Duggle Posted October 20, 2012 Report Share Posted October 20, 2012 Sorry to arrive on this thread so late with advice for improvement. It's about the cathode row drivers and the brightness of the display: At present each row is driven by a single pin of a HC595. Each pin may draw only 35mA and is driving up 16 LEDs at 1/16 duty cycle. I predict that the current design is dull (not bright) and has excessive flicker. I'm proposing two changes that will improve the display substantially: Place a ULN2803 driver in line each of with SR1 and SR2. This will increase the max current in each row to around 1A. Will result in much brighter. See SmashTV info pages on the DOUT module wiring options for the picture with ULN28xx soldered into the position of the current limiting resistors. Note that the GND pin on the driver needs to be connected to power GND on the PCB.Include an extra pair of shift registers (duplicate SR3, SR4 and RN1,RN2) so that the display is driven in two halves. This will double the refresh rate (1/8 duty) resulting in double brightness and much less flicker. If I am correct, all of TK's designs have a maximum of 1/8 duty cycle.I think I read that this is the authors first attempt at PCB design. Well done, its a very nice layout! Those of us that have been doing this for many years know that you never achieve perfection first up. ( But you DO on subsequent iterations!) Quote Link to comment Share on other sites More sharing options...
NorthernLightX Posted October 21, 2012 Report Share Posted October 21, 2012 Sorry to arrive on this thread so late with advice for improvement. It's about the cathode row drivers and the brightness of the display: At present each row is driven by a single pin of a HC595. Each pin may draw only 35mA and is driving up 16 LEDs at 1/16 duty cycle. I predict that the current design is dull (not bright) and has excessive flicker. I'm proposing two changes that will improve the display substantially: Place a ULN2803 driver in line each of with SR1 and SR2. This will increase the max current in each row to around 1A. Will result in much brighter. See SmashTV info pages on the DOUT module wiring options for the picture with ULN28xx soldered into the position of the current limiting resistors. Note that the GND pin on the driver needs to be connected to power GND on the PCB. Brightness is overrated :). No seriously, even on SmashTV's page the ULN drivers are specified for driving relays. Include an extra pair of shift registers (duplicate SR3, SR4 and RN1,RN2) so that the display is driven in two halves. This will double the refresh rate (1/8 duty) resulting in double brightness and much less flicker. If I am correct, all of TK's designs have a maximum of 1/8 duty cycle. In this example TK uses 11 LEDs in the exact same fashion that Fairlightiii used in his design. I can't judge if the duty cycle is important, but the brightness should be fine I guess. :flowers: Quote Link to comment Share on other sites More sharing options...
Duggle Posted October 21, 2012 Report Share Posted October 21, 2012 Brightness may not be that important but if the LEDs disappear in moderate ambient room light then it cant be a good thing. If the row driver (which is limited to 35mA) is divided by 16 ( all row LEDs lit) then we have 2.2mA per LED. With 1/16 duty cycle then we'd have the equivalent brightness of 0.13mA per LED. That's really dim! ULN2803 makes a really convenient row driver used with a DOUT shift register ( the pins line up very well) and work really well in this specific application. With the standard 1mS SRIO refresh rate and a column driver with 8 rows instead of 16 that' s 1/8ms=125Hz with the current design 75.5Hz. The human eye will start to notice flicker below 100Hz. Hence the suggestion of extra shift registers. I've proposed the two changes as they would have a pretty big impact on the appearance of the display with a modest increase in parts count. Quote Link to comment Share on other sites More sharing options...
rvlt Posted October 21, 2012 Report Share Posted October 21, 2012 How much current does this board draw now? The reason I ask is I want to put it in a controller (LPC17 with quite a few buttons and leds) which is powered by USB now, and I wonder if this would still be doable if I ad this PCB. Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 21, 2012 Author Report Share Posted October 21, 2012 (edited) New revision of the board: DIL connectors instead SIL ones. They are 1:1 with J8/J9 of Core8/Core32/LPC17 (ilmenator :thumbsup: ). Regarding the brightness/refreshing/current problem I let someone else with more background than me to resolve this question. But several thoughts: >In his schematic, Tk didn't use transistors arrays nor doubled serial shift registers. >Technicaly, it should be enough easy to implement 2 ULN2803 into the board but 2 ULN28036 + 2 RN + 2 SR will be sensibly different to route. >I don't know exactly how could be connected the 2 new SR but regarding the fact the core manage only 16 SR, we would not be able to use more than 2 serial boards (and I had planned to use 3 boards in my project). >Me and several future users eventually plan to use translucent acrylic (grey e.g.) and the brightness of the LED will be still more dimmed. Some help is welcome. Regards.MBHP_DINx4_DOUTx4_Rev2.1_layout.pdfMBHP_DINx4_DOUTx4_Rev2.1_schematic.pdf Edited October 22, 2012 by Fairlightiii Quote Link to comment Share on other sites More sharing options...
Duggle Posted October 21, 2012 Report Share Posted October 21, 2012 The current layout is a little tight to be adding too many chips, I agree. The biggest improvement to brightness would be to add the ULN2803 to the row drivers. In the case of "all LED on" the maximum current for a whole row goes from 35mA to 16*35mA (the limit now coming from the column driver). That's 16 times brighter! Actually there is a current limit of something like 70mA per SR but the brightness increase will still be huge! The flicker issue (if it is desired) can be overcome in software (without adding SR's) by scheduling extra SRIO scans to say double the default rate. (BTW: Long chains of SR's (much higher than16) can be used with MBHP if good routing and termination are done, it's quite simple. TK has stated the limitation to stop newbies getting into trouble.) Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 22, 2012 Author Report Share Posted October 22, 2012 This is the schematic with the two ULN2803. The 4 extra chips can not definitly be integrated because there is no more free space. Duggle, can you check the wiring of them? People that eventually won't want to use them could bridge pins. I made tests with the board I wired and in well lighted room the LEDs seem enough lighted for use as indicator with "nude" top. But with the use of translucent grey acrylic I plan to use as frontpanel, it seems (for me) it will be probably not enough. Regards.MBHP_DINx4_DOUTx4_Rev2.2_Schematic.pdf Quote Link to comment Share on other sites More sharing options...
smashtv Posted October 22, 2012 Report Share Posted October 22, 2012 Hi guys, Late to the party as usual, I'll quote/reply on a few things.... >Eagle say me that I have several "Zero length airwires" (yellow X-shaped cross) I still don't know how I can handle them (and if it is crucial for the building of the board).This is an artifact of auto routing on a different grid than placement and requires hours/days of manual cleanup. Not crucial for making the board as long as all the connections are complete in the fabrication files (trace meets pad/no gaps) - but it can make the DRC useless which can hide more serious problems. However, I think it would attract even more people if it was more modular - e.g. 4x2 instead of 8x2, or even 8x1 /4x1, because it is unclear whether other projects really have that much real estate.andYou could ask Tim/SmashTV, if he is willing to put it to his shop, I think many people would be interested in it, who are trying to improve the user interfaces of their old synths :).If you guys could come up with a basic spec for LED count and spacing something modular as a permanent shop offering would be workable. ...All seems ok except the layer 121 (_tsilk) which is not silkscreened!You have to specifically include it in the eagle CAM processor when generating the fabrication files, some of the default CAM processor settings are missing some layers vs. a 'full spec' board. I've asked to my PCB manufacturer and the price should be around 11-14 euros each (+shipping from me to you) depending the number of people interested.This is a good price for small quantity.... my supply/vendor chain is more suited for large runs with high NRE/setup charges that do not make sense at lower quantity. I propose to submit the board to Tim/SmashTV and if he is ok to sell this board in his shop and make a limited bulk order for (at least) non US people.I am happy to help with a bulk order for this without markup/at cost..... I can pack & ship to destinations where it saves cash or modify one of my ordering scripts to handle pre-payment and ordering, etc..... let me know how I can help. Place a ULN2803 driver in line each of with SR1 and SR2.+1. What works fine with 15mA LEDs without drivers might fail hard with 25mA LEDs....lots of factors involved like per chip fanout capacity variance between different manufacturers, trace width, scan rate, capacitance on the LED supply rails, etc. Those of us that have been doing this for many years know that you never achieve perfection first up. ( But you DO on subsequent iterations!)Absolutely..... this is a very good first effort. Some of the aspects of the design make me nervous, like 10 mil traces with 8 mil spacing..... but I live by manufacturing rules where I must consider issues like over-etch or a speck of dust on a photoplot affecting yield percentages. If you are stuck on the routing after the driver addition(s) let me know, I'll give it a try. Sometimes I can get the high score on trace tetris. :) Does a .sch/.brd version of this exist that has the schematic changes from your last post and the LEDs and encoders still in place on the board? Best regards Tim Quote Link to comment Share on other sites More sharing options...
Hawkeye Posted October 23, 2012 Report Share Posted October 23, 2012 (edited) Hi Tim :) If you are willing to improve it, and stock it in shop, I think everyone would profit - but please do not invest too much time or money - maybe nobody, but a few LED fanatics will be interested in this board? :-) Here is a personal wishlist, everyone might disagree, but I still have that wish for a FS1R controller and that synth has more controllable functions than an A380 airbus and thus needs a few encoders :): * If you ask me, I think for normal "MIDIbox" use, it is totally sufficient to have one lit LED per encoder, so the aggregated current consumption should not be so high? We would have only 16 LEDs on permanently, driven by four shift registers, flicker can surely be addressed by hacking the update routine speed, as Duggle said, but it should better be tested before... * 16 LEDs per wheel and a 16x16 LED matrix is a great design choice, as is the 8x2 encoder layout on the board. If you reduce the number of encoders per "modular board", you also have to reduce the LED matrix size and need more shift registers for the same encoder amount... also limiting the maximum encoder amount of the controller (even with good termination). * Regarding space - i´d say as small as possible with through-hole LEDs and common "pushable" encoders, don´t know what is possible with routing... * Shift registers and other components should be on the backside, if possible, so that the frontside is clean and can be used as a surface just as the fine MBSEQV4L, white PCB color preferred :). * SMD components for shift registers and diodes would be fine. * Also, an encoder "push acceleration"-bus would be really nice, if a pushable encoder is pressed down, it should just short the bus, so that a global "FAST" feature can be used - it is really nice for intuitive encoder use... Just two pins to be forwarded to the neighboring boards... Many greets, Peter Edited October 23, 2012 by Hawkeye Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 25, 2012 Author Report Share Posted October 25, 2012 (edited) Hi. This is the new revision of the board. Improvements: >C1, C2, C3 & C4 replaced by Wima cap footprint (bigger pad so easier to solder). >2 ULN2803 driver placed in line with SR1 and SR2. >16 PEC16-4xxxF-Nxxxx (without switch) replaced by 16 PEC16-4xxxF-Sxxxx (with switch). >1 pin of each switch is grounded and the other connected to 16 ways SIL connector (Peter/Hawkeye can explain how using them :rolleyes:). The new Eagle files have been sent to Tim. smashtv: I am happy to help with a bulk order for this without markup/at cost..... I can pack & ship to destinations where it saves cash or modify one of my ordering scripts to handle pre-payment and ordering, etc..... let me know how I can help. and Hawkeye: maybe nobody, but a few LED fanatics will be interested in this board? :-) An idea is to start a bulk order to see exactly how people are interested (...soon). Regards. Jerome.MBHP_DINx4_DOUTx4_Rev2.3_Schematic.pdfMBHP_DINx4_DOUTx4_Rev2.3_Layout.pdf Edited October 25, 2012 by Fairlightiii Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 25, 2012 Author Report Share Posted October 25, 2012 ...better views of top and bottom silkscreen.Bottom silkscreen.pdfTop silkscreen.pdf Quote Link to comment Share on other sites More sharing options...
Hawkeye Posted October 27, 2012 Report Share Posted October 27, 2012 Very nice! Thx for adding the encoder switch lines! By having 16 separate lines, it makes the board even more flexible, allowing to query 16 individual encoder switches, if needed. Otherwise, one can "short" them on the 16pin switch connector to get a single "acceleration" line. This design is good, because it reduces the total length of the switch "bus" - being starwired here, instead of using one long bus system with 70cm of trace on a single board... Many greets, Peter Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 27, 2012 Author Report Share Posted October 27, 2012 (edited) Some useful technical informations: LEDring: Board: Regards, Jerome. Edited October 27, 2012 by Fairlightiii Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 27, 2012 Author Report Share Posted October 27, 2012 (edited) BOM Coordinates PS: Someone know how inserting a table in a post? Edited October 27, 2012 by Fairlightiii Quote Link to comment Share on other sites More sharing options...
Fairlightiii Posted October 29, 2012 Author Report Share Posted October 29, 2012 The is now open. Regards. Quote Link to comment Share on other sites More sharing options...
FantomXR Posted November 6, 2013 Report Share Posted November 6, 2013 Hi, I want to create a panel-layout for your board. Do you have an Illustrator-file or something else to import it to Illustrator? It safes work and the risks to make mistakes are much lower ;) Thanks!! Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.