The new 'G474 DISCO misses the point, IMO

[waclawek.jan]

Discos have been valuable tools to get people start with the STM32. IMO, the newest 'G474 Disco misses this point (and it appearst this is a trend in the newer Discos):

• It is too expensive. One of the selling points of the Discos was the great value for great price. I know ST sells them at a loss, to be recouped at chip sales later; this does not appear to be the case here. Maybe it's the unnecessarily fancy 'F7-based STLink3, maybe the U-suffixed part used there, maybe ST doesn't want to sell at a loss anymore, maybe combination of all these factors; but... the 'G4 ought to be a replacement/upgrade on the 'F3 line; 'G474 in particular with HRTIM is an equivalent of 'F334. The 'F334 DISCO sells at a whopping 17$, showcasing mostly the same major features. Versus $59 introductory price for the 'G474 Disco. Go figure.
• The name does not sell it as a Disco; does not sell at all. B-G474E-DPOW1, what?
• UM is in the atrocious "modern look and feel"
• Schematics is missing from the UM. BAD BAD IDEA.

My 2 eurocents.

JW

PS. The other 'G4 DIsco- the 'G431 one - is a COMPLETELY different story (except the name (B-G431B-ESC1), sigh...)

[Ozone]

I tend to agree.

Far too few free pins for my taste, like most of the later Discos. Most hooked up by functionality very few users would actually look for.

The UM outlook is a matter of taste, but really no schematics. Does ST think it would overtax the target audience ?

And all at a higher price. Perhaps there is a more "pure" and cheaper Nucleo board on the way ?

And to be frank, I didn't purchase any new Disco/Nucleo board for almost two years.

The "Cube only" software package is a no-go for me. I have better things to do in my spare time then chasing Cube bugs.

But that's my personal 2 cents.

[Ozone]

> PS. The other 'G4 DIsco- the 'G431 one - is a COMPLETELY different story (except the name (B-G431B-ESC1), sigh...)

Except that I don't understand the name (it is really called "Discovery kit").

You can "discover" pretty much nothing except the motor driver / speed controller functionality.

Which is ok by itself - but it is a very specialized module.

[waclawek.jan]

> Far too few free pins for my taste, like most of the later Discos. Most hooked up by functionality very few users would actually look for.

At least there are jumpers (solder bridges) provided to free them up.

> And to be frank, I didn't purchase any new Disco/Nucleo board for almost two years.

I see the Discos as toys and bought some through impulse purchase, mostly to be added to the collection already on the shelf (yes then ocassionaly taken from there to do some little proof-of-concept work, or to play with some of the peripherals, think blinky++). My upper limit for those is somewhere between 20-30 eurobucks (I am not interested in the big displays stuff, my primary interest is in control.). Sitting in Central Europe I guess this is halfway between the "rich west" and "poor east"; however, my feeling is that even for the "rich west" for shelf-sitting toys like this, 60 may be a bit above the upper edge.

Definitively, in this particular case, the value is not there, compared to other Discos.

Nuclei as toys are uninteresting to me; IMO they are good only as a vehicle to build one-off systems around them.

> The UM outlook is a matter of taste, but really no schematics. Does ST think it would overtax the target audience ?

No; IMO this is just another bad design choice of ST, in line with the "modern look and feel". I'm just afraid this rot spreads.

> The "Cube only" software package is a no-go for me.

That's a deeper rooted plague. ST (and by that I mean the decision-making management) apparently completely does not understand what opportunity they are missing. But this is topic for another thread.

> But that's my personal 2 cents.

Cumulatively, that's 4 cents already... ;)

JW

PS. Let's leave the 'G431 erm disco or whatever to a different thread, too

[Tesla DeLorean]

Need to be able to take your points here and turn them into prizes...

Looking at the picture, and the schematic, it looks to be a pretty expensive to make board, only 4 layers though, but ~1600 drill holes. BGA's mean you can't get sloppy.

At $59 USD, probably not boards I'd want to burn through.

Feel I got more for my money on the H745I-DISCO's, but again it's one of those "Kitchen Sink" designs where everything went in, although better executed than the F429I-DISCO where no functionality escaped properly.

[waclawek.jan]

> Looking at the picture, and the schematic, it looks to be a pretty expensive to make board, only 4 layers though, but ~1600 drill holes. BGA's mean you can't get sloppy.

Exactly. This is a schoolbook example of several bad design decisions at one place.

JW

[waclawek.jan]

The G4 Nuclei IMO start missing the point, too.

https://www.st.com/en/evaluation-tools/nucleo-g431rb.html

https://www.st.com/en/evaluation-tools/nucleo-g474re.html

They are more expensive than the other 64-pin Nuclei; IMO it's the increased cost of the unnecessary STLink3.

The more expensive STLink is not detachable.

And they lost the so-far-standard 6-pin .1" debug header, to push ARM's harder-to-buy MIPI10 (R)(C)(TM)(whatever) .

Less value for more bucks, again.

I know ST's sales are not driven primarily by the garage-scale and hobbyist end of the market, but it may be still an interesting portion of it, which they managed to capture 10 years ago, and now work hard to loose them.

JW

PS. Oh, no schematics in UM2505 either. And it has 25 megabytes (versus one or two megabyte for the "traditional" boards' UMs), due to completely unnecessarily multimegapixel pictures.

Can ST please drop the "modern feel and look" managerial thinking and return to solid engineering?

[Ozone]

Well, saying ST's new manual look and feel is a "matter of taste" doesn't mean I like it.

I also prefer dull schematics (information) over true-color images (advertisement).

But I still can't see the target audience for this board.

It is too expensive for most hobbyists, tinkering once and a while in their garage.

With it's hooked-up pins and non-detachable ST-Link, it is uninteresting for most commercial evaluations.

The hardware thrown onto it (RGB-LED, buck-boost converter) would point to the lighting industry. But I've been there, and I know they use to go as cheap as possible, usually MC PIC18. And I know their stance towards space-wasting libs like Cube.

All in all, a new shelf warmer ???

[Amel NASRI]

Hi Jan,

This discussion with all the opinions expressed there are shared internally.

I would like to explain some points:

- Discovery boards that have an application focus (Ex: Digital power, IOT,...) are using a different codification rule for naming compared to usual ones ("DISCO" or "Discovery").

- That is true STM32F3348-DISCO is cheaper than B-G474E-DPOW1, but it can be explained by the difference of available features. B-G474E-DPOW1 price remains in the medium range of Discovery boards prices (< 100$).

- The schematics are no more in the user manuals of new ST boards for the purpose to ease the maintenance of both resources. The schematics remain accessible on their dedicated sections on board's page.

-Amel

[waclawek.jan]

Amel,

Naming and pricing may perhaps be a matter of discussion, but:

> The schematics are no more in the user manuals of new ST boards for the purpose to ease the maintenance of both resources.

It's a bad decision; together with the "modern feel and look" unnecessarily huge photos.

I repeat, a bad decision.

Maybe I should stress this, it's a bad decision.

It marks a general slipdown in the ST technical documentation quality.

Please do quote me. And, whomever talks through you, please may she/he come and talk under her/his own name.

JW

[Ozone]

> The schematics are no more in the user manuals of new ST boards for the purpose to ease the maintenance of both resources.

To dwell on this point, too:

Downloaded a Nucleo 144 manual yesterday, and was shocked. A size of 22 MByte for a 7-page document, with one picture.

A manual comes in handy when a decision has been made, and is expected to deliver proper and detailed information. Not pretty marketing look.

Engineers are not marketing guys.

[Nikita91]

It seems that ST has a big marketing budget, to the detriment of the quality budget for example.

As it seems that there are more graphic designers than software developers (see the Ethernet driver which still does not work on H7) or hardware (design flaws in STM32 devices).

For example, from a quality point of view why have almost identical peripherals, but with registers not quite in the same place (DAM1/DMA2 vs BDMA on H7) or slightly different names (APB1LENR/APB1HENR on H7 vs APB1ENR1/APB1ENR2 on L4 +)? It's horrible when you employ several families and you have to mutualize the code. It seems that everyone does what they want at ST, to the detriment of customers ...

About the table of contents at the beginning of the document and the B&W colors which are good ideas, how is it "modern" to put it at the end of the document and adopt colors that are not very readable? In fact from a marketing point of view it takes change to show that we are active. And editing the doc is easier than producing quality software!

Other subject.

Elsewhere on this site there are references to the hobbyist market that would no longer interest ST. On the contrary: ST puts a lot of resources on CubeMX which is only intended for hobbyists! For example in the company that employs me and designs industrial products (lifespan of 10 to 20 years), we cannot trust this kind of tool.

In my case, it was through my personal development and my knowledge of STM32 products that I was able to offer my company to use MCUs, which it is doing more and more. ST would be wrong to ignore the hobbyists, and not to offer efficient software libraries.

About the HAL, clumsy, buggy, with random sustainability, ST is not the only one to go astray in this way. Atmel/Microchip makes the same mistake with Advanced Software Framework (ASF) which tries to unify AVR + AVR32 + ARM.

Why we can't use this tool. Code quality: unqualified generators are frowned upon in the industry. Sustainability: what will CubeMX be in 10 years, we have seen ST abandon several software in recent years...

In reality we prefer to manage the obsolescence of MCUs than that of software, software is what costs the most to my business...

Too bad there is not an advisory committee made up of members from outside ST to guide developments right from the start (at the time of designing the MCUs peripherals).