electronics ethics

[Svart]

ok, here's the question:

more parts

or

less parts

in your designs?

I am working on a new design and have come to a crossroads. either I can do something with one part and make it work just fine, or use 6 parts instead of the single part and do this in a more proper way.

which would you choose?

 

[clintrubber]

What kind of circuit ? What's the intended use of what you're building ? BOM on a budget ?
Is it a design with potential in the LessIsMore-category ? Will those added components make it more reliable or can just give problems more quickly because there's more stuff involved ?




Things have obviously changed enormously, I first realized that when seeing the (integrated) component count for a chip that just did Teletext-ghost-image cancelling. I figure that function alone outnumbered the component count for complete earlier TV-sets...
So obviously parts-cost has changed the way of designing - especially when things become integrated. Doesn't mean they always become better, another 'obviousity'. Things done with lots of parts are often only done because it could be done.
Designs with parts serving just one function at a time are of course a lot less elegant than those triple-used expensive active components of longer ago. :thumb: Must have given those designers a good feeling.

 

[NewYorkDave]

Ask Madman Muntz :wink:

http://www.smecc.org/mad_man_muntz!.htm

 

[gyraf]

My own rules-of-thumb:

In the audio path, less is most often more.

If it's in support circuit like sidechain, psu, metering, and so on - go for maximum flexibility, independent of component count.

Jakob E.

 

[Peter Simonsen]

To me..theres only one way..It takes what it takes..no more..noo less..I dont give a rats ass as long as it sounds the way I want it to :grin: everything else imho is for marketing and such

Kind regards

Peter

 

[Svart]

ok, here's the scoop.

this is not an audio circuit, this is for work. it's part of a motor control circuit.

a little background on the circuit:

H-bridge current amp topology, sensitive SCRs sourcing and logic MOSFETs sinking. For this circuit there is no PWM control, just ON/OFF. the circuit is designed to drive 24vdc @ 10 amps in both directions from a DC-DC converter module. the gate control is optoMOS isolated and the MOS switches are powered from a lin Vreg straight from the 24vdc rail. this is the part that I am uncertain of. I can continue to use this design as it stands right now and go into production with a working unit or I can redesign the gate control circuits to handle rails up to hundreds of volts DC without too much fuss, but I also will need to go discrete and this will add about 6 parts to the MOSFET gate control.

so the question is, do I plan for the uncertain future and design a discrete gate drive stage for the MOSFET or do I just get the job done and have to tweak the design later if I need another one?

 

[clintrubber]

[quote author="Svart"]so the question is, do I plan for the uncertain future and design a discrete gate drive stage for the MOSFET or do I just get the job done and have to tweak the design later if I need another one?[/quote]

Hmm, for work, then other stuff enters the equation... the stuff that shouldn't matter but still needs to be taken into account :evil:

So as far as possible to judge from here & sorry for stating the obvious:

* You might be anticipating for an easier situation later on (good, people will thank you for that, but praise will not be extreme).

* Hmm, but then it turns out your anticipating gives a problem for the 'present' model. That might have been nice later on, but now there's a problem that wouldn't be there if you had kept it simple. (bad, 'you st*p%d #ssh*l@' might describe reactions best).

* How unconvenient is it to have the design modification later on ? Everybody will buy the story that you didn't want to risk the present situation but now have to adapt the design. (not glamorous, but likely the best compromise).

Too bad 'politics' need to be taken into account here, but we all know it can't be ignored for things like this....

Bye,

Peter

 

[alk509]

[quote author="NewYorkDave"]Ask Madman Muntz :wink:[/quote]

Nice reading. Thanks, Dave!

[quote author="NewYorkDave"]

[/quote]

RIGHT-FUCKING-ON!!! Glad it's back. :thumb:

Peace,
Al.

 

[tk@halmi]

Dude, where are the requirements for this poject?
What are the "ilities" and performance requirements? (reliability, maintainability, etc.)

 

[PRR]

Well, if I'd ordered you to design a 24V module, and later learned you wasted an extra day and $0.20/each designing a 24V-400V module, I'd call it insubordination and dock your pay for the wasted day and the increased production costs.

I'd feel better if you came to me last week for a 5-minute chat. "I can meet the spec as written, or for another day and $0.20 I can design a wide-range module; is that useful to the company?" He may say "Yah, that would be nice" or "NO! We'll never use hundred-volt rails!" He may be wrong either way, but that's his job. You done your job by alerting him to a possible low-cost improvement.

At another level: You can get paid to design the 24V module, and later get paid to design a 400V module, and still later get paid to consolidate the two inventories into a wide-range module. Paid in salary or, maybe better, in credit and fame for incrementially improving the company product. Don't give your boss something extra without getting something back.

At yet another level: it was "unethical" to ask this here without context. An answer appropriate for studio tracking may be very different from how you design a stamping machine. In the studio, if something fails, you find a work-around. We have lots of boxes, several alternatives, and few failures are dangerous. Heavy machinery really should never fail in a way that will injure workers or cause large factory downtime. Increased parts and wide-range design will, statistically, reduce reliability. 100 resistors with 100,000 hour MTBF will have a failure every 100 hours, 200 resistors will have a failure every 50 hours. High-voltage design is typically very different from high-current design, so wide-range has compromises. However specific added parts can actually improve reliability: limiting and protection sub-circuits sometimes do more good than harm, and putting 200V parts in a "24V" module might save your butt the day the line voltage rises to 30V or spikes to 200V.

 

[Svart]

awesome PRR.

You are right about everything. there are many ways to look at this and all are certainly correct in their context.

but another thought on context.. I asked the question in the way I did for a reason. I did not include context as this in my view would allow people to give a first impression without over analyzing the situation, but I see now that there needs to be a degree of analysis beyond just what a person would choose for themselves.

The design is not one that would "hurt" a person in the normal sense. This circuit interprets a digital control signal and then responds with relay and motor movement. The motor is actually a linear actuator that moves a camera head up/down on a robotic arm. the total weight of the structure is around 30lbs and is assisted by gas struts. the only harmful part of the structure is that it is a scissor jack-like device and could pinch a person's fingers, but the spacing and dullness of the finished metal parts would make this unlikely to do much more.

the microcontroller portion was completed by myself and a programmer, and now I am finishing up the current amplifier portion. High reliablilty and low electrical noise are key. I have also spec'd parts that are rated over 200% of the needed values.

:thumb:

 

[dale116dot7]

I design automotive engine controllers for a living, and I like to use the cheapest circuit that gets the job done. Or the fewest components. Here's my philosophies on design of this sort of circuit. My list is actually very long, each mistake I make (or someone else has) adds to the list. Take this as you wish....

A resistor costs a tenth of a cent, but it costs seven cents to put on the board. A resistor pack with four resistors costs maybe two cents but you get four resistors with a single pick-and-place operation (seven cents for all four). Board space costs money. Fewer parts and more integration = better.

For a 12 volt automotive system, the parts had better be rated for 60 volts at minimum because it'll see that. Even if someone questions why you are using a 60 volt FET for a 12 volt system. Doing anything less will ensure you have field failures. For a 24 volt system, look for 100 volts or adding a brute-force surge suppressor in there. Reverse polarity protection and reverse spike handling is not optional.

Similarly, the current ratings of self-protected parts (a protected MOSFET like the IPS021, for example) should be double the load. If it's driving an inductive load, ensure the repetitive avalanche energy is rated for the service you are looking for. In other words, if you are running a solenoid at 8 amps and cycling it at 50 Hz, you will need somewhere around a 75A MOSFET to take the abuse. Ensure that the MOSFETs are avalanche rated. Raising the cost of a MOSFET by 0.50 (to get a good avalanche rating) is better than trying to add 0.35 worth of external clamping. All outputs need short-circuit protection. MOSFETs always fail shorted.

 

[bcarso]

I agree with PRR and dale166dot7. The application of these criteria however will depend on anticipated volumes. Also, the cost of insertion can be widely variable, especially with Far East manufacturing, where it is much lower than for domestic automotive production.

When I was designing cheap powered speakers I had to look at the cost of my time and the so-called opportunity cost, if I wanted to be able to justify my decisions (I had a lot of autonomy based on track record and adherence to deadlines---normally one's boss or his/her boss would make these decisions). If we were going to ship 2 million a year, every penny saved represented 20k dollars. If it took another week, even, to eliminate a penny, it was worth it on the face of it---BUT only if you weren't running out of time. And of course, usually you were. The development cycles were typically 4-5 months from concept to volume production. Needless to say if you didn't know what you were doing and missed a failure mechanism, particularly one that takes a while to show up in life testing, you were in the proverbial World of Hurt.

The first speaker I did for Dell was a very conservative design, but was still very cheap, given its feature set of volume, bass, treble, and headphone output. Also, the EQ for the speaker changed when you used a subwoofer, and the headphone output was not subject to the midband EQ used with the speakers. But in some crucial areas the use of the power IC for example was pretty much out of the app circuit in the databook.

As time went on I got more creative in a way that reduced parts count and enhanced performance. In some cases the circuit got more obscure-looking and hence more intimidating to ripoff artists.

But then there is opportunity cost---what could I have been doing with other products and ideas if I weren't squeezing pennies out of other things.

 

[Svart]

good advice!

In testing, I have seen the inductive kickback from the 24vdc/3A motors create over 100v in picoseconds. I have also seen them pull 20A during startup and regen. given this, I use only 200v/+15A MOSFETs and actually plan on moving to IGBTs now that the number of small IGBTs can handle severe power situations have increased. flywheel diodes on all parts, TVS devices, picofuses and the like are all integrated into the system. the SCRs of course are much more hardy than the MOSFETs and the IGBTs, but they short CLOSED usually.

 

[CJ]

Dude, get a Gecko drive and be done with it.
One hundrud dollah.

 

[dale116dot7]

[quote author="Svart"]good advice!

In testing, I have seen the inductive kickback from the 24vdc/3A motors create over 100v in picoseconds. I have also seen them pull 20A during startup and regen. given this, I use only 200v/+15A MOSFETs and actually plan on moving to IGBTs now that the number of small IGBTs can handle severe power situations have increased. flywheel diodes on all parts, TVS devices, picofuses and the like are all integrated into the system. the SCRs of course are much more hardy than the MOSFETs and the IGBTs, but they short CLOSED usually.[/quote]

Here's some more stuff on what I've been dealing with, but like I mentioned before, I'm doing all automotive stuff.

IGBT's have a pretty high forward voltage drop, though, usually for anything less than 100 volts I like to use a MOSFET - but one that's avalanche rated. Then those 100V+ spikes get absorbed nicely without much effort on your part. IGBT's and MOSFETs also all blow up in an on condition. Also, I usually use logic-level MOSFETs when I can get away with a lower speed - most of my loads can. So then I can go from a logic gate or a micro through a resistor (47 ohms or so) and to the MOSFET. Mostly when I need an H-bridge I only need 3 to 5 amps so I use a TLE6209 single-chip IC (Infineon). The only place I use IGBT's is to drive ignition coils, and I use a self-clamped, logic-level, ignition IGBT.

I'm reluctant, though, to use a driver circuit that does not have hardware dead-time and cross-conduction prevention. Especially with any microcontroller by M****c**p. Bad experience. EMI + micro (or bug) may mean runaway code, which will often result in an explosion if you don't have hardware to deal with it. There was a watchdog but still an explosion. Oops. Put that on the list.

The other thing is even with 10 or 20 amp loads, I normally don't opto-isolate anything. The cost is too high in automotives, so creative grounding is done instead. The engine control module is the centre point of a star ground. That doesn't work in many applications, but that's sort of the norm that I need to work to.

 

[Svart]

good stuff!

Dale, I have found no other way but to isolate the two circuits because of the sensitivity of the grounding on the video circuit. when using a DC brushed motor like this, signifigant (current) noise is introduced to the video shield. the current source is also being sourced from another circuit, when needed, through a relay. the digital signal is interlaced into the video signal during the blank periods and sent to the receiver. this gets rid of noise during the video trace time and also uses fewer wires.

so the action works like this:

control signal given via control panel, signal transmitted, signal received, signal interpreted by MCU, relay actuated, wait 1 sec for contact bounce and DC-DC settling, MCU gives current amp direction signal, motor moves.

I am moving the direction of the design to twisted pair video (i already have working designs of this) so motor noise should not be a concern anymore.

concerning FETs, yes I've seen them blow up in the ON position but in my analysis it's more of the drain punching through to the gate and triggering it to stay ON rather than the body diode or D-S problems. same outcome though. :sad:

You are completely right about the low voltage Hbridge ICs. I should actually take a look at them in this application but I had initially hoped to use this design for more than one application, thus the reasoning behind making it as stout as possible. but a couple problems come to mind, mainly the problems I have seen with voltage/current spikes during regen.

but..

I just found the IRF3220 bridge and irf7484 FETS..

ordering some now to try out. I'll give them a try and see how it works.

thanks again!

 

[Svart]

got the ir3220 and matching FETs. they work very well. I proto'd a small board this morning and had it ready for the delivery of the parts. hooked it up and ran it through the normal tests. now i'm doing destructive testing.. my favorite part.. :green:

but i have yet to let out the factory smoke.

I think I am going to go with these parts instead, this will actually save money and PCB real-estate.

thanks for all the help!