How do I control this heater?

[nlis12]

So I have a coiled (Nichrome) heating element that's rated at 120VAC 60Hz. (Resistance of 14.4Ohms)
I am trying to use this heating element to warm up air blowing across it. I will have thermistor on the outlet end to monitor the temperature of the output air. This heater is coiled inside of a tube, so the air is funneled into the heating elements for maximum heat transfer.

My question is:
What is the best way to control this heater?
PID control with Pulse width modulation (PWM) seems to be an option, but turning on and off such huge currents seems like its not a good idea and could affect other electronics around through EMI.
Are there other options?

Thanks!

 

[CWatters]

How fast response time is needed? Could you use a slow PWM (say 1Hz or slower) with the switching synchronised to the 120V zero crossings? Not sure if there is off the shelf kit that will do this

 

[anorlunda]

I think you should be able to find a solid state relay capable of rapidly switching that much current.

 

[dlgoff]

Maybe a On Semiconductor UAA2016 zero voltage switch power controller. Very simple circuit.
From the attached datasheet:

The UAA2016 is designed to drive triacs with the Zero Voltage
technique which allows RFI−free power regulation of resistive loads.
Operating directly on the AC power line, its main application is the
precision regulation of electrical heating systems such as panel heaters
or irons.

 

[Asymptotic]

Nichrome heater wire doesn't have much thermal inertia. 'Wind speed' is a major factor in control quality - high air flow rates remove heat from the element faster than at low flow rates, and (if a form of digital on-off control is used) requires a short on-time period. I've seen successful implementations based around off-the-shelf solid-state, zero crossing power relays and PID controllers with voltage pulse output set to the minimum period (0.5 second = 100% on, 0.25 sec on time=50%, etc). In essence, a PID controller with voltage pulse output is PWM.

Generally, EMI becomes a concern when using instantaneous, non-zero crossing power relays (or discrete SCR pairs, or TRIACs) in a phase controller. Using the venerable Crydom series 1 panel mount as an example, D1210 (DC control input, 120V at 10 amps on the power side) is zero crossing while model D1210-10 changes output instantaneously.

If the 0.5 second period is too long and results in unacceptable temperature swings then the standard 'go-to' is a phase power controller with analog (4-20 milliamp, 0-10 VDC, etc) input. When commanded to 100% output a pair of back-to-back SCRs are triggered for the entire AC cycle, and turned on progressively later in the AC cycle as the controller is commanded toward 0%. This provides excellent power control, but generates a lot of EMI hash.

Payne Engineering offers heater power controllers using a hybrid approach (I don't know if anyone else offers these, but they've been around for awhile, and the patents may have run out). If for nothing else, visit their technical publications page, and download Henry Payne's paper Silicon Revolution for his insights on power silicon..

 

[nlis12]

How fast response time is needed? Could you use a slow PWM (say 1Hz or slower) with the switching synchronised to the 120V zero crossings? Not sure if there is off the shelf kit that will do this

Well, I need to reach ~50C outlet temperature from ambient room temp in about 2 min. If that gives you an idea of the response time.

 

[nlis12]

I think you should be able to find a solid state relay capable of rapidly switching that much current.

Hmmmm, sounds simple enough.
THanks!

 

[anorlunda]

Hmmmm, sounds simple enough.
THanks!

You're welcome. But @Asymptotic 's suggestion was better than mine.

Payne Engineering offers heater power controllers using a hybrid approach (I don't know if anyone else offers these, but they've been around for awhile, and the patents may have run out). If for nothing else, visit their technical publications page, and download Henry Payne's paper Silicon Revolution for his insights on power silicon..

 

[nlis12]

Nichrome heater wire doesn't have much thermal inertia. 'Wind speed' is a major factor in control quality - high air flow rates remove heat from the element faster than at low flow rates, and (if a form of digital on-off control is used) requires a short on-time period. I've seen successful implementations based around off-the-shelf solid-state, zero crossing power relays and PID controllers with voltage pulse output set to the minimum period (0.5 second = 100% on, 0.25 sec on time=50%, etc). In essence, a PID controller with voltage pulse output is PWM.

Generally, EMI becomes a concern when using instantaneous, non-zero crossing power relays (or discrete SCR pairs, or TRIACs) in a phase controller. Using the venerable Crydom series 1 panel mount as an example, D1210 (DC control input, 120V at 10 amps on the power side) is zero crossing while model D1210-10 changes output instantaneously.

If the 0.5 second period is too long and results in unacceptable temperature swings then the standard 'go-to' is a phase power controller with analog (4-20 milliamp, 0-10 VDC, etc) input. When commanded to 100% output a pair of back-to-back SCRs are triggered for the entire AC cycle, and turned on progressively later in the AC cycle as the controller is commanded toward 0%. This provides excellent power control, but generates a lot of EMI hash.

Payne Engineering offers heater power controllers using a hybrid approach (I don't know if anyone else offers these, but they've been around for awhile, and the patents may have run out). If for nothing else, visit their technical publications page, and download Henry Payne's paper Silicon Revolution for his insights on power silicon..

Is there a reason why you suggested a period of 0.5sec?
If I needed more fine control, could I use a period of something like ~1millisec?

Thanks for the explanation! Very helpful in choosing components.

 

[CWatters]

If you did PWM at 1Hz with zero crossing detection that would give you 120 half cycles per second that could be on or off (if the mains is 60HZ). That might allow you to set the long term temperature to within about 1%. However it sets a limit on how fast the temperature can change. With a simple fixed period scheme there could be a delay of up to one second before the heater power could be changed. Fast enough?

 

[CWatters]

If you want mS switching rates then you can't wait for the mains zero crossing to switch on/off because the mains frequency is too slow.

 

[NTL2009]

What are you using to measure the temperature? One msec updates don't sound reasonable, thermistors will have a far slower response time. If you adjust faster than you can measure, you will be under - damped and will overshoot.

 

[NTL2009]

If you did PWM at 1Hz with zero crossing detection that would give you 120 half cycles per second that could be on or off (if the mains is 60HZ). That might allow you to set the long term temperature to within about 1%. ...

Yes, you have about 1% resolution that way, but over time the system can duty cycle between say 50/120ths and 51/120ths and average to 50.x/120ths of max power.

 

[anorlunda]

A completely different approach is a controllable variable voltage supply; either AC or DC. But as @NTL2009 says, the temperature responses are much slower than 1ms. Are your sure that your ambitions are really required?

 

[russ_watters]

This sounds like a standard SCR duct heater control scheme.

 

[Asymptotic]

Is there a reason why you suggested a period of 0.5sec?
If I needed more fine control, could I use a period of something like ~1millisec?

Thanks for the explanation! Very helpful in choosing components.

0.5 second period came about because I'd been looking at an inexpensive PID controller (Solo 48xx series) offered with a voltage pulse output, and the cycle period on that particular model could be set from 0.5 seconds to 99 seconds. It depends on the vendor/manufacturer. If the developer chose to limit this parameter to integer values, a minimum period of 1 second may be the lower limit, while others offer a minimum cycle period of 0.1 second.

Upon further reflection, 0.1 second was the controller minimum period for the hot air application I'd been thinking about, built around an 18 kW Osram Sylvania Sureheat using type K thermocouples, an Omega Engineering PID controller, and a zero-crossing SSR. Worst case temperature swing was on the order of 2 or 3°F, and couldn't be tuned more closely in part due to this next thing ...

A 1 ms cycle period won't buy you anything for the reasons @CWatters and @NTL2009 have outlined.

A half-cycle of 60 Hz AC (from one zero cross to the next) is 8.33 milliseconds wide, so a zero crossing power controller won't do for precise control. So long as the PID controller can deliver pulses with adequate resolution and repeatability using a non-zero crossing SSR is an option, but then there's the EMI to contend with. In the above case, 8.3 ms is equivalent to an 8.3% output change (0.1 second at 100% output, 50 ms = 50% output, etc.), and prevented the PID controller from coming up with a more stable output solution. Our application was fine with a 2 to 3°F variation, but others may not be.

Sensor response time is a big deal in an application like yours. This gloss from Omega Engineering on thermocouple response times has the gist of it. Generally speaking, RTDs (Resistance Temperature Detectors) have the best accuracy and linearity, but also the slowest response, and are middlin' fragile. Thermocouples are physically robust and withstand higher temperatures than RTDs or thermistors, but have the lowest output of the three, are middlin' non-linear, and special attention needs to be paid to wiring and connectors. Thermistors have the highest output per degree of temperature change (but are even more non-linear than thermocouples), are more fragile than TCs or RTDs, and have the most limited operational temperature range. Response times between TCs and thermistors are roughly comparable.

I will have thermistor on the outlet end to monitor the temperature of the output air.

My experience with hot air control was limited, and I can't recall how the one I'm referencing was configured, but in process ovens and other similar applications it wasn't uncommon for a thermocouple to be mounted on the heater element. This doesn't control the temperature of the air itself, but has the fastest response time, and since air temperature and element temperature usually track closely with one another is often the better choice in terms of control stability.

 

[NTL2009]

... A half-cycle of 60 Hz AC (from one zero cross to the next) is 8.33 milliseconds wide, so a zero crossing power controller won't do for precise control. ...

Thinking about this a bit more, I don't think that statement is true. Precise control should achievable if the time constants of the system are relatively longer than the 8.33 mSec half-wave. I'll use numbers/examples to illustrate:

Say we want to heat to 140F +/- 0.1F in a room temperature environment. Let's say that when we are near steady state, the air flow will cool by 0.1F in ~ 3 seconds when heat is removed. And let's say we put in just enough watts to the heater so that a 0.1F temperature rise takes a bit longer, say 5 seconds. And to keep things simple, let's assume our thermistor responds fast enough to not be a significant drag on this (but it probably is a consideration in single digit second times, so check this out for the real system). We can also assume the heat rise/fall is linear over these short segments.

On the falling side, each pulse time represents an ~ 8.33mSec/3 Sec *0.1F = only 0.000278 F change. So the system can turn the pulses back on very close to the 0.1F limit.

The high side would be even more easily controlled with a 5 second ramp. Even if the above times were just 1 second each, 120 half-pulses would move the temperature 0.1F, so each time frame represents 1/120th of that. So the half-wave zero-crossing on/off control should allow extreme (relative*) precision, assuming real-world time constraints (we're not talking the tiny heating elements found in ink jet printers for example, but those are characterized and run open loop).

* Absolute precision will be a function of the thermistor precision/drift and any calibration, but relative precision will be extremely good over a small temperature delta. Noise and other system errors are likely to be a limit before the 8.33 mSec pulse width.Bottom line: As with any design project, the OP needs to understand and define the requirements, and characterize the system - heat rise and fall times, sensor response times, and controller response times. I suspect that once we see the actual requirements, none of these will be very challenging or critical, and rough assumptions will suffice.