EEVblog #1284 - How Bad Product Design Kills The Environment

[EEVblog]

Well, there you go, that's the difference between the two labs. Just lots of high frequency crap, otherwise identical.

[ali_asadzadeh]

you can use an ultra low cost Flyback chip in the non-isolated buck mode, ST have app notes on that, for example you can use Viper12 part, and it can supply in the order of 100mA or more,total cost would be under 1$

[GeorgeOfTheJungle]

Well, there you go, that's the difference between the two labs. Just lots of high frequency crap, otherwise identical.

Looks much nicer in LTSpice:

[GeorgeOfTheJungle]

Isn't this the power that's being fed back into the grid every 1/4 cycle?

[ckambiselis]

What is the failure rate of the crappy solution "they" have used compared to the ones you showed, could that be a reason they went with it?

Bordering on nothing in it really with a zener vs a HV regulator. I guarantee it's because of cost.

Was thinking more of a government regulation on safety systems, for them to be able to sell it as a smoke alarm, it must have a failure rate smaller than x%.

Oscar

[gf]

Looks much nicer in LTSpice:

Your current has double frequency of voltage. This was not the case in the posted oscillogram. Something must be different in your circuit, IMO.

[GeorgeOfTheJungle]

Your current has double frequency of voltage. This was not the case in the posted oscillogram. Something must be different in your circuit, IMO.

It's power V*I not current. Think about it: every cycle, the capacitor charges to +V, discharges to 0, charges to -V, discharges to 0, ergo 4 power peaks, two positive, two negative, and double the frequency.

[gf]

It's power V*I not current.

Sorry, misssed that. Reactive power alternates of course between positive and negative, with zero average (the average is not supposed to be zero here, as there is also a real power component in the play).

[tszaboo]

What is the failure rate of the crappy solution "they" have used compared to the ones you showed, could that be a reason they went with it?

Bordering on nothing in it really with a zener vs a HV regulator. I guarantee it's because of cost.

Was thinking more of a government regulation on safety systems, for them to be able to sell it as a smoke alarm, it must have a failure rate smaller than x%.

Oscar

They should require SIL for smoke or CO alarms. Honestly, they should also require gas detectors, for houses with gas heating, and the gas detector should have SIL, ATEX/IECEX. And all these alarms would need a remote alarm for the nearest fire brigade.

[TerminalJack505]

The capacitive loading is not a real problem. Usually there are extra capacitors near the distribution transformers to compensate for the inductive loads and the magnetizing current of the transformer. So in most cases they can just make those caps a little smaller. Smoke detectors tend to be on all the time, so no need to adjust that very often.

However the real power taken from the grid is still rather high. So he design is still bad. It is odd to see only half wave rectification - with full wave rectifier they could have reduced the expensive capacitor to half. - though with a little more loss.

The capacitor drop supply is just not good for something that usually needs low power, but sometimes needs much more.

BBM.

Maybe they add two more diodes and do full-wave rectification for the 120VAC version but keep all of the other components the same?

I don't see how that would be more economic than changing the cap but that's one reason I can think of for using half-wave rectification.

[SeanB]

Current waveform distortion is from all your local grid tie solar installations, feeding power back into the grid. If you want a higher capacitor current to use in the measurement simply use a 20-50uf motor run capacitor across the mains, it will have a very high current flow through it, and will survive direct connection for a long time. Higher current, better resolution on the higher current ranges on the current probe, and you can use a standard current transformer as well to measure, as the current will be in the order of 10A or so for the 50 uF unit.

The solar inverters probably have had some cost cutting done on the output filtering, removing the high cost LC filtering  by removing half of the LC sections, meaning they cause more AC current distortion.

[floobydust]

This is the schematic of capacitive-divider for the smoke alarm I felt recall-worthy. It's a special smoke alarm having an aux. relay for alarm-bell tie in, so it wastes over 1W. This cooked the pcb phenolic and plastics over a couple years. It is imported from china by "American Sensors"/First Alert and quite a POS with bodges etc. on the board. Note sim has the relay on, 120VAC mains.

You design a capacitive-divider for highest load. All of us use mains frequency for the math- which is not correct when you have HF harmonics present. So the zener and resistor cook even more because they waste the excess energy purely as heat.

[coppice]

This is the schematic of capacitive-divider for the smoke alarm I felt recall-worthy. It's a special smoke alarm having an aux. relay for alarm-bell tie in, so it wastes over 1W. This cooked the pcb phenolic and plastics over a couple years.

Did it cook because it was dissipating 1W, or because the voltage waveform in that place was quite distorted and the dissipation was way up? Well designed things with cap drop supplies use large resistors well spaced from the board and case, even when those resistors normally run very cool. That way when there is a lot of harmonic distortion on the supply waveform they don't overdissipate and cause damage or fires.

[floobydust]

Here's evidence smoke alarms contributing to global warming. The resistor is a steady 1W. Something started giving off a tiny bit of smoke and false triggering itself, seems to be high line, as harmonics only show up when a huge industrial phase-control heater is running. That may have aggravated it.

[David Hess]

It is fascinating to think of the various ways to make a suitable high efficiency switching regulator.  The high switching voltage increases charge pumping losses according to 1/2cv^2 which becomes overwhelming at low power.

A series resonate switching regulator might be suitable but one thing I have seen in very old blocking oscillator based inverter designs, which are commonly used for electronic ballast now, is a charge recovery circuit which extracts residual charge after the switching cycle and dumps it back into the supply.  I do not think it would help the power factor except through being orders of magnitude more efficient.  But just the output transformer would probably cost more than all of the parts in the existing poor circuit.

[kcbrown]

That TI part sounds interesting and all, but at $1 each in quantity, I can't help but wonder how a solution built around it can outcompete a low-power SMPS, which I'd expect to have roughly the same BOM cost in quantity (since SMPS controllers are cheap as chips these days, and the rest is jellybean stuff).

Of course, the SMPS might not have the reliability you want for a mission critical device like this.  Then again, isn't that what the battery is for?   

[GeorgeOfTheJungle]

No, the 80mA is real current that will have I²R losses in the transmission system until it is compensated for (lead/lag and harmonic). If you have say 5 in your home at 80mA a pop that's 0.4A that must be delivered from the grid infrastructure.
The lead/lag compensated part can be done locally at your home with other leading phase device, but the harmonic content usually cannot, so it will go back down the grid resulting in losses until it's filtered out.

I wonder, what would that power analyzer say if you connected it to the output of a solar inverter? Have you ever checked that?

[EEVblog]

Nice for a video down the track if you could build up those other two solutions (TI and OnSemi iirc?)  and then measure the three different options. Four if you count the full bridge.

Just for you!

[coppice]

That TI part sounds interesting and all, but at $1 each in quantity, I can't help but wonder how a solution built around it can outcompete a low-power SMPS, which I'd expect to have roughly the same BOM cost in quantity (since SMPS controllers are cheap as chips these days, and the rest is jellybean stuff).

Of course, the SMPS might not have the reliability you want for a mission critical device like this.  Then again, isn't that what the battery is for?   

Have you talked to TI about buying a million parts? If not, you have no idea whatsoever what the volume price really is. However, you know by inspection that its not as high as $1, as there would be no market for it at that price.

[Per Hansson]

70mA capable makes sense:
- 15mA peak as Dave stated, add 2x for safety factor, then add another 2x for degradation of the capacitor over time (10yr).
Big clive has seen a number of these capacitors where they have degraded to the point of not being able to supply enough current.

Enjoyed the video Dave, thanks.
No reason to put up with inordinate power usage when solutions exist, even if they initially cost a few cents more.
The idea of "built to a cost" and "cheap as possible" from mains power for safety equipment that should be reliable made me think of this pic.

This whole mains powered smoke alarms just sounds like a solution to a problem that does not exist to me.
There are so many ways it can fail: lightning strike, X2 capacitor degrading, or just overheating and catching fire as shown earlier in the thread because zeners are quite unreliable components!
But really I think the X2 capacitor degrading is the worst part, so you have this mains powered device with a weak battery.
And the X2 capacitor degrades so that it works fine in ordinary use but when the buzzer kicks in it just browns-out due to insufficient current.
Yup, great design for a device supposed to save lives!

[ogden]

Nice for a video down the track if you could build up those other two solutions (TI and OnSemi iirc?)  and then measure the three different options. Four if you count the full bridge.

Just for you!

You just demonstrated typical trap for young players of "Design By Inspection". Engineers who made 80mA dropper instead of 1mA, made it for for some quite obvious reason - ALARM. You definitely can't produce audible alarm using 15v @50uA supply, unless you are building "smoke alarm earbud". Only actually making whole device and measuring SPL, you will see how much power is actually needed. Could be so that 10.5mA @ 15V of NCP785A is not even enough.

[gf]

You just demonstrated typical trap for young players of "Design By Inspection". Engineers who made 80mA dropper instead of 1mA, made it for for some quite obvious reason - ALARM. You definitely can't produce audible alarm using 15v @50uA supply, unless you are building "smoke alarm earbud". Only actually making whole device and measuring SPL, you will see how much power is actually needed. Could be so that 10.5mA @ 15V of NCP785A is not even enough.

What about a low-current dropper (say 1mA or less) + a couple of Farad supercap which can store enough energy for 15min sounding?

[coppice]

You just demonstrated typical trap for young players of "Design By Inspection". Engineers who made 80mA dropper instead of 1mA, made it for for some quite obvious reason - ALARM. You definitely can't produce audible alarm using 15v @50uA supply, unless you are building "smoke alarm earbud". Only actually making whole device and measuring SPL, you will see how much power is actually needed. Could be so that 10.5mA @ 15V of NCP785A is not even enough.

It seems like he forgot that he has already measured the peak consumption during beeping as 15mA.

[coppice]

You just demonstrated typical trap for young players of "Design By Inspection". Engineers who made 80mA dropper instead of 1mA, made it for for some quite obvious reason - ALARM. You definitely can't produce audible alarm using 15v @50uA supply, unless you are building "smoke alarm earbud". Only actually making whole device and measuring SPL, you will see how much power is actually needed. Could be so that 10.5mA @ 15V of NCP785A is not even enough.

What about a low-current dropper (say 1mA or less) + a couple of Farad supercap which can store enough energy for 15min sounding?

You would need a 15V supercap. That gets pricy.

[ogden]

What about a low-current dropper (say 1mA or less) + a couple of Farad supercap which can store enough energy for 15min sounding?

You would need a 15V supercap. That gets pricy.

Right. Supercap is "no go", dropper as well. NCP785A  is excellent, only it has max 10.5mA specs As a solution I see either external power transistor (?) or electrolytic cap + sound "chirps" with let's say 20% duty cycle resulting peak ~50mA sound driver current. - Obviously if such sound chirps allowed by regulations.