Roboteq und PMG

  • @heiko:
    Jo, der hat nur zwei 1405 pro Brückenzweig, das habe ich aber bewusst in Kauf genommen. Die haben da angeblich ein zeimlich intelligentes Stück Software zur Strombegrenzung im Einsatz. Bei dauerlast natülich....schaltet das Dingen wahrscheinlich ab.

    @ holger: genau das ist das Problem, würde das Dingen von Siemens oder rockwell etc. produziert werden, würde es wahrscheinlich das 5 fache kosten. Die Stückzahlen dürften net das Problem sein, da immer mehr Hybridautos und akkubetriebene automatische Flurförderzeuge im Einsatz sind. Nur sind halt dort Preis und Qualität in einer anderen Dimension.

    • Offizieller Beitrag

    jo... ich habe gerade mal in deren propagandadaten gestöbert...

    120A bei 40V... anlaufstrom für kurze zeit (was auch immer das heissen mag... ich denke mal so 1-2s) 200A...
    das ist ein vernünftiges konzept... wenn die endstufe das auch wirklich aushält :-/
    danach strombegrenzung auf 120A, mit drosselung auf niedrigere werte, wenn die temperatur der mosfets steigt... auch das richtige konzept...

    allerdings sind 55V-FET's in einer leistungsendstufe, die für 40V angegeben wird fast schon kriminell zu nennen...
    (jaja... osmc und kees machens auch nicht anders... schon klar... aber das ist eben auch eine grosse schwachstelle aller regler... fets mit niedriger spannung, um den ON-widerstand und damit die verluste gering zu halten... und die spannungsreserve der fets ist dann gerade mal so 10V... da krisch isch plaque, wenn ich sowas sehe...)

    tja... wenn man die strombegrenzung einstellen kann, würde ich die auf 60A runterdrehen... und beten!!!

  • Nicht ganz aus den Augen verlieren sollte man dabei allerdings, dass zumindest die PMGs nur auf 24V laufen, was die Reserve vergrößert.

  • Zitat

    Original von UnskilledWorker

    tja... wenn man die strombegrenzung einstellen kann, würde ich die auf 60A runterdrehen... und beten!!!

    Mit ersterem kann ich leben, Meine "Primärwaffe" is eh net Schubkraft und Geschwindigkeit.... =)
    Mit zweiterem....hilft das auch bei Atheisten ?

    Da ich gerne zwei unabhängige Meinungen hören möchte: was sagst Du zu holgis idee mit den Transsildioden?

    • Offizieller Beitrag

    @ing:
    stümmt... allerdings sind die 200A anlaufstrom auch ne menge holz... bei der niedrigen induktivität eines scheibenläufers ergibt das eine heftige stromwelligkeit während des schaltens... (die 200A sind nur der MITTELWERT!!! des stroms... nicht vergessen!!!) und durch den hohen strom der dann geschaltet wird, sind die spannungsspitzen recht hoch... da reichen dann bei schlechtem design der endstufe auch keine 20-25V reserve...
    ...und so kompakt und optimiert sieht mir das design nicht aus (kondis links, fet-brücke rechts, doch recht weit weg und keine abblock-kondis an den fets zu sehen...) als dass man vermuten könnte, dass die spannungsspitzen vernünftig gedämpft werden könnten...
    nun gut... ich weiss ja nicht, was die noch so an snubber oder anderen schutzmaßnahmen eingebaut haben... wollen wir also mal nicht zu hart sein =)

    @mech:
    na ja... versuch macht kluch sach ich da nur... allerdings halte ich mich da lieber etwas zurück, da ich den aufbau der endstufe und die eventuell vorhandenen schutzschaltungen nicht kenne...
    generell helfen die transsorbs natürlich... wenn auch nicht garantiert... ein regler der ohne die dinger stirbt, wird durch die schutzdioden auch nicht unbedingt viel sicherer... und ohne eine last-messung ist es recht schwierig, die dioden vernünftig auszulegen... immerhin wird die gesamte in den spannungsspitzen enthaltene energie in den schutzdioden verbraten... selbst, wenn die das kurzfristig können, sind sie nicht für dauerbetrieb ausgelegt... kann also durchaus passieren, dass sie wegbrennen... dann haste eine fehlerquelle durch eine andere ersetzt...

  • I talked to the guy from behemoth in liverpool, who had already blown 3! robotecs... (I don't know what motors he uses, but I think bosches)

    According to him, the problem was with braking. He could load the motors very hard, pull cars with the robot etc. without any problems. It was the braking when the controller would blow.

    This could indicate a few problems, but it's hard to say what is happening without knowing how the fets are switched. Possibilities:

    - regen voltage is rising above the maximum fet voltage
    - regen current is going through body diodes
    - motor is shorted on braking, causing energy to be dissipated in the fets. (The very low internal resistance of pancake motors would be a problem here, since this would cause most of the energy to be dissipated by the FETs instead of the motor).


    Installing transil diodes doesn't seem useful to me unless more info is available about the nature of the problem. The IRF1405 is pretty rugged and is fully avalanche rated, so I expect transil diodes won't help here, but it probably doesn't hurt either.

    The picture does show nicely that there is space for a second set fo fets, which would transform the controller in the ampflow controller. (Although I don't know where the second set of fets would get their cooling from.) This might be a solution to the problem as well.

    Niels
    Hammerhead

  • Interesting.
    If the controller tends to die in braking condition, this would be a hint for a reasonable design problem - won't be easy to solve, because a different switching strategy would be the only cure.

    Adding Transils is indeed not more but a kind of safety issue, if the problem should be Vds caused - which doesn't need to be.

    • Offizieller Beitrag
    Zitat

    - regen current is going through body diodes


    --> cannot occour, because generated voltage has to be greater than battery-voltage+2*voltage of body-diodes to let generated current flow back to battery...
    only way to get this is running the bot down a hill or push it faster than normal maximum speed... =)
    (normaly you have to turn a motor around 8-10% over nominal rpm to generate a voltage over the nominal voltage=battery voltage)

    surely this case can happen if you use the controller for example in an alectric driven car... but there are not to many hills in the arena this times... so dont worry about that... =)


    Zitat

    - regen voltage is rising above the maximum fet voltage


    --> cannot occour, because before this would happen the current would flow through the body-diodes (battery voltage is normaly lower than fet-voltage :-O )


    EDIT: @ helmut
    sag mal, kannst du die aufschrift der beiden übereinanderliegenden IC's (im bild 'links') lesen?

  • UnskilledWorker

    Regen current and voltage IS an issue. The motor just converts mechanical energy back into electrical energy, and when there is no current flowing the voltage will rise to very high levels. There is no need for the motor to actually turn faster than the normal speed for this to happen. Even an unloaded motor will put energy back in the supply when braking. (This can easily be measured)

    The actual voltage spike depends on how quickly the battery can absorp the energy, the inductance of the supply, and the breakdown voltage of the FET.


    The switching strategy is probably done in software, so it should be pretty easy to change. (When looking at the manual it looks like there is a configuration item for this, but it hasn't been implemented yet. The current implementation only seems to support brake, which should short the motor). If it is really shorting the motor than this could very well be the problem. Because the resistance of pancake motors is MUCH smaller than that of other motors (PERM PMG080 is 0.024 ohm, bosch GPA750 is 0.15 ohm), the amount of energy dissipated by the controller is also much larger. If there isn't any current limiting on the brake current, this could possibly kill the controller.

    I'm also not a fan of the HIP4081A driver, because it's very sensitive to spikes. I personally don't use them anymore because of this, but with a good layout and proper protection these devices should be usable.

    Niels
    Hammerhead

    • Offizieller Beitrag

    @niels:
    lets do some definitions first:

    i am assuming that the nominal voltage of the used motor is equal to the nominal battery voltage...

    1. regen current (regeneration current):
    this is the current which is running through the motor, when the motor works as an generator.
    if the motor is running (with external voltage applied to the motor) the only way to let it work as generator is to feed mechanical energy TO the motor...
    (this means trying to turn the motor faster than it would turn normally with the external voltage applied without feeding in extra external energy... )
    the voltage generated by the motor has to be greater than the externally supplied voltage to change from motor to generator mode.

    2. freewheeling current: (freilaufstrom)
    this is the current which flows through the motor when its turning and no external voltage is applied to the motor and a closed current path is available...
    in a H-bridge this means for example right low fet (F3) is ON, left high fet (F2) (was ON to let current flow through motor and right low fet (F3) ) is OFF and left low fet (F1) maybeeON (current flow through fet) or OFF (current flow through body-diode)
    at least one fet has to be ON (together with one body-diode) to get a freewheeling current and in this case no
    current is flowing back to the battery.
    if the generated voltage exceeds battery voltage + 2*voltage of body-diode a part of the current flows back to the battery, so freewheeling current changes to regen current...

    3. the voltage applied to the motor:
    there are two possible scenarios (F3 is ON in both):
    3a. F1 is OFF and F2 switches:
    there is no regen current until the generated voltage of the motor gets above battery voltage + 2*Vdiode (see above) (and while F2 is ON) but there is a freewheeling current
    3b. F1 and F2 are switched alternatly:
    at the motor terminal (between F1 and F2) a voltage (VS) can be measured with a MEAN VALUE!!! dependant on the duty cycle of the PWM...
    if VS is higher than the voltage generated by the motor (VM) the motor works as motor, if VS > VM the motor works as generator and a (regen-) current is flowing back to the battery...
    (by the way:VS is allways < VBat!!!)

    4. freewheeling braking:
    F3=ON, F2=OFF, F1 switched
    there will be a freewheeling current (value dependant on duty-cycle of F1), no regen current back to battery and all power is dissipated in motor and fets ... a very bad idea if one is using an H-Bridge ;)

    5. regenerative braking:
    lets assume the motor is turning (applied voltage VS dependent on duty-cycle of PWM for example 10V) and generating a voltage VM of nearly 10V.
    -if the bot is 'pushed' externaly VM gets higher and the motor changes to a generator --> current flows back to battery (which means that the battery voltage rises at high currents!!!)
    -if you lower VS under the generated VM the motor changes also to a generator --> current flows back to battery, this is called regenerative braking

    if VM is nearly VBatt(24V) (--> motor at full speed) and VS is rapidly set from nearly VBatt to a low value (for example 10V) there will be an enormous regen-braking-current... its possible to brake very hard, but because of the high regen current the current limiter normally should start to work... (it should stop further lowering of VS) limiting the brake force...
    in a simple design the current limiter does what it does when the current is to high in normally (motor) mode...
    it lowers VS... in regenerative braking mode this is very bad because lowering VS means regen current to rise which means VS to be lowered until it reaches 0V... which means the motor gets shortend (no PWM, F1=ON, F3=ON) and regen current changes to freewheeling current... normally the fets are dieing at this point due to the high current and dissipated power...
    maybe this is an problem of the RTQ... a current limiter which cannot differ between driving and braking...

    conclusion:
    with a working current limiter the only way to let the controler die is to push the bot so fast, that regen current lets the current limiter starts to work AND the current limiter sets the PWM to full duty-cycle (the regen current is then flowing through motor and body diodes F2 and F3) AND the motor is generating a voltage higher then VBatt + 2*VDiode...
    AND the current ( (VM-VBatt) / involved resistance) gets high enough to let the bodie-diodes die... this normaly happens before VS exceeds the breakdown voltage of the fets!!!


    this means in any case that the bot has to be pushed high over its nominal speed... this should normally not happen to a bot in the arena, but may happen shurely to -for example- a car running down a hill which is braked to the limit...


    Zitat

    and when there is no current flowing the voltage will rise to very high levels. There is no need for the motor to actually turn faster than the normal speed for this to happen. .


    if there is no current flowing, there is no regeneration and no braking... meaning the motor is free-runing thus generating a voltage dependent on its speed... it has to turn really fast until the generated voltage exceeds breakdown voltage of the fets... much faster then nominal rpm's...

    Zitat

    Even an unloaded motor will put energy back in the supply when braking. (This can easily be measured)


    sure... but not very much energy because there is not much mechanical energy stored in an unloaded motor...

    Zitat

    The actual voltage spike depends on how quickly the battery can absorp the energy, the inductance of the supply, and the breakdown voltage of the FET.

    the undamped voltage spikes are dependant of the inductance and the switching speed (dI/dt) of the current...
    higher switching speed or higher switched current let the spikes getting higher...
    the spikes are NOT dependant of the breakdown voltage of the fets!!! if the breakdown voltage is exceeded the fet dies...
    the breakdown voltage has no damping effect on the spikes!!!

    Zitat

    The actual voltage spike depends on how quickly the battery can absorp the energy


    the battery or the used (damping and/or storing) capacitors!!! normally the capacitors do the work... not the battery!!!

    Zitat

    The current implementation only seems to support brake, which should short the motor).


    VERY BAD design!!!

    Zitat

    If it is really shorting the motor than this could very well be the problem. Because the resistance of pancake motors is MUCH smaller than that of other motors (PERM PMG080 is 0.024 ohm, bosch GPA750 is 0.15 ohm), the amount of energy dissipated by the controller is also much larger. If there isn't any current limiting on the brake current, this could possibly kill the controller.


    i fully agree...

    niels, please dont feel attacked, but in my opinion it is important not to confuse the causes and results if one wants to understand what happens in a H-Bridge... :-O

    so... ich hoffe, das war auch für alle anderen einigermaßen verständlich... wenn nicht, möge es bitte jemand übersetzen... ich mach das jetzt nicht mehr... mir raucht gerade der schädel... deutsch wäre wohl doch einfacher und klarer gewesen...
    (da merkt man doch erstmal wieder, wie eingerostet sein englisch ist...) :engel:

    @ta152: der beitrag den du gefunden hast passt recht gut ...
    Gizmo:
    jo... deshalb begrenzt ein guter 'speedo' neben dem motorstrom auch den batteriestrom (kann aus dem gemessenen motorstrom und dem duty-cycle der PWM berechnet werden)
    Das ganze führt allerdings dazu, dass ein vernünftig arbeitendes system ohne überlastung nur eine ganz bestimmte maximale bremswirkung hat... leider... aber dafür ist es sicher...

    • Offizieller Beitrag
    Zitat

    ... deutsch wäre wohl doch einfacher und klarer gewesen...


    äh....... ja =)

    (Siehe GRA-Satzung §3 Pkt. 4 =) )

  • UnskilledWorker

    I don't feel attacked, I like discussing electronics :)
    I have always build my own speedcontrollers, so I know some things about speedcontrollers :) I wasn't very clear in my last mail. The regen voltage itself indeed won't normally be higher than the supply voltage, but because we are switching an inductive load there will be spikes much higher than that. The motor inductance is what makes regen braking possible at all.

    I can read german without problems, so you can answer in german if that is easier for you.

    The H-bridge is normally switched according to 3b in most controllers, as this will give you regen braking automatically when decelerating. When you are decelerating you will feed energy back into the battery, and the supply voltage will rise. This will normally not be a problem with a 24 or 36V supply and 55V fets, but a 48V supply won't be possible with 55V or 60V fets.

    The difficulty comes when you have to brake to a full stop. option 1 is to leave all fets off. This doesn't actually brake, but doesn't supply any energy to the motor either. option 2 is to turn either both lower side fets or both high side fets on, which shorts the motor. This breaks very well, but all energy will be dissipated in the motor and the fets. option 3 is to switch between 1 and 2 on the PWM frequency. This is what the OSMC and kees controllers use. This will actually give you regen braking because of the voltage spike generated by the motor inductance. The disadvantage is that the regen current will flow through the body diodes of the fets. The roboteq seems to use option 2, which might be a problem with motors which have a very low resistance.

    Most controller designs I know only measure the current in one direction. This means that in most cases there is no current limiting on the regen current back to the battery, or the current when the motor is shorted for braking. I think only the 4QD200/300 actually have currentlimiting for the regen current.

    I think we agree on most points, and a controller with proper currentlimiting should never die. I don't think the roboteq is such a controller though.

    bugs has some nice pictures of a blown robotec from newark on his site.

    Niels
    Hammerhead

  • @mechaniac

    There are a couple of things that can be done to minimise the chance of controller failure until this problem can truly be fixed:

    - use slow acceleration and deceleration ramps. This will reduce large peaks in current and will make life easier for the controller.

    - Mount the controller isolated from the chassis, and make sure the d-sub connector cannot touch the housing. Apparently the grounded d-sub connector can easily make contact with the housing, which can indirectly cause problems. That's apparently what happened to this controller (ouch): http://www.bugs.nl/media/album/vi…UdfMDQ0Ni5KUEc=

    According to the documentation the roboteq only measures the current every 16ms, which won't protect against a short on the output, or a stalled perm.

    Niels
    Hammerhead