Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
Kai Klaas
08/25/03 22:29
Read: 3394 times
Germany


 
#53375 - Grounding scheme
Responding to: Stanley Lio's previous message
Hallo Stanley,

I have thought about your application and drawn a schematic to explain in more detail how to ground, filter and shield. As I do not know much about your application I have made some assumptions, which may differ from your situation. I assumed the use of H-bridge topology in order to achieve forward and backward running of motor. Perhaps even some low frequency PWM is possible. What you will see now is, what I would start with, when being in your situation.





And:





Please note, that circuitry shows only one half of H-bridge, and only for one motor.
One major problem in discrete H-bridge designs is the possibility of short circuits of upper and lower transistor of one side of H-bridge. This can happen, when microcontroller runs ill. In the circuit above, there is a protection scheme arround CD4011 and CD4025, which prohibits this malfunction. Protection scheme guarantees, that only ONE transistor can be turned-on at the same time. Either the upper transistor or the lower one, but not both.
It works like that: H-bridge transistor is turned-on, when according Port pin goes low. At the same time, due to according inverter this signal also becomes inverted and is locking the NOR-gate of the other transistor. So, at the same time, only one transistor is getting fired.
An additional protection is achieved by the help of MAX1232, a really state-of-the-art reset controller with brown-out detection. At power up and power down, when microcontroller is not yet running properly, any activity is removed from H-bridge transistors, by having all NOR-gates locked.
Only if microcontroller is regularily running, driver unit gets valid signal.

Keep in mind, that MAX1232 has a watchdog which must be served by microcontroller, here by PortX.1 line.

Vcc of MAX1232 is low pass filtered by the help of 4R7 resistor. This will help in the case not using solid groundplane. But as I stated so often, not using solid groundplane shall NOT be recommended here!!

Have a look at 5V regulator: MAX1232, like many other reset controllers, need a certain fall-down time of Vcc. Means, if power supply is switched-off, the fall of Vcc from 4.75V to 4.25V need to last some dozends or hundreds of microseconds. Here, supposing a current consumption of 30mA, this period will last for about t = C x U / I = 100µF x 0.5V / 30mA = 1.7msec, so being well enough. You see, this 100µF/16V electrolytic capacitor is necessary.

Add some series resistor. Under worst case conditon voltage drop across this resistor should be smaller than about 0.2V, or so. Assuming 30mA, R = 0.2V / 30mA = 6R8 would result.

A 'strange' power switch circuit can be seen in the schematic. It provides some 'soft start' function. The reason is, that certain accumulators can suffer from high inrush currents. If you use such accumulator type you can use this arrangement of two switches: At first, S1 is switched-on, which causes a certain limited current to flow into electrolytic capacitor. After about one half of second, S2 must be additionally switched-on. R can be in the range of about 100R.

'Soft start' can be neglected in most cases, but when increasing electrolytic capacitor to rather high values and using sensible accumulators, it can be of benefit.

In the schematic a choke can be seen, adding some high frequency blocking. It prevents cable to accumultor from acting like an antenna, radiating high frequency noise 'coming' from microcontroller circuit. Keep in mind, that most 5V regulators have no line or load regualtion capability for frequencies higher than about 1MHz. So, high frequency noise 'travelling' on Vcc can reach input side of regulator, and finally, when resonating with inductivity of battery cable, heavy emission of radio frequencies can occur. Only a wide band choke should be used here, means some windings on a ferrite body. Use the best a 'six-hole' ferrite choke having maximum impedance at about 100MHz.

Please put an eye on star point grounding scheme: Even if you use solid groundplane all these drawn ground connections should NOT be done elsewhere!!! Everything drawn in the schematic has to be connected to 0V-pin of 5V regulator. That's the only way to avoid common mode noise. Also, you rest circuitry like sensoring, etc., should be referenced to this point.

Be aware that the whole circuit is floating relative to chassis (metallic parts of your robot). For using metallic parts of robot for shielding tasks, central ground point must get a high frequency connection to metallic parts. This is done via 10nF capacitor.
If whole electronics isn't directly connected to chassis elsewhere in your robot, you can also connect central ground point (MCU_Ground) directly to chassis. But don't make more than one direct connection of ground to chassis. Otherwise you get ground loops, which can introduce annoying hum, when your robot is near any source of changing magnetic field, like motors (!), transformers, etc.

You can see, that signals 'upper-left' and 'lower-left' are low pass filtered. This is done by routing them via 10k resistors to 100pF capacitors, which are located as close as possible to central ground point. Why is filtering so necessary?

Assume output of NOR-gate is toggling. Then output shows a ramp with rise time of only some tens of nanoseconds. But this very fast rise time is not necessary at all. H-bridge transistors will show turn-on times in the µsec range. On the other hand, fast ramp represents a very high content of high frequency power. It's rather difficult to transmit such a signal over simple cables, because unavoidable inductivity will cause resonances and ringing, which causes much of high frequency noise on ground return connections. The result can be a big desaster, because some parts can misinterpret noise as being a valid signal and whole circuitry can become running ill.

So, it's always wise to reduce bandwidth of signals to the relevant limit, by low pass filtering them. Here, this is done by 10k resistors and 100pF capacitors. Time constant formed by these parts is 10kOhm x 100pF = 1µsec. So, slew rate is limited to 2.3 x 1µsec = 2.3µsec. This is a quite acceptable value.

What would happen, if 100pF capacitors would be connected to ground somewhere else on PCB?

Some high frequency noise would be added and low pass filtering would no longer yield common mode noise filtering. Also, slew rate limiting will work less properly. Finally, output signals again become noisy and signal quality would be eroded.

Now we come to next schematic:
One methode to reduce interference when connecting cables carrying high frequent signals, is to reduce signal current running back over ground return. Keep in mind, that for a noise 'free' system current running over connection between MCU-Ground and Power_Ground should be kept as small as possible. This is especially true for high frequent signals!!!
So, not only slew rate limiting of output signal of microcontroller board is essential, but also limiting of current running from mcu-modul to power-modul. That's the reason, why directly at input of power modul 10k resistors can be seen. Additional 100pF capacitors produce some low pass filtering.
One disadvantage of rather high impedant signal routing is, that additional transistor stages are necessary to fully turn-on BD675/BD676 power Darlingtons, which are capable of handling motor currents of up to 1.5A.

From basis of first transistor to ground additional 10k resistors can be seen. They do two things: Increase of signal voltage needed to turn-on transistor to about 2 x 0.7V = 1.4V, which improves noise margin, and to guarantee low level, in the absence of any signal, may be because mcu-board isn't yet switched-on, or power board is tested separately. Low level causes turning-off of both H-bridge transistors.
If, on the other hand, mcu-modul is powered, but power-modul not, nothing bad happens, only base emitter junction is turned-on and limited current is flowing back to mcu-modul.

By the way, used Darlingtons already have free wheeling diodes across collector emitter path, connected internally.

Schematic also shows soft start 'circuitry'. Again, in combination with big elcos they can show some benefit. But in this motor application only high current accumulators will be used and soft start is probably not necessary.

Again, have a look at star point ground routing. Only by using this scheme mixing of high and spiky motor currents with controll signals can be avoided! Again, use of solid groundplane would show superior performance. But do NOT route emitter currents of BD675 over it!!

Be aware, that Power_Ground is NOT connected directly to chassis, but via 10n capacitor.

A simple motor filter is drawn in the schematic: 100nF directly across contacts, then some common mode filtering via two chokes and two 10nF. Choke should also be that 'six-hole' type from above. Keep in mind, that ground points of the two 10nF capacitors must be located as close as possible to motor contacts. So, in most cases you will be forced to connect them to metallic chassis of motor. This is completely correct, because metall chassis of motor is acting as a shield at the same time, and so, there's no better point for ground referencing of these two 10nF capacitors! Keep in mind, that capacitors loose it's benefit, when terminal leads are not kept as short as possible! Inductivity of terminal wire is about hundred times bigger than that of a massive groundplane!

From the scheme you can see, that motor chassis should be directly connected to robot chassis. That's trivial, of course. Connect third 10nF capacitor (the one between chassis and Power-Ground) near that connection. More likely to robot chassis than to motor chassis.

When would an opto-coupler be necessary?

If low pass filtering is not possible and if high frequent current through stray capacitance is so desastreous, that ground routing is heavily eroded. Then, sub picofarad stray capacitance of opto-coupler is of high benefit. Another way can be the use of symmetrical signal routing in combination with extense common mode filtering.
But here, choosen measures are enough.

How to connect both modules?

Best way is to use flat band cable. First line becomes upper-left signal, for instance, next line becomes mcu-ground, next line becomes lower-left signal, next line again becomes mcu-ground, and so forth. Keep in mind, that only ONE of these mcu-ground connected lines becomes connected to power-ground! Other grounded lines will act as a shield, heavily decreasing crosscoupling between signals.
Press cable close to robot chassis in order to decrease resulting loop area. And: Keep cable as short as possible!!!

What, when two motors are to be controlled, sharing the same accumulator?

Add a second set of parts in such a way, that concept of ONE star point Power_Ground is still valid. So, do NOT simply build a second board, with finally connecting the resulting TWO Power-Grounds 'somehow'. This can introduce heavy ground loops, totally eroding ground routing.
Only the concept of ONE star point Power_ground with connecting signals in the way described above will satisfy.

How to introduce additional shielding between modules?

Because all local grounds are connected via 10nF capacitors to chassis a simple sheet of metal located between them and directly connected to chassis, will act as a good shield. E.g., stray field originating from mcu-modul will no longer end at other unshielded modul, but end at inserted shield, because mcu-ground and chassis are at nearly same potential, especially for high frequencies.

Finally a software hint: After turning-off certain H-bridge transistor, wait for about 10...20µsec before turning-on other H-bridge transistor. This will reduce probability of short lasting short circuits between H-bridge transistors, due to overlapp.


Good luck,
Kai

List of 51 messages in thread
TopicAuthorDate
ALE---indicator...failed MPU      Stanley Lio      08/22/03 10:15      
   RE: ALE---indicator...failed MPU      Kai Klaas      08/22/03 10:18      
      RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 10:48      
   RE: ALE---indicator...failed MPU      Oleg Sergeev      08/22/03 10:20      
   RE: ALE---indicator...failed MPU      Andy Neil      08/22/03 10:55      
      RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 10:50      
   RE: ALE---indicator...failed MPU      Raghunathan R      08/22/03 11:11      
      RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 10:47      
      RE: ALE---indicator...failed MPU      Abhishek Singh      08/23/03 10:57      
   RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 11:04      
   RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 11:05      
      RE: ALE---indicator...failed MPU      Oleg Sergeev      08/23/03 14:12      
         RE: ALE---indicator...failed MPU      Raghunathan R      08/23/03 20:35      
            RE: ALE---indicator...failed MPU      Kai Klaas      08/23/03 21:43      
               RE: ALE---indicator...failed MPU      Stanley Lio      08/23/03 22:53      
                  RE: ALE---indicator...failed MPU      Abhishek Singh      08/24/03 02:31      
                     RE: ALE---indicator...failed MPU      Stanley Lio      08/24/03 12:12      
                  RE: ALE---indicator...failed MPU      Kai Klaas      08/24/03 07:41      
                     RE: ALE---indicator...failed MPU      Stanley Lio      08/24/03 12:31      
                        RE: ALE---indicator...failed MPU      Raghunathan R      08/24/03 18:57      
                        Grounding scheme      Kai Klaas      08/25/03 22:29      
                           RE: Grounding scheme      Stanley Lio      08/26/03 09:43      
                           How to post pictures?      Stanley Lio      08/26/03 09:59      
                              RE: How to post pictures?      Kai Klaas      08/26/03 14:09      
                              RE: How to post pictures?      Ian Forse      08/27/03 13:22      
                                 RE: How to post pictures?      Raj Shetgar      08/28/03 23:55      
                                    RE: How to post pictures?      Stanley Lio      08/29/03 01:35      
   RE: ALE---indicator...failed MPU      Steve M. Taylor      08/24/03 18:35      
      RE: ALE---indicator...failed MPU      Stanley Lio      08/24/03 23:33      
         RE: ALE---indicator...failed MPU      Raghunathan R      08/25/03 01:56      
   RE: ALE---indicator...failed MPU      Stanley Lio      08/28/03 09:11      
      RE: ALE---indicator...failed MPU      Steve M. Taylor      08/28/03 10:51      
         RE: ALE---indicator...failed MPU      Stanley Lio      08/28/03 20:54      
      RE: ALE---indicator...failed MPU      Kai Klaas      08/28/03 14:07      
         RE: ALE---indicator...failed MPU      Hans Van Pelt      08/28/03 14:34      
            RE: ALE---indicator...failed MPU      Kai Klaas      08/28/03 19:16      
   RE: ALE---indicator...failed MPU      Stanley Lio      08/28/03 09:31      
   RE: ALE---indicator...failed MPU      Stanley Lio      08/29/03 07:33      
      RE: ALE---indicator...failed MPU      Kai Klaas      08/29/03 08:50      
         RE: ALE---indicator...failed MPU      Steve M. Taylor      08/29/03 09:07      
            RE: ALE---indicator...failed MPU      Kai Klaas      08/29/03 09:40      
               RE: ALE---indicator...failed MPU      Steve M. Taylor      08/29/03 10:35      
                  RE: ALE---indicator...failed MPU      Stanley Lio      08/29/03 13:49      
                     RE: ALE---indicator...failed MPU      Erik Malund      08/29/03 13:54      
                     RE: ALE---indicator...failed MPU      Steve M. Taylor      08/29/03 15:37      
      RE: ALE---indicator...failed MPU      Stanley Lio      08/29/03 13:52      
         RE: ALE---indicator...failed MPU      Raghunathan R      08/31/03 09:52      
            RE: ALE---indicator...failed MPU      Stanley Lio      08/31/03 10:28      
            RE: ALE---indicator...failed MPU      Stanley Lio      08/31/03 10:37      
               RE: ALE---indicator...failed MPU      Steve M. Taylor      08/31/03 10:41      
                  RE: ALE---indicator...failed MPU      Stanley Lio      09/01/03 08:13      

Back to Subject List