#6 -- More Timer Examples

We still got a couple of minutes yet, but I wanted to pass around one of these target boards This is an example of a board. This is a built version of the board you can build I supply the the green circuit board and I supply all the parts You build it and you can keep it or I'll take it back if you don't want it the this year's version instead of having a serial port here as a USB port and And next year's version will have this end cut off with just a header for either a USB port a serial port or a radio so If you wanted to this year, you could cut the end off but But these will have enough of these that everybody can build one or two if they want you have to budget it However, if you build it for the final project the budget cost of this completed is about $15 You don't pay $15. That's how much you have to subtract from your effective budget of $75 by the way, the If you go over budget on the final project your score Don't goes down as the square of the amount of money that you are over So going over 10 cents doesn't matter Going over $10. It turns out matters quite a lot So I'll pass this around pass it around the class I Like I like nonlinear relationships. It encourages certain kinds of behavior so for instance the there for the final project there's going to be a bonus I said everybody for finishing early and Especially the Friday group the slope day group The 30 people who are going to be there all afternoon on slope day So if you finish early You get a score boost which is proportional to the square root of the number of days early so it's it's it's It's a fairly large boost for a few days and a fairly small boost for a lot of days You finish 30 days early. I think it's worth 15 points 30 days early is you finish before you start? There's one group that's already said they're going to do that Any questions about lab one so far of the three lab sessions we've had Two groups have completely finished and checked out It's not bad pretty good Seems a pretty good progress Again any questions about lab one or homework one So there's three small topics That are sort of lingering with respect to lab one before we start lab two One of them is how do you generate a square wave? This which is answered in the code example by the way on the on the lab Another question is how do we use EEPROM exactly which is also answered? In a demo code on lab one and the third one is how do you write a debounce state machine? Which as of this morning has a code linked online? but Was actually linked online before but I didn't associate it with code with lab one So which of those three would you like to hear about first? show of hands Now wake somebody up sorry EEPROM square waves Debounce oh well, okay, it's a clear winner debouncing the EEPROM is actually embarrassingly easy to use The generating a square wave using timer two is Also quite easy in fact. It's so easy that that's what's confusing people look at the code and say that can't possibly work All you need is five lines of code, which is never repeated. It's in the header. It's in the init session section Before the while one there is no interrupt service routine associated with it You initialize the timer to make a square wave it makes a square wave forever Oops, how do you turn it off? And I know by the way it looks like the homework What's the easiest way of turning off the square wave if it's a if it's an autonomous timer? With no software that's generating square wave Disable the pre-scaler you can set the Z pre-scaler to zero it stops counting You turn it back on starts counting again. It's absolutely the simplest way Oh, you can mess you can mess with the with the pin There's three options for the pin set on match clear on match toggle on match If you set it to toggle on match it produces a square wave if you set it to either clear or set on match It does not so that would work also Okay, debounce Do I need to draw the debounce state machine again the diagram now that was the last thing I did last time I think so let's now go back to a version of the code which is very similar to sketch one Except that we're going to detect two buttons being pushed button zero is going to be not debounced button one will be debounced and Both of the buttons when pushed will increment a counter which is then displayed on the LEDs So I'm not going to write the interrupt service routine. It's the usual one millisecond interrupt service routine I'm not going to write main, which is just a Dispatcher that dispatches three tasks all I'm going to write is the task of that Absorbs the button push and turns it into an increment and then the two the two tasks one of which does a Debounced detect and the other does a non debounce to detect I do not have a preference on it, but People have a tendency to push really hard on buttons when they're trying to do it fast and The board so you have this circuit board like this which is supported by a foot at each end Button seven is here button zero is here and button three is in the center Which one do you think breaks the board the fastest when you push hard on it? Three so the flex on a board goes is what power of the distance from the support Mechanical engineer third power Use seven or zero Okay, so task one task one is going to Notice that a button has been pushed and increment the timer in increment a counter So or say if well Task two and task three are both going to set a flag called push flag so if push flag We're going to set push flag together to zero to So there so that we get one increment per button push We're going to then increment the LED variable and Then set port B to not LED because again the LEDs are active low and Then we're going to end the if and then we're going to end the task so task two then is going to do a button detect without Debounce no debounce and what we're going to do is we're going to Say if not pin D is Equal to 0 X 0 1 that is to say button 0 then flag equals 1 and Straightforward and the behavior of the system when you push the button is that it will be Impossible to get exactly one count per button push so for task 3 Will be the debounced version which will give you precisely one count of the LED every time you push and release the button We're going to enter task 3 every 30 milliseconds so it's going to We're going to get into task 3 every 30 milliseconds. We're getting into task 2 every 30 milliseconds also But with task 2 unless you happen to exactly hit The button at the time the task executes You miss it either miss the event if you have a very very fast finger Or you get two events or six events And what we're going to do over here is we're just going to build a big old state machine Which switches off some variable called push state? Push states going to have four possible values Corresponding to the names of the states in the state diet state transition diagram. I did last time which I think was Release may be pushed pushed may be released So in the switch statement in the switch statement, then we're going to start with a case of no push or release and that's a full colon and In this state so we fall into the into the task every 30 milliseconds We search for the state if we're in state not pushed then we we do a if pin D if not pin D is equal to Equal equals 0 X 0 to That being button 1 the value of button 1 then push state Goes to maybe else Where do we go? In other words, I wouldn't actually have to put an else in here because the state doesn't change But for symmetry, I'm going to say else push state Equals no push, which is probably is the state. It's already in and Then because of the way C handles case statements you have to put a break here Otherwise it falls through to the next case by default unlike MATLAB, which does exactly the opposite and Verilog, which does This does not fall through Well, so case the next case in the case statement is Going to be case maybe push. I Guess I just called it maybe down there maybe full colon and Then boringly enough if not pin D equals equals 0 X 0 2 now here we have to need two statements one is Push state equals What? We had two successful pushes in a row therefore we have a valid push and so we're going to go to state push and We have to set the push flag else If it's not pushed Where do we go? Where? Where what's the what's the consensus? No push break case pushed if Pin D not pin D is equal to equals 0 X 0 2 push state equals Pushed we stay in this state else push state equals Maybe release break and then case May release if not pin D equal equal 0 X 0 2 this this phrase here would be a perfect thing to put in a text macro a Define, but I didn't do that if pin D is equal to 0 X 0 2 push state Equals oh We got it. We got a we got a release up here, but now we're back to push Meaning that the button is still bouncing So where do we have to go? Say again Nope Go back to pushed else Else at this point now. We've gotten two releases in a row Which means that the button is truly and really? Released and so we go to No push and then we end the case statement, and then we end the task You don't need to put a break there because it's the last one so falls through all over symmetry I often do put a break there This is only detecting one button oh Yes, yes, it would fail. Yeah, no no you're quite right It would fail, and that's probably correct behavior to have it fail in this case Right so what that does is to increment the LED variable and then output it so it's counting So it's a counter and with the debounced version every ten every time you press button one you increment the count Precisely once with the undebounced non-debounced non-debounced button button zero every time you push the button you get a count someplace between zero and the amount of latency that your Finger has on the button divided by 30 milliseconds So the debounce code gives you a way of detecting Discrete events as push-button pushes or push-button pushes as discrete events For lab 2 you absolutely have to debounce the entire Keypad we're going to be using a 16 button keypad instead of the the Stk 500 buttons and you have to debounce them because you're going to be switching back and forth between Functions using the keypad for the digital voltmeter for this lab you absolutely have to debounce the release of the button Not necessarily the push, but the release Before you push again It'll keep counting Yeah, every 30 milliseconds like a cell phone or a Dialer of any sort you you don't want to if you if you're if you hold it down a little bit while you're thinking about The number you don't want to get six digits all the same so in this lab You push once to start the system if you're still pushed two seconds later You're either cheating or you've gone to sleep All right, so you do not have to detect the button up Unless you want so cheating could have two possibilities in this lab either which of which is okay either either alternative of which is okay One is you push and release and if you push again? before the Random timeout then you've cheated The other interpretation is you can push and release as many times as you want Before the random timeout as long as you are released at the moment that the timeout occurs See the difference Either one of those is a is okay one of them requires you to debounce the other one does not Once you've pushed and released to start the process going a second and a half between one second and two seconds later you're going to The bells going to go off you have to push the button as fast as you can That push button that push cannot be debounced because you can't tolerate the 30 millisecond ambiguity Because you're timing down to the millisecond The release of that button has to be debounced you have to have released it for at least two 30 millisecond Intervals in a row before you can press again to send the system back to the ready state So the only button transition that absolutely has to be debounced is the release after the timing after after you've timed your reflexes anything else Well, let's talk a little bit about Just a little bit about square wave generation, then this is it like I said embarrassingly easy The demo code that I produced the I think it's called timers GCC Uses timers in three different ways because I wanted to put as much as I could in in one code So the timer zero does it with the usual one millisecond time base? timer one is sitting there waiting for a positive transition and Every positive transition on its input pin it Records its own value into a register called the capture register, and it throws an interrupt the interrupt service routine then and goes and reads the capture register and gets a cycle accurate time of occurrence of the transition So the timer one is running continuously very fast 16 times a microsecond Takes five or six microseconds to get into the interrupt service routine The error of service routine then looks back and says oh, I know the exact time that occurred because it's frozen in a separate register So even though there's a software latency to read the the capture register The capture register has the exact time of the pin transition frozen in it Very handy so if I were to measure two transitions in a row I Could get a time between them or a frequency therefore Which you'll need for lab two because one of the things I want you to do is build a frequency meter So you'll have to use timer one in capture mode to build the frequency meter but for this lab The frequency that's being generated The square wave which is being generated that is being fed into timer one is being generated by timer two Timer two is sitting in an autonomous square wave mode generating a 200 cycle long square wave which is then being measured by timer one and guess what it measures 200 cycles They actually gave me some faith that it was cycle accurate It also meant that for purposes of demo I didn't have to hook up a square wave generator and try and figure out which was more accurate the square wave generator of the microcontroller So the square wave generator then first of all we choose the we choose the the half duration OCR to a output compare register Timer to channel a I'm going to set to 99 so a half cycle is a hundred cycles That's a hundred. That's a hundred. That's a hundred because I'm going to be toggling the I open I set TC CR to be to one Which is full speed? That's a pre-scaler divided by one I set TCC are to a to Turn on the Calm to a zero bit and or that with the Wg M 21 bit this This commands the timer This bit commands the timer to connect its compare register to the output pin this can turn it connects the timer output the compare register output to the To an IO pin so it physically makes IO This one causes the timer to clear on compare match so It's going to count up to a compare match. It's going to flip the IO bit from one state to another It's going to count up to an IO match. It's going to clear itself. It's going to flip the bit. It's going to Count up to timer match clear itself flip the bit and so on the only other thing we need to do is to set Ddrd the data direction register for Port D to turn on pin d7 as an output Why d7 It's the it's the pin It's the pin the designers decided to connect from timer to to the output There's no reason Avr The architecture we're using stands for Arnie and Verne's risk machine it's whatever Arnie and Verne decided to do and They decided to hook timer to the pin d dot 7 and so that's what you get to do There's no reason. It's just policy Once you execute these four commands Timer to starts to produce a square wave forever The easiest way to toggle it on and off is To toggle this between 1 0 1 and 0 turns it on and off 1 is no divide, so it's full speed So what's 200 cycles? Let's see 200 cycles is Divide that by 16 that's about Mmm There's a 16 millisecond 60 microseconds right 16 microseconds, so That's what you said that's about 600 kilohertz 800 kilohertz I don't think you'd hear that So you have to divide you're gonna have to set this to a different number to actually hear it of course What do you have to set it to? thousands billions 20 you can hear 20 kilohertz Yeah, that's enough to cause the th to come running over and say what's wrong with you I Can't hear 20 kilohertz anymore. I can't hear you know old to some televisions produce 15,725 Hertz because that's the high-frequency transformer that generates the high voltage. I can't hear that anymore In fact I can't hear above about 9 kilohertz unless it's really loud That's what happens when you listen too much loud music on your earbuds or You run a chainsaw for too long Or you fire a shotgun too many times without earplugs all of which I've done Not at the same time You Know when I was a kid nobody thought about ear protection you you went out, and you'd blast away with a chainsaw. Yeah, yeah You can't hear so good when you come back or go to the go to the drag races and sit in the pitch You know and you have these flames shooting six feet out of these v8 engines into your face you know and your whole body's shaking like this because it's about 180 DB's and The only ear protection is you put your you put your fingers in your ears So it doesn't blow your eardrums out when they hit the throttle Is this and then you wonder why you can't hear the traffic as you're driving home Well But it's worse with earbuds because it's constant and the damage is cumulative and it's never repaired Enough preaching Any questions on this Easiest thing to do generate a square wave is of course write four lines of code and listen to what comes out And tune it that way Well, you can calculate within point 0.005 percent what frequency the Square wave will be because that's about the accuracy of the crystal about 50 parts per million any questions If So yes, they do sell higher accuracy ones, but to get much higher than 50 parts per million You have to temperature regulate so you're one reason that your wrist watch is rather accurate is that your wrist is always about the same temperature and So and if the crystal is in thermal contact with your wrist, it always stays about 75 or 80 degrees or so and You can cut the crystal in a certain way so that it has a negative temperature coefficient below a certain temperature and a positive temperature coefficient above the temperature or vice versa and At the temperature you want to operate it has zero temperature coefficient. That's how they cut crystals for watches So they have a zero temperature coefficient at about 80 degrees the other way to do it is if you want extremely high accuracy is You run them really hot You run them at a temperature which is higher than any reasonable temperature you're gonna get in the environment say 50 C and you put them in an oven and you keep them hot at that temperature and you let them stabilize for a month and And then the clock you can get down to a part per million or so Still not no atomic clock But it's pretty accurate You can do that with a with a with a crystal clock it's it's probably not worth the trouble because you have to mechanically change the crystal and Radio amateurs found out decades ago that you could modify Slightly the frequency of a crystal oscillator by putting a set screw against it and torquing it Pushing physically on the crystal you can bend the resonant frequency But you don't want to get into that the control system for that's a mess Okay, we have a little more time You want to do you want to talk about EE prompt, or do you want to start on lab 2? Lab 2 EE prompt how many for lab 2? How many for EE prompt oh? Man, it's about so almost split Well, I'll spend 30 seconds to tell you everything. I know about EE prompt First is You can define You don't have to do this. You can just use integers, but I like to define Constants Let's say I want to store data in EE prom position address one I Can define a constant EE prom data, and then I can do an EE prom later in the program? Assuming ah assuming that I have included if I do it include of AVR EE prom dot H Then I can say EE prom right bite, and I'm doing a bite, so I'm going to cast this to a I'm going to cast this to a bite pointer a uint 8 star EE prom data So I'm going to cast that to the appropriate pointer to EE prom and Then write a data in it. Let's say 20 that's the data value and that will put a bite into EE prom and It will be there even after you cycle the power on and off even if you pull the chip out of the board put it Into another board it'll still be there as You might guess there's an EE prom right word There's also EE prom read byte which takes a uint 8 pointer to ROM prom It's a very easy user interface very easy to do Anything you want? That's correct, I was wondering so is Address one on the uint 16 pointer equal to address one on the bite pointer or two? In other words it should be two okay, so we're addressing the That Would be the consistent way that the that the that the the uint 16 should work I have to say I haven't tried that to see where the bits actually end up in this case And did it work So it might be in four Yeah, who cares right? Yeah? And tell you do we can tell you want to write dense stuff it doesn't matter yeah, yeah, I think you're doing it the right way Or you just define another constant Are you gonna write an array if you're gonna write an array? Yes, you're gonna increment the inner the pointer Yeah, yeah Typically you're only going to be putting a few values in EE prom There's also some tutorials that are linked up on EE prom which you might want to look at so for lab 2 For lab 2 we're going to build a Digital digital multimeter with just three functions. We've got to draw the line somewhere We could have done a lot more, but three is enough There's going to be a voltmeter We're going to have a resistance meter, and we're gonna have frequency and Keep things simple here We're going to make that the voltage that we're going to define that the voltage you need to measure is 0 to 5 volts That the resistance is going to be oh perhaps in the 100 ohm to 100 K. Ohm range So we're not going to get either ridiculously low or ridiculously high although. We could this is a little easier and the frequency is going to be in the Perhaps 10 Hertz to 100 K Hertz region and We're going to use a bunch of facilities Excuse me a bunch of the facilities of the microcontroller. We're going to use We're going to measure voltage. We're going to use an ADC an analog to digital converter That directly converts a voltage on port a Port a into a Number I'm going to ask you to auto range this It's going to have to be auto ranging Which means that you're going to need for a lower voltage you're going to need to Automatically get better sensitivity either by changing a reference voltage Internally or by using the internal amplifier, so there's going to be this so the voltage has to be auto ranging To measure a resistance Well, how do you measure resistance? Given that we have a voltmeter How do we measure a resistance? Jeremy Yeah, so we're going to have a current source of some kind We have a current source Which in this case is going to be just resistors hook the output ports and You're going to auto range the resistance by choosing which of the resistors happens to be an active output and which one is Turned off which outputs are turned off. How do you turn off an input output make it an input? Then it goes to high impedance The frequency meter Well, it would be nice to have that independent of amplitude. It would be nice to be able to Not use the ADC because if you have some ugly input waveform Then you have to do all this threshold detection and stuff But built into the microcontroller chip as you may already know or maybe not is an analog comparator There's an analog comparator that's quite fast nanosecond fast and It is pretty accurate a few millivolts So we're going to hook up a circuit that compares the input to some handy reference value like VCC over 2 and And Generates for this ugly input. It's the output of the comparator is going to be a nice clean square wave Of course, but transitioning on the zero transitions or the VCC over 2 transitions So we're going to be using the ADC for voltage We're going to be using The analog comparator to produce a square wave from ugly input to make a nice Easily counted waveform and then we're going to use the timer one capture Which very nicely can be triggered off of the output of the comparator? the output of the comparator can trigger the timer one capture and So we can capture every edge and once we have the positive edges or the negative edges of the waveform We've got the time of the waveform and therefore the frequency The period of the waveform and therefore the frequency Oh Okay You cannot set it up as a Schmitt trigger. There's no way to get The output signal from the comparator is not brought to the outside. So you cannot set it up as a Schmitt trigger It doesn't have very much hysteresis. And so you could have noise problems And that will be a limitation here This will automatically calculate the period as long as it's a as long as it is a cyclic waveform as Long as it's a cyclic waveform that crosses through VCC over 2 Then as long as periodic then the positive transitions should give you the period If the waveform is noisy that's going to produce some jitter in the period and therefore some jitter in the frequency But probably you'll be feeding in waveforms mostly from the signal generator, which will be beautiful and clean oh The chip is a Is not a wonderful arithmetic tip or a DSP chip But you can you can do an FFT at audio rate on this as long as you write it in integer arithmetic and Don't get too careful about scaling the output for optimum accuracy. You can get about a hundred and twenty eight points per per per window at 8 kilohertz sample rate But you'll be able to measure frequency much more accurately by timing the period than by doing FFT Now I'm going to talk about this a lot more. I'm just giving you the the nosebleed level overview of this right now so There's going to be three functions You're going to be using you need to read up on as soon as possible you need to read up on the analog digital converter section the More about I open a little bit about the angle analog comparator And I want to tell you the piece of the data sheet that describes the 80d converter is the most complicated piece of writing I have read this side of Umberto echo It is a mess is really hard at least for me to understand It was so hard to understand that I couldn't tell if I got it right until I wrote the code When the code ran I knew I understood it in fact in general. That's the way my brain works I don't understand stuff until I write the code myself And Your brain is probably the same way so plow through some of the ADC documentation But more importantly plow through the examples. I've seen a couple of people who have gone through the example codes very carefully and By very carefully. I mean they've written more on it than I did So there's more of their words written in red than I wrote in Than characters I wrote in Of source code that's about the level at which you have to read these codes You can't just say oh yeah four lines You really have to take it apart Why did he set this stupid bit? Anything else any other questions? Okay, see a bunch of you in the lab in a few minutes