| star882 |
I'm designing a power supply for a refrigeration compressor (about 1/3-1/2 ton) used to cool computer equipment. Since the load can vary a lot, the compressor speed must be varied in order to regulate the temperature. The traditional way to do it is with a VFD (Variable Frequency Drive), which is very similar to a class D amplifier. However, they cost a lot.
Therefore, I would like to build my own VFD. I have already thought about the design.
First, the incoming 120v AC is rectified using a voltage doubler to 340v DC.
The 340v DC powers a buck converter that outputs 0-175v based on inputs from the control circuit. The 0-175v output is modulated into a unipolar sine wave of the proper voltage for a given frequency.
Since the sine wave from the buck converter is unipolar, a H bridge inverts the wave every zero crossing to make a bipolar sine wave, which powers the compressor.
A small flyback converter, also running from the 340v DC, generates 5v and 3.3v for the control logic and 12v for the MOSFET drivers. Separate windings supply 12v for each high side of the H bridge. A pulse transformer drives the high side of the buck converter.
The controller of the buck converter uses multiplying DACs as part of the voltage reference. The H bridge driver and one DAC is connected to a Flash ROM chip, which stores the voltage and polarity values needed to make a sine wave.
The address lines of the Flash chip are sequenced by a counter, stepping through the values. The counter is clocked by a PLL.
A microprocessor performs the task of reading the inputs (temperatures, maybe pressures, current, etc.) and adjusting the PLL and the second multiplying DAC that determines the output voltage. The PLL and multiplying DAC adjust only on zero crossings to ensure a smooth transition. The microprocessor is programmed to adjust the frequency and voltage in steps.
Some more logic provides fault protection. If a slight fault is sensed (enough to indicate a problem but not enough to stop the unit), it turns the status lights blinking yellow. If a more serious fault is sensed, it powers down for 5 minutes (while turning the status lights red) then tries again. If it attempts to restart 3 times and fails, it locks out, blinking the status lights red and sounding a buzzer. If it successfully restarts in 3 tries or less, the status lights alternate between red and yellow. In normal operation, the status lights are green. In standby, the lights are yellow. A separate row of lights indicate the code for the last fault encountered. |
|
|
| wine&dine |
| Industrially built VFDs would normally employ a H-bridge driven directly by a PWM waveform generated by a microcontroller. The sine waveform you generate doesn't have to be that perfect, post-filtering will take care of most issues (and your load will spoil the rest ;-)) |
|
|
| star882 |
But by using a H bridge, all 4 MOSFETs would have to run at the carrier frequency, which makes it a little harder to design. The isolation circuits have to be built to function at high frequencies even with 340v peak to peak of high frequency AC between the sides and pulse transformers cannot be (easily) used since they will saturate at high duty cycles. All 4 MOSFETs would also have to handle high frequencies.
With a H bridge after a buck converter, the main 4 MOSFETs only need to handle low frequencies while the MOSFET in the buck converter is the only large MOSFET that runs at high frequencies.
Or maybe I can switch the H bridge in a "magic sine wave" fashion, requiring only a simple filter for the output and use the buck converter only for voltage regulation. (I'm not sure if common magic sine wave algorithms can regulate the output voltage without requiring a more complicated output filter.) |
|
|
| ak_47_boy |
Hey star 882, i have seen you in the phase change section on XS :D
I have actually been experimenting with an 80watt amplifier playing sine waves, square waves ect. into a small AC motor.
I have found that the motor will run with any shape wave, it just vibrates with some.
A HUGE problem is that the independence of the windings inside the motor changes with frequency. At 400hz my motor needed double the voltage to draw its rated 20watts. When i turned it down to around 30hz the amp overloaded and kicked out. I was using a 10:1 transformer to step up the voltage from the amp.
So i don't really know what to do, i will think about it. |
|
|
| star882 |
| quote: | Originally posted by ak_47_boy
Hey star 882, i have seen you in the phase change section on XS :D
I have actually been experimenting with an 80watt amplifier playing sine waves, square waves ect. into a small AC motor.
I have found that the motor will run with any shape wave, it just vibrates with some.
A HUGE problem is that the independence of the windings inside the motor changes with frequency. At 400hz my motor needed double the voltage to draw its rated 20watts. When i turned it down to around 30hz the amp overloaded and kicked out. I was using a 10:1 transformer to step up the voltage from the amp.
So i don't really know what to do, i will think about it. |
That's the reason why the voltage is reduced at reduced frequency. Otherwise, the core saturates and the motor overheats.
Try using a 5:1 transformer when running at 30Hz.
Have you tried measuring the efficiency with different waveforms? I think square waves are very inefficient (which is why a 0.8A ECM fan is more powerful than a regular 2A fan of the save voltage).
As for frequency range, I think down to 30Hz is as low as it can ever go (maybe not even that low). Not sure about highest frequency but I think it can go to 70Hz without issues. (That's particularly true if a R22 compressor is used with R134a.) Not sure what range common TXVs can regulate over but I don't think it would be by much. |
|
|
| ak_47_boy |
It does work, it seems like is has more torque also. I think thats just voltage issues because i cant find a smaller tranny.
I would get a cheap car amp off eBay and run a sine wave through through it, use the volume to adjust the voltage. Since distortion does not matter really at all when running a motor it should be alright.
Use a microwave oven transformer with a rewound secondary to give 12v to power the amp. You could take all the pieces and put it in a nice box. They sell signal generator kits some ware also. |
|
|
| ak_47_boy |
How many watts do you need to need? I would love to see a controller designed like that. If you only need a hundred watts you could use a tubes :D
Microwave oven transformers make me so happy. Could use one to supply 2000 volts to the tubes and another brings it back to 120V. Pulse triode maby?
Haha the mods at XS would **** their pants :D |
|
|
| star882 |
| quote: | Originally posted by ak_47_boy
How many watts do you need to need? I would love to see a controller designed like that. If you only need a hundred watts you could use a tubes :D
Microwave oven transformers make me so happy. Could use one to supply 2000 volts to the tubes and another brings it back to 120V. Pulse triode maby?
Haha the mods at XS would **** their pants :D |
A typical 5000BTU compressor uses maybe 500w, although startup current is much higher. And MOSFETs are surprisingly cheap. The MOSFETs should take the inrush current fine if sized right, as it's a very short time.
BTW, I have seen a prototype of a cooling unit for 2U and 4U rackmount servers. There was a small inverter drive compressor (codenamed "Little Kim") about the size of 4 to 5 stacked tuna cans lying on the side. The compressor is a rotary that runs at 90-160Hz to reduce size. The refrigerant used is R404a. The inverter circuit only has one 3 phase IGBT module and a few small inductors with a few tiny SMD chips for control. |
|
|
| ak_47_boy |
There might be ideas inside a cheap power inverter from walmart. I have one il take a look inside.
edit: Complicated insides. |
|
|
| star882 |
| quote: | Originally posted by ak_47_boy
There might be ideas inside a cheap power inverter from walmart. I have one il take a look inside.
edit: Complicated insides. |
Power inverters usually just have a DC/DC converter to convert 12v to 170v and a H bridge to turn DC to AC. Most UPSes are similar.| quote: | Originally posted by d2134
Hi star882,
Consider using a dedicated microcontroller like MC3PHAC in cojunction with some MOSFET driver IC. Take a moment and read this datasheet http://www.freescale.com/files/micr...eet/MC3PHAC.pdf
It may be the solution to your problem. |
Looks interesting. How well does it work for a refrigeration compressor?
Oh, and what chipset is used in a variable capacity Carrier Infinity A/C unit? |
|
|
| d2134 |
I know nothing about your compressor. If it is driven by a 3 phase induction motor it can work . If is driven by a 1 phase motor with starting device it probabily not work. At 5000 BTU can be eighter.
By the way, is it not more easy to use an on/off methode to regulate the cooling power? like in household refrigerators. |
|
|
| star882 |
| quote: | Originally posted by d2134
I know nothing about your compressor. If it is driven by a 3 phase induction motor it can work . If is driven by a 1 phase motor with starting device it probabily not work. At 5000 BTU can be eighter.
By the way, is it not more easy to use an on/off methode to regulate the cooling power? like in household refrigerators. |
Yes, it's easier to just turn it on and off, but then a pretty large buffer tank would be needed to avoid excessive starting and stopping. Of course, I suppose it could be designed so the water/glycol solution can flow though copper tubing placed in a tank of plain water, so ice formed in the tank tends to regulate the temperature and also perform better since the heat of fusion of water is being used instead of just the specific heat. |
|
|
| d2134 |
| For 5000 BTU/h and 6 min between starts and 10 F temperature sweep you need about 15 lb of water/glycol mix. How much liquid contains your instalation now? |
|
|
| star882 |
| quote: | Originally posted by d2134
For 5000 BTU/h and 6 min between starts and 10 F temperature sweep you need about 15 lb of water/glycol mix. How much liquid contains your instalation now? |
Currently, I have a very big 48000BTU unit cooling the air in the entire building. However, it's not all that efficient overall (air is not a good thermal transfer medium) and since it's very big, it would be very impractical to move to another building. Therefore, the equipment is considered part of the building, so I'll have to obtain more equipment when I move, as I most likely will later this year. I would like to make a small, easily transportable unit that is used only to cool the CPUs. I might have to put the unit in the bedroom so I would not want it to start and stop too often (noise doesn't bother me as long as it's not too loud, not too annoying, and relatively constant).
Ideally, the unit should be just a little larger than an average briefcase and be relatively quiet when in operation. I'll probably use a buck converter running at 21kHz to avoid noise issues. I can install special mufflers and vibration absorbers on the compressor, then engineer some way of wrapping the compressor in soundproofing while keeping it cool. I suppose I could use a thermosiphon using R152a from a duster can. The condenser fan will probably be a "squirrel cage" type fan driven by a variable speed brushless DC (ECM) motor to minimize noise. |
|
|
| TechGuy |
>A microprocessor performs the task of reading the inputs (temperatures, maybe pressures, current, etc.) and adjusting the PLL and the second multiplying DAC that determines the output voltage. The PLL and multiplying DAC adjust only on zero crossings to ensure a smooth transition. The microprocessor is programmed to adjust the frequency and voltage in steps.
Some more logic provides fault protection. If a slight fault is sensed (enough to indicate a problem but not enough to stop the unit), it turns the status lights blinking yellow. If a more serious fault is sensed, it powers down for 5 minutes (while turning the status lights red) then tries again. If it attempts to restart 3 times and fails, it locks out, blinking the status lights red and sounding a buzzer. If it successfully restarts in 3 tries or less, the status lights alternate between red and yellow. In normal operation, the status lights are green. In standby, the lights are yellow. A separate row of lights indicate the code for the last fault encountered.
If I am correct, the load on the compressor is probably nearly static, except for start up. You may be able to cheat by avoiding using a DAC/feedback loop and set up a discrete set of drive speeds and voltages. This would allow you to preprogram a set of PWM (sine) tables into a microcontoller equipped with an PWM output for a set of frequencies and drive voltages and avoid using a PLL (perhaps reduce your development time). In a normal UPS or inverter, the load is variable and there needs to be tight feedback loop. The advantage with a static load, you can use a set of preprogramed PWM tables would likely reduce the complexity and part count. You can either use the same microcontroller to manage the drive speed using an adc input connected to a temperature sensor, or configure a set digital inputs to select the drive speed.
>First, the incoming 120v AC is rectified using a voltage doubler to 340v DC. The 340v DC powers a buck converter that outputs 0-175v based on inputs from the control circuit.
Also why are you stepping up to 120 AC to 340VDC and using a stepdown buck back to 175VDC? Why not just rectify to 175VDC and use a commerical PFC controller to regulate voltage? I believe that your probably going to have a pretty high ripple current running across those caps if the output is 500Watts. I believe a PFC controller (Power Factor Correction) would the better way to go and afford you better efficiency (less heat) and it should reduce your part count to boot.
>The H bridge driver and one DAC is connected to a Flash ROM chip, which stores the voltage and polarity values needed to make a sine wave.
The address lines of the Flash chip are sequenced by a counter, stepping through the values. The counter is clocked by a PLL.
Your probably better off using a digital sine wave via PWM instead of driving the H-bridge using an analog output. If you operate the output transistors in a non-saturated state, they are going to disspate a huge amount of heat. A better method is to set up a table of discrete set of pulse duration that varies the duty cycle to recreate a sine wave using PWM (ie duty: 0%, 2%, 8%, 16%... 100%... 16%., 8%, 2%, 0% = half-cycle sinewave). You also need to drive this output into a BTL( Bridge Tied Load) H-Bridge configuration or a second H-BRIDGE to Flip the voltage because AC output is 340V Peak to Peak.
FWIW: I would recommend that when you test, that you start off with a cheap test load (perhaps a set of 5 (or less) 100watt bulbs). While these are resistive loads as appose to inductive loads of a AC motor, it would allow to you tweak your PWM\Drive output before testing on a relativily expensive air conditioner.
Over all I think it would be more efficient to use a storage medium to serve as a coolant reservor. There is a chance that you might decrease the lifespan of the compressor motor by operating it at lower speeds. I suspect that compressors are probably equiped with vibration dampeners that are optimimized for a fixed frequency.
Hope this was useful. |
|
|
| Eva |
| You can just rectify 120V AC and use a plain low frequency full bridge, controlling both duty cycle and frequency, to drive the motor with a square wave and get rid of most of the complexity. The drawback is that the motor will run a bit hotter than with a pure sine wave. A three-phase motor would be great btw. |
|
|
| star882 |
| quote: | Originally posted by TechGuy
>A microprocessor performs the task of reading the inputs (temperatures, maybe pressures, current, etc.) and adjusting the PLL and the second multiplying DAC that determines the output voltage. The PLL and multiplying DAC adjust only on zero crossings to ensure a smooth transition. The microprocessor is programmed to adjust the frequency and voltage in steps.
Some more logic provides fault protection. If a slight fault is sensed (enough to indicate a problem but not enough to stop the unit), it turns the status lights blinking yellow. If a more serious fault is sensed, it powers down for 5 minutes (while turning the status lights red) then tries again. If it attempts to restart 3 times and fails, it locks out, blinking the status lights red and sounding a buzzer. If it successfully restarts in 3 tries or less, the status lights alternate between red and yellow. In normal operation, the status lights are green. In standby, the lights are yellow. A separate row of lights indicate the code for the last fault encountered.
If I am correct, the load on the compressor is probably nearly static, except for start up. You may be able to cheat by avoiding using a DAC/feedback loop and set up a discrete set of drive speeds and voltages. This would allow you to preprogram a set of PWM (sine) tables into a microcontoller equipped with an PWM output for a set of frequencies and drive voltages and avoid using a PLL (perhaps reduce your development time). In a normal UPS or inverter, the load is variable and there needs to be tight feedback loop. The advantage with a static load, you can use a set of preprogramed PWM tables would likely reduce the complexity and part count. You can either use the same microcontroller to manage the drive speed using an adc input connected to a temperature sensor, or configure a set digital inputs to select the drive speed.
>First, the incoming 120v AC is rectified using a voltage doubler to 340v DC. The 340v DC powers a buck converter that outputs 0-175v based on inputs from the control circuit.
Also why are you stepping up to 120 AC to 340VDC and using a stepdown buck back to 175VDC? Why not just rectify to 175VDC and use a commerical PFC controller to regulate voltage? I believe that your probably going to have a pretty high ripple current running across those caps if the output is 500Watts. I believe a PFC controller (Power Factor Correction) would the better way to go and afford you better efficiency (less heat) and it should reduce your part count to boot.
>The H bridge driver and one DAC is connected to a Flash ROM chip, which stores the voltage and polarity values needed to make a sine wave.
The address lines of the Flash chip are sequenced by a counter, stepping through the values. The counter is clocked by a PLL.
Your probably better off using a digital sine wave via PWM instead of driving the H-bridge using an analog output. If you operate the output transistors in a non-saturated state, they are going to disspate a huge amount of heat. A better method is to set up a table of discrete set of pulse duration that varies the duty cycle to recreate a sine wave using PWM (ie duty: 0%, 2%, 8%, 16%... 100%... 16%., 8%, 2%, 0% = half-cycle sinewave). You also need to drive this output into a BTL( Bridge Tied Load) H-Bridge configuration or a second H-BRIDGE to Flip the voltage because AC output is 340V Peak to Peak.
FWIW: I would recommend that when you test, that you start off with a cheap test load (perhaps a set of 5 (or less) 100watt bulbs). While these are resistive loads as appose to inductive loads of a AC motor, it would allow to you tweak your PWM\Drive output before testing on a relativily expensive air conditioner.
Over all I think it would be more efficient to use a storage medium to serve as a coolant reservor. There is a chance that you might decrease the lifespan of the compressor motor by operating it at lower speeds. I suspect that compressors are probably equiped with vibration dampeners that are optimimized for a fixed frequency.
Hope this was useful. |
Reciprocating compressors are not constant loads. The motor more or less freewheels half the time (suction) and is under changing load the other half (discharge). Rotary compressors should be similar except there will be less freewheeling and the load will resemble a sawtooth wave. Scroll compressors should be fairly constant. However, operating pressures also determine load so even a scroll compressor will need some form of feedback control.
The buck converter is used to shape the waveform as well as step down the voltage. The H bridge only reverses polarity. The buck converter is the PWM while the H bridge merely changes the current direction. The voltage doubler gives the buck converter a greater input voltage. A normal bridge rectifier will require a buck/boost converter which is more complex and harder to design. They are also often less efficient. Even if a 240v compressor is used, a buck/boost converter is still needed since the PWM output must cover 0-350v of range to make a sine wave. But if magic sine waves are used...
I think compressor motors are more sensitive than most motors so a square wave probably will not work for long. |
|
|
| TechGuy |
>The buck converter is used to shape the waveform as well as step down the voltage.
Perhaps you thought this through but I will toss it out anyway. When your driving at a lower or higher frequency than the mains 60 hz, the available current from the AC mains will be out of sync. In some situations, the Main Sine output will be near zero crossover where the input current is virtually zero, while your current output demand may near its maximum. In this situation you may not have sufficent current available to properly drive your motor. This is why I assumed that you were using a buck to regulate the the input voltage. I suspect that in order to compensate you need a big bank of caps ($$$). This isn't probably going any more efficient than a PFC that produces a stead output voltage. Its likely that the caps you need to compenate will have a significant ESR, and you'll disapate a lot of heat from them. At 240 volts the ESR of the caps will be even higher. In my opinion big caps aren't the way to go. First they can be costly and they have a reliatively short operation life under high ripple loads (~3000 hours).
> The voltage doubler gives the buck converter a greater input voltage. A normal bridge rectifier will require a buck/boost converter which is more complex and harder to design.
http://www.irf.com/product-info/dat...data/ir1150.pdf
Goto mypower.irf.com/pfc (need to create a login account). Enter your Input/Output voltage, power requirements, efficiency, and tolerances, and the site will calcuate everything and provide you with a list of parts, values and schematic. Takes less than two minutes (including registration)
Take Care |
|
|
| star882 |
| quote: | Originally posted by TechGuy
>The buck converter is used to shape the waveform as well as step down the voltage.
Perhaps you thought this through but I will toss it out anyway. When your driving at a lower or higher frequency than the mains 60 hz, the available current from the AC mains will be out of sync. In some situations, the Main Sine output will be near zero crossover where the input current is virtually zero, while your current output demand may near its maximum. In this situation you may not have sufficent current available to properly drive your motor. This is why I assumed that you were using a buck to regulate the the input voltage. I suspect that in order to compensate you need a big bank of caps ($$$). This isn't probably going any more efficient than a PFC that produces a stead output voltage. Its likely that the caps you need to compenate will have a significant ESR, and you'll disapate a lot of heat from them. At 240 volts the ESR of the caps will be even higher. In my opinion big caps aren't the way to go. First they can be costly and they have a reliatively short operation life under high ripple loads (~3000 hours).
> The voltage doubler gives the buck converter a greater input voltage. A normal bridge rectifier will require a buck/boost converter which is more complex and harder to design.
http://www.irf.com/product-info/dat...data/ir1150.pdf
Goto mypower.irf.com/pfc (need to create a login account). Enter your Input/Output voltage, power requirements, efficiency, and tolerances, and the site will calcuate everything and provide you with a list of parts, values and schematic. Takes less than two minutes (including registration)
Take Care |
A local electronics shop sells 1100uF, 450v capacitors for about $20. They're intended mainly for variable capacity A/C units so they should be able to take the load. However, I'll probably design it to use cheaper caps, since the large ones are made to be used in 2+ ton systems and are no doubt overkill for a small chiller. But I suppose they won't hurt other than by increasing the total cost...
Come to think of it, I should check if they sell motor inverter units. I do remember seeing what appears to be an elaborate motor control module the last time I was shopping at the store...
And don't other power supplies of that wattage range (about 500w) face the same problems? How do they deal with the problem in large audio amps and PC power supplies?
If PFC is used, shouldn't it modulate in response to load current when possible to reduce ripple current? |
|
|
| TechGuy |
>A local electronics shop sells 1100uF, 450v capacitors for about $20
I think you need may need more than that to drive a 500W variable frequency load, but I haven't done the math to check.
>And don't other power supplies of that wattage range (about 500w) face the same problems? How do they deal with the problem in large audio amps and PC power supplies?
The advantage of using a PFC is to maximize the power of an AC cycle. I will defer to others about large audio amps, but for a very long time PC power supplies output was modest. I believe the new large PS'ss (+500W) use PFCs. My guess is that large audio amps have lots of caps coupled to the output of a linear power supply.
>If PFC is used, shouldn't it modulate in response to load current when possible to reduce ripple current?
The PFC regulates output voltage. When the voltage begins to drop, either do to current demand or fallling input voltage it uses the Buck boost to maintain output voltage. The PFC output caps would have a significant lower ripple current, than a simple rectificed output coupled with caps. The PFC would also aid in brown out conditions that might plague your area during summer highs.
Another alternative, is to see if you can find a compressor that uses a high voltage DC motor (~90VDC). With a DC motor you could regulate its speed with a simple PWM controller. Try http://www.rparts.com. You might need to use a belt driven compressor since I think Rparts only sells low voltage DC compressors. |
|
|
| gearheaddruid |
Star, I just recently repaired my basement refrigerator (compressor locked up) using Peltier devices. M.P. Jones sells 225 watt modules for about $30.00. I removed all of the freon parts and cut a hole in the back of the thing to mount a flat heatsink and fan through. The same frige that drew about 500w when running now maintains 38 degrees F on less than 60 watts. Regulation is done by a simple closed loop feedback system. On high demand, I have 540 watts available to cool warm food. When temp is stable, the system varies the voltage to the Peltier chips to mantain constant. No cycling on and off, no temp fluctuations.
Just a suggestion.
I use inverter drives on my three phase machine tools in my garage. Single phase in, pfc into filter caps, voila! I am in process of designing a 10kw PFC to run a plasma cutter, hope I can do it. Used drives are not expensive if you are doing a one time build. I also have a large amount of 680ufd/450volt low ESR caps from SMPS if you need some. I bank up 10 0r 12 inside my inverter drives and they give plenty of reserve. Good luck with the design. Best regards |
|
|
| star882 |
| quote: | Originally posted by gearheaddruid
Star, I just recently repaired my basement refrigerator (compressor locked up) using Peltier devices. M.P. Jones sells 225 watt modules for about $30.00. I removed all of the freon parts and cut a hole in the back of the thing to mount a flat heatsink and fan through. The same frige that drew about 500w when running now maintains 38 degrees F on less than 60 watts. Regulation is done by a simple closed loop feedback system. On high demand, I have 540 watts available to cool warm food. When temp is stable, the system varies the voltage to the Peltier chips to mantain constant. No cycling on and off, no temp fluctuations.
Just a suggestion.
I use inverter drives on my three phase machine tools in my garage. Single phase in, pfc into filter caps, voila! I am in process of designing a 10kw PFC to run a plasma cutter, hope I can do it. Used drives are not expensive if you are doing a one time build. I also have a large amount of 680ufd/450volt low ESR caps from SMPS if you need some. I bank up 10 0r 12 inside my inverter drives and they give plenty of reserve. Good luck with the design. Best regards |
A normal refrigerator actually has very little heatload. It's completely different when the objects being cooled actually give off very significant amounts of heat. For my application, compressor-based cooling is still far cheaper than Peltiers. Peltiers have the most advantage for small heatloads, although small scroll compressors are narrowing the gap.
They're cheap enough at the right place, though. In my area, a local electronics parts store sells 100w Peltiers for $13 each including heatsinks on both sides, although some specialty HVAC contractors often buy them all... I did manage to actually buy one and it's at least very fun to play with. I think I'll make a tiny A/C with it to precool the air into the computer, although I'm not sure how well it would work. I'll use a modified ATX power supply to power it.
I think it would most likely be a few years before I actually build the compressor-based cooler. By then, I hope inverter drive compressors like the "Little Kim" become common and cheap. |
|
|
|
