Electric Motor Secrets

[Jetijs]

Lighty, do you mean that I should connect both scope probes across the output battery terminals? If so, then would I even see any spikes? Because when I put the scope across the charging battery on a Bedini SSG, I could see only a more or less straight line of 12 or so volts and very small peaks (in milivolt range) at the SSG produced frequency. I mean, the battery absorbs all the spikes and I only can see the full spike voltage if the charging battery is removed, but as we know, that can damage the transistors or in this case the MOSFETs.

So, I set my scope software to a demo mode with simulated sinewave at 2kHz on both probes. Now I set the math to A-B (I think) and a third waveform appeared in red (see the picture) it is almost flat. Is this the right way?
http://www.emuprim.lv/bildez/images/...nti_1/math.jpg
Also, I went through the help file of the software and the word RMS was mentioned only in one part. If I go to the FFT menu, I have this window:
http://www.emuprim.lv/bildez/images/...unti_1/rms.jpg
And the "help" explains what the different measurements mean:

-SNR (Signal to Noise Ratio): The ratio of the amplitude of the fundamental frequency to the Noise.

-ENOB (Effective Number of Bits) : The number of bits in an ideal converter that would be required to give the same SNR performance.

-SINAD (Signal to Noise and Distortion):The ratio of the amplitude of fundamental frequency to the Noise, but Noise include Harmonics.

- THD (Total Harmonic Distortion): The ratio of the RMS sum of the harmonics to
the RMS value of the fundamental.

-SFDR (Spurious Free Dynamic Range): The ratio of the RMS signal amplitude to the RMS value of the peak spurious spectral component. The peak spurious component may or may not be a harmonic.
: The RMS value of the peak spurious spectral component.

-Total Power: The RMS value of the sum of all spectral components.
: Sum of Noise excluding DC and Nyquist.

Which one do I need to look at to get the RMS value?
I haven't dealed with these kinds of measurements before
Thanks.

[lighty]

You cannot have any load connected to recovery diodes. You must measure unaffected signal because battery, capacitor or any other device will effectively "suck up" all of the released energy and you won't get any usable measurement of the voltage spike itself. So, disconnect everything from the diodes and measure signal between free end of recovery coil and recovery diode.

Also, check out that both ground clips on your probes are connected together and not with anything else. If possible use laptop on battery power in order to prevent any ground influence on measurement.

I'm not familiar with your scope software since I usually use Tektronix DSOs but I will help as much as I can based on the screenshots you've posted.

So, forget about FFT as for the moment we don't need to know the spectral content of your inductive discharge impulse.


So do the following:

1. Disconnect battery or anything else from the recovery coil.
2. Start up everything and measure signal between drain and source of MOSFET. If the voltage spike seems like it's suddenly cut out between 450-500V that's because protection diodes are doing their job. Also, if the protection diodes starts to get even lukewarm it means they're doing their job. So, the voltage in this case is above 450-500V and in that case we'll have to worry about not burning your scope. However, this doesn't mean protection diodes will interfere with our intended recovery because those protection diodes will have to act only in case your recovery circuit doesn't do it's job. So, nothing to worry about for the moment if the voltage spike is under the protection diodes turn on threshold and in that case proceed to point 3.
3. Connect differential probe between free end of recovery coil and free end of recovery diode.
4. Set up probes voltages to the highest possible value (we'll probably deal with few hundred volts) in software. Also, set voltage divider on your probe to x10 or higher (just to be on the safe side- we don't want to burn your scope).
5. Set frequency to 20MHz (2kHz is way too low for such short impulses). I hope that's the frequency of bandwidth limiter because as I said I'm not familiar with your software controls.
6. When you're measuring if you get all three signals represented your software should be capable of omitting representation of both channels on screen and showing only the differential signal (the red one). So try un-checking CH1 and CH2 selections. If the software is standard one you should now see only the red waveform. If not you'll just have to ignore the green and yellow ones.


Also, could you please send me User manual for your software or if you don't have it could you then send me a Help file from your installation (CHM file most probably) so that I could check out other inner settings of your scope we could possibly set up like sampling rate and possibly RMS measurement of the represented math form.

BTW- thx for posting screenshots- please do that in the future as well so that I can see what you're getting. I'm pretty certain Peter would be interested as well.

[Jetijs]

Thanks lighty for the guidance
I just sent you the whole software so you can install it and check everything out by yourself.
I will make those measurements this evening and then post the results.
Thanks.

[Jetijs]

Hi all
Lighty, I did everything as you instructed. I used a laptop on battery for the measurement time. I got two scope shots, the first one is from the probes across source and drain of one of the MOSFETs. The other shot is from the recovery coil and recovery diode:





I set the probes to the highest voltage divider setting there was - x10. Also in the software I set everything to x10 and the highest possible voltages. I then unchecked the green and yellow waveforms leaving only the red one, as you said. You can see, that the voltage across the MOSFET source and drain goes off the limits. If every square is set to be 50 volts, then this means that the voltage goes way over 400v. At the second image we can see in the upper part some sort of ringing. For this test I used the one IN4007 diode on the recovery coil. If I run the motor without load, the amp draw increases a little and the protection diodes start to get slightly warm, so they are doing their job

So is there any way how I could see the full voltage spike? Maybe I should use a resistor in series with the probes?
Thank you
Jetijs

[lighty]

It's a good thing protection diodes are doing their job or you would now have instantly fried MOSFETs.

Anyway- simply adding resistor would very probably compromise frequency response of your probes and with such sudden impulses and short periods that could mean erroneous measurement. The only real solution would be to acquire higher attenuation probes.

For now you could go to your software and Util->Measurement->Voltage->Peak to Peak and the soft will do measurement for you. You can also use other functions there like RMS measurement and other stuff. At time domain measurement you can measure rise time and other factors if you ever plan to test another coil configuration.

In my experience the voltage you're getting is probably in the range between 470-550V because otherwise protection diodes would get very hot in short time.

So for now leave everything as is and have fun with recovery circuit. In case you're using capacitor it would be advisable for it to have it's breakthrough voltage rated on 600V or more just to be on the safe side.

As for the protection diodes sinking excess energy- don't worry as soon as you route energy anywhere else they won't open anymore. Check out and you'll see that they'll be stone cold.

[Jetijs]

Hello,
Today I made some tests before removing the rotor for balancig. Firstly I measured how much energy I am getting back using the same 40 degree gap on each commutator side. I put a better battery on the output that that what you saw in video. Here is the input current waveform measured across the pinput section amp meter, like I did with my first motor.



We can see that the overall ON time is a little smaller than the OFF time. This is because we have two 40 degree gaps, that is 160 degree per revolution. So the ON time is 160 degree and the OFF time is 200 degree.
And here is the waveform of the output pulse measured across the amp meter on the positive line of the output. This is using the IN4007 diodes on the output:



And here is the same only this time I am using the ultra fast diodes on the output, you can see that there is almost no difference:



After some crude current/voltage measurements I got the following results.
Input 12.24V, 1.81A
Output 13.5V, 0.6A
That is 36.6% recovery.

Then I switched to the commutator wheel with six smaller gaps on each side. All these gaps fill about 65 degree of the 70 degree firing window. Now the input current looked like this:



And this is the output current:



This time the current draw on the input was just 0.75A and the motor run very slow, loud and was easy to stop, I mean it was easy to stop in the NO fire zone, but it was not possible to hold the rotor in the firing position with hands.
This time the crude calculations showed 45.5% recovery.

That is it for now, I will now remove the rotor for balancing.
Any comments?
Thank you

[lighty]

The difference between ordinary diodes and hyperfast diodes can be easily noticed in the voltage spike peak when the reversion of flux on current turn off is very fast. It's hard to notice when voltage impulse is already converted to current. There are some advantages to using hyperfast diodes but I won't go into that in this thread.

Also, you can use math function of your software to see power graph directly. Use CH1 probe to measure voltage on input (or output) and use CH2 probe to measure current with shunt resistor (you should use 0.1 Ohm carbon resistor-you should never use wirewound type). Then use math function to multiply CH1 with CH2. The red graph you'll get is the power graph. Then turn on RMS measurement and you'll get RMS values of the power graph (it will be in red letters). In this way you can directly see any change in power in real time when you change something in motor configuration (recovery, load etc.).

You can also measure time periods with your software by using various time measurements (duty cycle and period comes to mind) or by using measurements cursors you can move over the scope screen. Either way you won't dealing with estimations then but rather with precise values.

Ah, and one more thing- whenever possible use differential probe configuration with appropriate math function (CH1-CH2) in order to reduce noise at least to some level (when measuring just one value at the time, of course). Also, I just noticed that when you using Channel setting you can turn Bandwidth limit Off. That's a great function to capture fast impulses (as much as your DSO sampling rate and input circuitry allows) although you might get some additional noise.

[Jetijs]

well,
I removed the rotor and put it on two parallel sharp blades to see the balance. I must say that there is a serious unbalance in the rotor, no wonder the motor vibrates so much.

It might also be that the shaft is bent Because all the parts are more or less symmetrical for a disbalance this big. You can see how many magnets I need to put as a counterweight to get the rotor more or less in balance.
here is a video:
YouTube - Unbalanced rotor or bent shaft?

[Jetijs]

Ok, here's an update.
I sanded off the "hump" on the heavier side of the rotor, only that gave a good balancing.



I had also to sand off a little corner. Anyway, I assembled the motor again and this time it did not have any vibration and ran alot more quiet. But ocasionally the motor got louder like every 10 seconds or so and then again it got quieter. I checked everything and saw that one of the bearings tends to rotate into the bearing holder, when it gets to a certain position, the motor gets louder and then, when it goes past that position, the motor is quiet again. This is a good indication that the shaft in fact IS bent. Also there is no way that the plates were cut so inprecise that there is need to sand off so much. It's a shame I did not check this when I got my shaft made. I will have a serious talk to the engineer who made this shaft for me

I guess I will have to make a new rotor. Must check if I have enough rotor plates left to do that. Also, I verified the motor speed. It indeed is about 1600 RPM, I counted the output pulses on a scope and measured the speed on different places with the laser tachometer. Also I guess that we will need to move up to 24V eventually.
Ok thanks,
Jetijs.

[sykavy]

Ok, here's an update.
I sanded off the "hump" on the heavier side of the rotor, only that gave a good balancing.



I had also to sand off a little corner. Anyway, I assembled the motor again and this time it did not have any vibration and ran alot more quiet. But occasionally the motor got louder like every 10 seconds or so and then again it got quieter. I checked everything and saw that one of the bearings tends to rotate into the bearing holder, when it gets to a certain position, the motor gets louder and then, when it goes past that position, the motor is quiet again. This is a good indication that the shaft in fact IS bent. Also there is no way that the plates were cut so inprecise that there is need to sand off so much. It's a shame I did not check this when I got my shaft made. I will have a serious talk to the engineer who made this shaft for me

I guess I will have to make a new rotor. Must check if I have enough rotor plates left to do that. Also, I verified the motor speed. It indeed is about 1600 RPM, I counted the output pulses on a scope and measured the speed on different places with the laser tachometer. Also I guess that we will need to move up to 24V eventually.
Ok thanks,
Jetijs.

Hi Jetijs,
I know you probably want to perfect this design you are using but could I make a suggestion that may help? If your going to make a new rotor maybe an "S" shape rotor could help. It may give you an easier time setting-up the timing. You then could grind down your electromagnet lips making coil replacement easier as Lighty suggested earlier. Your plate construction would also seem good for the "S" design. Or maybe wait till you do another one. I don't know as much as all of you on here but Peter did suggest this design once in his video. If anyone hasn't bought the video I would suggest you should ASAP. It is so important for this topic.

Anyway great job so far keep up the great work

[nali2001]

Hi there Jetijs,
Too bad about the rotor.
I doubt that it is bend, I mean why would it be, did you drop it or anything?

Are you sure the axle itself in precise? I mean when you machine one side of an axle and then take it out of the chuck, turn it around and machine the other side you already have an off center axle since a lathe chuck is generally not all that precise unless you are using a collet chuck. That is why I always machine the whole axle in one pass without taking it out and such.

Regards,
Steven

[Jetijs]

sykavy,
Peter already said once in this thread that an S shaped rotor did not perform very well in practice. In the video it is just an idea

Steven, I did not made the shaft by myself, I got someone else to do this for me so I don't know how he made the shaft and of if he dropped it on the floor or something. I already found another guy who can make a new shaft for me and I will instruct him to make the shaft the way you suggested, in one pass.
Thanks,
Jetijs

[lighty]

If you're re-doing whole rotor this time you can have smaller airgap without a need for sanding. It seems it was bent axle that was giving you hard time.

[Jetijs]

I agree to that, the bent shaft explains all those problems

[Jetijs]

Hello,
today I installed a 60v 10000uF across the power supply like Peter suggested some posts ago. Now I can freely change the output as I wish, either to a battery or right in the input section I did some measurements with two commutator wheels - one with just one 40 degree gap on each side and other with six small gaps on each side. I connected the output to the capacitor and measured the current drop. With the first commutator wheel, I observed a current drop of 36.2% (just as estimated before). Then I tried the second commutator wheel the current drop was 51.25%.
Now I will try to switch to a commutator wheel with only one 65 degree gap on each side and to flash the optotrigger LED's with my function generator to find the frequency with the best recovery. This type of MOSFET triggering will make the motor slower and less current will be able to flow through the windings, that is why we will have to increase the input voltage to get a decent current through the winding using the new triggering way. This will require some changes to the circuit. And as long as my new rotor is being made, I will experiment with this one
Thanks,
Jetijs