Chips/parts you hate working with...

[tszaboo]

[/i], datasheets which show the pinout looking at the chip from below

Oh yes. Mr Smoke there you go, revision 2 here I come.
I also dislike the SOIC, where the orientation is where the mold is thinner (becasue just printing a dot is so 20 century), but my favourite are the ICs where the datecode or some other BS is on the bottom.
For the Infineon haters, I totally agree. I refuse to design with anything infineon which has more than 2 PN junction or three pin.

[Alex Eisenhut]

Chip-like parts with clockwise pin numbering from the top-right.

That was a Mini-Circuits part, years ago...

[wraper]

Minor gripe:  TI's TPS54226 switching regulator.  There's an Enable pin, and a Vreg pin that supplies 5v for control circuitry.  The Enable pin is diagonally opposite from the Vcc pin (with a ground pad between) but in my layout, the Vreg pin was fairly easily accessible, so I thought "OK, I'll tie Enable to Vreg".  Predictably, it turns out Vreg is ALSO disabled when Enable is LOW, so my prototype board had a bit of a chicken and egg thing going on.

To be fair, I realized this was a possibility, and combed through the datasheet looking for confirmation one way or another -- and therein lies my chief complaint.  There isn't one word about this either way.  In the block diagram, Enable goes to a single isolated block labelled "Enable Logic".  In the application circuits, Enable goes to an external header.  In the timing diagrams, the sequencing is shown between Enable, Vout, and PowerGood, but not Vreg.  In the text, not even a single paragraph is devoted to Enable's function.

For an essential (albeit trivial) signal, there's not much attention given to how it works.

Actually there is written quite enough. Part of it:

TPS54226 discharges the output when EN is low, or the controller is turned off by the protection functions (OVP,
UVP, UVLO and thermal shutdown). The output is discharged by an internal 50-? MOSFET which is connected
from VO to PGND. The internal low-side MOSFET is not turned on during the output discharge operation to
avoid the possibility of causing negative voltage at the output.

In the specs max EN voltage = Vin
And how do you imagine that EN would work at all if there is no high voltage on it. Voltage cannot appear on the output in principle with EN disabled.

[SirNick]

Yes, the datasheet at least tells you the acceptable voltage range of En, and that it will disable the output.  However, I don't follow how your quoted text defines the relationship between En and Vreg -- being the main point of my gripe.  (To be clear, Vo != Vreg.  They are two distinct regulators.)

On one hand, you might assume Vreg will be disabled -- it's a regulator, and you might want to have control over whether it's enabled.

On the other hand, you might assume Vreg will NOT be disabled -- it exists for control of the switching regulator, so it should be available to... control the regulator.

It took a failed prototype to determine that Vreg is in fact disabled when Enable it de-asserted.  No biggie, but that would have been nice to know before wasting ~$20 on a board and some parts, and a couple weeks for turnaround.  In hindsight, I should've just broken out the pins so I didn't have to re-spin the board if I guessed wrong.  C'est la vie.

[wraper]

Now understood that you talked about separate regulator. The thing is, seems that is not intended to connect anything to it at all except capacitor and PG resistor. Basically it exists as separate pin only because of the capacitor.

[David Hess]

I dislike large lead count parts which do not come in PLCC (J-Lead) packages and anything smaller than small outline packages except for sot-23 style packages.

I also dislike the 555 because it suffers from noise induced by cross conduction when switching.

[alanambrose]

>>> Curious which chip is that ?

The LT6017 in DFN - theoretically it should have top side marking but even under a microscope you can't see it so you have to look at the copper underneath - where there's a corner cut-off. One of the diodes I was talking about is the ZHCS400 which has BD written on it - B is the anode....

A.

[Odysseus]

When I add a 555 to my breadboard, I always do so begrudgingly.  The active-low reset and VCC (often shorted) are at opposite corners, the control and ground (decoupling cap) are at opposite corners, and the trigger and threshold (often shorted) are on different sides. All this means I can never satisfy my desire to make very neat and compact breadboard layouts with a 555.  I really wish there was a more breadboard friendly variant.

[Psi]

Any SMD part that doesn't have exposed pins. Can't measure anything, can't remove it (easily) and can't visually inspect the connections.

+1

[tszaboo]

LCDs and I2C devices. In both cases you can guarantee that the designer was suffering from severe brain damage when he created the interface. With LCDs they usually require you to use annoying image formats and poorly documented commands, mostly to save money in the controller hardware. With I2C devices there isn't even a standard for basic stuff like register access, so everyone comes up with their own slightly different and incompatible system.

I remember once working with an I2C LCD. It had a 16 bit register named as "contrast". After suffering for 2 days, I realised that this register has to be written with the correct value, and that is why nothing else seems to work. Yeah, why would be the reset value useful?
So then I started randomly writing numbers to it... nothing. Finally I wrote a code, which incrementally wrote every single value to this 16 bit, and I was sitting there with a stopwatch in my hand to figure it out which value was almost good. Ther was about 30 or so values which made anything visible on the screen.
Datasheet is useless chinglish.

[Alex Eisenhut]

LCDs and I2C devices. In both cases you can guarantee that the designer was suffering from severe brain damage when he created the interface. With LCDs they usually require you to use annoying image formats and poorly documented commands, mostly to save money in the controller hardware. With I2C devices there isn't even a standard for basic stuff like register access, so everyone comes up with their own slightly different and incompatible system.

There isn't even a good hardware spec, I2C is too old to account for all the various voltages used these days so you need I2C level translators liberally spread around your design.

[daqq]

On one side I love STM32xxx processors for their rich features, on the other side I hate them for their pinout from hell - seriously, how can you distribute one ~16 signal interface over THREE sides of a 4 sided 100 pin mixed up pretty much with everything right around the analog pins? There's a fair amount of similar examples - the SDIO interface is on two sides with pins not next to each other... seriously... WTF? 

[Davesaudio]

Anything small (QFN etc) with a thermal pad

[Gribo]

Antennas from reputable manufacturer without at least one radiation pattern in their datasheet.   

[pyrohaz]

On one side I love STM32xxx processors for their rich features, on the other side I hate them for their pinout from hell - seriously, how can you distribute one ~16 signal interface over THREE sides of a 4 sided 100 pin mixed up pretty much with everything right around the analog pins? There's a fair amount of similar examples - the SDIO interface is on two sides with pins not next to each other... seriously... WTF? 

+1

[Artraze]

On one side I love STM32xxx processors for their rich features, on the other side I hate them for their pinout from hell - seriously, how can you distribute one ~16 signal interface over THREE sides of a 4 sided 100 pin mixed up pretty much with everything right around the analog pins? There's a fair amount of similar examples - the SDIO interface is on two sides with pins not next to each other... seriously... WTF? 

I second the sentiment, but will say it's a pretty common problem for the larger microcontrollers... A couple years ago I was working with the LPCxxxx (don't recall) and its external memory bus was almost literally every other pin around the QFP.  It's pretty clear the when the manufacturers figure everyone is using a 4+ layer board and will just dump things on their own layer so the pinout isn't that vital .  I've actually found the STM32 F0 and F1 to be not that bad, in terms of at least bunching the GPIO in groups of 8 and generally keeping alternate functions close.  I've definitely seen much worse.

That said, the STM32F030F4P6 is a chip I (love to) hate working with.  It's an ARM in a TSSOP20 for ~$1 (esp via AliExpress).  The problem: 9 of I/O are basically just GP and all the dedicated functions are grouped on the other 6.  To add insult to injury, the bootloader ROM uses the only locations for I2C (and other funcs) for its UART, when the UART could be multiplexed on pins that have no other functions .

And with that I will add a more general gripe: datasheets that advertise a lot of functionality without mentioning that not all of it is actually usable, either because of shared resources or because it's simply not present on some versions of the device.

[SirNick]

I second the sentiment, but will say it's a pretty common problem for the larger microcontrollers... A couple years ago I was working with the LPCxxxx (don't recall) and its external memory bus was almost literally every other pin around the QFP.  It's pretty clear the when the manufacturers figure everyone is using a 4+ layer board and will just dump things on their own layer so the pinout isn't that vital .

Agreed!  I've been working with a few ICs lately that seem to be designed by people who enjoy making your life difficult.  "Keep the Clk pin as far away as possible from analog signals and power planes."  That would be the Clk pin between Vcc and A+?  Got it.  I'll try and steer clear of those two adjacent pins, then. 

My previously griped TI switching regulator has this problem too.  A much cleaner layout could've been achieved by juggling a few pins around.

[Prizmatic]

Toroid/binocular cores. I've noted the number of wire turns achieved before losing count is age related. 1,2,3,4,5,6.... 7,8,9...
Balls!

[microbug]

Tiny packages. I just added two ICs to my design which are only available in 1.5mm by 1.5mm WSON6 packages... (2xBQ29700)

I haven't tried yet, but I'll probably inhale them before I can solder them.

[nctnico]

Anything from Atmel. If their datasheet states in bold that a device runs from 3.3V it doesn't. It's more like 3.6V but that is only in the fine print.

[Whales]

Anything from Atmel. If their datasheet states in bold that a device runs from 3.3V it doesn't. It's more like 3.6V but that is only in the fine print.

It can, but at degraded performance/clock?

[nctnico]

Anything from Atmel. If their datasheet states in bold that a device runs from 3.3V it doesn't. It's more like 3.6V but that is only in the fine print.

It can, but at degraded performance/clock?

Not even that.

[rob77]

Anything from Atmel. If their datasheet states in bold that a device runs from 3.3V it doesn't. It's more like 3.6V but that is only in the fine print.

It can, but at degraded performance/clock?

Not even that.

the whole world is using atmel mcus at 3.3V running at reduced clock of 8 MHz or less... so think about it a bit

[nctnico]

Anything from Atmel. If their datasheet states in bold that a device runs from 3.3V it doesn't. It's more like 3.6V but that is only in the fine print.

It can, but at degraded performance/clock?

Not even that.

the whole world is using atmel mcus at 3.3V running at reduced clock of 8 MHz or less... so think about it a bit

You should read more carefully. I wrote IF an Atmel datasheet says the lowest voltage is 3.3V then you shouldn't run the device at 3.3V or you will get burned sooner or later. It is an example of what you can encounter. Other manufacturers like NXP or TI are way more honest about their devices.

[Alex30]

That's interesting, I have always considered Atmel to be pretty good with their datasheets. Very descriptive if you read them thoroughly