Silly designs that made it into production

[NiHaoMike]

I was examining a motherboard out of an old Thecus file server when I noticed a second GD75232 on the board with a HIN232E next to it. The latter turns out to be a MAX232 clone. So the TTL level signal from the I/O controller gets translated to RS232 and right back to TTL level less than an inch away before going to the front panel LCD controller. I suspect different design teams worked on the base system and the front panel controller, then the unnecessary parts wasn't picked up after the designs were combined into one board.

[vk6zgo]

One place I worked specialised in such things!

A control generated a PWM signal,which they wanted to input to their "mainboard".
Their answer was to feed it to a  passive integrating network,then to an ADC.
When I asked,"Why not sample the PWM directly?",the "powers that be" were not impressed!

In another instance,we used to spend about half an hour painstakingly soldering a diode network to a group of pushbutton swirches.
When they finally let us see the mainboard schematic (worried about IP),I noticed that none of the other input switches used these diodes.

They were,in fact,a last vestige of some previous idea which had been dropped.

I actually had a win there,as they stopped using them,cutting the preparation time of these switches to about 5 minutes.

There were a number of other silly ideas which made it into production.

[coppice]

I was examining a motherboard out of an old Thecus file server when I noticed a second GD75232 on the board with a HIN232E next to it. The latter turns out to be a MAX232 clone. So the TTL level signal from the I/O controller gets translated to RS232 and right back to TTL level less than an inch away before going to the front panel LCD controller. I suspect different design teams worked on the base system and the front panel controller, then the unnecessary parts wasn't picked up after the designs were combined into one board.

This kind of thing is quite common when a draughtsman has been told to merge 2 PCBs into one. The only motivation an engineer would have to look for possible simplifications is if management is demanding a cost reduction.

[westfw]

There was a cisco 16/32-port serial card (for 2600/3600 series routers) that had 6 signals worth of max232 for each port.
Maybe we had a good deal on those chips, but it seemed to me that if you've got 16 ports, you really ought not have a separate voltage converter for each one...  I think the main power supply had +/-12V, too.

I'd also claim that the Arduino has some rather silly aspects to its design...

[tszaboo]

OK how is the design?
It is in the prototype phase.
Does it work already?
No that is just a status LED.
So it is working?
No, half the function is not tested.
So most of the stuff is working, and we can start the production.
No, that would be a big mistake.

..few weeks later...

OK, so we made an entire batch of the boards but they don't work, can you tell us why?
There was 150 change request after we tested the prototype, that is why.
Well we are already making this, making a change would be too expensive, so just make this work.

[T3sl4co1l]

One place I worked specialised in such things!

The Ministry of Silly Circuits?

"Design this way, please!  [trots off ludicrously]"

Tim

[German_EE]

The AGFA Horizon A3 scanner used to have all sorts of intermittent errors but when an engineer came along and plugged his laptop into the test port all the problems went away. Eventually they discovered that some of the timing loops in the software were 'off' and fitting the test lead (which had RTS and CTS shorted altered the timing just enough to make the problems vanish.

From then on part of the preventative maintenance procedure was to check for that RTS-CTS link and if it was missing then fit one.

[tszaboo]

I saw a product with a very elaborate set of logic gates that implemented some multiplexing and inversion. The output was a single digital signal that went to a microcontroller... Which had more than enough free pins to implement the entire thing in firmware. The designer really didn't trust the firmware developers. He would do things like fit a comparator to the board when the micro already had one, just so that he could control it instead of relying on the software engineer to program the micro correctly.

Honestly, I'm starting to think that the designer was right.

[SeanB]

You find a lot of industrial instrumentation that take a digital value, convert it to a voltage and then to a 4-20mA current, only to pass it to the box next door where it is converted back to a voltage then fed into an ADC. Just because each part is made by a different manufacturer ( or is designed to interface to a standard externally if they are actually the same manufacturer) to an old standard, or the next step of a digital interface is too steep a step up price wise so they talk the same data digitally. 

Only good thing is you can take that 4-20ma signal, feed it into 5km of dry loop telephone line and use it at the other end with no problems.

[SirNick]

I've done stuff in HW that I could've done in SW because I felt latency or bugs might impact the performance unacceptably.  I was a little miffed at myself for choosing the HW solution over my SW, but it turns out "your mom" jokes and hair-pulling aren't very effective when you're arguing with yourself.  I just stopped talking to myself for a while to teach me a lesson in diplomacy.

[tszaboo]

Honestly, I'm starting to think that the designer was right.

Is that an ad hominem or just a general comment on embedded software engineers? Many of us are both SEs and EEs, so I'm wondering if I should be half or fully insulted.

Sorry I've been a bit harsh. I've had a meeting today discussing why our product is still not working. Turns out, because the company's awesome code monkeys software engineers still are nowhere with the code.
They just cannot solve really simple things. Like if the analog input is half the full scale, then please set the programmable amplifier to have twice the gain (and they whine that we are not according the specification). Or calibrate multiple things together.
But oh, no, they decided that they are not going to do that. Or having products where the bootloader takes 20 minutes to finish, and you need CAN, a server and a fracking spaceship to do it. It has USB, and ethernet, they will not bother with it.
Every time the product is finished, I feel like I gave them a full toolbox, and only the hammer gets used. Sight.

[vk6zgo]

You find a lot of industrial instrumentation that take a digital value, convert it to a voltage and then to a 4-20mA current, only to pass it to the box next door where it is converted back to a voltage then fed into an ADC. Just because each part is made by a different manufacturer ( or is designed to interface to a standard externally if they are actually the same manufacturer) to an old standard, or the next step of a digital interface is too steep a step up price wise so they talk the same data digitally. 

Only good thing is you can take that 4-20ma signal, feed it into 5km of dry loop telephone line and use it at the other end with no problems.

I've certainly had to do stuff like this when connecting various incompatible interfaces where you can't (or aren't given the time to) get inside stuff.

On one occasion we modified an old Marconi Transmitter for Automatic operation.
It had never been intended for such service,& had separate Sound & vision Tx in the same box,each with their own control supply,one being 60v AC,the other 60V DC.

A PLC was configured as a "Virtual Technician" to perform all the actions a real Tech would do in the morning,at closedown,& under fault conditions.
We discovered the cable we had bought to do the job was only rated at 30v,so we made up a relay box to switch the 60V control circuits.

Add to that,video & sound fail alarms interfacing into vision & sound switching which was never intended to have a second control input,plus a lot of other stuff,& it was a complete "sh##t mix".
Worked well for around 7 years,though!

[T3sl4co1l]

Moral being, assuming that an average engineer (in any discipline) is as capable as yourself* is usually a bad idea.

*Dunning-Kruger effect notwithstanding.

Or, do as much as you can, as well as you can, even if it extends beyond your scope or responsibility.  Of course... most people won't appreciate your overreach, and it may cause negative rather than positive response from management.  Not that you'll be compensated for your initiative, anyway.

Tim