If you change the positions of the 10k and 115k resistors, feedback will try to maintain 4.6V across the 100Ω resistor, and 4.6/100 = 46 mA. Not what you want!
Both the 0.4V reference and the 100Ω resistor are referenced to the positive supply. I should have noted on the schematic that the 0.4V is measured across the 10k resistor, not wrt ground. Negative feedback is connected to the bottom of the 100Ω resistor and acts to keep 0.4V across the 100Ω, and therefore 0.4/100 = 4mA flowing in both the emitter (source) circuit. The same current also flows in the collector (drain) circuit through the sensor. There is a small base current error if you use a PNP transistor, but not much if it is a superbeta, and none if it is a p-mosfet.
Here is how the feedback works. Suppose the current is greater than 4mA. Then there will be a drop greater than 0.4V across the 100Ω and the (-) input of the op-amp will be at a lower voltage than the (+) input. That causes the output of the op-amp to slew upward, which turns off the pnp transistor, which in turn lowers the emitter to collector current back toward 4mA. The stable condition is 4mA.
The amp and mosfet sink in ok, the divider not being wrt ground is hard to chew.
If the greater drop of 4.6 across the 115k is closest to ground, then the voltage would be kept up to 4.6 across the non-inverting input.
But maybe this will get us over semantics: If I replace the divider with a pot and dial in .4 volts to the non inverting input this would satisfy the requirement. Right?
And if so is the total resistance of 125k a requirement or would getting close to that value with a pot suffice?
Well, the divider, 10k & 115k, is hooked up between ground and the +5.0V regulated supply to divide that into 4.6V + 0.4V. Yes, the (+) input is kept at 4.6V above ground, but it is more meaningful to think of it as being kept at 0.4V below Vdd. It is not semantics. Feedback is comparing the voltage across the 100Ω resistor with the voltage across the 10kΩ resistor, and both are referenced to the +5 supply, not to ground. Feedback transfers the 0.4V from the 10K over to the 100Ω and in so doing regulates the collector (drain) current at 4 mA.
Yes, you can use a potentiometer. The value does not matter so much. Too high it may be subject to noise, too low it draws excess current.
If you were to use the 7.2V unregulated supply, instead of 5V regulated, you could could no longer use a resistor divider, and would have to include a reference diode such as the LT6650, to fix the voltage at 0.4V between Vdd and the (+) input.
OK, That put it all in Focus.
Now the sensor. In my circuit it is in series with the CC Diode.
Is that true for this as well or would it now be in parrellel with the diode?
This circuit replaces the CC diode. The sensor is the box connected between the collector (drain) of the transistor and ground. Am I understanding your question correctly? The CC diode is gone and the new circuit serves the purpose of regulating the sensor current at 4 mA, without temperature dependence.
Thanks again Tracy,
Yes, you understood. I think mind head has finally wrapped it's way around this circuit.
Now to order some of those reference diodes so I can keep the higher voltage.
Cheers,
You won't find a 0.4 V reference diode that can be used exactly as shown in that circuit. The LT6650 is a 0.4V reference+op-amp in an SOT23 package but it would need additional components to use in that circuit. If you want to use the unregulated supply directly, the easiest might be a bandgap shunt reference like the 1.2V LM385-1.2 in series with a resistor to bias it at about 100 µA. Then use a 300Ω resistor on the other side to give 1.2V/300Ω = 4 mA.
What voltage range do you expect across the sensor in operation with the 4mA bias? The dynamic range during the impact? That information is probably back in somewhere in the thread, but I'm too lazy to look!
If I run it from the battery(7.2 to full charge of 8.4) minus the reverse protect shotkey diode(which I think is 0.4) we get 6.8 to 8.0 volts on the sensor. The sensor at the positive end adjusts the voltage somehow(don't understand piezo electric too well) down to around 2.5 to 3 volts for 0 G's.
Then 500 G's(max) is plus 1 volt or 2mV per G
The compliance range of the current source circuit on a 5 Volt supply ranges from 0 Volts up to around 4 Volts, so it should not have any problem in holding the constant current over your 1 to 3 Volt range. I guess I don't see why you need to "order some of those reference diodes so I can keep the higher voltage". Wouldn't that be an complication? The important thing is to choose a good rail to rail op-amp. The buffered output from the sensor goes into an ADC and then into the Prop at about 57k samples per second--is that right?--And the impact signature might last something like 11ms total?
All true, I am baseing the need on the performance I got from going with the higher voltage(in the current configuration) and in the thread here where mentioned that a higher voltage was optimal. I think that was mentioned for use with the CC diode however.
I like simplicity, and will stick with regulated voltage if you don't see any problems. The fewer componants is certainly a welcome site.
If it doesn't work we could always try the more complicated circuit later, I guess.
The voltage for a supply of 5 is 2.5 volts for 0 G's and +/- 500 G's is 1.5 to 3.5 volts!
The cc diode needs quite a bit of voltage across itself before it really hits the constant current part of its curve, and even with the 7.2V supply it is on the edge, on a "soft" part of the curve where there is still quite a bit of dependence on both voltage and temperature. It's best performance would require a >15 Volt supply.
On the other hand, the circuit using the op-amp maintains the constant current using feedback. The current source is very stiff, as stiff and temperature independent as the 5 Vdd supply. The compliance goes all the way from a short circuit on the output (0 Volts) up to a limit at the top, which will be about Vdd - 0.4V - 0.6V = 4V. The 0.4V is dropped across the 100Ω resistor, and the 0.6V is allowed for the emitter to collector voltage of the pnp transistor. The rule of thumb is not to drive a pnp transistor too far into saturation where the base current error increases. Saturation is not an issue if you use a pmosfet. With a pmosfet, the compliance on the top end can go pretty much up to Vdd-0.4V. The pmosfet has to be a logic level type with Vgs threshold of -1 or -2 Volts with respect to Vdd. The op-amp is going to pull the gate down below Vdd to the point where the transistor turns on just enough to deliver 4mA.
Leon, I can agree that it is a good idea to simulate the circuit, yet it is almost as fast to breadboard it. I would certainly recommend doing that, and to try it with the impact sensor itself or a good simulated load. I've used that circuit 100s of times and have found it to be quite stable. Is stability your main concern? Capacitive loads are not a problem. There might be an issue with an inductive load. I am not sure how that impact sensor works.
P-mosfet it is then and I like the 5 volt source.
I have 1uf load caps at ever major chip on the board as opposed to one large one at the reverse protect in front of the battery. Do you think, being downstream from the 5 V regulator, this circuit should have a load cap as well?
@leon I do wish now, and in times past, I had been able to understand spice more than I do. It would often have saved some burnt thumbs.
My schedule right now doesn't allow for much study, so I agree with the breed boarding suggestion.
This is the mosfet I've been looking at, and since I have them in stock allready I thought of using the same op amp as I use for the voltage follower.
Would the mosfet here satisfy!
That vertical channel pmosfet should work fine. Vgs is the parameter to look at, also the second graph, "saturation characteristics". The op-amp will be able to pull the gate down to at least 4.6 Volts below the source, and that is well above the worst case 3.5V Vgs threshold, and also sits at a good point on the saturation curve for Vgs=-4V. What op-amp is it? It has to be able to go pretty much rail to rail on both input and output o the 5V single supply.
Unfortunately, while it is rail to rail at the output, the input common mode range of the ADA4851 on a 5V supply only goes up to 2.8V. Not enough to work at 0.4V below the 5V rail. This circuit requires rail-to-rail at both input and output.
In Analog devices product line there are plenty to choose from. For example, in the SOT23 package they have the AD8515, available from Digikey for about a buck, and it does rail-to-rail at both input and output, offset 1mV, and slew rate 2.8V/µS capable of driving the 50pF gate of the pMosfet.
OK, Circuit is built and in place except for the sensor. I am testing current before hooking it up!
Problem, With two different DMM's from mosfet source to ground I am measureing 40 to 41.5 mA, not 4.
Is breaking the circuit at this point not the proper way to measure just the current through the sensor. Or does the sensor have to be in place on one leg of the DMM?
The DS says greater than 4mA will damage the sensor, and at 350 bucks a pop I really don't want to go pop.:shakehead:
The AD8515 op-amp is rail to rail at both input and output, ~1mV input offset, so that should be fine. The ZVP1320 pmosfet is a vertical channel type with low Vgs, nice characteristics for this purpose. The pmosfet is used instead of the PNP shown in the diagram.
The statement that bothers me in your report is,
"Problem, With two different DMM's from mosfet source to ground I am measureing 40 to 41.5 mA, not 4."
Uhhh, however, the pmosfet should have its source connected to the 100Ω resistor, and the current should be coming out of the drain toward ground. I suspect the pmosfet is in backwards--The source-to-drain diode would be conducting and you would see about (5V - 0.6V)/100Ω = 43 mA. That is suspiciously close to your result.
Good idea to test it with the meter and breadboard before applying it to a $350 sensor! It is the same with laser diodes. Too many mA==PoP!!
All Right, Steady 3.93 mA....!
I appriciate the help on this one. Especially since folks want a demo of it tomarrow. I'll let you know how it went.
Thanks again.
just a thought... have you tried this new sensor WITHOUT the wireless interface? how much output power is the wireless pushing? possible energy/frequency RFI screwing with your readings?! Is the ADC/sensor/signal/power wires shielded from the RF source? Shielded twisted pair silver wire is the best bet for conducting miniscule voltage differences while rejecting outside EMI.
A capacitor acts as a high pass filter when used in series R/C network and a low pass when used in parallel R/C network. Capacitors connected across the power supply act to couple stray AC on the power supply rail to gnd, using a few different values of capacitance near every amplifacation stage is something i have seen in alot of high end tube amp schematics and uber high end headphone amp schematics. Has something to do with filtering different AC frequencies on the power supply rails more effectively.
OK, Things didn't go as planned, but I did finally make it to the lab today.
Started at ambient room temp. and dropped 5 times. Then stuck the head in the enviromental chamber for 40 min. At -4 celcius. It performed like a chorus line. No change in the readings where I would have seen a 20 to 25 G rise in the readings with the constant current diodes.
Tomarrow, we plan to get an ambient reading and then heat it up to the other end of the scale. I got my fingers crossed!
@RC If I where still seeing distortion in the curve of data representing the drop, I would look into that a little more. The capacitor change took care of the squiggles in the data that I was getting. Good point though, In fact, that has crossed our minds a few times. The wire is twisted, but not shielded other than being on opposite sides of the circuit board. The radio uses an antenna that is hopefully carrying all those issues outside the encloseure for us.
Well here is the results for the hot end.
From ambient to 110 degrees the device registered an 8.4 G drop.
Much better than the diodes, but I was hopeing it would totally eliminate the problem. What next to suspect, ADC, Regulators, Batteries?
The device was accepted as is, but I would like to know what the weak link is!
Thanks for the circuit Tracy, I'll be designing a board tonight for this circuit. It will upgrade the existing boards, and then I'll redesign the whole board to include this circuit.
My device was approved and used today. You will be happy to know that the Atlanta Falcons game tomorrow got a thumbs up from the unit and will go on as scheduled. And my corner of the galaxy has been saved
Comments
Both the 0.4V reference and the 100Ω resistor are referenced to the positive supply. I should have noted on the schematic that the 0.4V is measured across the 10k resistor, not wrt ground. Negative feedback is connected to the bottom of the 100Ω resistor and acts to keep 0.4V across the 100Ω, and therefore 0.4/100 = 4mA flowing in both the emitter (source) circuit. The same current also flows in the collector (drain) circuit through the sensor. There is a small base current error if you use a PNP transistor, but not much if it is a superbeta, and none if it is a p-mosfet.
Here is how the feedback works. Suppose the current is greater than 4mA. Then there will be a drop greater than 0.4V across the 100Ω and the (-) input of the op-amp will be at a lower voltage than the (+) input. That causes the output of the op-amp to slew upward, which turns off the pnp transistor, which in turn lowers the emitter to collector current back toward 4mA. The stable condition is 4mA.
If the greater drop of 4.6 across the 115k is closest to ground, then the voltage would be kept up to 4.6 across the non-inverting input.
But maybe this will get us over semantics: If I replace the divider with a pot and dial in .4 volts to the non inverting input this would satisfy the requirement. Right?
And if so is the total resistance of 125k a requirement or would getting close to that value with a pot suffice?
Yes, you can use a potentiometer. The value does not matter so much. Too high it may be subject to noise, too low it draws excess current.
If you were to use the 7.2V unregulated supply, instead of 5V regulated, you could could no longer use a resistor divider, and would have to include a reference diode such as the LT6650, to fix the voltage at 0.4V between Vdd and the (+) input.
Now the sensor. In my circuit it is in series with the CC Diode.
Is that true for this as well or would it now be in parrellel with the diode?
Yes, you understood. I think mind head has finally wrapped it's way around this circuit.
Now to order some of those reference diodes so I can keep the higher voltage.
Cheers,
What voltage range do you expect across the sensor in operation with the 4mA bias? The dynamic range during the impact? That information is probably back in somewhere in the thread, but I'm too lazy to look!
Then 500 G's(max) is plus 1 volt or 2mV per G
I like simplicity, and will stick with regulated voltage if you don't see any problems. The fewer componants is certainly a welcome site.
If it doesn't work we could always try the more complicated circuit later, I guess.
The voltage for a supply of 5 is 2.5 volts for 0 G's and +/- 500 G's is 1.5 to 3.5 volts!
On the other hand, the circuit using the op-amp maintains the constant current using feedback. The current source is very stiff, as stiff and temperature independent as the 5 Vdd supply. The compliance goes all the way from a short circuit on the output (0 Volts) up to a limit at the top, which will be about Vdd - 0.4V - 0.6V = 4V. The 0.4V is dropped across the 100Ω resistor, and the 0.6V is allowed for the emitter to collector voltage of the pnp transistor. The rule of thumb is not to drive a pnp transistor too far into saturation where the base current error increases. Saturation is not an issue if you use a pmosfet. With a pmosfet, the compliance on the top end can go pretty much up to Vdd-0.4V. The pmosfet has to be a logic level type with Vgs threshold of -1 or -2 Volts with respect to Vdd. The op-amp is going to pull the gate down below Vdd to the point where the transistor turns on just enough to deliver 4mA.
Leon, I can agree that it is a good idea to simulate the circuit, yet it is almost as fast to breadboard it. I would certainly recommend doing that, and to try it with the impact sensor itself or a good simulated load. I've used that circuit 100s of times and have found it to be quite stable. Is stability your main concern? Capacitive loads are not a problem. There might be an issue with an inductive load. I am not sure how that impact sensor works.
I have 1uf load caps at ever major chip on the board as opposed to one large one at the reverse protect in front of the battery. Do you think, being downstream from the 5 V regulator, this circuit should have a load cap as well?
@leon I do wish now, and in times past, I had been able to understand spice more than I do. It would often have saved some burnt thumbs.
My schedule right now doesn't allow for much study, so I agree with the breed boarding suggestion.
This is the mosfet I've been looking at, and since I have them in stock allready I thought of using the same op amp as I use for the voltage follower.
Would the mosfet here satisfy!
Since I have these on hand now I thought of using it. I am pretty sure it will do rail to rail, but haven't looked over the data sheet in a while!
In Analog devices product line there are plenty to choose from. For example, in the SOT23 package they have the AD8515, available from Digikey for about a buck, and it does rail-to-rail at both input and output, offset 1mV, and slew rate 2.8V/µS capable of driving the 50pF gate of the pMosfet.
Problem, With two different DMM's from mosfet source to ground I am measureing 40 to 41.5 mA, not 4.
Is breaking the circuit at this point not the proper way to measure just the current through the sensor. Or does the sensor have to be in place on one leg of the DMM?
The DS says greater than 4mA will damage the sensor, and at 350 bucks a pop I really don't want to go pop.:shakehead:
Going to bed.......
from this earlier post
http://forums.parallax.com/showpost.php?p=931548&postcount=90
The AD8515 op-amp is rail to rail at both input and output, ~1mV input offset, so that should be fine. The ZVP1320 pmosfet is a vertical channel type with low Vgs, nice characteristics for this purpose. The pmosfet is used instead of the PNP shown in the diagram.
The statement that bothers me in your report is,
"Problem, With two different DMM's from mosfet source to ground I am measureing 40 to 41.5 mA, not 4."
Uhhh, however, the pmosfet should have its source connected to the 100Ω resistor, and the current should be coming out of the drain toward ground. I suspect the pmosfet is in backwards--The source-to-drain diode would be conducting and you would see about (5V - 0.6V)/100Ω = 43 mA. That is suspiciously close to your result.
Good idea to test it with the meter and breadboard before applying it to a $350 sensor! It is the same with laser diodes. Too many mA==PoP!!
PS, Yes the circuit above is the one I'm using.
I appriciate the help on this one. Especially since folks want a demo of it tomarrow. I'll let you know how it went.
Thanks again.
A capacitor acts as a high pass filter when used in series R/C network and a low pass when used in parallel R/C network. Capacitors connected across the power supply act to couple stray AC on the power supply rail to gnd, using a few different values of capacitance near every amplifacation stage is something i have seen in alot of high end tube amp schematics and uber high end headphone amp schematics. Has something to do with filtering different AC frequencies on the power supply rails more effectively.
Started at ambient room temp. and dropped 5 times. Then stuck the head in the enviromental chamber for 40 min. At -4 celcius. It performed like a chorus line. No change in the readings where I would have seen a 20 to 25 G rise in the readings with the constant current diodes.
Tomarrow, we plan to get an ambient reading and then heat it up to the other end of the scale. I got my fingers crossed!
@RC If I where still seeing distortion in the curve of data representing the drop, I would look into that a little more. The capacitor change took care of the squiggles in the data that I was getting. Good point though, In fact, that has crossed our minds a few times. The wire is twisted, but not shielded other than being on opposite sides of the circuit board. The radio uses an antenna that is hopefully carrying all those issues outside the encloseure for us.
From ambient to 110 degrees the device registered an 8.4 G drop.
Much better than the diodes, but I was hopeing it would totally eliminate the problem. What next to suspect, ADC, Regulators, Batteries?
The device was accepted as is, but I would like to know what the weak link is!
Thanks for the circuit Tracy, I'll be designing a board tonight for this circuit. It will upgrade the existing boards, and then I'll redesign the whole board to include this circuit.