I think everyone at least once faced with fake inexpensive, very similar to the originals of things from China. At first, these were "abibasy" and "Nokly" - now iPhones and chips. However, if earlier the forging of the microcircuit was limited either by incorrect marking (when the usual cheap microcircuit was marked as expensive) or by the absence of the microcircuit in the case (at best there is nothing, at worst all the outputs are shorted ) - now everything has become more interesting.

Recently, exp131 and X4ZiM noticed that some batches of the popular USB-RS232 interface chip are all the same suspiciously buggy: they are defined in the system as regular FT232RL, “recognized” by the official software, but the data sent is not transmitted. Of course, we are all accustomed to the fact that the error in the first place you need to look for yourself ... But what if the problem is in the chip?

Photos of microcircuits: on the left it works fine, on the right it is buggy You can notice the difference in the marking - it is laser engraved in the working chip, in the non-working one - printed (however, this is not a universal rule, it happens and vice versa). First we look at the photos, conclusions at the end.

')

Original FT232RL

After etching the plating:


You can look in more detail at the individual elements, rows of standard cells from which the logic is auto-synthesized:


ROM:


Static memory:

Chinese clone FT232RL

But the chip is completely different! Immediately you can note that the contact pads are much larger than necessary. On the edge you can see the inscription "SR1107 2011-12 SUPEREAL"


After etching the plating:


More details on individual parts, standard cells:


In the other part of the chip, the standard cells were significantly different. Not often this can be found - the microcircuit was assembled from large “pieces” that were available only in the form of topology:


The first type of static memory:


The second type of static memory:


And finally, ROM - it was programmed at the polysilicon level in the production of the chip, so that we can directly see the recorded data:

Summary

The Chinese in this case didn’t just make a “fake”, but took a ready-made programmable mask in the production of a microcontroller (only one mask needs to be changed - it is much cheaper and explains the extra contacts on the chip), we ordered the production of a batch of these chips at the factory. but apparently there was a mistake somewhere, and even though the chip was defined as real, it did not work as it should. This did not bother anyone, and these chips began to sell as real FT232RL. However, it is also possible that the amount of ROM memory just was not enough for full emulation.

Update: The resulting clone worked fine until FTDI released a driver update that could distinguish the original from the clone via USB (on newer drivers, the clone transmits only zeros). If you now install driver version 2.08.14 or earlier, the clone also works. It is almost impossible to foresee all possible future checks without full restoration of the original electrical circuit - and this forethought saved FTDI.

What is the economic sense of making a software "copy" of a known chip, and not releasing it under your own name? No purchase of USB VID is required, advertising costs are not needed. A copy will be used in numerous high-volume products. With a new microcircuit - the development of final products from scratch would be required - the sales growth would start only in 2-3 years. The cost of production of microcircuits is 10-15 cents in the case of both the original and the Chinese version. The manufacturer could also reduce the cost - but does not do this, they earn their well-deserved profit.

Hence the moral - the farther, the more carefully you need to apply to suppliers of microcircuits, and cheap microcircuits from unknown Chinese suppliers can go sideways and guarantee long hours of fascinating debugging.