-
EDIT 21.08.19: scroll to the bottom to read about the recent consideration on the topic.
Hi All
----disclamer----
first time posting on the blog and i am not sure this is the right section for this topic. Feel free to point me to the more appropriate section or to a post that already discusses the topic.
General / open question (Below more context is given):
Can anyone point me to documentation or suggest me some DC voltage sensing cirquit? I am working on a project where i need to measure a voltage of roughly 270V.
Design restriction:
- No microcontroller, all analog
- Low budget
- Accuracy of ~1% with the voltage between 270 and 320V
- Low power consumption
context:
I need to design a voltage sensing cirquit for a DC bus of 270V. Multiple AC motors and the respective driver are connected to the DC bus. The circuit needs to "regulate" the dc voltage by powering a "braking resistor" (Pnominal = 500W) in case that the dc voltage raises above 300V. The option that i found to be the best at the moment is the "typical application Circuit" proposed on the datasheet of the ACPL-C87 sensor with some sort of schmit trigger or comparator stage at the end to drive the Mosfet of the braking resistor (see the attachements).
My concern with this design is precision. To measure the voltage i first need a voltage divider which is already pretty inaccurate. The ACPL-C87 accepts an imput voltage of 0-2V. The variation of the DC voltage is about 10% (((300-270)/270)*100) and since Vin is proportional i have roughly 200mV to work with.
Now,to me, it sounds a bit unrealistic to have a reliable measurement in an environment with high AC Noise.
Should i not be worried? Is there a valid alternative?
if something is unclear(maybe i should delete the "if") just ask, i'll be glad to give further explanation.
Thanks!
-
A rssistor divider is as accurate as the tolerance of its individual resistors.
The 1% tolerance resistors are cheap as dirt nowadays, and you can also get 0.1% easily (at least in major countries).
Using exclusively 0.1% resistors, and by selecting the proper values, the resulting divided down DC sample will retain that accuracy level. -
There's no need for the ACPL-C87, as the circuit being switched, isn't isolated for the voltage being measured.
I would go for the TL431 controlling some power transistors, with most of the power dissipated by a resistor. -
I'd call it a bleeding resistor since the bus voltage is already providing the breaking.
You can use a TL431, but it will need a few auxiliary components. How about something like this? Switching won't be very fast.
-
There's no need for the ACPL-C87, as the circuit being switched, isn't isolated for the voltage being measured.
you are definitly right! Don't know why I came up with the isolated OP amp.
I would go for the TL431 controlling some power transistors, with most of the power dissipated by a resistor.
I am not very familiar with the TL431, is the circuit proposed by Marco the same you had in mind? or would you propose something different? i'd really appreciate some schematicsI'd call it a bleeding resistor since the bus voltage is already providing the breaking.
You can use a TL431, but it will need a few auxiliary components. How about something like this? Switching won't be very fast.
I think i haven't done a good job at explaining the situation.
the AC/DC converter actually generates a DC Voltage of 540V which is then divided in two with capacitors in series. this gives me virtually 2 dc lines: 270-->"0" and "0"--> (-270). I then have several motor units on the "High"dc bus and the same number on the "low" dc bus. in normal operation every motor has the same mechanical, and therefore electrical, load. this keeps the bus and the "0" stable. so far so good.
it can occour that one or several motors need to stop for a while. This causes an unbalanced load and the vitrual "0" drifts. which leads the "high" dc bus to go up to 350V and the "low" one to go as low as 190V. Is there some sort of circuit used for this applications? does anyone have an idea on how to adress the problem or which approach should i take? -
Ah, that’s interesting. So you’re more interested in balancing than braking?
In principle you could use a synchronous buck converter to balance the upper and lower halves, but it’s a lot of work. It’s not very useful for braking the whole system because you can’t dump bulk energy.
Using two simple voltage-triggered dump resistors has the advantage of simplicity but wastes power whenever load imbalance occurs. If that’s infrequent, no problem. If it happens often and for a long time, you’ll be making lots of wast heat.
Dump resistors are also good for braking mechanical loads which might be helpful for your process control. -
Ah, that’s interesting. So you’re more interested in balancing than braking?
In principle you could use a synchronous buck converter to balance the upper and lower halves, but it’s a lot of work. It’s not very useful for braking the whole system because you can’t dump bulk energy.
Using two simple voltage-triggered dump resistors has the advantage of simplicity but wastes power whenever load imbalance occurs. If that’s infrequent, no problem. If it happens often and for a long time, you’ll be making lots of wast heat.
Dump resistors are also good for braking mechanical loads which might be helpful for your process control.
As you said, i'd exclude the buck converter since it seems a bit overkill.
the dump resistor seems an appropriate choice to me since the inbalance will only rarely occour.
Do you have any suggestions for the voltage sensing circuit? -
I would say, utilize a variation of Marco's circuit, but with hysteresis added.
Otherwise, the power Mosfet *could* be biased in its linear region, and would dissipate excessive amounts of power.
When I say *could*, is a circuit operational mode that requires to be verified experimentally. -
the AC/DC converter actually generates a DC Voltage of 540V which is then divided in two with capacitors in series.
Do the motors have grounded metal casings you don't want to have at high voltage when they are driven reverse or something? If you drive them with a full bridge you could just have a single DC rail.Quotedoes anyone have an idea on how to adress the problem or which approach should i take?
Here's a rethought version of the shunt reference based circuit. You should probably use some micropower shunt reference, not a standard TL431 ... for the cost of the circuit it doesn't matter, that will be dominated by the P-MOSFET, a IXTA10P50P is 4 Euro for a single.
PS. it already has hysteresis, that's what R5 is for. R5 would be relatively large relative to the resistor divider, so when the P-MOSFET starts conducting it pulls up voltage on the divider a tiny bit, creating hysteresis.
PPS. re-rethought. This circuit now puts more power on resistors rather than Q2 and the shunt regulator, this allows you to use higher current through the shunt regulator and a smaller gate resistor, speeding up switching.
-
You are correct about R5. Thanks for pointing it out.
-
That circuit looks like a good start. I would add an LED which lights when the dump is on, a flyback protection diode across the dump resistor (it’s likely to be inductive) and a bit of noise filter capacitor (maybe 1uF film across the rails?).
The trip point of the clamp should be set with care. You should do some calculations to see how high the DC link might get if your incoming AC supply is 10% high. You don’t want the dump load trying to pull the AC line voltage down, it’s wasteful.
Hysteresis is good, and can be quite large. The power MOSFET should switch at quite low frequencies (100Hz or slower is my guess). Indeed, seeing the LED go blink blink would be fine. You should also look at how much margin you get between clamp voltage and over voltage for the drives.
Finally, I recommend an earthed metal box for this one in case something goes bang. -
Many thanks to all of you!
i'll run some simulations, prototyping and then i'll come back and post what configuration worked for me.
Very much appreciate all your contributions! -
Soo.. I am back as promised.
A couple of months have passed and things have changed slightly in the meantime.
I succesfully built and tested the small prototipe board. The circuit works exactly as intnded. Perfect, Right? Not really...
Unfortunately, it was only once the circuit was running, that I realized it wasn't what I actually needed. The circuit I need has to accomplish a slightly different
task than the one I initially designed with your help.
The balancer doesn't have to dump energy while the dc voltage is higher than a set voltage since the total 540V are regulated by the main power supply.
What I need instead is a way to keep the virtual "0"(middle Voltage of the DC bus) from drifting.I think i haven't done a good job at explaining the situation.
the AC/DC converter actually generates a DC Voltage of 540V which is then divided in two with capacitors in series. this gives me virtually 2 dc lines: 270-->"0" and "0"--> (-270). I then have several motor units on the "High"dc bus and the same number on the "low" dc bus. in normal operation every motor has the same mechanical, and therefore electrical, load. this keeps the bus and the "0" stable. so far so good.
it can occour that one or several motors need to stop for a while. This causes an unbalanced load and the vitrual "0" drifts. which leads the "high" dc bus to go up to 350V and the "low" one to go as low as 190V. Is there some sort of circuit used for this applications? does anyone have an idea on how to adress the problem or which approach should i take?
I searched for such a solution and i found the following paper: Active Voltage Balancing of DC-Link electrolytic capacitors.
https://www.pes-publications.ee.ethz.ch/uploads/tx_ethpublications/ertl_IET-PE_2008_.pdf
The proposed topology (attachment "DC_Voltage_Balancing.png") in the paper is not exactly designed for my application. I substituted the BJTs with two pairs of MOSFETSs and added a breaking resistor in series (see attachment"DC_Voltage_Balancing_modified"). I tested this design and it works. The only thing is that I don't like the way it works. On the first design the voltage would reach a certain level, switch the MOSFET On and all the energy would be dumped by the Resistor. With the second design 30- 60% of the power is dissipated by the "Switching" element, which is not actually switching. In fact the mosfet mostly works in the linear region. While there is (almost) nothing wrong with that, i still feel unsatisfied by the design, because that's not what MOSFETs are meant to do or be used for.
Can anyone propose some improvements or even alternatives to this design?
Elia
-
The capacitors will be automatically balanced as the one with excessive voltage is bled, the one with a lower voltage will get charged from the rails and the virtual 0 gets pulled back to mid voltage. You just need one circuit for each capacitor.
You could create a version of the circuit I proposed which is more elegant that just copying it, but still dumps the energy into a resistor rather than the MOSFET. Leaving that up to you. -
Ah, I think I see what happened, eliafavero. Did you just build one voltage clamp that went across the whole (-270V) <= (+270V) rail? That wouldn't help with balancing.
I was thinking you could build 2 voltage clamps, i.e. one for the lower half (-270V) <= (0V) and the other for the upper half (0V) <= (+270V).
The ETH paper has a couple of nice features: few components and (hopefully) good tracking. They do good work at ETH (especially papers by Kolar). However, it will dissipate power in the transistors. This is fine if it's about balancing off capacitor leakage currents, which are quite low. If you have an unbalanced load, you could wind up dissipating a lot of power in the balancing transistor.
So I'm going to ask very clearly: are ALL loads connected across the full (-270V) <= (+270V) supply, with NO connection to the (0V) mid-point?
I suggest you answer in the form of a sketch of your complete system, showing (as basic blocks) what loads are connected and where. -
I clearly did a poor job at explaining myself.
Find attached the mock up that I should have posted since the beginning.
I'll try to re-formulate what i want to achive. During the operation of the whole system it can occour that some of the many motors has to stop. (There are arround 200 motors connected to the DC Bus). Since the load won't be balanced anymore the tension V+ -> "0" and "0" -> V- (will abbreviate to V+ and V- for semplicity's sake) will have a different absolute value. If just a couple of motors stops, the shift in load is negligiable. If instead, an important amount of motors stops, we can come to a situation where we have for example V+ =300 and V-=240. The electrolytical capacitor on each Driving board ("drive" on the mock-up) has a maximal voltage of 350V. so if V+ is allowed to drift above that value it's easy to see what happens. BANG!
Up untill this point the originally proposed circuit could be used to do the trick and limit the voltage across each cap. Of course, there is a BUT...
During operation we will mainly use the motors as "motor" and not as generator, BUT the ultimate goal of the project is to be able to operate the drivers as bidirectional so that we can, occasionally, generate power back. Assuming we have an equal number of motor spinning on both sides and all working as generator, the voltage on the DC Bus will raise equally on V+ and V- e.g. V+=320 , V-=320. Now you see that if we simply put a limit on the maximum tension of each rail we would dump precious energy that could be otherwise be kept and used to re-accellerate the motors. Therefore what i actually need is some sort of circuit that keeps the DIFFERENCE between V+ and V- inside a certain limit.
hopefully this claryfies my point. I'll be glad to answer further questions.
-
If efficiency matters the rail splitting is very counterproductive, you never did answer why all the motors couldn't be run from single rails.
If efficiency matters you should probably use DC-DC conversion for balancing.
If efficiency matters even when using resistive bleeding you should not try to preserve the mid point, you should start bleeding the higher voltage capacitor when the lower capacitor gets too low to operate the motors. Imbalance should not be an inherent problem assuming the motors have some kind of closed loop control, it only becomes a problem when the lower charged capacitor can no longer be charged from the rails. -
If efficiency matters the rail splitting is very counterproductive ...
If efficiency matters you should probably use DC-DC conversion for balancing ...
Efficency matters a lot for the project. The other thing that is even more important is the cost of the motors and drives. That is why we decided to split the dc bus so we could work with lower tension and have an easyer time designing the driving boards. Cost is also the reason why i am trying to avoid the DC-DC Solution.If efficiency matters even when using resistive bleeding you should not try to preserve the mid point, you should start bleeding the higher voltage capacitor when the lower capacitor gets too low to operate the motors. Imbalance should not be an inherent problem assuming the motors have some kind of closed loop control, it only becomes a problem when the lower charged capacitor can no longer be charged from the rails.
The goal isn't to preserve the exact mid point, infact we just want to limit the difference between V+ and V- by some amount (still to be determiermined). All the driving boards use a control loop to drive the motor.
Of course the goal isn't to operate a permanently unbalanced system constantly dumping energy. There is a central plc that comunicates with each drive and the plan is to stop the whole system if the unbalance persist after a set periode of time. this means that the balancing circuit will notify the plc when is dumping. -
I assume the only reason for the split rails is a motor voltage limitation + the desire to avoid a bulky transformer.
Even an autotransformer across the AC supply with a center tap to the DC 0V would keep the rails reasonably well balanced except during regeneration.
If you want to implement active balancing, drive a half H-bridge at exactly 50% duty cycle to switch one end of a large, high current inductor between V+ and V-, with the other end tied to 0V. If the voltages are balanced, the average DC current through the inductor will be zero. If they are not balanced, it will transfer charge to correct the imbalance. Its probably no more expensive than braking resistors as you are switching similar currents, but don't have to dissipate a lot of energy as heat. There are a few more refinements needed - it would be advisable to add current sensing to take the half H-bridge hi-Z if the inductor current approaches saturation, letting the current flow through the MOSFET body diodes until its decayed enough to restart active balancing, and if more than half the motor load on can ever enter regeneration at the same time, you, probably still need a braking resistor across V+,V- to dump energy into, that cuts in at say 10% overvoltage and drops out at 5%, as previously discussed. -
Even an autotransformer across the AC supply with a center tap to the DC 0V would keep the rails reasonably well balanced except during regeneration.
We are facing the unbalance issue exactly with the center tap...If you want to implement active balancing, drive a half H-bridge at exactly 50% duty cycle to switch one end of a large, high current inductor between V+ and V-, with the other end tied to 0V.
That's quite a clever idea! This is the kind of circuit i am looking for. Did you just come up with this or have you employed the technique somewhere else?
I forgot to mention that the regeneration will only take place in case of a blackout. At that point all the motors (which all have the same load) turn into generator and the energy produced is directly used to drive some auxiliary motors. The regeneration and the correlated voltage increase is something i still have to investigate on, for now my focus is to get a stable and balanced DC Bus.
Thank you very much!