HVAC design questions

[GoingFastTurningLeft]

I put off my HVAC thermostat project since I barely ran the barely working A/C this summer. I've had to run the heat a few times, time to get this thing going. I have some questions I haven't found an answer to or want opinions on:

*I am replacing one of those generic round honeywell heat/cool auto/fan bi-metal thermostats*

#1. What should I use as a minimum on cycle time? Maximum on cycle time? I think you're supposed to run it for a few minutes at a time at the minimum to avoid stressing it, and probably should give it some rest before restarting it.

#2. To rectify 24VAC down to 5VDC, would the route to go be a transformer to drop it down to about 7VAC then bridge rectify it to DC and use a 5VDC regulator?

#3. HVAC control signals are reasonably low amperage (under an amp), correct?

#4 wasn't there a full HVAC controller somebody made in N&V within the last 2 years? I looked briefly through what I have around but didn't see it.

[dtief]

#1 Running an A/C less than a few minutes would mostly waste electricity. A delay or lockout from restarting too soon is a very good idea. The compressor can't start with backpressure. It takes a few minutes to equalize the pressure in the system. Maximum on isn't really a problem - if it is sized correctly, it should run most of the time on the hottest day.


#2 Depends on the current needed. If a linear regulator is going to dissipate more than a watt or 2 consider a switcher, or transformer down so there is less drop across the regulator. For low power, low loss I have used dual winding transformers in this configuration:

wire the two 120V primaries in series. Wire the secondaries in series (or paralel for lower voltages) Apply 120V to the two primaries. The output will be half the voltage, so pick the secondary voltage with this in mind. The magnetizing current will be FAR less than if you run the transformer the conventional way with the two 120V primaries in paralel. The output will be more load sensitive (higher output impedance). This trick works for low power(compared to the transformer's rating) applications.
The Tamura 3FD-nnn style from Digikey are excellent for this.


#3 Most should be under an amp.

#4 Don't know.

National Semiconductor has some temperature sensor ICs that output degrees F or C directly in millivolts.

[GoingFastTurningLeft]

Thanks dtief for shedding some light.

#1. I agree. The A/C in my apt pretty much just wastes electricity to begin with... I could never get my place colder than 76 or so degrees, and that took a long time. I utilized fans and opened windows at night, somehow I went from $100 for July to $60 for August. Somehow September ended up being $80

However, being that cold weather is fast approaching, I am more concerned with cycling the heater. I have gas heat and the thing heats up my place no problem.

#2. This is an HVAC system... the control signals are 24VAC stepped down from 120VAC... I am not dealing with 120VAC. Though if this thing I'm building is hooked into a computer and eventually will be USB, I might as well just run it 5VDC off the USB hub. I don't even know if the original installers even ran a 24VAC wire to the thermostat.

#3. Cool, I have Tyco/Potter & Brumsfield T77S1D3-05 relays... 5 volt relays that handle 3A max. Should be fine.

#4. Might have been in Circuit Cellar?

-I am using an LM35DT 10mV/C sensor... easy to hook up, but watching on my o-scope it's very noisy, (and had a weird frequency spike), adding a 3.3uF cap from GND to output helped a lot. My AVR is setup to read voltages up to 1.1V and to take 10 readings then use the average. I've tried using these sensors with no filtration or averaging... the display was constantly flickering +/- about .3 degrees!

New Questions:

#5: What is a good +/- envelope value for the temperature set point?

[dtief]

#2 The thermostat systems I've seen have only a closure at the thermostat, no real way to get full time power from the wires available at the thermostat.

Using a PC to power it would use a noticable amount of power - unless the PC is already on 24/7 for other reasons.

For the temp sensor, be sure to bypass the power supply pins with a .1uF ceramic or stacked film. I never had trouble with them oscillating. Avoid a long cable run ( sheilded cable = capacitive load) Only load the output lightly. (check data sheet for max output current)

I think typical thermostats are around a few degrees. I would make it adjustable & see how it goes. 1 degree would probably be about the limit to avoid things cycling too much.

[Bob Scott]

-I am using an LM35DT 10mV/C sensor... easy to hook up, but watching on my o-scope it's very noisy, (and had a weird frequency spike), adding a 3.3uF cap from GND to output helped a lot. My AVR is setup to read voltages up to 1.1V and to take 10 readings then use the average. I've tried using these sensors with no filtration or averaging... the display was constantly flickering +/- about .3 degrees!

I've used a lot of those LM35's. Make sure the ground is good at that sensor. I mean, run an independent ground just to the sensor if there is any other current presently on that ground line to the IC. Just a few millivolts on the ground line caused by current to some other device can affect the reading. You shouldn't need such a large capacitor on the output, but if it works, hey, why not?

As for adding hysteresis in your temperature sensitivity, my furnace thermostat uses a couple of degrees F difference between turn-on and turn-off, but it also has a "heat anticipator" resistor built-in to speed up warming of the thermostat.

[CeaSaR]

#4 wasn't there a full HVAC controller somebody made in N&V within the last 2 years? I looked briefly through what I have around but didn't see it.

Do you receive the E-version of N&V? If you do, open one up and look at
the address. At one point you will see (for Nov. 2008) this:
/nutsvolts/200811/
The 200811 denotes the year and month. Just change them to go back to
previous editions and search the indexes / Table Of Contents. As for the
controller, I remember an Air Conditioner controller project. Maybe this is
what you recall.

CeaSaR

[Bigglez]

#4 wasn't there a full HVAC controller somebody made in N&V within the last 2 years? I looked briefly through what I have around but didn't see it.

(1) Beat the Heat with this External AC Controller [Relevance: 2.39]
March 2007 - Al Jaszek

(2) Make The Smart Room Air Conditioner Controller [Relevance: 1.83]
October 2008 - Tom Fox

(3) Household Thermostat Heating/Cooling Control [Relevance: 1.47]
April 2008 - Chuck Irwin

[bodgy]

#4 wasn't there a full HVAC controller somebody made in N&V within the last 2 years? I looked briefly through what I have around but didn't see it.

(1) Beat the Heat with this External AC Controller [Relevance: 2.39]
March 2007 - Al Jaszek

(2) Make The Smart Room Air Conditioner Controller [Relevance: 1.83]
October 2008 - Tom Fox

(3) Household Thermostat Heating/Cooling Control [Relevance: 1.47]
April 2008 - Chuck Irwin

What? I didn't get a mention?

My view is that almost any controller/ heat sensor project would work, your biggest challenge is actually hooking it up to the air conditioner and deciding where the sensor was to go.

As a PS, I've been surprised at the amount of people, who have wanted to use my original project in a different way, I'd never thought about mushroom farming control or egg incubation as a use.

[Bigglez]

I am using an LM35DT 10mV/C sensor... easy to hook up, but watching on my o-scope it's very noisy, (and had a weird frequency spike), adding a 3.3uF cap from GND to output helped a lot. My AVR is setup to read voltages up to 1.1V and to take 10 readings then use the average. I've tried using these sensors with no filtration or averaging... the display was constantly flickering +/- about .3 degrees!

How far is the LM35 form the AVR?

As this part is analog and low level it requires careful
power supply decoupling and handling of the output
signal. With analog there's only one chance to do it
right - once a signal is degraded there's little room to
get back the original.

You might consider pairing your LM35 with a LDO
linear regulator to get a clean regulated supply at
the device (a small PCB or perf board perhaps).
While you're at it, you could add an Op Amp to
drive the cable between the two. Single rail Op Amps
are available in small packages (SOT or SOIC/DIP).

Another approach is to use a combination temperature
sensor and A2D, that sends the digital data into your
AVR. You can do this on only two wires (so called
one-wire ICs) or three wires such as I2C bus,
depending upon the sensor type. Here's the DS18S20 datasheet.

Check back issues of N and V for Michael Simpson's
excellent weather station articles, they cover this part
in great detail.

[Bigglez]

What? I didn't get a mention?

Who are you?
I searched the on-line archives for "Thermostat"
I just searched again for "bodgy" and got
Your search did not produce any results. Please try again.

[bodgy]

Who are you? "I am a name not a number" well I did forget to sign this post, but normally you'll see Colin at the bottom.

I searched the on-line archives for "Thermostat" Well, it was actually Climate Controller and it was waaay back around the beginning of the century

I just searched again for "bodgy" and got -
Your search did not produce any results. Please try again.that seems to be my lot in life

1. I updated the code, I think I posted that on the forum earlier this year.

2. The "I'm a name not a number", is paraphrased from the 70's TV series of The Prisoner, starring Patrick McGoohan and filmed in Portmerrion in Wales. The actual phrase is 'I am not a number! I am a free man!

Colin

[Robert Reed]

Coincidence - I have just completed working out a design for a room thermostat- heating system only. I was curious why you chose the LM35 (Celsious) instead of the LM34 (Farenheit). If you arent concerned with a calibrated readout on your control then it may just be a moot point. I also noticed in your previous post that you bypassed the LM35 out put lead to ground (3.3MF) and I am surprised that did not cause more problems in regards to instability, as these devices are vulnerable to capacitive loading. The spec sheets guarantee stable operation only up to 50 PF loads. A better way is to isolate cable capacitance with a simple RC filter. Say 10K of resistance right off the out put lead and a capacitor to ground of your choosing for your situation. Of course with that resistance you need to be feeding into a hi-Z input (>100K) or if it is lower reduce the value of the isolation resistor. And as others have mentioned use a dedicated ground for long runs. I am at the point where I need to know the best placement for the sensor in relation to the heater, so I will probably be posting in the near future with a room layout asking for suggestions. Any HVAC engineers out there?

[Bigglez]

Portmerrion in Wales.

I haven't been to "The Village" for quite a few years...
Is it still there? We've probably lost this audience, so
you can explain what ROVER is/was...

I thought Danger Man was a better show, or was
I thinking of Callan, oh well. Cursed with a long
memory of late '60's TV dramas from my childhood.
Orange Alert!

[Bigglez]

I am at the point where I need to know the best placement for the sensor in relation to the heater, so I will probably be posting in the near future with a room layout asking for suggestions.

You didn't say what the heat source was. If its forced air
the temp sensor is best placed near the return air register.
(Above a floor return, below a ceiling return).

[GoingFastTurningLeft]

Coincidence - I have just completed working out a design for a room thermostat- heating system only. I was curious why you chose the LM35 (Celsious) instead of the LM34 (Farenheit). If you arent concerned with a calibrated readout on your control then it may just be a moot point. I also noticed in your previous post that you bypassed the LM35 out put lead to ground (3.3MF) and I am surprised that did not cause more problems in regards to instability, as these devices are vulnerable to capacitive loading. The spec sheets guarantee stable operation only up to 50 PF loads. A better way is to isolate cable capacitance with a simple RC filter. Say 10K of resistance right off the out put lead and a capacitor to ground of your choosing for your situation. Of course with that resistance you need to be feeding into a hi-Z input (>100K) or if it is lower reduce the value of the isolation resistor. And as others have mentioned use a dedicated ground for long runs. I am at the point where I need to know the best placement for the sensor in relation to the heater, so I will probably be posting in the near future with a room layout asking for suggestions. Any HVAC engineers out there?

The reason I am using an LM35 is that's what I have on hand. I think I originally got it because my original temperature projects were temperature displays for my kegerator, and it was just conveinent to have 0V = 0C... Also i think it was rated for a higher precision than the LM34... Though at this point i think the noise takes away most of the precision.

But for all intents and purposes it doesn't really make a difference, since I have to do math on the ADC input to convert the digital value into how many degrees it is - its just a matter of combinging the C->F and ADC->Temperature equations.

The main sensor is going to be located on the controller, so the wires will be a few inches at the most.

I am running it to ADC0 on an AVR... pretty sure its a high impedance input when the ADC and port registers are set correctly.