This is a about 30HP DC motor, likely powered by rectified 3 phase 460 VAC. Motor current draw spectrum and TWF has been acquired. No histrical data is available.

Amperage measured with an analog meter showed AC component being about 37% of DC. Looks too high to me!

A matter of concern is the fact that with such a significant AC amps component, torque fluctuation will result.

Attached plot shows dominating 360 HZ with harmonics, little bit of 120 HZ and modulation of 360 HZ by the run speed. The two latter are probably not problematic.

There is a tachometer sitting on the motor shaft. Being part of a feedback loop it is maintaining set constant speed, but obviously it does not maintain constant torque.

No mechanical vibration data is available.

Isn't the above picture a sign of trouble?

Thanks,
David

Mot_current_360HZ.doc (30 Kb, 30 downloads)

6xFL is normally the dominant peak for all DC motor current spectra (due to the six SCRs for full-wave rectification), so that in itself should not be a cause of concern.

However, the 37 % AC ripple is high. I believe Siemens recommends installation of smoothing reactors if the ripple is more than 30 %.

I have attached a couple of DC motor cases' spectra with confirmed problems. The dominant frequencies were FL & 2FL though.

Regards,

Aditya

DC_motors.ppt (205 Kb, 36 downloads)

quote:

Originally posted by Aditya:
I believe Siemens recommends installation of smoothing reactors if the ripple is more than 30 %.

I have to admit that I have no information on recommended percentage value for ripple, but even 30% seems too much to me. 30% of torque magnitude is variable???? Shouldn't all shafts in the power train be designed with this in mind? At least it may not be desirable for certain applications where high magnitude torque oscillating at 360 HZ may excite serious torsional vibration or resonance.

David

Aditya,

It is hard to see from the attachment, but the first plot appears to show 6xFL. If so, was there a confirmed problem?

Thanks,
David

I don't have many answers on the above questions. Just a few thoughts fwiw.

I agree it seems possible there may be torque oscillations on the same order of magnitude as the current ripple.

Even if it is so, it may not be a huge cause for concern. Look at the single phase motor. In absence of any auxiliary winding, it's current and flux go to 0 twice per cycle, at which time torque must go to 0 also. So if average torque is 100%, we might imagine torque oscillating between 0 and 200%. One saving grace is that the frequency of oscillation is constant, so chance of exciting a resonance is reduced (compared to start of sync motor where the oscillating frequency sweeps through a wide range as the machine starts up). Another factor is that some single phase motors do have an auxiliary winding that remains in the circuit after it is connected (for example cap-start/cap-run motor) and the oscillations would be reduced.

I have never measured current spectra on a dc motor. I started to try to figure it outo from theory, but I don't think I come up with much conclusive.

Attached is analysis of the unfiltered voltage of an unregulated 3-phase-input half and full wave bridge rectifier.

For the unregulated 3-phase full wave bridge rectifier output, the rms of the ac voltage ripple is approx 4% of the dc voltage (RatioAcToDc:=ac_fwr_rms/dc_fwbr). And the peak of the ac waveform is on the order of 10% of the dc voltage (see plot titled FWR_Voltage_AC_Portion)

As a point of information, for an unregulated half wave bridge rectifier output, the rms of the ac voltage ripple is approx 18% of the dc voltage and the peak is above 30% of the dc voltage. But half-wave bridge rectifiers are not often used, and if they were used here the ripple frequency seen would be 3*LF instead of 6*LF.

All of above assumes no filtering.

We also should mention that we have to be carful to say what the 4% or 30% means. Are we talking about ratio ac rms / dc ? Or something else. I assume we're talking about ratio ac rms / dc and I'll call it ripple (I know others may have different definitions for ripple).

So far we're talking about 4% ripple... how do we get up to 30% or beyond ? I don't really know the answer, but note two aspects of the 4% number:
#1 – It is a voltage not a current.
#2 - it is based on an unregulated supply (an unregulated supply is a constant voltage supply using diodes, a regulated supply is variable voltage supply using SCR's or other gated devices).

Let's look at #1 – current vs voltage. If I have 4% ripple in my voltage and hook it up to a resistive circuit, I would get 4% current ripple. Now I might think that hooking this up to a circuit with series inductance present tends to decrease the ripple in the current below that in the voltage, and hooking it up to a circuit with parallel capacitance tends to increase the ripple in the current above that in the voltage. I tend to think the parallel capacitance is not large enough to make a big difference (even at these high frequencies), but series inductance is large enough to play a role.

The above logic leads me to the conclusion that a dc motor with 4% ripple in the applied terminal voltage would draw current with 4% or less current ripple. BUT, I am not positive about this conclusion - I am a little leery of modeling a dc motor as a simple constant-value R-L-C circuit, when we know that a motor is a little more complicated than that (effective resistance changes with load from steady-state to steady-state, for one thing). Measurement of voltage ripple along with current ripple would help clarify the situation.

Now let's look at #2 – regulated vs unregulated supply. If the supply adjusts the output dc voltage using gated SCR type switches which operate for variable portion of each cycle to control the voltage, then the waveform will be more complicated than presented above, and the voltage ripple can be much much higher. (assuming no filtering).

This message has been edited. Last edited by: electricpete,

FullWaveRect_HalfWaveRectR2.pdf (38 Kb, 23 downloads)

As shown above by Pete and assuming a no regulated full 3 phase bridge rectifier, voltage ripple (defined as AC/DC), is relatively small - 4%. However, and this is the surprise part, current ripple is high - 37%

How to explain it? Here is one possible scenario. Assume a shunted DC motor.

1. Magnitude of electrical current circulating at 360 HZ in the rotor core is defined only by its R-L-C and, therefore, may be unproportionally to AC ripple voltage high.

2. Torque due to interaction of AC current and AC voltage is unidirectional although slightly oscillating, but of little magnitude because of small AC voltage supply component.

3. On another hand this AC current at 360 HZ interacting with DC supply voltage creates significant torque oscillation at 360HZ. It may not cause significant angular vibration or a resonance, but it definitely will create oscillating stress in the shaft, IMO. The more moment of inertia of the rotor train, the more stress.

David

This message has been edited. Last edited by: David_G,

quote:

Originally posted by David_G:
Aditya,

It is hard to see from the attachment, but the first plot appears to show 6xFL. If so, was there a confirmed problem?

Thanks,
David

On the first slide; the LHS motor signal shows a very small peak at line frequency & the dominant peaks at multiples of SCR firing frequency. For the RHS motor, line frequency is quite strong & the dominant peaks are multiples of 3xFL (or half the SCR firing frequency).

Yes, this was a confirmed analysis. The plant engineer replaced one of the cards & the speed fluctuations stopped.

Regards,

Aditya

quote:

Originally posted by Aditya:
For the RHS motor, ... the dominant peaks are multiples of 3xFL (or half the SCR firing frequency).

Yes, this was a confirmed analysis.

Aditya,

How does the spectrum look now?

I still wonder if dominating 6xLF signifies a problem. Also, again, what is the acceptable ratio between the AC and DC components at 360 HZ ?

Thanks,
David

David
Can you provide a zoom-in of the current time-waveform over a shorter interval like 25-50 msec. I am interested in looking at the shape and seeing if it looks like the shape of a full-wave-rectified voltage waveform.

The ac reads about 15A rms I think. Is the dc around 40A?

I came accross the quote below which implies a 360hz torque ripple of 10-20% of rated torque is typical for some dc motors fed from 6-pulse converters, which I assume includes simple 3-phase full wave diode bridge rectifiers as well as regulated SCR bridge power supplies. It's suprising to me that the number is so high.

quote:

"Control Techniques drives and controls handbook" by Drury ISBN 0 85296 793 4
Torque ripple is inherent in almost all electrical variablespeed drives. The frequency and magnitude is dependent upon the type of converter and control applied.

Consider first the D.C. motor fed from a six-pulse converter (so named because the resultant motor armature current has ripple comprising six peaks for every cycle of mains frequency). On a 50 Hz mains supply this ripple has a fundamental frequency of 300Hz (60Hz=~360Hz). In a separately excited D.C. machine the torque is proportional to armature current, the torque ripple has a frequency of sixtimes mains frequency. The magnitude of this ripple is typically in the range 10-20 per cent rated torque. The frequency of the torque ripple is a function of the mains frequency and is independent of operating speed, so provided 300 Hz (360 Hz) is not close to the natural resonant frequency of the mechanical system no problems should result.

Torque ripple due to the commutation process in the motor is also important where a small number of commutator segments are used. This component of torque ripple has a frequency which is proportional to speed.

Consider now the situation with a PWm inverter system. Although the theory of torque ripple calculation is complex, torque ripple at six and twelve times the output frequency of the drive are of most practical importance. The magnitude of the torque ripple is dependent upon the magnitude of the current harmonics which is in turn very dependent upon the drive and control type and the demands of the application. It is not possible to give exact calculations of torque ripple but a general comparison by drive type is helpful.

Pete,

Attached are some more details. I have not measured DC component myself but they say that ACrms / DC is about 37% when measured with their instrument. I have entered scaling 0.005 V/AMP in the in analyzer box as my amp-clamp says.

Theoretically still puzzled by the high AC content.

David
PS
Looks like we posted the last two simultaneously. I have not read it yet.

This message has been edited. Last edited by: David_G,

Motor_amps_360HZ_zoomed_TWF.doc (53 Kb, 16 downloads)

Attached is a ppt file showing the waveform of the current on a 250 Hp, 500V DC motor. The average current, measured with the clamp on was 248 Amps. The current actually fluctuated between 160 and 300 amps. This ripple current caused a very poor commutation and heavy arcing on the comm.
Adding the reactor in series with the armature decreased the ripple substantially and eliminated the arcing despite the average current being higher (290 Amps).
This motor rated at 500 V was fed by 6 SCR 3-phase full wave rectifier. The input was close to 500 VAC. Rated speed 2500 RPM, real speed less than 1800 RPM. So the SCR drive had to scale down the voltage considerably. The price is, of course, a high ripple.
You can read about NEMA power code letters for DC motors in: http://www.usmotors.com/PB630/gen_appl.pdf beginning at the page 3.

jank

ripple_current_in_DC_motor.ppt (146 Kb, 15 downloads)

It strikes me that David's time waveform (ac portion) and Jan's time waveform resemble very closely the ac portion of the full wave rectified voltage time waveform that I showed in my attachment.

Dave - notice on yours that the sharp negative peak of the ac component is roughly twice the distance from zero as the round positive peak at top - exact same pattern as on the full wave rectified voltage waveform (if you ignore that slower variation superimposed).

Jan - you said yours was due to a commutation problem, but the frequency is 6*LF. But the pattern looks just like power supply ripple, right? What is the connection between 6*LF ripple and poor commutation? The high current in peak of the ripple created problems in commutation?

So it looks to me in both cases the shape of the current ripple (ac component of the current) is very similar to the shape of the voltag ripple. What I don't understand is (as David asked) why the current ripple (compared to dc current) would be much higher than the voltage ripple (compared to dc voltage).

This message has been edited. Last edited by: electricpete,

Thinking some more, I am pretty sure that due to the resistive/inductive nature of the motor load, the motor cannot have higher ripple than the voltage.

So I'll bet that if you measured voltage on those machines with the high 6*LF ripple in current, you would see similarly high ripple on voltage (did you measure voltage Jan?)

The fact that it closely resembles the waveform unregulated/diode type 3-phase full-wave rectifier (except for the magnitude) may be just a coincidence (that type of supply would have much lower ripple).

Maybe it is a regulated/SCR type 3-phase full-wave rectifier. That may explain why the magnitude of the peaks tends to wander up and down in magnitude every few peaks, also. I don't picture the voltage waveform of a regulated SCR supply would look so close to that of an unregulated diode supply, but perhaps after it has been low-pass filtered by the inductive load it does.

I checked with a friend whose company is specifying transformer rectifier sets for aluminium plants right now. They specify that the AC ripple component in the voltage should not exceed 5 %, seems there is an IEEE document on it. Not sure if this would be applicable for DC motors.

In my experience, the voltage waveform does not follow the current waveform (similar period but different waveshape). See attached data.

Dumb question now - how do you get the ripple value from the time waveform? Is it peak to average (the single largest peak, or av. of all the peaks)?

Regards,

Aditya

Ripple_Patterns.doc (290 Kb, 11 downloads)

Ripple = (RMS value of the AC component)/(DC value)
jank

The current should resemble a low-pass-filtered version of the voltage. So the higher harmonic content of the current should be equal or less than the voltage. Your posted waveforms obey that much as expected. Your voltage waveform looks to me like a little strange and I expect it is a regualted waveform. Clearly not the output of a simple diode rectifier.

Now the interesting part - your current waveform does give a similar current appearance to the other current waveforms posted, which is interesting. To me, it seems to confirm that low-pass filtering can transform a regulated voltage waveform into a current waveform that has the same general shape as the unregulated/diode-bridge votlage waveform... but with a lot higher ripple. In other words just a coincidence that the current looks like that... doesn't mean that it's fed from an unregulated diode supply. In my view it tends to support my most recent post, instead of earlier posts where I thought it must be fed by diode rectifier.

Thanks for posting that Aditya. It takes a lot of guesswork out of the picture to be able to see the voltage.

There is a lot of different terminology used as far as ripple. I think the way I defined ripple above (ac rms / dc) was unfortunate. Ripple is more commonly defined as difference between peak and valley, divided by average. Form factor appears more commonly used to characterize the harmonic content. Form factor = total rms / dc component. Close to 1 is pure dc. Higher has more ripple and harmonics.

In my pdf above I had computed for full-wave 3-phase diode bridge:
acrms/dc = 0.04
total = sqrt(acrms^2+dc^2)
So based on the voltage, form factor = total/dc = sqrt(acrms^2+dc^2)/dc = sqrt(0.04^2+1)/1 = 1.004. Current form factor would be even lower (for this type of 3-phase unregulated/diode full wave bridge rectifier) due to the filtering effects of the inductive load.

This message has been edited. Last edited by: electricpete,

quote:

Originally posted by electricpete:
I am pretty sure that due to the resistive/inductive nature of the motor load, the motor cannot have higher ripple than the voltage.

I think current ripple magnitude and voltage ripple magnitude ( percentage wise) do not have to be equal. The reason is as follows.

AC current in a DC motor depends on: AC voltage and R & L values of the rotor core curcuit. It is not dependant on mechanical load (which is true for DC current at given DC voltage).

That is why, as Jan showed, inserting a reactive load in series lowered AC current and did not affect the DC portion.

In general I guess people do not worry too much about oscillating torque unless there is a torsional resonance.

Aditya,

What kind of power supply is in your example?

David

[

quote:

Originally posted by David

quote:

Originally posted by electricpete:
I am pretty sure that due to the resistive/inductive nature of the motor load, the motor cannot have higher ripple than the voltage.

I think current ripple magnitude and voltage ripple magnitude ( percentage wise) do not have to be equal. The reason is as follows.

The current ripple must be equal or less than the voltage ripple (provided we are using my original definition of ripple voltage as the ratio of ac rms to dc).

The reason comes exactly as you said - the model of the motor as a series R - L circuit.

If we view the motor R-L admittance (admitance Y is inverse of impedance Z) as a filter which converts voltage into current, it is a low-pass filter. The low-frequency components of the voltage pass through to become current bettter than the high frequency components.

Mathematically the admittance Y as a function of frequency whose magnitude is given by
Y(f) = 1/sqrt(R^2 + (2*Pi*f * L)^2)


At low frequency Y(f) has it's highest value. (i.e. the filter passes dc easily from it's input voltage to its output current). As f goes up, Y(f) goes down and the filter magnitude decreases (i.e. as the frequency increases, the filter passes less and tends to block these frequency components from appearing in the output current).

If you have a certain ratio ac/dc present in the voltage (input of the filter), then the ratio ac/dc present in the current (output of the filter) will be smaller than it was in the voltage because the ac is attenuated by the filter more than the dc.

Adding a series inductance as Jan mentioned increases the L and makes the low-pass filtering effect even more pronounced (more of the ac is removed).

This message has been edited. Last edited by: electricpete,

The voltage waveform posted by Aditya looks like the output from a "Line Frequency Phase Controlled Rectifier" as described by "Power Electronics" 2nd ed by Mohan/Undeland/Robbins.

It is a subset of what I would call voltage-regulating SCR type supplies.

The output voltage waveform is controlled by varying parameter alpha which is a firing angle of the SCR. Figure 6-21 of the above textbooks is attached and shows how the output voltage changes as we vary alpha.

The figure looks a little messy at first. It is helpful to separate the dark line from the light lines. The dark line is the output voltage waveform. There are 6 lighter-curve lines which represent the three input phase-to-phase voltages and their inverses (separated by 60 degrees between each of the six). The dark output curve is formed by selecting a 60 degree window along each input curve, and then shifting to the same 60 degree window along the next input curve.

When alpha = 0 (figure 6-21.a) , the SCR fires at the same time a diode would have started conducting, and the output voltage looks just like a full-bridge diode rectifier. It traces out the upper-most portion of the six available curves and gives the highest possible dc output voltage. As we increase alpha, we trace out a delayed portion of each of those curves, which gives a lower average (dc) voltage and a different waveform. The waveform from Aditya looks very similar to figure 6-21.c where alpha=60 degrees. Although since Aditya's voltage didn't go to zero, I would say that particular alpha angle is somehwere between between 30 and 60.

This message has been edited. Last edited by: electricpete,

LF_PhaseControlledRectifier.pdf (826 Kb, 6 downloads)