Cost reduction : component count VS unique part count

[ratatax]

That's a thing i'm wondering now. The schematic of my product is almost finished and I have some components used only 1 time, Eg. a single 220 ohm resistor, 47uf, 22uf, 39nF cap...

I plan to externalize PCB production and assembly.
I know pick'n'place machines have a limited number of slots, and the setup cost done by hand is probably not free.

Should I try approximating odd component values with more common ones to reduce the number of unique parts my BOM has ?
It will of course increase the total component count. I feel it's worth having less different reels, probably for cost AND simplicity, but I'd like to have your opinion on this.

I'm also tempted to simply round some values in my circuit, but most of time with design examples found in datasheets it's not very clear if the recommended values are critical or not.

[SWR]

That depends on how your EMS does the calculation?
Most EMS we deal with use open calculations, and you can often get an Excel sheet with the formulas used.
It will often be a tradeoff between NRE and VC depending on the series volume you are ordering.

At the one we're using now we pay NRE according to the number of different components to pick from stock and set up the machines.
Other NRE costs are not related to amount of different components (like P&P or ICT programming).

We pay VC per component (both purchasing cost and placement cost).
Often the component types are grouped like:
  - 2 pin SMD without alignment.
  - 2 pin SMD with alignment.
  - Multi pin SMD with alignment.
  - Fine pitch with accurate alignment.
  - Etc.
Often the cost is calculated from a machine hourly rate and the capacity in each of the above categories.
The EMS usually have slightly different ways of calculating this.

Once you have the Excel sheet you can try out different alternatives.

[SMTech]

Broadly speaking SWR is on point here, and many contractors may indeed share precisely how they reach their costing, but equally if they are doing the scouring, they could be using savings in one area to subsidise something else.

As a contractor who does a lot of small batch work, increased BOM line count has a huge effect on the time and hence cost of a job, 20 boards could easily only be twice the cost of 5. With batches @ this scale we wouldn't break down the cost for someone because we have to include a fudge factor (using intuition) to account for the fact in low volumes you can end up with parts you need to fit by hand, packaged badly, bent legs etc and if the client was trying to apply some kind of rigid formula to predict costs it wouldn't be sufficiently flexible.

However once you scale up past smaller batches of say 500 or so (its going to depend on how complex your board, how fast the line is, how busy the line is) that setup time, that plays such a big part with a small job becomes almost irrelevant. Every line on a BOM needs checking in, verifying (at some level even if its reading the label), loading onto a feeder, checking again. If its a new part or package to the manufacturer that part will need teaching to the machine, if its a short strip, they might need to add header, footer or cover tape. All that stuff on its own is the work of  few minutes and required once but it can easily add up.

[ratatax]

Thank you for your replies. This would be for 100-unit production. Board has a medium-high complexity ( ~300 components, ~35 unique SMD parts and ~10 unique through-hole parts)

I used some online calculators as most PCBA companies provide a quick estimation of the price, and I was surprised that the number of BOM lines didn't make a huge difference in price.

I still think about optimising a few things, just for the sake of reducing complexity, time spent on sourcing and potential risk of errors..

[SMTech]

35 unique SMD parts really isn't that many. I would stick to very basic simplification, making sure you don't request the same value in two different specifications or picking convenient tie-up values that overlap with other values you need to use anyway