Post by thread:[EE]: How to calculate linear power intervals for

I have tried to figure this out but it seems to be just beyond my mathematics level. I am implementing a PID loop to control the temperature in a small greenhouse. The heating will be provided by a typical household portable heater that has been modified so I have control of the AC applied to the heater elements via an opto-isolated triac. Phase angle control will consist of interrupting on the zero cross and loading the power setting to Timer1. The triac will be gated on when Timer1 overflows and interrupts. Running at 8 MHz there will be 16,666 instructions between each zero cross, so lots of resolution for the power setting. My understanding of PID loops is that they work better if the controlled variable has a linear response to the affect on the plant. Obviously if I change the gate time 1 ms, this will have a far different effect if the change was from 1 to 2 ms or from 3 to 4 ms. So, the question is, how would I calculate the offset times from the zero cross, for say, 256 equal power intervals of the half cycle. I may then do a linear interpolation at this point for any further resolution. I am sure that I might be able to totally ignore the non-linear response of the control variable and still get the loop to work, and I would like to hear suggestions to that end, however, I will still be curious as to the math required to do this calculation. regards, Bob Bullock -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

The power in chopped mains into resistive load is the defined integral on the domain between the firing angle and the end of the semiperiod. This is a defined integral of the sine squared function (Pchopped = 2*SumOf(Pi = {Upk*sin(ti)}^2/R, i=firing_angle to pi). You can neglect the start and the end where there is little power (and elegantly avoid the problem of when exactly the triac fires and extinguishes). You can reverse this function using a computer program to obtain a lookup table of firing times (convert radians in ti into usec/period) wrt power desired. I don't think that you need to interpolate at this level (256 steps on a 2kW heater is about 10 W per step - you'd have to have pretty good thermometers to see that amount of heat difference in a usual sized box or room). At most the output will jump between two adjacent values from time to time. Peter -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

> I don't think that you need to interpolate at this level (256 steps on a > 2kW heater is about 10 W per step - you'd have to have pretty good > thermometers to see that amount of heat difference in a usual sized box or > room). At most the output will jump between two adjacent values from time > to time. Even better, why not screw the phase angle and switch in (or not) complete power line cycles? At 60Hz that gives you 13 different power choices and still drive the heater with at least 5Hz to reduce stress. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

On Mon, 14 Oct 2002 17:56:38 -0400, you wrote: >> I don't think that you need to interpolate at this level (256 steps on a >> 2kW heater is about 10 W per step - you'd have to have pretty good >> thermometers to see that amount of heat difference in a usual sized box or >> room). At most the output will jump between two adjacent values from time >> to time. > >Even better, why not screw the phase angle and switch in (or not) complete >power line cycles? At 60Hz that gives you 13 different power choices and >still drive the heater with at least 5Hz to reduce stress. ....as well as having the lower noise of zero-cross switching. -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

> >Even better, why not screw the phase angle and switch in (or not) complete > >power line cycles? At 60Hz that gives you 13 different power choices and > >still drive the heater with at least 5Hz to reduce stress. > > This would be yet another good application of a Bresenham or phase > accumulator algorithm. Yes, that kind of algorithm would allow much finer control over the average heat output. However, it would also lower the fundamental frequency of the input. It's hard to say how much this may matter. You want to keep the temperature of the heater elements reasonably constant to reduce mechanical stress on them as they expand and contract. I used 5Hz just as an example that was admittedly pulled out of the air. I don't know what a reasonable number is, but the bigger the heating elements, the lower the frequency. 5Hz is probably more appropriate for thin elements not much thicker than large lightbulb filaments. Most heaters with thermostats do a good enough job regulating room temperature with two power settings (off, full on). I bet the greehouse heater of the original poster would work fine like that with the right control algorithm. Even if not, 4 settings would almost certainly be fine, with anything over 8 effectively "continuous" for the purpose. Given that, I'd prefer to optimize for low noise and low heater stress. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

>I used 5Hz just > as an example > that was admittedly pulled out of the air. I don't know what > a reasonable > number is, but the bigger the heating elements, the lower the > frequency. We have done exactly what Peter describes for an 8-bit res. phase angle controller. While the math is certainly fun, it turned out to be amazingly simple. Likewise a table was created to linearize the phase angle timing - While this works great, it was needed because the design is used to apply PID control to multi-kW heater loads in a *very* fast process where the PID cycle time can be as short as 100 mS. The heaters have very light filaments so as to be able to respond so quickly - hence they emit lot's of visible light as a by-product, and it was considered to be aestheically negative for any "flicker" to exist. I would never go through this kind of effort for a space heater, which needs only to be a veerrry sslloowww process, and can be often switched on & off in intervals of seconds, or even minutes. Phase angle firing can get messy when you realize that your resistive heater has some inductive quality that the phase angle firing makes evident in some funky unexpected way. > 5Hz is probably more appropriate for thin elements not much > thicker than > large lightbulb filaments. Well, not exactly... Short-wavelength heaters with low filament density will visibly flicker at low frequency. The thermal stress on the lamp end seals (rather than the filament) is the greater concern. I recall going through this on an old thread wherein it was demonstrated that phase angle firing allows for the longest life from low density filaments. more dense loads, such as cal-rods can be turned on and off in intervals of seconds with the only ill effect being inrush current on the control switch. Since there are no thin moly foil parts squeezed thru a thin section of quartz glass, the life of the rod heater is not as severely affected as would be an IR quartz "lamp". I have collected data on all sorts of heaters and many switching methods over the years - quartz SW heaters like phase-angle firing, high density elements will only have their life slightly compromised by being hid hard & often. I have been able to zero-cross fire IR lamps with a proportional control algorithm that mixes 1/2 cycles in such a way as to minimize flicker - different values require differing numbers of half cycles. The key here was to reduce the duration of off-time at low values. I would think it best to use a higher-density heater for space heating and to use zero-cross firing proportionally, perhaps with a 1 sec or longer time base. If you want to go simple and skip the math, use an MOC3062 zc optocoupler that's made to fire triacs or SCR pairs with minimal parts count. They can be connected directly to an I/O pin and there'll be less stress on the thyristors and virtually no noise. While this isn't recommended for IR lamp filaments, it's a sure bet for heater coils, cal-rods, even medium wavelength quartz-tubes. > > Most heaters with thermostats do a good enough job regulating room > temperature with two power settings (off, full on). I bet > the greehouse > heater of the original poster would work fine like that with the right > control algorithm. Bingo! look at the myriads of commercially available on/off PID controllers out there that work exatcly this way. This kind of process if fast enough for processing plastics, and can be 100's of times speedier than is needed to heat a greenhouse. -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Somebody said: > > >Even better, why not screw the phase angle and switch in (or not) > complete > > >power line cycles? At 60Hz that gives you 13 different power choices and > > >still drive the heater with at least 5Hz to reduce stress. Bob Ammerman commented: > > This would be yet another good application of a Bresenham or phase > > accumulator algorithm. Olin added: > Yes, that kind of algorithm would allow much finer control over the average > heat output. However, it would also lower the fundamental frequency of the > input. and now Bob Ammerman rebuts: Assuming you are switching on half cycles, the Bresenham/Phase accumulator approach will more evenly spready out the on/off cycles. Given any output power the Bresenham will always do at least as well as conventional PWM at spreading the on-time around. Bob Ammerman RAm Systems -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Controlling the heater by switching complete half-cycles does sound like the way to go for this situation. That hadn't occurred to me so that in itself is a big help. Bob, if I understand what you are saying, the Bresenham/Phase accumulator algorithm will somehow determine a sequence of half-cycles to skip that gives you a reasonable range of power control with a more evenly distributed half-cycles over some period. Could you provide just a little more detail in implementing this, or point me in the right direction. I have heard of Bresenham in reference to line drawing and don't see the immediate connection. thanks Bob Bullock At 10:21 AM 10/15/2002 -0400, you wrote: {Quote hidden}>and now Bob Ammerman rebuts: > >Assuming you are switching on half cycles, the Bresenham/Phase accumulator >approach will more evenly spready out the on/off cycles. Given any output >power the Bresenham will always do at least as well as conventional PWM at >spreading the on-time around. > >Bob Ammerman >RAm Systems > >-- >http://www.piclist.com hint: The list server can filter out subtopics >(like ads or off topics) for you. See http://www.piclist.com/#topics -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

On Mon, 14 Oct 2002, Olin Lathrop wrote: *>> I don't think that you need to interpolate at this level (256 steps on a *>> 2kW heater is about 10 W per step - you'd have to have pretty good *>> thermometers to see that amount of heat difference in a usual sized box or *>> room). At most the output will jump between two adjacent values from time *>> to time. *> *>Even better, why not screw the phase angle and switch in (or not) complete *>power line cycles? At 60Hz that gives you 13 different power choices and *>still drive the heater with at least 5Hz to reduce stress. That's the way the little project I posted to the piclist once works. But most people who use a closed loop have concerns about the response time of such a scheme. With some good reason, since even 10 power steps on an air heating system implemented like this could cause a loop to oscillate (with the integrating time of the heater proper you get time constants of the order of magnitude of two seconds or more). You need a HUGE I component in the regulator to get it steady and iron patience to tune it (it takes ages to respond to user input). However for large loads, like water heating, it is ok. Also synchronous motor fans (the mass of the rotor does the integration here). Peter -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

My point was that if you decide to do zero-cross switching, then the minimum 'quantum' of power is that which you get out of one half cycle. You use phase accumulation, or equivalently the Bresenham algorithm, to decide which complete half-cycles are turned on. This will provide a 'smoother' distribution of power than conventional PWM. It also allows for higher resolution of the average power. For example: assume we have decided that we want to be able to go down to 1/16th power. For a simple PWM, that means that we turn on the first cycle in 16 for 1/16th power, the first 2 cycles in 16 for 2/16th power, ..., the first 8 cycles in 16 for 8/16th power, ... The phase accumulator approach can generate smoother output, for example at 8/16th power the phase accumulator will turn on for every other half cycle (instead of 8 half-cycles on then 8 half cycles off). The phase accumulator can also result in more resolution. For example to generate 3/32 power the phase accumulator would generate 3 roughly equally spaced half cycles for each 32 half cycles. Bob Ammerman RAm Systems {Original Message removed}

>The phase accumulator approach can generate smoother output, for example at >8/16th power the phase accumulator will turn on for every other half cycle >(instead of 8 half-cycles on then 8 half cycles off). This may cause you problems, especially where iron cored inductances are involved. It is much nicer if it can be made to have a positive half cycle, and then a negative half cycle rather than a number of positive half cycles in a row without any negative half cycles to balance the core magnetic fields, which is what you seem to be suggesting here by "every other half cycle". As the unit is being suggested for a green house, i am envisaging the possibility of an isolating transformer in the picture. It may be that it is to be run on an RCD, but I would be wary of having something in a greenhouse being run directly off the mains without some sort of fault protection. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

While I found the need with my homegrown phase accumulator (thanks guys, now I know what to call it!) to balance the number of pos & neg 1/2 cycles, it, like any zero-cross firing scheme will not work well with inductive loads for several reason:. First, any imbalance on phases will create DC on the line, secondly, the inductance will create a current phase shift, so the zero-voltage switching will actually be self-destructive since there will be current flowing at the voltage zero points. transformer loads mean phase-angle control that you can adjust for the shift, otherwise watch the thyristors fizz.... > > > >The phase accumulator approach .... > This may cause you problems, especially where iron cored > inductances are > involved. > As the unit is being suggested for a green house, i am envisaging the > possibility of an isolating transformer in the picture. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

>While I found the need with my homegrown phase accumulator >(thanks guys, now I know what to call it!) to balance the >number of pos & neg 1/2 cycles, it, like any zero-cross firing >scheme will not work well with inductive loads for several >reason:. First, any imbalance on phases will create DC on the I did not mean to imply that the scheme will not work, just that the way that particular post had suggested may give you problems with any isolation transformer that may be there because repeated half cycles all on the same "side" (positive or negative) may introduce a DC magnetisation in the transformer. Just make sure what ever zero crossing scheme you use makes some attempt to use both positive and negative cycles in roughly equal amounts. Second, your heater should be a resistive load anyway. This should mean that any triac will have close to zero phase angle in its current waveform, and switching it should not be a problem. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

> Given any output > power the Bresenham will always do at least as well as conventional PWM at > spreading the on-time around. True, but Bresenham allows additional power settings that are not achievable with finite time slice PWM. It is these settings that contain the low frequencies. You can sortof think of the PWM as realizing a finite set of rational numbers with a fixed denominator. Bresenham can implement arbitrary rational numbers, including some that expand to repeating series with very long periods. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

> may give you problems with > any isolation > transformer that may be there There are two possibilities that come to mind: 1. If a transformer primary is controlled via the thyristor(s) you then have an inductive load, regardless of whether or not you have a resistive heater load on the secondary. This is often done for isolation and to be able to rectify the secondary as needed in different applications. My point was simply a suggestion that if you intend to isolate with a transformer, that this is not the easiest place to put it, since you will then have to deal with the current phase shift (fun!). 2. > because repeated half cycles > all on the same > "side" (positive or negative) may introduce a DC magnetisation in the > transformer. Just make sure what ever zero crossing scheme > you use makes > some attempt to use both positive and negative cycles in roughly equal > amounts. This ^ will apply to a source transformer only. At a few thousand watts, the local utility serice will handle the imbalance no problem. If you isolate the supply at this point, a small transformer can be made to smoke easily from the induced DC from even a small imbalance. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

> > because repeated half cycles > > all on the same > > "side" (positive or negative) may introduce a DC magnetisation in the Just to state the obvious, the beauty of bresenhams and similar systems is that the size of the "chunk" really doesn't matter. You could do a COMPLETE cycle, ie a + half followed by a - half, then do the bresenham based on the complete cycle being the "chunk" size. This would always maintain an average of zero, keeping the power company happy. Heating a greenhouse is going to give a massive thermal lag and hot/cold areas anyway, a PID system sounds like overkill that might not work any better than a $10 air conditioner thermostat available from any appliance repairer. -Roman -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.

----- Original Message ----- From: "Olin Lathrop" <spam_OUTolin_piclistTakeThisOuTEMBEDINC.COM> To: <.....PICLISTKILLspam@spam@MITVMA.MIT.EDU> Sent: Wednesday, October 16, 2002 8:13 AM Subject: Re: [EE]: How to calculate linear power intervals for phase angle > > Given any output > > power the Bresenham will always do at least as well as conventional PWM at > > spreading the on-time around. > > True, but Bresenham allows additional power settings that are not achievable > with finite time slice PWM. It is these settings that contain the low > frequencies. Exactly what I said. > You can sortof think of the PWM as realizing a finite set of rational > numbers with a fixed denominator. Bresenham can implement arbitrary > rational numbers, including some that expand to repeating series with very > long periods. An excellent way to put it! Bob Ammerman RAm Systems -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.