power supply design: AC three phase to DC

I'm curious about the variations and their effects in the design of power supplies intended to utilize three phase AC power to produce DC output. In particular I am wanting to understand the harmonic effects of choosing between wiring the phases in parallel vs. series. All the schematics I have seen on the net show a parallel design. But what about doing a series design?
By a series design, what I mean is that each of the three phases is wired from separate transformer secondaries with a full wave bridge rectifier, and the DC side of each of those is wired in series.
In the parallel design, it would seem to me that the current waveform would be affected by the fact that different phases are at different voltage, and all the current would flow from just one phase at a time resulting in harmonic currents. A series design could have current in all three phases all the time.
I'm also wondering about the effect of the series voltage on the rectifier switching where the voltage from the other two phases could bias the rectifiers the wrong way, and put reverse current flow on the transformer secondaries (backfeed). Could controlled thyristor designs avoid this? But it would seem to me that even a parallel design could have such issues.
Do you know of any information source that would show a series type design to be workable, or unworkable? The ideal goal would be to have all the continuous power that multiphase AC can provide in the form of DC output with a power factor of 1.0 and no harmonics (however that is formally measured), with reasonable efficiency.

Just of the top of my head:
1. The voltage ripple is the same as a fullwave triphase rectifier (6 diodes), that is ~ 5%, but your configuration requires 12 diodes!
2.> (snip) A series design could have current in all three phases all the time. I think this is an incorrect statement. Let's assume a normal triphase system with R, S and T phases. When phase R (0 phase shift) is at +60 degree, the voltage of phase T (+240 degree phase shift) is zero, therefore the return current impressed into the load by the sum of the voltage contributions of R and S (which is Vsum ~ 2*0,86*Vpeak) doesn't pass throught the T winding, but through the diodes of the rectifing bridge connected to Phase T.
Gene

| Just of the top of my head: | | 1. The voltage ripple is the same as a fullwave triphase rectifier (6 | diodes), that is ~ 5%, but your configuration requires 12 diodes! | | 2.> (snip) A series design could have current in all three phases all the | time. | I think this is an incorrect statement. Let's assume a normal triphase | system with R, S and T phases. When phase R (0 phase shift) is at +60 | degree, | the voltage of phase T (+240 degree phase shift) is zero, therefore the | return current impressed into the load by the sum of the voltage | contributions | of R and S (which is Vsum ~ 2*0,86*Vpeak) doesn't pass throught the T | winding, but through the diodes of the rectifing bridge connected to Phase | T.
OK, so it would just be a big bucking contest to see which phases would be the ones to conduct.
What about a thyristor switching system where each phase was switched 1/3 of the time to each of 3 capacitor (banks), such that at any one time, each phase was always connected to one of the banks, and which bank it is depends on its voltage level. If you draw the three sine waves of three phase power at 120 intervals, there would be two levels where voltages cross each other. By switching at the right times, you get one bank always fed by the phase currently at the lowest voltage, another fed by the phase currently at the middle voltage, and the last fed by the phase currently at the highest voltage. These three banks would thus each have a different voltage. But wired in series, now, they add up.

  Quoted Text Here  
I think you miss my point. The way I see it, there is no "bucking" contest anywhere. The return current finds two couples of two forward-biased diodes in series (the bridge connected to the phase when its instanteneous voltage is 0 Volt).
  Quoted Text Here  
But the lowest level is 0Volt!
  Quoted Text Here  
But this is what diodes do naturally in a rectifing bridge. Perhaps I don't follow your argument, sorry.

| |> OK, so it would just be a big bucking contest to see which phases would |> be the ones to conduct. | | I think you miss my point. The way I see it, there is no "bucking" contest | anywhere. The return current finds two couples of two forward-biased | diodes in series (the bridge connected to the phase when its | instanteneous voltage is 0 Volt).
It's not at 0 volts all the time. But there will be times when the voltage is higher and lower. The question is, when will the rectifiers conduct from the transformer secondary? They would have to at least some of the time.
|> What about a thyristor switching system where each phase was switched |> 1/3 of the time to each of 3 capacitor (banks), such that at any one |> time, each phase was always connected to one of the banks, and which |> bank it is depends on its voltage level. If you draw the three sine |> waves of three phase power at 120 intervals, there would be two levels |> where voltages cross each other. By switching at the right times, you |> get one bank always fed by the phase currently at the lowest voltage, | | But the lowest level is 0Volt!
Yes, it drops down to 0 volts, but it goes up to (I have yet to calculate the exact value) some point, and averages somewhere in between.
|> another fed by the phase currently at the middle voltage, and the last |> fed by the phase currently at the highest voltage. These three banks |> would thus each have a different voltage. But wired in series, now, |> they add up. | | But this is what diodes do naturally in a rectifing bridge. | Perhaps I don't follow your argument, sorry.
But there is an amiguity. If the diodes conduct current from the transformer then that voltage is present to the DC side. However, if they don't, then the voltage is not present. If it is not present at all times, then none of the phases can present a voltage, and DC has 0 volts. But how can they not conduct at least some of the time since each phase is itself just a typical full wave bridge. The issue is whether the diodes will conduct given the voltage in series from the other phases. Do you think they will or won't?
Normally only 2 diodes conduct at a time, and they switch at zero cross so the transformer is effectively flipped from the view of the DC side. The series voltage, though, is directed at the diodes in the direction to make all 4 conduct. But 2 of them see voltage from the transformer in the other direction. So will they conduct or not then?

  Quoted Text Here  
-------
http://ham.org/ |
  Quoted Text Here  
http://ka9wgn.ham.org/ |
  Quoted Text Here  
-------
What you are attempting to create is a twelve pulse system. The part you didn't mention is that there must be a phase shift of 30° between the two secondaries. The phase shift is achievable by using either a delta and a wye winding, or by using zig-zag windings. The choice between series and parallel is a matter of economics. For high voltages, series is used. (X-Ray, e bombardment, electrostatic precipitators & HVDC transmission). For more reasonable voltages, parallel is used because it is cheaper in terms of diode cost. When parallel is used, there must be some form of forced current sharing between bridges to achieve good waveforms (both for the AC & DC sides). One common method is to use a balancing transformer, and the other is to use separate filter inductor for each six pulse bridge. It is not practically possible to produce to matched voltages with a 30° shift. When parallel is used, this voltage imbalance will cause the entire load to be carried on one bridge at light load. At full load, the impedance of the two windings can be chosen to get a reasonable balance. The high THD at light load really isn't a problem when you consider that it's total harmonic current that causes problems.
As you mention, SCRs could be used to balance the current in parallel rectifiers. In the case where SCRs are already required for voltage control, there usually is an additional control loop to help balance the current. Where voltage control isn't necessary, the additional cost isn't justified.
To achieve 5% THD 18 pulse is required. This is achieve by having a shift of 0°, -20° and +20°. For applications like drives where no isolation is required, one bridge can be connected direct to the line and the other two phase shifted using a autotransformer. In this case the rating of the transformer is something like 30% of the total power. I believe this method applied to VFDs is patented.
A method used where there are a lot of drives at a location is to direct connect ~ 1/3 of the HP of drives, run 1/3 of the HP through a +20° autotransformer, and the remainder through a -20° autotransformer. Again, 5% THD is achievable if the drives have a optimal value of DC link inductor. (Cheap drives don't have DC link inductors and have real bad THD. Some cheap drives bring the terminals out so a optional inductor can be added.)
One reference is: http://ece-www.colorado.edu/~pwrelect/book/slides/Ch16slide.pdf For more detail: http://scholar.lib.vt.edu/theses/available/etd-08202003-104831/unrestricted/thesis_carl1.pdf
I would recommend getting a power electronics text and read up on rectifiers. In this case, older is better! The designs were worked out in the years following the invention of the mercury arc rectifier.
Matthew

On Tue, 31 Aug 2004 03:00:59 GMT Matthew Beasley
| What you are attempting to create is a twelve pulse system. The part | you didn't mention is that there must be a phase shift of 30? between | the two secondaries. The phase shift is achievable by using either a | delta and a wye winding, or by using zig-zag windings. The choice
The secondaries are simply isolated (6 leads), so they are neither in delta nor wye configuration.
I didn't mention a phase shift, because it seems not relevant. The phase angle of concern is what comes from the transformer.
| between series and parallel is a matter of economics. For high | voltages, series is used. (X-Ray, e bombardment, electrostatic | precipitators & HVDC transmission). For more reasonable voltages, | parallel is used because it is cheaper in terms of diode cost. When | parallel is used, there must be some form of forced current sharing | between bridges to achieve good waveforms (both for the AC & DC sides). | One common method is to use a balancing transformer, and the other is to | use separate filter inductor for each six pulse bridge. It is not | practically possible to produce to matched voltages with a 30? shift.
Again, I don't see where the 30 degree phase shift relates. Yes, I do know a delta-wye transformer will do a 30 degree shift. But in this case, I'm only dealing with the secondary side and the windings are isolated from each other (not wired in either delta or wye).
| When parallel is used, this voltage imbalance will cause the entire load | to be carried on one bridge at light load. At full load, the impedance | of the two windings can be chosen to get a reasonable balance. The high | THD at light load really isn't a problem when you consider that it's | total harmonic current that causes problems.
Is that all on one bridge during the time that bridge is the highest of the three voltages? That would be the harmonics I would want to avoid. But this goes away at high load?
| As you mention, SCRs could be used to balance the current in parallel | rectifiers. In the case where SCRs are already required for voltage | control, there usually is an additional control loop to help balance the | current. Where voltage control isn't necessary, the additional cost | isn't justified. | | To achieve 5% THD 18 pulse is required. This is achieve by having a | shift of 0?, -20? and +20?. For applications like drives where no | isolation is required, one bridge can be connected direct to the line | and the other two phase shifted using a autotransformer. In this case | the rating of the transformer is something like 30% of the total power. | I believe this method applied to VFDs is patented.
I can clean up the DC with ordninary filtering, so I don't think I need to go to 18 pulse.
| A method used where there are a lot of drives at a location is to direct | connect ~ 1/3 of the HP of drives, run 1/3 of the HP through a +20? | autotransformer, and the remainder through a -20? autotransformer. | Again, 5% THD is achievable if the drives have a optimal value of DC | link inductor. (Cheap drives don't have DC link inductors and have real | bad THD. Some cheap drives bring the terminals out so a optional | inductor can be added.)
FYI, this isn't for a motor application (although that might be fun to play around with at some point, too). What I am looking at doing is just powering an arc lamp with DC from a generator, and thinking about how to scale that up.
| One reference is: | http://ece-www.colorado.edu/~pwrelect/book/slides/Ch16slide.pdf | For more detail: | http://scholar.lib.vt.edu/theses/available/etd-08202003-104831/unrestricted/thesis_carl1.pdf
Thanks.

wrote:
  Quoted Text Here  
part
between
a
sides).
is to
  Quoted Text Here  
shift.
The phase shift is what gives a good AC side waveform. Take for example, a rectifier with a "infinite" DC inductance. The input will be a step function. A 6 pulse rectifier will draw over one cycle 0, 1, 0, -1 and back to zero, with 1 equal to the positive peak value. Run this through a Delta-Wye or Wye-Delta and this becomes a waveform of 1/2, 1, 1/2, -1/2, 1, -1/2. Combine another rectifier without a phase shift, and now the input is a step function of 1/3, 2/3, 1, 2/3, 1/3, -1/3, -2/3, -1, -2/3, -1/3. With the greater number of steps the input now better approximates a sinewave. Of course a real rectifier would have a real value of DC inductance that will introduce a ripple to each step, but the input is still far better with 12 pulse than 6. Another way to look at this is to look in the frequency domain. The largest (AC side) harmonic of a 6 pulse rectifier is the 5th. The 30° phase shift reverses the phase of the 5th harmonic. When combined with the non-phase shifted rectifier, the 5th harmonic is canceled out. Harmonics from the 7th up are not affected. To combine to equal 6 pulse rectifiers with no phase shift will result in no harmonic cancellation.
  Quoted Text Here  
load
impedance
high
of
avoid.
Yes, as a percentage of current load the harmonics will be higher at light load. The harmonic current at light load will be less than the harmonic current at full load. Its the total harmonic current that causes problems. This point is often dwelled upon, but it really isn't an issue.
  Quoted Text Here  
parallel
the
line
case
power.
need
The point of the 18 pulse is getting a good AC side waveform. Particularly in AC drives, the savings in the filter is of negligible benefit because the filtering cap bank required for the inverter side dwarfs the current from the rectifier. For really large rectifiers or high voltage rectifiers, the savings in the DC side filter becomes significant.
  Quoted Text Here  
direct
real
just
to
If you are looking at an arc lamp, I am assuming that the voltage is low (>200V?). In this application, you definetly are looking at a parallel rectifier. Can you get the generator wound for the desired output voltage? This can be done for less than you might think if you can actually get in contact with the right person at one of the generator manufactureres. It will cost more, but less than the cost of a transformer for the full output of the generator. You are pretty much on your own for a regulator, but you want to have the regulator regulating on the DC side not the AC side anyways.
You may want to look at getting the generator wound with either 6 phase star, or wound with two stators, phase shifted by 30°. If the 6 phase star is used, balancing transformers or inductors will be required in both the positive and negative legs to ballance the current between the two sets of rectifiers because of the common neutral. With two isolated windings, only one ballancing transformer or one set of ballancing inductors is/are required. As a side note, locomitives that still use DC motors use two isolated windings on the AC generator and switch between series and parallel rectifiers using SCRs to "shift gears" between low speed and high speed operation.
I have been involved in a design where about 30kW at 48V DC was desired from a backup generator. The cost of a custom generator, custom controls and a custom regulator was cheaper than a stock generator and controlled rectifier. The customer avoided the cost of a backup controlled rectifier because the generator now became the backup if the rectifier died. The application was used at a bunch of cell sites. The only downside was the long lead time for new or replacement generators. I believe local repair shops have rewound generators by reverse engineering the windings. Even though we included a automatic monthly "exercise" in the controls moisture still took its toll on the generator windings, particularly in mountain top sites.
  Quoted Text Here  
http://scholar.lib.vt.edu/theses/available/etd-08202003-104831/unrestricted/thesis_carl1.pdf
-------
http://ham.org/ |
  Quoted Text Here  
http://ka9wgn.ham.org/ |
  Quoted Text Here  
-------

  Quoted Text Here  
> You may want to look at getting the generator wound with either 6 phase
  Quoted Text Here  
Matthew, doesn't 'six-phase star' provide 60° phase shift "steps" {instead of 30°?}
—s falke

|
|> | > You may want to look at getting the generator wound with either 6 phase |> star, or wound with two stators, phase shifted by 30?. |> If the 6 phase star is used, balancing transformers or inductors will be |> required in both the positive and negative legs to ballance the current |> between the two sets of rectifiers because of the common neutral. | | Matthew, doesn't 'six-phase star' provide 60? phase shift "steps" {instead of | 30??}
Yes, there are 6 phase vectors at 0, 60, 120, 180, 240, and 300 degrees. However, you can connect between 0-120, 60-180, 120-240, 180-300, and 240-0, to get "in between" phase angles at 30 degrees (and times sqrt(3) voltage).
Now I need to find some tools to create some graphs to show the kind of setup I've been thinking of, which none of the aforementioned resources have hit upon. I was only going to do things in 3 phases for now, but they could be adapted to higher phases and could well be more practical that way.

wrote: -snipped-
  Quoted Text Here  
of
resources
practical
Believe it or not, I have used excel to do the phasors, using polar to cartesian conversion and then X-Y graphing to create the various phasors. Once I had the windings figured out using phasors in excel, I used pspice to calculate the currents of the complete rectifier design.
When I was doing a autotransformer for creating both a +20° and -20° phase shift with no voltage transformation, I made such a file. This was for the input of a large switching supply. I wish I still had the file, but it appears that it is lost. We had evaluated using a delta connected autotransformer, a wye connected autotransformer with a delta tertiary, and a zig-zag autotransformer. In the two cases with a neutral point, there was no connection between the source and the wye point. The delta wound ended up with the smallest PU rating of each of the designs. In the wye connected autotransformer the point of the delta tertiary was to establish a stable neutral point. When calculating out the harmonics, the delta tertiary ended up carrying a significant amount of harmonics, giving the wye connected the highest PU rating. The zig-zag connected was in between as far as a PU rating, but we felt that the possible circulating currents were best understood in the zig-zag so this is the design we went with (since only one winding is connected to each line, the winding current can't be higher than the line current; with delta there can be circulating currents in the delta). IIRC the PU VA rating of the autotransformer was .35 of the converter W rating.
Matthew

wrote...
  Quoted Text Here  
phase
will be
  Quoted Text Here  
current
{instead of
  Quoted Text Here  
Yep, you got me on that one. In the generator application we used two isolated wye windings with a 30° phase shift. Six phase star is usable to provide 6 pulse rectification with one DC terminal connected to the wye point.
Matthew

  Quoted Text Here  
Please pardon me; I was not trying to be rude or to "corner you." I recently made a similar mistake in thinking that a "6-phase star" winding with 60° "spacing" for 12-pulse rectification would be easier to implement than a delta-wye / delta-delta transformer pair with 30° "spacing." Apparently such reasoning is in error for both of our circumstances.
—s falke

  Quoted Text Here  
two
usable
the
recently
60°
a
Apparently such
  Quoted Text Here  
No offence taken. Once you pointed it out to me, I had a big DUH!
  Quoted Text Here  

On Tue, 31 Aug 2004 14:38:02 GMT Matthew Beasley
| The phase shift is what gives a good AC side waveform. Take for | example, a rectifier with a "infinite" DC inductance. The input will be | a step function. A 6 pulse rectifier will draw over one cycle 0, 1, | 0, -1 and back to zero, with 1 equal to the positive peak value. Run | this through a Delta-Wye or Wye-Delta and this becomes a waveform of | 1/2, 1, 1/2, -1/2, 1, -1/2. Combine another rectifier without a phase | shift, and now the input is a step function of 1/3, 2/3, 1, 2/3, | 1/3, -1/3, -2/3, -1, -2/3, -1/3. With the greater number of steps the | input now better approximates a sinewave. Of course a real rectifier | would have a real value of DC inductance that will introduce a ripple to | each step, but the input is still far better with 12 pulse than 6. | Another way to look at this is to look in the frequency domain. The | largest (AC side) harmonic of a 6 pulse rectifier is the 5th. The 30? | phase shift reverses the phase of the 5th harmonic. When combined with | the non-phase shifted rectifier, the 5th harmonic is canceled out. | Harmonics from the 7th up are not affected. | To combine to equal 6 pulse rectifiers with no phase shift will result | in no harmonic cancellation.
OK, I see what you are saying. You are combining waveforms from each side of the delta-wye so you have more phases. But that would be 12.
| The point of the 18 pulse is getting a good AC side waveform. | Particularly in AC drives, the savings in the filter is of negligible | benefit because the filtering cap bank required for the inverter side | dwarfs the current from the rectifier. For really large rectifiers or | high voltage rectifiers, the savings in the DC side filter becomes | significant.
Any schematics online?
What about 24 pulse?
| If you are looking at an arc lamp, I am assuming that the voltage is low | (>200V?). In this application, you definetly are looking at a parallel | rectifier.
I was figuring it would be somewhere between 200 and 600 at no load. Of course, while operating, the voltage across the arc drops way low and you have lots of current. I really don't know what voltage I'd need because I don't know that much about it, yet. I just want to do in with DC because I don't want it to be AC modulated.
| Can you get the generator wound for the desired output voltage? This | can be done for less than you might think if you can actually get in | contact with the right person at one of the generator manufactureres. | | It will cost more, but less than the cost of a transformer for the full | output of the generator. You are pretty much on your own for a | regulator, but you want to have the regulator regulating on the DC side | not the AC side anyways.
I'll probably be getting some used generator and going from there. As the rectifier circuits need 3 isolated secondaries, I either need a transformer, or have to break out the winding leads from the generator.
Do you think maybe driving the exciter level from the DC voltage would be the best way to regulate the generator?
| You may want to look at getting the generator wound with either 6 phase | star, or wound with two stators, phase shifted by 30?. | If the 6 phase star is used, balancing transformers or inductors will be | required in both the positive and negative legs to ballance the current | between the two sets of rectifiers because of the common neutral. | With two isolated windings, only one ballancing transformer or one set | of ballancing inductors is/are required. As a side note, locomitives | that still use DC motors use two isolated windings on the AC generator | and switch between series and parallel rectifiers using SCRs to "shift | gears" between low speed and high speed operation.
Interesting.
| I have been involved in a design where about 30kW at 48V DC was desired | from a backup generator. The cost of a custom generator, custom | controls and a custom regulator was cheaper than a stock generator and | controlled rectifier. The customer avoided the cost of a backup | controlled rectifier because the generator now became the backup if the | rectifier died. The application was used at a bunch of cell sites. The | only downside was the long lead time for new or replacement generators. | I believe local repair shops have rewound generators by reverse | engineering the windings. Even though we included a automatic monthly | "exercise" in the controls moisture still took its toll on the generator | windings, particularly in mountain top sites.
So I take it these were not epoxied windings.

Matthew Beasley wrote:
  Quoted Text Here  
http://scholar.lib.vt.edu/theses/available/etd-08202003-104831/unrestricted/thesis_carl1.pdf
One good book is "Rectifier Circuits" by Johannes Schaefer. It's out of print, so happy hunting. Another is "Power Converter Handbook" by the Canadian General Electric Company. This one was out of print, but I have some info at work of a company that was publishing this again. I'll check. There are ANSI standards covering various rectifier circuits as well. I'll look for those as well.