Transient Voltage measuring

[seyyah]

I'm trying to measure a transient voltage with dspic. Sensor output shows that the transient continues more than 500msecs. It starts with 300mV and decays exponentially to zero in a time of 500msecs. There is no problem with a continous signal but i can't measure this transient. I only get values at the very beginning of the transient (at the first several tens of usecs.) If i try to measure in the msec region i get zero. According to the values i get, it seems that voltage decays to zero in several times of ten usecs. But sensor output on the scope view shows that it continues for atime period of ~500msecs. What may be the reason? Why can't i get correct results?

 

[seyyah]

An addition:

Multimeter also can't measure the transient. And when i measure with the scope in dc coupling mode, it verifies my meaurement with dspic. But in ac coupling mode, it doesn't. And in theory it must be like in the ac coupling mode.

I added this because it may help.

 

[artem]

try to increase decoupling capacitor (between input and ADC ) to match requested timing .
You signal conditioning circuit should pass about 1 Hz just to catch transient . For precision measurement you should decrease it much lower than 1 Hz.

 

[seyyah]

There is no capacitor between the output of the sensor and adc if you mean.

 

[artem]

So i get you wrong when you said scope in ac coupling measure.)

What is the sampling rate of your ADC ?
What do you see from scope when probing dspic ADC input directly?
Is it same signal (same form) as on sensor output ?
Do you use analog low pass filter before ADC and what it is order and cut frequency ?
Could you post your schematics with exact component values ?

 

[seyyah]

So i get you wrong when you said scope in ac coupling measure.)

What is the sampling rate of your ADC ?
What do you see from scope when probing dspic ADC input directly?
Is it same signal (same form) as on sensor output ?
Do you use analog low pass filter before ADC and what it is order and cut frequency ?
Could you post your schematics with exact component values ?

At every ~11usecs a sampling/conversion is completed.
The signal which i posted and described is the waveform on the ADC input.
Yes i use lpf but before the sensor which is first order and cut off frequency is ~700Hz. The output of the sensor has an opamp to shift the signal +2.5V (to make 2.5V sensor output equal to zero voltage in real) and output of the opamp is connected directly to the ADC.

I'm thinking in this configuration that, i must be able to measure both ac and dc or instant values. I've been able to measure both dc(not very pure), low and high frequency sinusoidal signals until now.

I'm confused at this coupling issue. Whay scope shows the signal so differently, i didn't understand. I was expecting only a 2.5V shift in dc coupling setting.

 

[seyyah]

I added two waveforms. AC coupling is as expected the other is not. Am i wrong anywhere?

 

[seyyah]

...

??? I'm creating the transient myself. There is no protection issue here.

 

[silvio]

I'm confused at this coupling issue. Whay scope shows the signal so differently, i didn't understand. I was expecting only a 2.5V shift in dc coupling setting.

Are you sure that your sensor output level never goes to zero ?
Otherwise, after opamp conversion level, you'll get -2,5v and scope shows what found at his input. If for 2,5v at sensor output you get 0v at ADC input, then for 0v at sensor output you'll get -2,5v at ADC input; obvious if the opamp has -Vcc rail.

Regarding Ac coupling, the scope shows correctly as long as his input capacitor is charged and discharged at signal edges.
During the idle states of +2,5v and -2,5v the scope shows 0 which is correct in AC coupling mode.

Don't expect to measure the transient voltage during 500ms. As I told you, this is due to scope running in AC coupling.
In fact the voltage above +2,5v at slope edge is very short, even less than your 11usecs sampling/converson time.
So you won't be able to see it. Don't be cheat by the AC coupling scope showing a transient of 500ms.

 

[artem]

DC pulse at sensor output is about 3 seconds duration and is longer than scope's internal timing in AC coupling mode. Therefore you get such transient (from scope in AC coupling mode) which does not actually exist on sensor output.

You said :
"According to the values i get, it seems that voltage decays to zero in several times of ten usecs".

It is a square pulse fall part you measure by ADC, has about several times of tenth usec falling time duration.

 

[seyyah]

Thank you. You're right, i've mistaken. The problem wasn't with the measurement. It was the excitation. I'm modifying my question. From the beginning; i'm trying to apply a voltage pulse to the three phase inductive load via an igbt based inverter. My first attempt was wrong i think. The test configuration is shown in "trans_test_1.JPG". I first made A=B'=C'=ON as seen in "trans_test_2.JPG" . After the load is saturated i switched off the A, as seen in "trans_test_3.JPG". This was an error. I'm measuring U-V voltage. If we assume switches are ideal, i get zero voltage reading. I made this error because i forgot the presence of the diodes. Nevermind, what must i do to achieve my purpose? What must be the states of the switches before and after? I tried a second way which is shown in "trans_test_4.JPG" instead of the operation held in "trans_test_3.JPG" and i think the current path after the signal is cut down, is shown in "trans_test_5.JPG". This method gives accurate results in simulation but in real i couldn't verify it. The result is different than the first method but it still fades very quickly, i think. Please help me.

 

[artem]

Could you consider using thyristors instead of igbt's?

Added after 22 minutes:

or any switch (including igbt) connected through diode bridge (assuming control signal levels are properly applied)

 

[seyyah]

You did understand wrong. IGBTs exist. I didn't show them to indicate that they are in OFF position and diodes are in charge.

 

[artem]

i know that igbt do exist . You have a problem when coil is released after saturation and induced voltage gets reversed polarity and protection diode in igbt conducts induced voltage - that is the reason you get fixed square 400 mV pulse (voltage drop across igbt protection diode). You can avoid this effect if igbt will be put into diode bridge or single diode as in picture . d7 is the protection diode in igbt , it is not show in left variant .

P.S. idea to use thyristors is not appropriate as you have a DC . That was my mistake .

 

[seyyah]

To artem:

Either you didn't understand me or i didn't understand you. 400mV pulse is the pulse which is applied to the winding. What important thing is the second phase of the operation when the signal is removed. My first operation was wrong but i think the second one is right. I also couldn't understand your last post. Can you clarify it more?

 

[artem]

May be i get you wrong .

Could you be more specific?

if pwm switches in bridge are turned ON ( as no any winding has floating connection) you can not get the result different than range :

from Vdc - Vsat of IGBT till Vdc + Vdrop (drop on forward biased freewheeling diode builtin in igbt) when upper PWM bridge is ON ,

and from -Vdrop(same as above) till Vsat of ON igbt if lower PWM bridge is ON.

Said is true for one particular winding voltage measured relative to ground.

If measuring voltge between 2 windings , voltage will be sub of combination of above voltages .
Is this the part you need to measure ?

Or do you need to measure winding voltage transient during PWM switch deadtime ?

Or do you need to measure Vsat drop voltage on igbt and roughly estimate winding current derived from igbt saturation voltage ?

Or induced motor voltage or "back emf"?

I suppose yo are about to control 3 phase AC motor and trying to implement sensorless handling (from your last posts).

 

[seyyah]

May be i get you wrong .
...
Or do you need to measure winding voltage transient during PWM switch deadtime ?
...

I'm trying to do a similar thing. But not during dead time. I'm applying a voltage pulse to the winding for a time period (here 3secs), then removing it. Then i'm watching the transient voltage on the winding. During applying voltage, current flows through U windings and reaches to ground through V and W windings. Since B' and C' are ON, V and W are in parallel and they are in series with U. So total winding impedance is 1.5 times of the one winding. Is this OK? Voltage is measured between U and V. (Since V and W are in parallel, we can also measure from U and W). So the question is after applying this voltage pulse, what must be the switch states that allows the proper measurement of the transient on the winding. In my first attempt i made a mistake and kept B', C' ON after pulse is removed (by making A and A' OFF). In my second attempt, i switched OFF all switches to watch the transient and showed the current path in the "trans_test_5.JPG" file. Do you understand now? I'm asking now, is my second attempt true?


Freewheeling operation in "trans_test_5.JPG" seems right but i couldn't still understand how B and C diodes begin to conduct? They seem reverse polarized. So i'm a bit confused :roll:

 

[artem]

OK , i got the question .

Are you sure that B and C are conducting ?
Suppose you switched off all switches (after applying 3 seconds duration voltage to windings through A' to B and C ). The induced voltage in windings will have a reverse polarity (i dont take transformator effect in windings into account)then U get positive and V and W negative voltages . So it seems, when all switches are OFF, A to B' and C' diodes are to blame for clamping .
You should get transient (not clamped) when induced voltage (sum of U and V or W) declines below level Vdc + 2*Vforwdiode.

Are you interrested to see very early phase of transient when voltage is much above threshold Vdc + 2*Vforwdiode?

Please also pay attention to fact that when all switches are off, U V W are floating and you should provide current path to get proper measurement .