Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Parts count vs. parts cost - Manufacturing question

Status
Not open for further replies.

mdwebster

New Member
This is, I suppose, more of a manufacturing question than an electronics question per se, but it is electronics manufacturing and I thought there might be a few here with some insight and/or rules of thumb that they've developed over the years that might be willing to share some of their wisdom.

The question is summarized in the title, so let me give a more concrete example that actually triggered this search:
I can build a switchable, constant-current source with 3 transistors and 3 resistors for ~$0.25.
Or, I could use a custom IC with a single set resistor for ~$0.50.

Assuming a large production run (100k+ units) and further assuming board real-estate is not an issue, how do you determine the break-even point for parts count vs. parts cost?

I appreciate any insights on this,
Mike
 
You have to take into account the assembly costs don't you? I'm not sure if on such a large production run if the cost of the extra PCB real-estate would be a problem (which I would think is different from real-estate being an issue which I feel has to do more with compactness and formfactor).
 
Yeah, I just meant that there was extra room on the board we were planning on using. I don't think the cost of manufacturing the PCB itself is very much affected by parts count as long as it all fits so I'm really just focusing on the cost to manufacture the final product (pick and place machines, wave soldering, etc.) I know that mixing surface-mount and through-hole increases price since you have to send the part through two different "lines". I was just hoping someone had an idea of the incremental cost of increasing parts count during final assembly.
 
Really depends on your manufacturer. Best to get some quotes.

But I really can't see how placing 4 extra parts would cost anywhere near $0.25. Are there only 2 to 6 parts on the board, or is this part of a larger design. I used to figure half a cent per lead, but this has almost certainly gone down.
 
The placement cost is, obviously, highly dependent on volume. Assuming you are doing at least 50,000 pieces, then the placement cost might be in the region of 0.015 to 0.035 per part. But I must be honest, I'm not a manufacturing specialist and these figures are a guess based on seeing other people's numbers. The highest numbers I have seen per part in decent volume is about 5.5 cents per part, and I think the lowest figure can go below a penny, but that might be above 500K pieces. So, if these figures are roughly correct, the extra four parts might cost you a dime plus or minus.

There will also be a cost related to purchasing labor, usually a small percentage surcharge on the cost of the part which should be factored in. I've met many operations managers who hold overall BOM reduction up as a religion. They argue that fewer unique parts in an assembly has a multiplicative affect on product cost thanks to the reduction of hidden administrative costs on stocking and tracking and purchasing parts. I don't get it myself, but thought I should pass that on.

There is another issue that should be considered. The predicted yield in production is partly based on the number of solder joints and partly on how reliably the design will work with component variations (among other things). With 3 transistors and 3 resistors you have 15 solder joints, while with an IC (assuming, say, 8 pins) we have 10 solder joints. These aren't that far apart, so next I wonder which circuit will work the more reliably and tolerate the variation in components and suppliers. Often, use of an IC implies that the circuit is more complex, that is with greater design sophistication, and has been more thoroughly engineered. The individual transistors in the IC will be better matched, and it might have better temperature compensation. But I'm guessing now. I wonder how good your simpler discrete circuit will perform. Have you done any Monte Carlo on it? If there is any doubt that the circuits aren't equivalent in function and performance, then consider carefully that the downside cost of this difference could overwhelm the other costs.
 
Thanks Ron, your message helped a lot. I was really just looking for a real ball-parkish type estimating routine just to get an idea of when to start considering a single IC over multiple parts.
As to the circuit vs. IC, the IC certainly has better and tighter specs than the circuit I've come up with, but the circuit only needs to limit current below a certain point to meet the 100mA low-power USB current draw limit and it doesn't matter a whole lot if it comes in a bit low. So I'm actually targeting ~60mA and, yes, I've done some Monte Carlo type analysis on it to make sure that it comes in less than 70mA in the worst case (30mA worst-case budget to the rest of the circuit (micro, some IR receivers, and some indicator LED's)).
Just as an aside, if anyone is interested, I'm using this to charge a pretty sizeable cap (3300uF) because this device has a couple of high-intensity IR LED's on it that are set to draw ~350mA per pulse. Just using a current-limiting resistor wasn't feasible as the time to charge from, say, 3.0V back to 3.3V was pretty damned lengthy through a resistor that limited the current below 70mA when the capacitor was fully discharged.
So I'm using a PNP to regulate the current through a PFET with an NFET down at the bottom to turn the whole damned thing off when not in use to prevent electrolytic leakage current from killing the batteries (this is a dual-use device, can be powered from USB or batteries). All credit to "The Art of Electronics", pg. 167 Fig. 3.72c.
 
I would agree, though it depends on where manufacturing is and how it is done. In the US or, I assume, Canada you are looking at $0.25-0.50 for every hand insertion. In Mexico, $0.10-0.15. In China, it used to border on free.

For automatic equipment it varies from $0.02-0.05 depending on the part and how many times it is used (they still have to change out empty feeders)

Optimum efficiency is obtained on automated equipment in panels optimized for the manufacturer's machines. If they do a lot of surface mount that fits in a 7 x10 solder stencil, and they are ordering the boards they are likely to order the most that will fit in that area per panel.
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top