm absolutely sure the EEs at that time tried to make the same argument, that significantly lowering the standby power would never be cost effective or competitive, because of the increased cost of the designs.  Fuck that kind of statist thinking: wasting power is silly, and sooner or later people will understand its effects on their finances, and will switch.

Before any of you respond with "but Nominal, making a better but slightly more expensive product makes no business sense, so you're wrong", I'll just say that unless you have earned millions from your EE designs, your opinion has zero weight in real life.  The trajectory I am seeing has happened several times in the past (different technologies, from steam forwards), and most recently, just over a decade ago with standby power.  So I am not proposing this is anything out of the ordinary; I am saying this has happened before, will happen again, and Dave pointing out one possible (albeit small) way we could take right now along this trajectory, is a very good point.



I only wish Dave would attack USB and LiPo power supplies to 3.3V projects, because there is very little talk of this right now on the internet, we're talking about 34% of losses (5V to 3.3V using LDO is only 66% efficient; step-down converters with low enough ripple and noise are black magic to us hobbyists), and even a bumblefuck like I can find chips and datasheet designs that can reach 90%-95% for ≲ 5€ total cost per unit for a set of ten, using resources available for us hobbyists.  Yes, small Chinese step-down modules are ubiquitous, but because their ripple and noise are unknown factors (and one really needs an oscilloscope to determine them in practice), so many avoid them, and opt to use wasteful and hot LDOs instead, because at least their characteristics are easy to understand and determine, needing just a multimeter, really.

Yet, thousands of us hobbyists are doing these projects, and would love to know how to do it better/properly; and realize that just like standby power or bad product design that wastes power, this too is one way we can reduce the overall energy costs of our projects: many do not even realize how many of the practical problems of their projects are due to waste heat or too high current draw.  (The latter is more of an issue with single-board computers; desktop machines and laptops tend to have better filtering and power budgets for USB power lines.)  For these USB gadgets, the vastly reduced waste heat means one can use cheap 3D printed closed enclosures -- which was kinda the way I stumbled on this myself.

My phone has a charger that probably offends even the most ambivalent EE designer. And how has my phone manufacturer dealt with it? A big fat message that takes up two thirds of the screen "Unplug Charger!". When I followed this up, cos it's really annoying when I just want to glance notifications while it's in it's little cradle, they don't mean the plug going into the phone.

There is actually an animated graphic depicting the user pulling the plug out of the wall socket. I suspect that this 'solution' has been shoved into the phone OS to obviously appease somebody, wouldn't it have been easier to include another 20c worth of parts to make the charger go to sleep when nothing is connected at the USB end?

[TerminalJack505]

Like it, or not, it will likely take legislation to change this.  That's why you no longer see big, inefficient wall warts included with new devices.  The US has minimum efficiency requirements of external power supplies.

Because they are internal, these capacitor dropper power supplies don't fall under any of the requirements.  The only incentive manufacturers have to use anything more efficient is cost.

Sadly, I don't imagine it would increase the cost by more than 5% to 10% to include a switching regulator.  You can get something like a UCC28881 for $0.54 in volume, for example.  This is a high voltage AC-DC non-isolated buck converter with fairly low external part count.  It could easily power a smoke detector.

[Nominal Animal]

Like it, or not.  It will likely take legislation to change this.

Sure, that's how the standby power problems were solved as well.

But first, the enthusiasts, hobbyists, and ordinary users must realize and understand the situation.  Then it may become a political issue, and as it is such a simple thing, it may go through quite rapidly, just as the standby power stuff did.  The only reason legislation was necessary, was because the companies and their EEs didn't want to spend the extra to produce devices users wanted.  Think about that for a second, when rereading the comments in this thread.

Like I said, this kind of step forward is not new; it has happened dozens of times in the last two centuries already, and will happen again.

What is important, is someone like Dave making these easily understood videos so more people realize this is an issue.
(And my USB-to-3.3V thing is related, so I'm hoping Dave or one of the other youtubers with EE design experience could show how the same applies to current 3.3V USB gadgets, and how hobbyists can avoid that pitfall.)

Let me elaborate a little.

I have an Odroid-HC1 that I'd like to use as an LTE firewall, with a Huawei ME909s-120 LTE modem (a real modem, not an embedded Linux system!) as the uplink.  The modem is one of many modules that use a miniPCIe interface -- or, actually just the USB and SIM card pins on the interface.  These modems run on 3.3V, and tend to be a bit power-hungry: the average current consumption of this one is only 200mA, but it can be a bit spikey.  (Older 3G/4G/LTE modems and dongles, like the ZTE 823 embedded USB model I also have, is very power-hungry, but I'm not exactly sure of its consumption as it has been in continuous use for over a year already.)

MiniPCIe-WWAN adapters (an USB connector and a SIM card slot, connecting to such a module via MiniPCIe; most modules have very good support in Linux with ModemManager -- in my case, plug-and-play) are easy to obtain at e.g. Ebay, and there are even companies making their own for use in ad displays, indoor advertisements, et cetera, but they all have the same issue: they use an LDO to drop the 5V to the required 3.3V, and the waste heat is wreaking havoc with the enclosures.  (I even know of a Finnish company that is having this issue with this in their own designs, having observed the change in LDO to one with a serious heatsink on it!)

I now have four of these adapters (plus one either in, or lost in the mail).  Only one has a reasonable step-down DC-DC converter, a Silergy SY8009B (in SOT23-6).  It's not bad for this particular application, as it is 90% efficient at the nominal 200mA (5V to 3.3V), and can do 2A max, but the datasheet doesn't even list the recommended inductor values, so I am quite suspicious of it.

In fact, I am considering making my own.  Which is, given my level of competence in these matters, rather ridiculous.  And that just shows how too stuck to their ways most EEs designing these jellybean modules are, not considering the issues Dave highlighted here -- and the issues like nasty waste heat due to silly component selection causing real use risks.  And that is why this kind of videos -- showing the issue in plain terms, then a few suggestions on how to do better -- are so good: it is food for thought.  That thought leads to better designs and wider understanding of the issue, and may lead to the legislation that eventually enforces the change, as it did with wasteful standby power a decade or two ago in most Western countries.

(And lots of traps for us new players, too.  Which is why I'm going slow and careful, trying to work out the reasonable design, before I commit to testing it in real life, risking my 60€ modem.)

The point BigClive makes about apparent power -- that eventually, we will be charged by apparent power use, instead of the actual power use --, is simply a continuation of the same trajectory.  It is just much further in the future; I'd say a decade at least, perhaps two, but it will come.
It would be interesting to see what kind of a circuit you could use right now to create a LED light with a reasonable (say, 0.95 or higher) power factor, and how much it would cost.

Or, better yet, how much a 2.4A (12 watt) USB charger would cost, if it had efficient components and actual power factor correction.  You know: useful information, food for thought.

[nctnico]

I only wish Dave would attack USB and LiPo power supplies to 3.3V projects, because there is very little talk of this right now on the internet, we're talking about 34% of losses (5V to 3.3V using LDO is only 66% efficient; step-down converters with low enough ripple and noise are black magic to us hobbyists), and even a bumblefuck like I can find chips and datasheet designs that can reach 90%-95% for ≲ 5€ total cost per unit for a set of ten, using resources available for us hobbyists.  Yes, small Chinese step-down modules are ubiquitous, but because their ripple and noise are unknown factors (and one really needs an oscilloscope to determine them in practice), so many avoid them, and opt to use wasteful and hot LDOs instead, because at least their characteristics are easy to understand and determine, needing just a multimeter, really.

It is not that simple. A switching DC-DC converter uses a few to several tens of mA for itself. Often this means it is actually less efficient compared to an LDO in that situation. If you want to power a microcontroller from USB an LDO is the most efficient solution in many cases.

[Nominal Animal]

It is not that simple. A switching DC-DC converter uses a few to several tens of mA for itself.

What you are claiming is that TI Webench is lying, and that you know better.  For some reason, I'm not convinced.

The particular circuit I am talking about is the one straight off the TPS82084 datasheet, except with a 510k:162k voltage divider to get the 3.3V instead of the 200k:162k to get 1.8V shown in the datasheet.  The figures I quoted are from the TI webench design report that I first found; I only later noticed it is really a direct copy of the datasheet circuit.

So, which one is it, then?  TI straight out lying and you right, or you too stuck in your familiar patterns to admit there might be something here?

[nctnico]

It is not that simple. A switching DC-DC converter uses a few to several tens of mA for itself.

What you are claiming is that TI Webench is lying, and that you know better.  For some reason, I'm not convinced.

The particular circuit I am talking about is the one straight off the TPS82084 datasheet, except with a 510k:162k voltage divider to get the 3.3V instead of the 200k:162k to get 1.8V shown in the datasheet.  The figures I quoted are from the TI webench design report that I first found; I only later noticed it is really a direct copy of the datasheet circuit.

So, which one is it, then?  TI straight out lying and you right, or you too stuck in your familiar patterns to admit there might be something here?

Well, this didn't stand out in your wall of text. But this converter looks like it (finally) addresses the downside of switching converters at low load. And it seems cheap too (which is also new to me when it comes to integrated switching modules). Nice catch!

[Nominal Animal]

Well, this didn't stand out in your wall of text.

Yeah right.  You're very talented in avoiding admitting any error or culpability, instead hiding it inside a snipe at the accuser.  The trick with the compliment at the end was a masterful stroke; it makes you look like a reasonable person, if one isn't very observant.

Have I mentioned I absolutely detest people who try to manipulate others like that?

[ogden]

Well, this didn't stand out in your wall of text. But this converter looks like it (finally) addresses the downside of switching converters at low load.

Power Save Mode (PSM) of TPS82084/5 is similar to what pulse-skipping/burst_mode converters do for long time already. Check TPS62743 for example (360 nA Operational Quiescent Current, Up to 90% Efficiency at 10-µA Output Current).

[nctnico]

Well, this didn't stand out in your wall of text.

Yeah right.  You're very talented in avoiding admitting any error or culpability, instead hiding it inside a snipe at the accuser.  The trick with the compliment at the end was a masterful stroke; it makes you look like a reasonable person, if one isn't very observant.

Ofcourse it is a masterfull stroke. Carefully designed to piss you off so the next time you write a shorter text and put a link in there with an actual part we can check out (instead of a writing a long rant nobody is interested in). Take an example from how I can do so much with so little text. 

[Nominal Animal]

Ofcourse it is a masterfull stroke. Carefully designed to piss you off so the next time you write a shorter text and put a link in there with an actual part we can check out (instead of a writing a long rant nobody is interested in). Take an example from how I can do so much with so little text. 

I am a self-confessed unexperienced hobbyist with a cheap multimeter, an Analog Discovery 2, and a few odd projects.

My point in the post was not to advertise existing products, but to explain why this video by Dave was very on point.
You disagreed, and were proven wrong.  You didn't even admit to that, just stated you were only wrong because I write too much. 

When you need people like me to point out what the reality is and what the existing products can do, you proved you are stuck in your incorrect "undestanding" (beliefs), and need these videos to get you to do your job properly.  I showed why it is so important Dave makes these videos for us ordinary people, you showed why it is so important Dave makes these videos for you who work in this field.

Stuck like fly in amber, you are, and still smiling like an idiot.

Perhaps you should learn to read more.

[Nominal Animal]

Reported to moderators as rude hijacking idiot ^

Try convincing Simon to ban me.  He already hates me for pointing out how he blatantly lied in one thread.  That might work.

[David Hess]

It is not that simple. A switching DC-DC converter uses a few to several tens of mA for itself. Often this means it is actually less efficient compared to an LDO in that situation. If you want to power a microcontroller from USB an LDO is the most efficient solution in many cases.

There are many ways to lower the quiescent and operating current of a switching regulator.  Quiescent and operating currents of 10s of microamps or lower are possible:

1. Hysteretic regulators have no oscillator and because they require no frequency compensation, they also have no error amplifier.  Note that because hysteretic and constant off-time regulators require no frequency compensation, they do not care about operating in discontinuous conduction mode which is common at low output current.

2. Shut down the oscillator and error amplifier between "burps".  This is a combination of a conventional switching regulator and hysteretic regulator.  Linear Technology calls this "burst mode".

3. Bootstrap the regulator's supply from its low voltage output after startup and then disconnect the startup supply.  Bootstrapping from the output or a dedicated winding is common in off-line regulators which start from a small capacitor charged from the high input voltage.  This technique is as old as transistors.

Switching regulators can be competitive in micropower applications but often a linear regulator is used for lower cost and ease of use.

[EEVblog]

Like it, or not, it will likely take legislation to change this.

Almost certainly.

[floobydust]

Engineers gone down the rabbit hole. The power supply that is say 25% more efficient is better for the environment?
You have to compare a high and low efficiency PSU from end to end. The fact that the capacitors fail after 2 years while the low efficiency PSU is still running for 5+ years, consider the manufacturing, recycling, disposal costs and the coal burned to do that.
You save 25% on energy and global warming during the PSU's running life - but had to use two power supplies after the high efficiency one failed due to cheap parts.

The local Eco Electronics Recycling Center is full of cheap consumer electronics that had a short life. I dumpster dive, put in new electrolytics and it's good for years.
I think longer lifetime electronics benefits the planet more than a bit higher efficiency.

[EEVblog]

The local Eco Electronics Recycling Center is full of cheap consumer electronics that had a short life. I dumpster dive, put in new electrolytics and it's good for years.
I think longer lifetime electronics benefits the planet more than a bit higher efficiency.

Probably, but full proper life cycle analysis is hard to calculate.
Any good engineer will shoot for both though.

[Nominal Animal]

Small USB gadgets using an LDO to drop 5V to 3.3V and consuming 200-400mA (pretty common for anything with a radio) in small enclosures definitely have shorter lifespans due to heat issues.  (I don't have any details, but I seem to recall many of the early USB WiFi dongles had issues with that too, and quickly switched to buck converters.  Not so much for energy savings, but because in such small plastic enclosures (no airflow inside, with the plastic acting essentially as an insulator), the heat was causing them to fail.  I could recall wrong, though.)

[thm_w]

Engineers gone down the rabbit hole. The power supply that is say 25% more efficient is better for the environment?
You have to compare a high and low efficiency PSU from end to end. The fact that the capacitors fail after 2 years while the low efficiency PSU is still running for 5+ years, consider the manufacturing, recycling, disposal costs and the coal burned to do that.
You save 25% on energy and global warming during the PSU's running life - but had to use two power supplies after the high efficiency one failed due to cheap parts.

The local Eco Electronics Recycling Center is full of cheap consumer electronics that had a short life. I dumpster dive, put in new electrolytics and it's good for years.
I think longer lifetime electronics benefits the planet more than a bit higher efficiency.

I'm not sure why you instantly imply the high efficiency PSU is less reliable. The low efficiency has two clear wear items, main capacitor and electrolytic. A high efficiency version may be more or less reliable, there is no way to say for sure without going into more details.

BTW I tried to get electronics from my city recycling center, but they won't let you remove anything from there, sadly.

[amyk]

More components, more complexity, more points of failure.

My experience with LED lamps shows the ones with capacitive droppers and otherwise minimal circuitry have outlasted the more expensive ones with a full SMPS, and the failure mode has always been some active component in the latter.

[Alti]

I'm not sure why you instantly imply the high efficiency PSU is less reliable.

In the context of a smoke detector: power, power factor, reliability, installation cost - these are all factors that allow customers to decide whether they want Kwality or Miele, based on their subjective oppinions. Whether they are aware of power factor or reliability or installation cost at purchase time - that is another topic for another discussion. The undisputamle fact is that every smoke detector draws some power, has some reliability, etc.

From systems engineering we know that (with fixed reliability of the designs considered) an increase of design complexity imposes using more reliable components, w.r.t. a base design. So the net effect is that with smoke detector the BOM cost includes a factor that is a square function of complexity. First you pay for higher count of components and then you pay for higher grade of those components (usually achieved by underrating).

So the outcome is that (for unaware customers that are purely interested in purchase price) this pushes the smoke detector market in the direction of simple, less efficient designs that barely meet warranty period.
Surprised?