Some of the old carbon-composition resistors suffered from "excess noise," which was noise above the Johnson-noise formula, under the condition of a substantial current flowing through the resistor (i.e., a substantial voltage across the resistor). A Google search shows rumors of excess noise for (some) thick-film resistors, and perhaps less so, for some thin-film resistors... At any rate, this should be an easy parameter to measure, if indeed it exists. I've made modest efforts to take such measurements from time-to-time in the last few years, hoping to see something, but I've not actually observed any excess noise with modern parts. Perhaps I was trying the wrong brands, or the wrong resistance values?
All resistor types have the same Johnson noise, the spontaneous noise that you see with no bias. If DC is flowing, you get added shot noise. Metal-film (thin-film) resistors are supposed to have less shot noise for quantum-mechanical reasons; Phil Hobbs has commented on this here and in his book.
Some really nasty old carbon resistors have excess noise, as Win notes.
So, use low-value metal-film resistors and don't have DC across them if you can help it.
I've been looking spec for resistors, rather boring, but what's the difference between thick film and thin film noise, the data sheets are not very forthcoming
It seems that thick film are noisier, so I started looking for 1% thin film stuff (for audio), but although they seem to be available all the distro's, Digikey/farnell just seem to stock .1% types, a bit excessive for audio purposes in most cases, the only exception would be RIAA EQ (joke)
Current flow that originates from a voltage difference across an electrical conductor does not have shot noise. This includes all types of resistors, not just film types. Note, a diode junction is not a conductor for purposes of this statement.
If one has a shot-noise current source signal, such as photo- diode current, and uses a resistor to measure this current, the Johnson noise from the resistor will be less than the signal shot noise if the voltage drop exceeds ~ 50mV. This can be seen by comparing the Johnson noise power 4kT/R to the signal shot noise 2qI, and getting I*R = 2 kT/q = 2 Vt = 50mV at room temp. This little observation would not be possible if resistors had their own shot noise. The excess noise exhibited by some types of resistors at high voltages is another matter, of course.
"...enhance without compromising, the base of the Cradle of Silence is made of high-gloss polished Corian, one of the least resonant materials available today - but unfortunately one of the most expensive."
Hilarious. I get my Corian from scraps left over from making kitchen countertops. It's great for pogo fixtures and such. But "least resonant"?
My designs are about 80% lead free now, and I admit to some 'teething' trouble. The newer components are rated (well, so the mfrs say) at 260C typical for reflow (rather than the more typical 220C - 225C for Sn/Pb components).
The unfortunate thing about lead free components (or RoHS if you prefer) is they are far more sensitive to thermal variations than the components they replace, so you have to get within 5C or so everywhere during reflow (rather than the more liberal 10C or so one could get away with with leaded parts). This makes layout a real problem, because thermal shadowing now becomes a primary issue at layout.
Another issue is the contract manufacturers are having to experiment with various solder paste types to even have a good idea of what works best in their ovens. Ah, echoes of yesteryear. I have a board contract manufacturer I use who seems to have figured out what works (lead free) in their ovens and the appropriate profiles, but I haven't got anything over 8 months old since starting with lead free parts. I'll see what the failure rate is at a year and maybe have an idea of reliability issues.
I have yet to see any utility software than could, given a 3D model of my unit and the dimensions of the oven, come up with a thermal profile. Now that would be really useful. I haven't needed it in the past (because the latitude was a little wider), but I am getting to the point of wanting it now.
As others, I don't have the money (or the spare boards that can be destroyed), or the time to get full thermal profiling done, but given the size contraints I have, SMD is my only realistic option for the majority of components I use.
As I get more experience with the process (and less hair) I'll post my 'experiences' here (I can think fo more appropriate words).
On my very small "production line", we hand-assemble in batches of about a dozen. We've been doing SMT, hand-soldered using .015" solder and a fine-tip soldering pencil. It's actually turned out to be slightly faster than through-hole (because one doesn't have to bend and trim leads or flip the board over to solder), but I will say that it's taken some training to get good soldering quality; and that's with SnPb solder. I'm staying away from lead-free for as long as I possibly can, because some initial forays convinced me that it's just not going to work for hand assembly.
We've only just started with SMT, though, so reliability issues (e.g., is our soldering any good) remain to be seen.
What pushed me over the edge was the need to cram more components into the same board area. But it does seem like I see a lot of newer ICs, even analog audio-related ones like opamps, coming out only in SMT. An example of a nice-looking chip that is only available in SMT is Analog's AD8671/2/4.
I certainly agree it's easier to work on through-hole components. One of the things that pushed me into SMT was that I was starting to get a number of jobs involving modifications to, or repairs of, SMT boards; and I found that once I had the right tools (mainly, a hot air gun), it was actually not very hard. So I figured that the rest of the world would probably make the same discovery before long :-)
One big difference, though, is that SMT components can't generally be identified in the field. Caps don't typically have their values printed on them at all, and semiconductors have short manufacturer-specific codes that may or may not be identifiable in the field. Without a schematic and a board layout, it's pretty hard to work on SMT boards; the same is not true for through-hole.
We had exactly the same sort of 'loop' in system useage. Some time ago, I was tending to work on products where large runs were the 'norm', and basially went SMT. On the current products though, the volumes involved are smaller, and savings almost non-existent. The big problem though is that some of the kit goes into remote, enviroments, where servicing is needed, without having to send back to the UK for a new board. By deliberately keeping component types low (which in some cases 'costs' extra in real estate), the units become far more field seviceable. An engineer in the middle of India or China, can carry a relatively small kit of parts, and have a good chance to service the equipment 'on site'. We are now using technology, that looks 'old fashioned', but with design for service (test points, diagnostics in the processors etc.), can be kept running much more easily. The savings even work in less remote enviroments (engineers in the middle of the US, and in Europe etc., can do the same). In many cases, the kit can even become 'self diagnostic', sending with it's recorded data, information on any errant behaviour. There is a very big difference in this regard between 'consumer' electronics where cost per unit is the key priority, and industrial electronics, where the ability to keep equipment running 24/7 is the winning formula.
First off - it's clear that SMT is great for large scale high quantity manufacturing. It involves quite a high investment in tooling, jigs etc for a product made that way to make sense though and the big economies only really result when a line can be set up to run full tilt for a while.
We tend to make products in rather more modest quantities - hundreds per run rather than thousands. That simply doesn't make sense for SMT.
Also, many key components we use are only available in 'conventional' technology anyway so there's quite a large amount of leaded components in any event.
Bench serviceabilty is greater for the product using non SMT. We have many customers ( or their techs ) who'll open up a product and fix it themselves once it's out of warranty. They're unlikely to be able to do that with SMT. We gain from offering a service friendly product. In comparison other brands at the entry level in our sector are frankly now making products that make more sense to throw away than repair.
I'm also looking ahead to lead-free. That's going to make pcb processing so much more critical and especially so for SMT. We don't have the size of runs to muck around playing with temperature profiles to make the soldering reliable.
What will the likely consequences of that stress be, do you think?
Does it not affect through-hole components soldered in the same manner? If so, how come?
I've been thinking about trying toaster-oven reflow as an alternative. It would speed the assembly, I think, but solder paste and stencils are expensive (for my quantities, ~5% increase in CoGS - ouch).
There's little stress if the components were hand soldered (I have a local guy who does my prototypes / low quantities for me by hand).
to load the bottom of the board first and get it reflowed, then do the top.
There is always stress on a component in reflow because it will not contract at the same rate as the solder during cooling (the cooling phase is usually 1 - 6 degrees C per second). There is an inherent tradeoff here: the faster the cooling, the better the solder joint (finer granularity) but the more stress on the component (because it can't contract in that time due to thermal mass). The slower the cooling, the lower the stress, but a poorer solder joint occurs. For an excellent overview of the stresses involved, see this TI app note, which covers various lead free issues as well:
formatting link
The components on the bottom of the board get this twice, of course (held in place by a little piece of glue).
I have, in the past, had failures due to thermal stress when compounded by other stresses. In particular, small SM caps on the bottom of the board would crack and go low capacitance. In that case, however, the issue was compounded by a defective nozzle on the chip shooter which exacerbated the issue. I did not have the problem with top side components of the same type, however. Once the nozzle was fixed, I did not see the problem again.
I've had a number of dual sided (component wise) boards made (and some had a large amount of thermal mass to dissipate) and never had a problem (apart from the above) *only* because of the added stress of going through reflow twice. I have had components on both sides fail on other boards due to defective loading equipment, of course.
So my experience is there may be a long term reliability issue with the components on the bottom due to two reflow cycles, but the products have been out there for years now without such failures showing up with that root cause. i.e. the reliability issue is there, but beyond the normal service life of the equipment. (Not a good thing for medical stuff, though, I would be the first to admit).
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.