I was reading a manufacture manual and noticed they had a table for total harmoic distortion (current). The table showed that ithd for their power supplies were inversely related to total load. As the load dropped to only 20 or 30%, ithd significantly increased. I don't understand the reasoning behind this. Can anyone explain? I get why the true power factor decreased, because the equation has harmonics in the denominator, and since this value is getting larger your true power factor decreases. But I'm not sure why ithd exactly decreases. My understanding is that it's simply a summation of the fundamental current in the denominator and higher order harmonic current in the numerator squared and then square rooted. So I guess if the nominal current is small, it accentuates the harmoic currents in the numerator?
Harmonics in server power supplies
- Thread starter minesh21
- Start date
- Status
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180525-2015 EDT
minesh21:
Get a scope and a Hall device current probe. A shunt would work, but to use a shunt you have scope grounding problems.
Look at the current waveform for a pure resistance load and judge harmonic content.
Next look at a small power transformer, 100 W is fine. With no load estimate % harmonic content. Now fully load the transformer with a resistive load. Again make an estimate. Probably much lower with the full resistive load. Quite logical reason. You should be able figure out why.
The server power supply is much more complex, but you can expect a similar reason.
Some waveforms, but not complete to answer your question, are at my website
http://beta-a2.com/EE-photos.html
In particular look at P3 for a resistive load, and P8 for an unloaded transformer.
.
minesh21:
Get a scope and a Hall device current probe. A shunt would work, but to use a shunt you have scope grounding problems.
Look at the current waveform for a pure resistance load and judge harmonic content.
Next look at a small power transformer, 100 W is fine. With no load estimate % harmonic content. Now fully load the transformer with a resistive load. Again make an estimate. Probably much lower with the full resistive load. Quite logical reason. You should be able figure out why.
The server power supply is much more complex, but you can expect a similar reason.
Some waveforms, but not complete to answer your question, are at my website
http://beta-a2.com/EE-photos.html
In particular look at P3 for a resistive load, and P8 for an unloaded transformer.
.
junkhound
Senior Member
- Location
- Renton, WA
- Occupation
- EE, power electronics specialty
basically correct, especially if one assumes the server power supply does not have a power factor correction circuit At low load, the rectifier circuit turns on onlynear the very peak of the sine wave, at higher loads, as the load discharges the ripple capacitor further each cycle, the diodes turn on earlier thus widening the diode conduction duty cycle and reducing harmonics relative to the load.
Harmonic content
Harmonic content
The harmonic content will almost remain the same with different load. since ithd or vthd is ratio to fundamental value being denominator, ithd will be lower in higher load and higher in lower load. Practically i came accross on UPS output where ithd on neutral us uch higher upto 360%for a balanced load (where neutral current is close to zero) than unbanced load of same total load(where neutral current is significant more thsn zero)..
Harmonic content
The harmonic content will almost remain the same with different load. since ithd or vthd is ratio to fundamental value being denominator, ithd will be lower in higher load and higher in lower load. Practically i came accross on UPS output where ithd on neutral us uch higher upto 360%for a balanced load (where neutral current is close to zero) than unbanced load of same total load(where neutral current is significant more thsn zero)..
Yes ithd is higher in lower load..
Yes ithd is higher in lower load..
The harmonic content remains same or amost not linearly changing. Since ithd is a ratio with its denominatir being the fundamental unit, the higher the load is the lower the ithd.
I have a similar/ practical issue where the ithd on neutral line is much higher for balanced load (where neutral current is close to zero) than unbalanced load (where neutral is having significant current/load)
Yes ithd is higher in lower load..
The harmonic content remains same or amost not linearly changing. Since ithd is a ratio with its denominatir being the fundamental unit, the higher the load is the lower the ithd.
I have a similar/ practical issue where the ithd on neutral line is much higher for balanced load (where neutral current is close to zero) than unbalanced load (where neutral is having significant current/load)
Ok so what your saying essentially is at higher loads the capacitor has a greater discharge cycle effectively widening the pulses. By widening the pulses there is greater cancellation of harmonics current. Is this correct?
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180527-1443 EDT
minesh21:
You are listed as an electrical engineer. Thus, at some point you should have had some exposure to Fourier analysis.
Any infinitely long periodic waveform can be broken down to the sum a fundamental sine wave plus sine wave harmonics of that fundamental with specific phase and amplitude relationships between the harmonic sine waves and the fundamental.
For a non-infinitely long waveform you can get a reasonable good estimate by assuming it is infinitely long.
As soon as you do any distortion to a sine wave, then harmonic content is added.
Generally the more peaked or square a waveform is the more harmonic content there is.
I have already referenced you to waveforms from a transformer.
I don't have plots of current of a capacitor input power supply. But as you increase the power load on that supply you do broaden the input current pulse width and that does increase the fundamental current relative to the harmonic current components. Everything shifts around to produce this different current pulse.
.
minesh21:
You are listed as an electrical engineer. Thus, at some point you should have had some exposure to Fourier analysis.
Any infinitely long periodic waveform can be broken down to the sum a fundamental sine wave plus sine wave harmonics of that fundamental with specific phase and amplitude relationships between the harmonic sine waves and the fundamental.
For a non-infinitely long waveform you can get a reasonable good estimate by assuming it is infinitely long.
As soon as you do any distortion to a sine wave, then harmonic content is added.
Generally the more peaked or square a waveform is the more harmonic content there is.
I have already referenced you to waveforms from a transformer.
I don't have plots of current of a capacitor input power supply. But as you increase the power load on that supply you do broaden the input current pulse width and that does increase the fundamental current relative to the harmonic current components. Everything shifts around to produce this different current pulse.
.
So wouldn't it be misleading for UPS manufacturers to publish %thd with respect to increasing load? For example, the 5% THD is at full load for a UPS manufacturer, but this is only because its a percentage with the denominator growing. So at full load it's not that the actual distortion is decreasing, its just that the fundamental is increasing, correct?
You mentioned the harmonic content itself stays the same regardless of loading. So what we really need is a better way to represent harmonic distortion to compare it to different UPS technologies. Or I guess we could just compare thd at full load for all UPS units, so that it is a more fair comparison.
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180626-0743 EDT
minesh21:
If you have a simple capacitor input filter on the output of a rectifier, and for a constant capacitance, then as output load is decreased the input current will become more peaked. More peaked means that relatively speaking the harmonic content increases.
If you add inductance between the diodes and the capacitor, inductive input filter, create a low pass filter, then you create conduction over the entire cycle or a greater portion of the cycle depending on loading. Thus, input harmonic content is less for a specific load because the input current is less peaked. With continuous current flow the current is close to sinusoidal, and near no harmonic content. The diode drop creates some error.
Harmonic content of input current of a UPS or other power supply vs loading will be greatly dependent on the design of the power supply. Simple capacitor input filters are low cost, provide moderately constant output voltage vs load, and are lower in weight, when compared to an inductive input filter. Therefore, capacitor input power supply filters are more common.
.
minesh21:
If you have a simple capacitor input filter on the output of a rectifier, and for a constant capacitance, then as output load is decreased the input current will become more peaked. More peaked means that relatively speaking the harmonic content increases.
If you add inductance between the diodes and the capacitor, inductive input filter, create a low pass filter, then you create conduction over the entire cycle or a greater portion of the cycle depending on loading. Thus, input harmonic content is less for a specific load because the input current is less peaked. With continuous current flow the current is close to sinusoidal, and near no harmonic content. The diode drop creates some error.
Harmonic content of input current of a UPS or other power supply vs loading will be greatly dependent on the design of the power supply. Simple capacitor input filters are low cost, provide moderately constant output voltage vs load, and are lower in weight, when compared to an inductive input filter. Therefore, capacitor input power supply filters are more common.
.
As a percentage or in real terms?
FionaZuppa
Senior Member
- Location
- AZ
- Occupation
- Part Time Electrician (semi retired, old) - EE retired.
switching PSU vs linear PSU. switching is "noisy". switching PSU's are however more compact. pros-cons to everything.
now, if the noise in interfering with stuff connected to the BC, or Rf related issue, then make sure the PSU is properly certified.
now, if the noise in interfering with stuff connected to the BC, or Rf related issue, then make sure the PSU is properly certified.
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180626-2041 EDT
I have the feeling that total harmonic distortion as a numeric measuring tool was invented for use in evaluating audio amplifiers. Possibly as early as the early 1920s. In this application it probably provides a useful comparison tool.
Applied to other applications it may not be so useful.
In some way a THD value for a power supply may provide a comparison between somewhat similar power supplies, but may not be useful to compare different designs.
Someone mentioned that harmonics remain constant as load on a power supply is varied, and that only the fundamental changes. That will be an incorrect statement. What is certainly changing with load is the input waveform. With changing load what will happen is that the fundamental and all harmonics have to change magnitude and phase to whatever combination is required to represent the current waveform.
.
I have the feeling that total harmonic distortion as a numeric measuring tool was invented for use in evaluating audio amplifiers. Possibly as early as the early 1920s. In this application it probably provides a useful comparison tool.
Applied to other applications it may not be so useful.
In some way a THD value for a power supply may provide a comparison between somewhat similar power supplies, but may not be useful to compare different designs.
Someone mentioned that harmonics remain constant as load on a power supply is varied, and that only the fundamental changes. That will be an incorrect statement. What is certainly changing with load is the input waveform. With changing load what will happen is that the fundamental and all harmonics have to change magnitude and phase to whatever combination is required to represent the current waveform.
.
ron
Senior Member
- Location
- New York, 40.7514,-73.9925
I suspect the UPS manufacturer listing its %THD at full load is a marketing strategic spec and not because of the engineering characteristic nature of it.
The UPS listing, is typically the input THD what the source "sees". Not the load PSUs.
The UPS frontend made up of a SCR or IGBT rectifier, depending on its age, a will have varied THD values depending on it design, load and if it has a filter and whether it is employed or not.
Andy Delle
Senior Member
- Location
- Los Angeles CA
Any UPS must re-create a 60hz sine wave, or try to :huh:. The distortion spec on UPS systems is how clean the output sinewave is. Really no different than measuring an audio amplifier. When feeding a largely computer or other electronic load, waveform distortion isn't that important within reason because switch mode power supplies don't need a clean sinewave input. In fact most can even run on pure DC. Its when you have motors and 60hz transformers that waveform distortion becomes an issue, typically when over 10%, which is a lot.
ron
Senior Member
- Location
- New York, 40.7514,-73.9925
The newer large "transformerless" UPSs that I work with in the data center market are IGBT rectifiers, so there is no filter to have relatively low input THD. These are generally from 100kW to 1200kW per module, then we parallel them to get larger systems where needed.
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180627-1950 EDT
minesh21:
How did we go from a server power supply in your first post to a UPS in a later post?
Server power supply implies input current.
UPS would need a definition of whether it related to input or output.
.
minesh21:
How did we go from a server power supply in your first post to a UPS in a later post?
Server power supply implies input current.
UPS would need a definition of whether it related to input or output.
.
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180627-2406 EDT
Some general comments on waveform and harmonics.
Pure sine wave --- no harmonics.
Sawtooth that has the symmetry of a sine wave --- fundamental + odd harmonics that fall off quickly.
Square wave --- same as sawtooth but more and stronger odd harmonics. Diminishing more slowly.
Rectangular pulse --- same as square wave. Fundamental gets smaller, harmonics get stronger, and diminish more slowly as the pulse width shortens.
Diminish more slowly is in reference to the harmonic order.
.
Some general comments on waveform and harmonics.
Pure sine wave --- no harmonics.
Sawtooth that has the symmetry of a sine wave --- fundamental + odd harmonics that fall off quickly.
Square wave --- same as sawtooth but more and stronger odd harmonics. Diminishing more slowly.
Rectangular pulse --- same as square wave. Fundamental gets smaller, harmonics get stronger, and diminish more slowly as the pulse width shortens.
Diminish more slowly is in reference to the harmonic order.
.
gar
Senior Member
- Location
- Ann Arbor, Michigan
- Occupation
- EE
180628-2422 EDT
By what I said above I don't want to imply that some even harmonics don't occur, nor that there are always all odd harmonics. What the harmonics are is dependent upon the wave shape.
.
By what I said above I don't want to imply that some even harmonics don't occur, nor that there are always all odd harmonics. What the harmonics are is dependent upon the wave shape.
.
- Status