EEVblog 1679 - Why Does This Uni-T AWG Cost $5800 - BOM Cost Analysis

[EEVblog]

Why does the Uni-T UTG9504T abitary function generator cost US$5800?
A teardown and detailed BOM cost analysis of how much high end product designs costs, and the margins involved.#

[tszaboo]

I think the costing is mostly spot on. The only thing underestimated is the mechanical parts, and firmware. Injection molds that are this big are going to be expensive, even in China. And you need a lot of fine tuning to get it all right, no warping, things snapping properly, no flow marks.
30-50K per mold, for two molds, even amortized is significant, plus you need someone working on that for "a year". So I think 20 for the mechanical is underestimated, but I'm happy to be corrected if someone knows more about these.
I think assembly cost was also on the lower end, it feels low for such a PCB. They might be doing XRAY by default, and it's many many parts. Plus yield on such complex boards is going to be significant to mention, adding to the cost price per board sold.
I think the ~x2.5 rule will apply to the lowest end of this product range (same with scopes), and higher frequency is going to be proportionally more expensive. You pay for the extra bandwidth extra.

[Simmed]

relay price on LCSC is $0.5
on mouser it is about $3.1
(1000pc)
gee weez

[Kleinstein]

There are quite a few small parts in the power supply (e.g. all the impressibe filtering).

Even if not really worth it for the bare PCBs, it makes still sense to look at the failing units. Knowing what sometimes goes wrong helps to optimize the manufacturing. Much of the repair costs can be finding the fault. If they later actually do repairs is a different point - some repairs are easy, while reflowing the FPGA and maybe already the DACs could be too complex.

For the final product price the availability of competing products (same manufaturer and competition) can be a large factor. If there is no real competition they can earn more on a product.

[SiliconWizard]

Beyond BOM cost, there's of course all the R&D, production, marketing, sales costs, plus all the distribution chain and taxes up to the final consumer. So that adds up very quickly and the margin is probably not nearly as high as one may think.

Not sure I would personally shell out $5800 on a Uni-T product, but that's just me and probably a lot psychological.

[asmi]

I think assembly and custom manufacturing costs are grossly underestimated. First of all, I don't think they do all 5k units in the single batch - that's a lot of money to lock in the inventory, so they most likely manufacture it in batches of 100-500 units at a time, and run additional batches as previous inventory getting sold). That means that will have to pay PnP setup fee multiple times - once for each batch run. And the same goes for other custom parts - not many companies are prepared to lock in $5M (5k units x $1k per unit - I think this is more reasonable cost estimate than $800 featured in a video) in hope that they will actually be able to sell it all in a reasonable amount of time (time is important too because warehouse space/time cost money).

[Simmed]

I think assembly and custom manufacturing costs are grossly underestimated. First of all, I don't think they do all 5k units in the single batch - that's a lot of money to lock in the inventory, so they most likely manufacture it in batches of 100-500 units at a time, and run additional batches as previous inventory getting sold). That means that will have to pay PnP setup fee multiple times - once for each batch run. And the same goes for other custom parts - not many companies are prepared to lock in $5M (5k units x $1k per unit - I think this is more reasonable cost estimate than $800 featured in a video) in hope that they will actually be able to sell it all in a reasonable amount of time (time is important too because warehouse space/time cost money).

on the distributor page
there is something like 22 countries
i suppose there is a discount ladder for qty
but i think nobody will order 10 units to keep in stock
maybe the first batch is only 50 units ? 

[Kleinstein]

The production batches would likely be smaller than 5 K units. The 5 K is more the number expected to sell overall before a major revision is needed. So the one time costs would be divided by this.
For such a product the warehouse space is less of an issue. The point is more like the capital and expected price drop for the FPGA over time. Chances are it will get cheaper next year or so. One the other side the one time costs have to be spend upfront anyway, even if only producing to order. So it makes sense to separete to true BOM costs for the parts used and the one time costs.
At least for the start there is also the possibilty that HW issues come up only after more actual use. So they may not want the initial batch to be that large.  Ideally they have prototypes for testing, but some issues may still only show up later.

Some of the metal and plastics parts (e.g. the backside) could be shared between different products. Similar quite some of the software may be shared or use a common base.

[thm_w]

I think using 1k quantity prices is fine, not necessarily because they'll order 1,000pc at once, but, its China. Much easier to get bulk pricing from various parts brokers than it is in the west.

Though you sometimes have to be careful with LCSC pricing, that price could be a clearout of some surplus reel they got hold of.

30-50K per mold, for two molds, even amortized is significant, plus you need someone working on that for "a year". So I think 20 for the mechanical is underestimated, but I'm happy to be corrected if someone knows more about these.

The PCB is dated Jan 2022, so they've been working on this for at least 2-3 years.

The rear mold is shared with UPO2000/3000 oscilloscope, so no design needed there.
The front mold is unique but they will probably try to re-use that on a newer scope if they are able to (since its designed like a 4CH scope at the front).

https://instruments.uni-trend.com/products/waveform-generators/UTG9000T

[EEVblog]

The PCB is dated Jan 2022, so they've been working on this for at least 2-3 years.

The rear mold is shared with UPO2000/3000 oscilloscope, so no design needed there.
The front mold is unique but they will probably try to re-use that on a newer scope if they are able to (since its designed like a 4CH scope at the front).

https://instruments.uni-trend.com/products/waveform-generators/UTG9000T

Nice spotting ,thanks.

[EEVblog]

The production batches would likely be smaller than 5 K units. The 5 K is more the number expected to sell overall before a major revision is needed. So the one time costs would be divided by this.

My 5k guess was total lifetime units.

[EEVblog]

First of all, I don't think they do all 5k units in the single batch - that's a lot of money to lock in the inventory, so they most likely manufacture it in batches of 100-500 units at a time, and run additional batches as previous inventory getting sold).

Correct, that's 5k is my guess at lifetime units for long term amortisation reasons. You then handle batch runs however you need to.

Brymen for example manufacture to dealers orders, they keep zero manufacturing stock.
If I order 500 meters then they slot it into the next run in two months time.

[asmi]

Correct, that's 5k is my guess at lifetime units for long term amortisation reasons. You then handle batch runs however you need to.

My point was that because of multiple batches, they will have to pay setup costs multiple times, which increases the amortized cost of just about each assembly and manufacturing job. Which is why I think $1k is a more reasonable cost estimate than $800 featured in the video.

Brymen for example manufacture to dealers orders, they keep zero manufacturing stock.
If I order 500 meters then they slot it into the next run in two months time.

Of course they don't have to hold any stock, because you (and other resellers/white label customers) are doing that for them - and assume all associated risks and pick up the tab for all associated costs.

[EEVblog]

Correct, that's 5k is my guess at lifetime units for long term amortisation reasons. You then handle batch runs however you need to.

My point was that because of multiple batches, they will have to pay setup costs multiple times, which increases the amortized cost of just about each assembly and manufacturing job. Which is why I think $1k is a more reasonable cost estimate than $800 featured in the video.

I do too, but that's what popped out as a minimum, which is what the costing was always about, a realistic minimum. There is always more overhead.

Brymen for example manufacture to dealers orders, they keep zero manufacturing stock.
If I order 500 meters then they slot it into the next run in two months time.

Of course they don't have to hold any stock, because you (and other resellers/white label customers) are doing that for them - and assume all associated risks and pick up the tab for all associated costs.

But as you mentioned above, that increases their setup costs for each run.
If they know a particular meter is going to sell say 25k units over the next 6 months, and your production is set up ready to run, why not make those 25k units?

It's a rhetorical question BTW, I know the answer, but that answer is not always the right one.

Just like I don't save cost by sea freighting my meters.

[coppercone2]

sometimes when i hear about injection molds I think please just put it in a junction box

I feel like there is alot of slop hiding behind these mysterious injection molding costs. It is partially true but you can hide the biggest dumpster fire behind injection molds.

As a EE sometimes you hear something along the lines of "get out of here peon, we are dealing with injection molds !!!"

[tszaboo]

sometimes when i hear about injection molds I think please just put it in a junction box

I feel like there is alot of slop hiding behind these mysterious injection molding costs. It is partially true but you can hide the biggest dumpster fire behind injection molds.

As a EE sometimes you hear something along the lines of "get out of here peon, we are dealing with injection molds !!!"

Have you seen a mold IRL?
Its several hundred KGs of tool steel, custom made just for, you, with moving parts, heating and cooling channels cut into it, precision made down to the fraction of the mm. Then surface treated selectively, measured, assembled, taken apart and reassembled several times. It takes hours to take it apart if a change is necessary, and it is. Because you design the part smaller, and increase it in small steps to get a critical dimension or "clicky- ness" right.

[SiliconWizard]

sometimes when i hear about injection molds I think please just put it in a junction box

I feel like there is alot of slop hiding behind these mysterious injection molding costs. It is partially true but you can hide the biggest dumpster fire behind injection molds.

As a EE sometimes you hear something along the lines of "get out of here peon, we are dealing with injection molds !!!"

Have you seen a mold IRL?
Its several hundred KGs of tool steel, custom made just for, you, with moving parts, heating and cooling channels cut into it, precision made down to the fraction of the mm. Then surface treated selectively, measured, assembled, taken apart and reassembled several times. It takes hours to take it apart if a change is necessary, and it is. Because you design the part smaller, and increase it in small steps to get a critical dimension or "clicky- ness" right.

Yep. And designing those molds properly takes serious skills too.