Not another evalboard for STM32 - do it right in XFR-L3 "KYRNN"
Hello, Habr! I would like to talk about creating the XFR-L3 "KYRNN" debugging board for the STM32F4 series MK:

It uses components with quality that are orders of magnitude higher than everything that is on such devices, as well as some of the fact that many evalboards often do not get at all . The board file with the circuit is available for download - everyone interested is welcome to cat.
Now there are a lot of ARM-based debugging boards, especially for STM32 - it affects the availability of these stones and related documentation, an abundance of sample programs, the availability of IDE sets and a fairly adequate price. Whoever did not make boards for them - there are, among others, varieties of the well-known Arduino, which are based on the mentioned microcontrollers. Variants of devices based on MK from different STM32 families (both commercial and amateur) are widely available, with a rather different set of functions (from a bare board with seats for headers, to samples with screens, cameras, various sensors, etc.) and in various form factors. And it’s logical to ask: KYRNN, is that yet another board? Before answering this question, a few words about how commercial evalboards are made, what they are,
Probably, many of those who once felt the need for a debug board for a particular platform noticed an unusual similarity of such products from completely different manufacturers. If you haven’t noticed, then there’s a small hint - they are really similar in many respects: picking up several boards with a more or less matching set of functions (for example, Ethernet + RS232 + USB + SD + a pair of sensors like an accelerometer) the similarity will be as if they are different minor revisions of one device . At the same time, partnumbers of components (or are direct replacements / analogues), switching schemes, and, of course, operational characteristics, which will be discussed later, will most likely coincide.
From this phenomenon of surprising similarity, a completely reasonable question arises about yet another board, since its relevance increases slightly more slowly than the inadequate increase in the number of Asian (especially Chinese) debugs and their clones. It would seem, well, even if they are almost the same, what's wrong with that? The catch is that the vast majority of evalboards on absolutely any platform (except MilSpec / RadHard / Space and some RF devices) are made in many of the lowest-grade components (not counting the stone itself), although they allow, at the very least, to launch this very platform (as comrades of a certain category like to say on this score: “Well, it worked somehow, then everything is fine”). Pure capitalism - it’s necessary to “do it cheaply,” therefore, Asian noname components are used, again the origin and quality of which is often not known to Asians themselves. Surprisingly abnormal switching schemes of various kinds of microcircuits are often used (for example, powering the paper displays) - some samples are very striking that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY). power supply of paper displays) - some samples are very striking in that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY). power supply of paper displays) - some samples are very striking in that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY).
Of course, it’s especially worth noting the wiring of such boards: when I come across another piece in which the ground is looped along narrow boards along the circuit, the power circuits are separated by “tracks” - and the components and vias from these circuits are connected through thermal barriers, USB is not spread out in a two-layer diffparames and many similar phenomena from the indicated semantic series, I understand that I could write a small book (pages 100-200) about all the techniques of the curve, wretched and frankly crap wiring, which are classics, the standard for fact in evalboard (and in quite a few commercial and even industrial grade devices). The exception, perhaps, is the motherboard for the top FPGAs - also imperfect, but from this point of view it is a cut above other similar devices. The fact completes the picture, that some such evalboards are made in Cadence Allegro - a powerful CAD system in which you can design the most complex boards. A funny sight: miserable arrangement, scary components and footprints, vile layout, but high-speed signals are surprisingly well divorced and alignment is done (the last two points are one of the foundations of the "strength" of packages like Allegro).
Much more can be written about these miracles, but this will be a completely different article, and therefore we will return to the topic of debugging for STM32 and actually XFR-L3 “KYRNN”. The implemented functional set (including for comparison) is taken as follows: USB + Ethernet + CAN + RS485 + uSD. A few words about the implementation of each of them:
USB - taken from the stone itself (in this case it is STM32F439VGT6, while the signal lines are protected by two ESD5V3U1U-02LRH. Also, the MAX14575 power switch with filters and suppressor is installed, the connector costs 65100516121.
Ethernet - implemented on TLK106 enabled via MII. In addition to PHY itself, it allows cable diagnostics using TDR. Protected by assemblies LC03-3.3, connector 7498111001.
RS485- MAX13451E is responsible for it, which allows you to terminate the line with different resistances, invert channels and much more. It is protected by SMA6J suppressors and TBU-CA assemblies, connector 09551666817741.
CAN - made on MAX13041, a rather successful and reliable transceiver. It is protected by SMA6J suppressors and TBU-CA assemblies, connector 09551666817741.
SD - removed from the stone itself. There is an IP4252CZ16-8 assembly with filters and suppressors, socket 0475790001.
In addition, there is an external MRAM MR25H256 - an exceptionally reliable kind of memory that can be used both as RAM and as ROM. Power supply, designed for stable + 12V, organized on two LD1086. An external watchdog STWD100 is installed for MK.
In fact, this is a project of an industrial-grade device based on premium components operating in the industrial temperature range and having sufficiently reliable sets of protections. Although the hardware set is, from a certain point of view, rather start-up, nevertheless, its implementation is a cut above all that is offered in the overwhelming majority of debugs (and KYRNN is positioned just like evalboard).
I want to emphasize several important points:
- the components are selected exclusively top-end and expensive, they can not be replaced;
- footprints on many of them also exclude the installation of anything that does not fit into the ideology of this board;
- performance first, even if it requires tougher manufacturing standards (as it happened with IP4252CZ16-8, for example).
At this point, it is logical to proceed to the board layout itself. It was created using Altium Designer 15 and PCB Library Expert PRO 2015, which I fortunately have licenses for. Below are layered screenshots:






The board is made on a standard 4-layer stack, which is especially logical due to the presence of USB and Ethernet, which, of course, are separated by diffpairs - while the suppressors from all interfaces have their own ground, not connected to the signal ground (and this ground should always be connected). All power circuits are laid out polygonly, for the main supply voltages corresponding polygons are allocated in the layers of the board, thermal barriers are completely absent. Components with diffpairs are located on one side, most signal paths are on the opposite. Techno norms were taken from standard debugs, though the situation was somewhat spoiled by IP4252CZ16-8, in which the case is designed for a "different" board. All connectors are designed for surface mounting, including DB9 for CAN and RS-485. The final board size is 105x74mm.
A bit of 3D:



“KYRNN” is easy to understand - not arduino or regular debugging. Of course, this is never the ideal PCB design, but a project with a special ideology (performance first) and rather specific solutions taken from specialized industrial applications. The project is not positioned as an “answer” and all the more as a replacement for “ordinary” evalboards, but it is, without false modesty, their much more advanced, secure, reliable and high-quality option - while, of course, it focuses on a fairly narrow audience, interested in solving / practicing tasks with a level other than "just play with the stone."
I must say that this is not the last board made with such a key. "KYRNN" is the basic version of a series of devices that will have many other interesting features (isolated interfaces, high-end measuring chips for 2 and 3 phase power supply, eInk displays, Qi and much more) - some of them will be in the public domain as well as "KYRNN ".
Predicting a number of questions:
1. The name is made based on its own system of naming conventions.
2. The above said about the established approach to creating evalboards is not a manifest or an appeal to change something; on the contrary, in its current form it is excellent for the tasks of "making a penny board for everyone." All my criticism of the "ordinary" debugging is applicable only to specialized tasks for which this board is designed.
3. A set of files in the archive is more than enough to make any production documentation for the plant and make minor modifications - it was not planned to lay large opportunities.
4. Footprints are called “in a simple way”, without IPC naming conventions, so that they can understand amateur radio and just beginners.
The board file, the scheme in PDF, BOM and STEP model are available for download . If anyone succeeds in finding errors, please report them.
Happy New Year, everyone!

It uses components with quality that are orders of magnitude higher than everything that is on such devices, as well as some of the fact that many evalboards often do not get at all . The board file with the circuit is available for download - everyone interested is welcome to cat.
Now there are a lot of ARM-based debugging boards, especially for STM32 - it affects the availability of these stones and related documentation, an abundance of sample programs, the availability of IDE sets and a fairly adequate price. Whoever did not make boards for them - there are, among others, varieties of the well-known Arduino, which are based on the mentioned microcontrollers. Variants of devices based on MK from different STM32 families (both commercial and amateur) are widely available, with a rather different set of functions (from a bare board with seats for headers, to samples with screens, cameras, various sensors, etc.) and in various form factors. And it’s logical to ask: KYRNN, is that yet another board? Before answering this question, a few words about how commercial evalboards are made, what they are,
Probably, many of those who once felt the need for a debug board for a particular platform noticed an unusual similarity of such products from completely different manufacturers. If you haven’t noticed, then there’s a small hint - they are really similar in many respects: picking up several boards with a more or less matching set of functions (for example, Ethernet + RS232 + USB + SD + a pair of sensors like an accelerometer) the similarity will be as if they are different minor revisions of one device . At the same time, partnumbers of components (or are direct replacements / analogues), switching schemes, and, of course, operational characteristics, which will be discussed later, will most likely coincide.
From this phenomenon of surprising similarity, a completely reasonable question arises about yet another board, since its relevance increases slightly more slowly than the inadequate increase in the number of Asian (especially Chinese) debugs and their clones. It would seem, well, even if they are almost the same, what's wrong with that? The catch is that the vast majority of evalboards on absolutely any platform (except MilSpec / RadHard / Space and some RF devices) are made in many of the lowest-grade components (not counting the stone itself), although they allow, at the very least, to launch this very platform (as comrades of a certain category like to say on this score: “Well, it worked somehow, then everything is fine”). Pure capitalism - it’s necessary to “do it cheaply,” therefore, Asian noname components are used, again the origin and quality of which is often not known to Asians themselves. Surprisingly abnormal switching schemes of various kinds of microcircuits are often used (for example, powering the paper displays) - some samples are very striking that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY). power supply of paper displays) - some samples are very striking in that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY). power supply of paper displays) - some samples are very striking in that they generally work. In essence, such devices allow limited play with the target platform in a rather narrow range of operating conditions (without EFT / ESD, at room temperature, that is, essentially only in a laboratory setting), while the implementation of the board itself often does not make it possible to test some of the key stone features (for example, the lack of some interfaces like CAN - but with an abundance of RS232, or the implementation of Ethernet through an interface converter instead of a normal PHY).
Of course, it’s especially worth noting the wiring of such boards: when I come across another piece in which the ground is looped along narrow boards along the circuit, the power circuits are separated by “tracks” - and the components and vias from these circuits are connected through thermal barriers, USB is not spread out in a two-layer diffparames and many similar phenomena from the indicated semantic series, I understand that I could write a small book (pages 100-200) about all the techniques of the curve, wretched and frankly crap wiring, which are classics, the standard for fact in evalboard (and in quite a few commercial and even industrial grade devices). The exception, perhaps, is the motherboard for the top FPGAs - also imperfect, but from this point of view it is a cut above other similar devices. The fact completes the picture, that some such evalboards are made in Cadence Allegro - a powerful CAD system in which you can design the most complex boards. A funny sight: miserable arrangement, scary components and footprints, vile layout, but high-speed signals are surprisingly well divorced and alignment is done (the last two points are one of the foundations of the "strength" of packages like Allegro).
Much more can be written about these miracles, but this will be a completely different article, and therefore we will return to the topic of debugging for STM32 and actually XFR-L3 “KYRNN”. The implemented functional set (including for comparison) is taken as follows: USB + Ethernet + CAN + RS485 + uSD. A few words about the implementation of each of them:
USB - taken from the stone itself (in this case it is STM32F439VGT6, while the signal lines are protected by two ESD5V3U1U-02LRH. Also, the MAX14575 power switch with filters and suppressor is installed, the connector costs 65100516121.
Ethernet - implemented on TLK106 enabled via MII. In addition to PHY itself, it allows cable diagnostics using TDR. Protected by assemblies LC03-3.3, connector 7498111001.
RS485- MAX13451E is responsible for it, which allows you to terminate the line with different resistances, invert channels and much more. It is protected by SMA6J suppressors and TBU-CA assemblies, connector 09551666817741.
CAN - made on MAX13041, a rather successful and reliable transceiver. It is protected by SMA6J suppressors and TBU-CA assemblies, connector 09551666817741.
SD - removed from the stone itself. There is an IP4252CZ16-8 assembly with filters and suppressors, socket 0475790001.
In addition, there is an external MRAM MR25H256 - an exceptionally reliable kind of memory that can be used both as RAM and as ROM. Power supply, designed for stable + 12V, organized on two LD1086. An external watchdog STWD100 is installed for MK.
In fact, this is a project of an industrial-grade device based on premium components operating in the industrial temperature range and having sufficiently reliable sets of protections. Although the hardware set is, from a certain point of view, rather start-up, nevertheless, its implementation is a cut above all that is offered in the overwhelming majority of debugs (and KYRNN is positioned just like evalboard).
I want to emphasize several important points:
- the components are selected exclusively top-end and expensive, they can not be replaced;
- footprints on many of them also exclude the installation of anything that does not fit into the ideology of this board;
- performance first, even if it requires tougher manufacturing standards (as it happened with IP4252CZ16-8, for example).
At this point, it is logical to proceed to the board layout itself. It was created using Altium Designer 15 and PCB Library Expert PRO 2015, which I fortunately have licenses for. Below are layered screenshots:






The board is made on a standard 4-layer stack, which is especially logical due to the presence of USB and Ethernet, which, of course, are separated by diffpairs - while the suppressors from all interfaces have their own ground, not connected to the signal ground (and this ground should always be connected). All power circuits are laid out polygonly, for the main supply voltages corresponding polygons are allocated in the layers of the board, thermal barriers are completely absent. Components with diffpairs are located on one side, most signal paths are on the opposite. Techno norms were taken from standard debugs, though the situation was somewhat spoiled by IP4252CZ16-8, in which the case is designed for a "different" board. All connectors are designed for surface mounting, including DB9 for CAN and RS-485. The final board size is 105x74mm.
A bit of 3D:



“KYRNN” is easy to understand - not arduino or regular debugging. Of course, this is never the ideal PCB design, but a project with a special ideology (performance first) and rather specific solutions taken from specialized industrial applications. The project is not positioned as an “answer” and all the more as a replacement for “ordinary” evalboards, but it is, without false modesty, their much more advanced, secure, reliable and high-quality option - while, of course, it focuses on a fairly narrow audience, interested in solving / practicing tasks with a level other than "just play with the stone."
I must say that this is not the last board made with such a key. "KYRNN" is the basic version of a series of devices that will have many other interesting features (isolated interfaces, high-end measuring chips for 2 and 3 phase power supply, eInk displays, Qi and much more) - some of them will be in the public domain as well as "KYRNN ".
Predicting a number of questions:
1. The name is made based on its own system of naming conventions.
2. The above said about the established approach to creating evalboards is not a manifest or an appeal to change something; on the contrary, in its current form it is excellent for the tasks of "making a penny board for everyone." All my criticism of the "ordinary" debugging is applicable only to specialized tasks for which this board is designed.
3. A set of files in the archive is more than enough to make any production documentation for the plant and make minor modifications - it was not planned to lay large opportunities.
4. Footprints are called “in a simple way”, without IPC naming conventions, so that they can understand amateur radio and just beginners.
The board file, the scheme in PDF, BOM and STEP model are available for download . If anyone succeeds in finding errors, please report them.
Happy New Year, everyone!