What are the component brands i should use or avoid?

[RossK]

A wiki search hasn't given me the answer so I'll just ask - what are the good and the bad brands for the basics such as capacitors, resistors, diodes, etc? I imagine there is a list somewhere or an opinionated blog ready made but I cant find it at the moment.

[Kremmen]

Of the components most likely to fail, electrolytic caps are the prime suspects. They have an entire web site and forum dedicated just for that: http://www.badcaps.net/forum/index.php
After that it is not quite as clear. There are outright fakes floating around such as (originally) Motorola MJ series bipolar transistors, but you can hardly blame the genuine components for the shortcomings of the fakes.
You are less likely to encounter fakes if you only acquire your parts from reputable sources such as Digikey, Mouser, Farnell. Of course there are other reliable sources as well.

[Simon]

grey market parts are apparently a huge problem and as a trader myself I wonder sometimes and try to go to the most reliable sources, then again I've never had complaints. I expect the cheap generic parts are not a problem, it may be the more expensive ones that are being faked, but poor quality is something you will have to deal with and as stated electrolytic caps are the one things that you get right or get wrong.

[free_electron]

capacitors ,in particular electrolytic , are the main problem. stick to Panasonic , Chemicon , and other big brands. stay away from unknown brans lake shaxon and other

[Kremmen]

Panasonic is my #1 choice as well, and i have had good luck with Nippon Chemi-Con, Nichicon, as well as Epcos and Cornell Dubilier in high-ripple PSU applications. One thing to look out for is that all of the mfgs have different series of caps designed for different applications. The cap should match the application regardless of who manufactured it.

[HardBoot]

Electrolytic capacitors are the only big problem generally... bad brands are terrible and I've had better brands die too.

Price is a big problem though... cheap caps are 1-5 cent each and good ones are 20-80x more expensive, it's ridiculous.

Anyone know where to buy good caps for cheap? When messing around I use cheap ones just because stocking up on a variety of good ones cost so much.

[RCMR]

I love buying no-name components from Chinese sellers at bargain basement prices.

It's like lotto -- you never know what you're going to get!

So far, most of my purchases have been pretty good -- although I did get some 7805 regulators that most certainly weren't regulators at all.  I ground away some of the epoxy and they actually seemed to be an NPN power transistor.  In a test circuit they had an hfe of about 30 but trying to pass more than 1A caused them to fail.   Never the less, they were (re?)marked as 7805s and they only cost $0.08 each (that ought to have been a clue) :-)

I also bought a "series" of 0806 ceramic caps from one vendor -- only to find that they were *all* about 220pF, regardless of the claimed values.

But hey, they didn't cost much and were ultimately worth even less :-)

If I really need parts for a project that I don't want to fart around with (tracking down problems due to bogus parts) then I'll buy from Farnell or a vendor I trust.  If I see a bargain and don't really care too much if it turns out to be a scam, I'll sometimes take a punt.

[free_electron]

Now here is another point of consideration.

Selecting a known brand is one thing , but you also need to understand the application of the capacitor ! There is a reason that companies like Panasonic Rubycon, Chemicon have at least 20 families of electrolytic capacitors and you will find a 10uf 16 volt in all these families. You need to look at the specifics of the family...
some caps are good for low cost stuff , some are high temp , some are high reliability , some are made for pulse loading, some are built to handle extreme ripple currents etc ... Depending on the construction a cap wil excell on one parameter but may stink on another parameter.

Someone mentioned Cornell-Dubilier. There is also Mallory and Sprague. Those caps are litterally coke-cans. they are oversized , built like a tank , have rubber venting holes , screw on terminals and are intended for immense ripple current without giving a crimp. attempt to swap such a beast with an equivalent wingpangpong with same capacitance and voltage rating and it will be 1/10 the size and die a horrible death after half a year...

All those tv SMSP dying is becuse the WRONG type cap was selected becasue of cost reasons. i have systems with SMPS in them that are in the 'high-rel' category and there is a notably different type of cap being used there... We had a bunch of expensive Samsung Monitors (DVI input + YcrCb so they could do picture in picture with an analog HD source ). they ALL died within a year. in  spare moment i swapped all the 'shitcon' caps with a carefully selected panasonic intended for 5000 hours life at 105 degrees. i folded them away from the heatsinks , drilled a couple of holes in the pcb to get them away from the bulk heat.

the end result is all these monitors still work 4 years later ... not a single hiccup. that shitcon was simply not capable of handling the pulse currents from the switcher.

[cwalex]

I just don' understand why people are buying smt parts from ebay 

The prices in 100qty or even single qty are so cheap from places like element14 and other good distis. You get a datasheet and you don't have to worry about the quality and if your project is malfunctioning due to dodgy parts!

[pickle9000]

I'd say it takes time. You need confidence in a supplier but bad components will creep in. Even worse changes to components over time that go unnoticed until it's too late.

If you buy components from ebay be aware that the guy supplying them probably has no idea what he has. Having said that, it is a good exercise and you can test and compare to see for yourself how they get on. Make up a test bed. I can't say if they are good or bad, regardless it's inexpensive. 50 bucks worth of ebay stuff, that's a pretty cheap lesson. Try putting those parts in a real world product and you are taking an unknown risk (probably a big one), that is an expensive lesson to learn and not worth the risk.

[cwalex]

I'd say it takes time. You need confidence in a supplier but bad components will creep in. Even worse changes to components over time that go unnoticed until it's too late.

If you buy components from ebay be aware that the guy supplying them probably has no idea what he has. Having said that, it is a good exercise and you can test and compare to see for yourself how they get on. Make up a test bed. I can't say if they are good or bad, regardless it's inexpensive. 50 bucks worth of ebay stuff, that's a pretty cheap lesson. Try putting those parts in a real world product and you are taking an unknown risk (probably a big one), that is an expensive lesson to learn and not worth the risk.

True, for $50 you could get heaps of different values from element14 or RS components in qty of 100 each. If it turns out later you need a different value you can bodge the value with combinations of what you have on hand while you wait (1day delivery forme here in aux) for the actual values you need.

You will end up with a vast collection of different values in stock and you get datasheets!

[helloworld922]

Another +1 for choosing a trusted supplier, I mainly use Digikey but I also trust RS Components, Farnell/Newark, and Mouser (might be a few others, I'm based in the US).

They offer datasheets, have good parts searching capabilities, and know how to properly handle/ship components (especially ESD sensitive ones). I even trust some of their "bargain-brand" components for less critical projects.

[rollatorwieltje]

I love buying no-name components from Chinese sellers at bargain basement prices.

It's like lotto -- you never know what you're going to get!

I did the same thing, bought a load of TO220 LM7812 regulators that were only useful as random noise generator. They actually did sort of regulate, but had a massive output ripple. I don't know the price anymore, but it was in the order of 10 for the price of one.

Took another gamble with a load of those cheap red LEDs, but they turned out fine. Out of 50 LEDs or so only one or two were dead, not bad for a few cents. The envelope they came in was more expensive than the LEDs

[RossK]

Thanks for all the input. I buy from Mouser and Digikey myself 

As this discussion has really focused on caps i'll mention i do need to pick a good electrolytic cap to absorb the induction-induced ripples out of a half H-Bridge driving DC motors ranging from about 10A to 50A from an input voltage of between 10 and 18 volts (13.5v to 15v typical). I'm still working out how much ripple the power supply can take but the rest of my control system needs to stay under 20v (so 11% ripple). I expect to be driving the bridge at 30kHz to 40kHz. Considering the 20% cap tolerances and a 'guess' at the inductance of one such motor of 2.3uH (anyone have a good way to measure inductance with a VOM  ) my calculations put me into the 65uF range - assuming my calculations are correct (which they may not be) would this cap be a good choice http://www.mouser.com/ProductDetail/Panasonic-Electronic-Components/EEH-ZC1V680XP/?qs=sGAEpiMZZMtZ1n0r9vR22WlgczIAtDiIYdlExSigoug%3d ?

Figuring out this drive circuit is proving to be a little maddening to my lack of experience but that's for another topic 

[Kremmen]

Ok, let's approach this from the beginning, so to say.
A smoothing cap sitting between a source and a load is itself a load and a source at different times of the operating cycle. If it wasn't, nothing useful would happen. So part of the source cycle the "true" source is supplying the "true" load, as well as charging the cap. The rest of the source cycle the cap is supplying the "true" load. The load cycle for its part affects the details of this charge/discharge cycle.
To estimate/calculate the ripple, a number of things must be known:

1. What kind of source is supplying the DC rail of your bridge. You say the rail voltage is btw 10 and 18 V but not from where. For the sake of argument i assume an iron transformer secondary with a full bridge rectifier. That means your smoothing caps sitting across the rail will charge rather spiky.

2. There are 2 kinds of ripple, a) voltage ripple and b) current ripple. Both will be influenced by the capacitance of the smoothing caps. The capacitance you have arrived at is very low for a medium current PSU. A more plausible value would be several 1000 uF. But that depends entirely on the nature of the supply. I am assuming point 1 above.

Wait a minute. Is this the same project that you said would be fed from a battery? In that case you won't need much of a smoothing cap at all - the battery is one huge cap in itself. There is nothing you can do about the motor coil inductances, but they are not a problem. In fact the inductance limits the di/dt and that way also  ripple currents and voltages. If you have very long feeder cables then a cap in a strategic place might do some good, but please throw away the pitiful piece of mouseshit and shop in this department instead: http://www.digikey.com/scripts/dksearch/dksus.dll?FV=fff40002%2Cfff80009%2Cfffc0152%2Cfffc01ef&vendor=0&mnonly=0&newproducts=0&ptm=0&fid=0&quantity=0&PV16=11813.
These are caps for grown-ups and they are what you want in a PSU. The SMD cap while from a quality mfg that i use myself, just isn't in this league. I am not saying you must have one of these caps in the circuit, but you might want one - depends on the details of the circuit and app.

[RossK]

Ok, let's approach this from the beginning, so to say.
A smoothing cap sitting between a source and a load is itself a load and a source at different times of the operating cycle. If it wasn't, nothing useful would happen. So part of the source cycle the "true" source is supplying the "true" load, as well as charging the cap. The rest of the source cycle the cap is supplying the "true" load. The load cycle for its part affects the details of this charge/discharge cycle.
To estimate/calculate the ripple, a number of things must be known:

1. What kind of source is supplying the DC rail of your bridge. You say the rail voltage is btw 10 and 18 V but not from where. For the sake of argument i assume an iron transformer secondary with a full bridge rectifier. That means your smoothing caps sitting across the rail will charge rather spiky.

2. There are 2 kinds of ripple, a) voltage ripple and b) current ripple. Both will be influenced by the capacitance of the smoothing caps. The capacitance you have arrived at is very low for a medium current PSU. A more plausible value would be several 1000 uF. But that depends entirely on the nature of the supply. I am assuming point 1 above.

Wait a minute. Is this the same project that you said would be fed from a battery? In that case you won't need much of a smoothing cap at all - the battery is one huge cap in itself. There is nothing you can do about the motor coil inductances, but they are not a problem. In fact the inductance limits the di/dt and that way also  ripple currents and voltages.

I'm building a variable speed DC motor controller - where the speed will be changing drastically several times a second. As for the power source there are no batteries involved, it will be driven from high end power supply (which i am still trying to get the specs for ). The supply is itself is adjustable output but the voltage will be fixed while in use to drive the motor (the supply voltage may be changed between 'uses'). I know the supply can take generator ripple but i'm not sure how much - so, for the moment, i'm basing my calculations on the needs of my driver circuit.

Perhaps i'm using the wrong calculation, i used
C = 1/64 * (Vbat/Vmax_bat_ripple / Lm) * tcycle^2

from http://modularcircuits.tantosonline.com/blog/articles/h-bridge-secrets/sign-magnitude-drive/ in the "Input Capacitor" section

This section does state that the lower the frequency the larger the cap that's needed, running at 500Hz would require a large cap indeed
(300,000uF per this calculation).

If you have very long feeder cables then a cap in a strategic place might do some good, but please throw away the pitiful piece of mouseshit and shop in this department instead: http://www.digikey.com/scripts/dksearch/dksus.dll?FV=fff40002%2Cfff80009%2Cfffc0152%2Cfffc01ef&vendor=0&mnonly=0&newproducts=0&ptm=0&fid=0&quantity=0&PV16=11813.
These are caps for grown-ups and they are what you want in a PSU. The SMD cap while from a quality mfg that i use myself, just isn't in this league. I am not saying you must have one of these caps in the circuit, but you might want one - depends on the details of the circuit and app.

now that's a cap 

[Kremmen]

Attached is a photo of an unregulated PSU in progress, that i made for a small cnc mill a couple years ago. The smoothing caps are Epcos PEH series 4700 uF / 63V units that will take a lot of ripple to kill.

You say your motors will do high accelerations in both directions. What kind of inertia are they driving? It sounds like you might need full 4 quadrant operation where the supply must be able to produce as well as absorb power. Unless you do it the hard way inverting the excess to the grid (probably illegal as well), you need braking resistors to dissipate the excess power. That will complicate the PSU somewhat. Not to mention that you do need smoothing caps. Based on your post, if you have a "high end supply" then that should most likely come with a rather good filtering already. Although in my opinion such a source is wasted on a motor. You do not even need a regulated source, really. Depending on how smart your pwm driver is, it can easily compensate for any ripple in the PSU voltage provided said ripple is significantly lower frequency than the pwm freq. If it was me, i'd consider either a traditional transformer/rectifier/caps supply or if that won't cut it, then a robust SMPS with PFC, but not a lab quality one by any means.

As to the smoothing capacitance - the values can get pretty significant. Check the simulation that illustrates; the scope trace A is 10V/div, all current probes are 50A/div. At 20VAC50Hz supply and 1 ohm load the rail sags 5 volts at 40 000 uF caps. The 33 mohm resistors represent a typical ESR of these kinds of cap.

[HardBoot]

Mmm, beefy caps, what software is that?
I use LTSpice, but I don't like the controls much, works great though.

[Kremmen]

Ain't they now
The sim is NI Multisim 11. The UI is OK, and i like the way the integrated grapher autostores the analyses so that you can refer back to earlier ones without disturbing anything. The grapher lets you reformat the graphs freely even after the analysis so you can zoom in and out, and scale and switch the axes at will. It has its quirks - you have to watch out for singularities in the network, and it often spends some time with larger circuits to come up with a working set of calc rules but nothing you can't live with.