I still suggest physical placement of the two sections on the PCB should match their isolation functions. The SSRs have their isolation in the middle between second and third pins.
PS: The schematic could be a tad more readable, eg: Divide it logically. In fact I'd go so far as to say readability of the functional idea is the primary purpose of drawings like schematics. Things like netlists and auto-generate are just a bonus.
I still suggest physical placement of the two sections on the PCB should match their isolation functions. The SSRs have their isolation in the middle between second and third pins..
Ahh, Yea, I like it, that would be easy to do too. Thanks I will incorporate that!
Having now put you through all that and even though the INA219 has a wide common mode volt range I'm not completely comfortable with separating the grounds with the INA219 still in non-isolated position. I'm thinking your earlier single ground pour layout was superior to this latest. Your currents aren't all that severe and you do have wide pours to hold things together.
If you'd like to continue down path of electrically isolating power circuit from control circuit, which would certainly be good experience, then the next step is sorting out isolation of the current sense. The usual solution is data rated optocouplers which is usually just functionally dropped in-line, but because it interfaces with the processor via I2C serial it's a little more involved than the more basic SPI type signalling. A quick google produces this http://www.nxp.com/documents/application_note/AN10364.pdf
PS: The annoying part is the 3.3 volt (or 5.0 volt) supply rail for the INA219. This needs another regulator and smoothing and whatnot feeding from the motor's 12 volt supply.
I favor my current layout over the first as there is more isolation and I would like to make one to just see how it performs.
Just as an FYI I have a prototype of this circuit working and I have been able to get the noise out that was causing my reboot (see Post #20 in this thread) . Over last week I simulated a full year of motor cycles and not one single reset or problem of any sort. Before I made the changes it was rebooting after just a few cycles. Since the SSRs are optically isolated I did think about isolating the INA219 data lines but as you pointed out the two way communications that occurs over a the SDA line complicates that so I put it on the back burner. Its an option for the next round but for now I feel comfortable with this layout. I also like your idea of a dedicated power regulator for the INA219 but I don't think I need it at this time. I have to draw the line at some point as cost and complexity is a consideration. But I will keep it on the table for the second round if needed.
I favor my current layout over the first as there is more isolation and I would like to make one to just see how it performs.
Just as an FYI I have a prototype of this circuit working and I have been able to get the noise out that was causing my reboot (see Post #20 in this thread) . Over last week I simulated a full year of motor cycles and not one single reset or problem of any sort. Before I made the changes it was rebooting after just a few cycles. Since the SSRs are optically isolated I did think about isolating the INA219 data lines but as you pointed out the two way communications that occurs over a the SDA line complicates that so I put it on the back burner. Its an option for the next round but for now I feel comfortable with this layout. I also like your idea of a dedicated power regulator for the INA219 but I don't think I need it at this time. I have to draw the line at some point as cost and complexity is a consideration. But I will keep it on the table for the second round if needed.
Thanks for the input!!
To simplify current sense isolation it would make sense to give it its own itty-bitty micro which collects the data and transmits it as a aynch serial stream at regular intervals. That way you only need a single opto. I have used little PICs in the past but my new favourite is Silabs EFM8BB10F8G in a SOIC16 package for under 50 cents but there are so many to choose from.
Self-clocked bit-streamed ADCs also work cleanly with a single opto. No config/programming and can even be monitored with simple R-C filter and multimeter/scope.
Self-clocked bit-streamed ADCs also work cleanly with a single opto. No config/programming and can even be monitored with simple R-C filter and multimeter/scope.
Do you have part numbers? In fact I have used the internal ADCs of the micros I have used. However in this case I forgot about power although in the case of current sense it should be possible to regulate off that line.
Since you mentioned monitoring with an RC filter then does that mean that they simply output PWM? I suppose that would be accurate enough anyway especially if the opto used a strong pullup or logic output for symmetry.
Since you mentioned monitoring with an RC filter then does that mean that they simply output PWM?
Bit-stream (1-bit) ADCs/DACs are equivalent to PWM signalling but there is no defined resolution or cycle interval. The modulation cycle time is defined by a balanced charge/discharge of an integrator/differentiator in a tight feedback. And usually they stack extra stages for steeper slope tracking. So, fastest cycle times occur at centre balance 50/50 duty - which is one clock high. one clock low.
You can consider them as a seriously oversampled integrating ADC, sacrificing bit depth for increased sample rate. It's up to digital processing to recover whatever bit depth deemed desirable. Needless to say they aren't used for super fast ADCs nor DACs but are plenty good enough for the best audio work, due mainly, I think, to the no-aliasing feature.
One part that has been discussed on the Parallax forum a few times is the AD7400/AD7401. It has the isolation built right in, but, again, doesn't supply it's own power on the isolated side.
There is alternative non-isolated parts that can be piped through a single optocoupler but you probably won't find much cheaper. AD7720 is offered. TI offers many more of both. Two examples: AMC1305 is an isolated one targeted at current shunts. ADS1208 is a dual non-isolated ADC with additional constant current source. Two optocouplers would be needed for this, one for each channel.
As for the digital processing required, minimum required is a single gated counter. The Propeller can used it's counters for this, two per Cog, and it's no biggie to mux between modulation sources so more than two ADCs can be handled on a single Cog. Obviously, you aren't monitoring all bitstreams at all times then though.
Standard circuits for FPGAs and the likes have a more complex multistage counter setup (rule of thumb being one more counter stage than modulator stages) ... but I don't know how vital that is for optimal results.
Here is a video of the new circuit with object detection. In the video you will see several attempts to extend the actuator. The first two attempts are with super sensitive object detection (Stop if current is 7% over normal). The third attempt is with a slightly less sensitivity setting (I have to push a little harder to stop it.). After the 3rd attempt to close, the firmware stops trying to extend the actuator and waits for an override signal. When the override command is received it will do a full current close (no object detection) and as you can see I cant stop it by hand.
Just finished two weeks of testing with this new board. Works like a camp all the spikes are gone at-least my old scope cant pick them up and I haven't had one single reboot with over 1000 cycles of the actuator. The current sense circuit is working extreamly well.
Looks great. Very smooth & powerful actuator, where from? Do you have a guess what the most important changes were to solve your problem?
The actuator is a Duff-Norton LT25-1-300 and is a little jewel I have another one that has been in use outside for 3 years and it works great in both freezing and hot summer temperatures. It is a bit pricy for some projects at $278 so I will be trying some more affordable models to see how they compare. I would like to see one under $100 if possible.
As for the most important change that solved my problem I would have to say it was all software and had nothing to do with the new board. On the previous board that was having the problems I made a change to the firmware that took the spikes out. It was so simple. Since I'm controlling the motor with 4 SSRs I was just turning off the pair of SSRs that were driving the motor to stop it. That would cause a small spike on my power buss that caused the problems. I added a second command after the off command to turn on the two SSRs that would connect each of the motors leads to ground. At 80Mhz the propeller executes those two commands so fast it grounded out my spike.
However, I did want to have a good clean drive circuit as I want to use other actuators that may not be as clean as the Duff-Norton so I went ahead and made all the changes (I thought would help) suggested in this thread. And I think the two that helped the most are isolating the ground plane and the fly-back diodes across my SSRs in an H bridge configuration.
Comments
PS: The schematic could be a tad more readable, eg: Divide it logically. In fact I'd go so far as to say readability of the functional idea is the primary purpose of drawings like schematics. Things like netlists and auto-generate are just a bonus.
Ahh, Yea, I like it, that would be easy to do too. Thanks I will incorporate that!
If you'd like to continue down path of electrically isolating power circuit from control circuit, which would certainly be good experience, then the next step is sorting out isolation of the current sense. The usual solution is data rated optocouplers which is usually just functionally dropped in-line, but because it interfaces with the processor via I2C serial it's a little more involved than the more basic SPI type signalling. A quick google produces this http://www.nxp.com/documents/application_note/AN10364.pdf
PS: The annoying part is the 3.3 volt (or 5.0 volt) supply rail for the INA219. This needs another regulator and smoothing and whatnot feeding from the motor's 12 volt supply.
Just as an FYI I have a prototype of this circuit working and I have been able to get the noise out that was causing my reboot (see Post #20 in this thread) . Over last week I simulated a full year of motor cycles and not one single reset or problem of any sort. Before I made the changes it was rebooting after just a few cycles. Since the SSRs are optically isolated I did think about isolating the INA219 data lines but as you pointed out the two way communications that occurs over a the SDA line complicates that so I put it on the back burner. Its an option for the next round but for now I feel comfortable with this layout. I also like your idea of a dedicated power regulator for the INA219 but I don't think I need it at this time. I have to draw the line at some point as cost and complexity is a consideration. But I will keep it on the table for the second round if needed.
Thanks for the input!!
To simplify current sense isolation it would make sense to give it its own itty-bitty micro which collects the data and transmits it as a aynch serial stream at regular intervals. That way you only need a single opto. I have used little PICs in the past but my new favourite is Silabs EFM8BB10F8G in a SOIC16 package for under 50 cents but there are so many to choose from.
Do you have part numbers? In fact I have used the internal ADCs of the micros I have used. However in this case I forgot about power although in the case of current sense it should be possible to regulate off that line.
Since you mentioned monitoring with an RC filter then does that mean that they simply output PWM? I suppose that would be accurate enough anyway especially if the opto used a strong pullup or logic output for symmetry.
Bit-stream (1-bit) ADCs/DACs are equivalent to PWM signalling but there is no defined resolution or cycle interval. The modulation cycle time is defined by a balanced charge/discharge of an integrator/differentiator in a tight feedback. And usually they stack extra stages for steeper slope tracking. So, fastest cycle times occur at centre balance 50/50 duty - which is one clock high. one clock low.
You can consider them as a seriously oversampled integrating ADC, sacrificing bit depth for increased sample rate. It's up to digital processing to recover whatever bit depth deemed desirable. Needless to say they aren't used for super fast ADCs nor DACs but are plenty good enough for the best audio work, due mainly, I think, to the no-aliasing feature.
There is alternative non-isolated parts that can be piped through a single optocoupler but you probably won't find much cheaper. AD7720 is offered. TI offers many more of both. Two examples: AMC1305 is an isolated one targeted at current shunts. ADS1208 is a dual non-isolated ADC with additional constant current source. Two optocouplers would be needed for this, one for each channel.
Standard circuits for FPGAs and the likes have a more complex multistage counter setup (rule of thumb being one more counter stage than modulator stages) ... but I don't know how vital that is for optimal results.
Thanks all for the help!!
The actuator is a Duff-Norton LT25-1-300 and is a little jewel I have another one that has been in use outside for 3 years and it works great in both freezing and hot summer temperatures. It is a bit pricy for some projects at $278 so I will be trying some more affordable models to see how they compare. I would like to see one under $100 if possible.
As for the most important change that solved my problem I would have to say it was all software and had nothing to do with the new board. On the previous board that was having the problems I made a change to the firmware that took the spikes out. It was so simple. Since I'm controlling the motor with 4 SSRs I was just turning off the pair of SSRs that were driving the motor to stop it. That would cause a small spike on my power buss that caused the problems. I added a second command after the off command to turn on the two SSRs that would connect each of the motors leads to ground. At 80Mhz the propeller executes those two commands so fast it grounded out my spike.
However, I did want to have a good clean drive circuit as I want to use other actuators that may not be as clean as the Duff-Norton so I went ahead and made all the changes (I thought would help) suggested in this thread. And I think the two that helped the most are isolating the ground plane and the fly-back diodes across my SSRs in an H bridge configuration.