Subject:
|
Re: pneumatic cylinder: why not hydraulic ??
|
Newsgroups:
|
lugnet.technic
|
Date:
|
Sun, 11 Apr 2004 17:41:05 GMT
|
Viewed:
|
3557 times
|
| |
|
|
In lugnet.technic, Kevin L. Clague wrote:
> In lugnet.technic, Mark Bellis wrote:
> >
> > A XOR gate has two inputs, one with 1 switch and one with 4 switches. The
> > 4-switch one is a standard reverser and the 1-switch one has the air supply on
> > the middle and the two outer ports go to the input ports of the reverser (the
> > middles of switches 1 and 2 from the earlier description).
>
> Hmmmm....
>
> When talking about pneumatic gates in the past, I came to realize that different
> people defined gate inputs and outputs differently.
>
> Mark Tarrabain's single switch AND gate has two individual pressure inputs; one
> forces the piston to expand, and the other goes into the center port of the
> switch. There is one single port that provides the A AND B output. This output
> can make the next piston expand, but there is no provision to make the next
> piston contract.
>
> By pairing Mark's single switch AND gate with one of my two switch multiplexors,
> I can create a dual polarity AND gate. Dual polarity meaning that it has
> pressure outputs for each of the ports of the load piston being driven by the
> gate.
>
> Using a setup similar to the AND gate, I can create a dual polarity OR gate in 3
> switches.
>
> I can create a dual polarity XOR gate using 4 switches (basically just two muxes
> driven by the piston).
I class the input as the position of each bank of switches. A bank of switches
may be driven by one cylinder (1 or 2 switches), two cylinders (3-5 switches) or
more (6 cylinders in the case of 9 switches in my octopus arm logic). The
number of cylinders driving depends on the load of the number of switches in
each bank.
All my gates have what you call dual polarity - I always provide outputs for
both nozzles of a cylinder. I class the output of a gate as being the position
of the cylinder connected to the output hoses after a suitable time delay.
> >
> > An AND/OR/NAND/NOR gate has 2 banks of 3 switches. As I don't have pictures
> > yet, please try this from text.
> > Create the two banks on beams with the levers connected together (with 7-hole
> > beams) so that they're all at the middle position at the same time and travel
> > fully both ways together.
> >
> > Make the banks of switches have their nozzles facing each other. Number the
> > switches A1 (left bank, furthest from you), A2 and A3 (nearest) and B1 (right
> > bank, furthest from you), B2 and B3 (nearest), with nozzles U, M and L for Upper
> > (furthest away) Middle and Lower (nearest).
> >
> > Connect A1U to A2U and B1U to B2U
> > Connect A1M and B1M to a T junction. This goes to the top of the output
> > cylinder.
> > Connect A1L to B3U and B1L to A3U
> > Connect A2M and B2M to a T-junction. This goes to the air supply.
> > Connect A2L to B3M and B2L to A3M
> > Connect A3L and B3L to a T-junction. This goes to the bottom of the output
> > cylinder.
> >
> > Set both banks of switches to the lower position, pump some air in and the
> > cylinder should be contract.
> > Move only one bank to the upper position and the cylinder should remain
> > contracted.
> > Set both banks to the upper position and the cylinder should extend. This is
> > now an AND gate.
> > To get a NAND gate, swap the connections to the output cylinder.
> > To get a NOR gate, turn both banks of switches round or swap connections to the
> > cylinders that drive them.
> > To get an OR gate do both the above swaps.
> >
> > This gate has the same air leakage properties as a single switch i.e. it leaks
> > only the air of the side of the cylinder that is meant to be expelling air. If
> > you have both banks of the AND gate in the lower position, with the cylinder
> > contracted, and it is lifting something, and you then move one bank to the upper
> > position, the amount of leakage from the cylinder is only the amount in the
> > pipes. Therefore I suggest keeping the pipes short and/or using flex tubing as
> > much as possible. That way the cylinder won't move much when only one bank of
> > switches is moved. If it becomes a problem, drive a single switch with the
> > output cylinder of the gate and let this switch control the cylinder that does
> > the work. This is effectively a buffer.
>
> I'll try to build up some LDraw images of these gates so I can better understand
> what you are saying here.
I just doodled it again on a piece of paper before typing in the connections :-)
> >
> > To control the AND gate, you might need two cylinders per bank of switches.
> >
> > Looking at your article on pneumatics, I think this gate is what you need for
> > your D and E problem. I would suggest using the 3x5 bracket to hold the
> > cylinders that move the switches, and putting "1" joints on the top of the
> > switches, rather than using 3x3 brackets and longer levers on axle extenders.
> > These new switches mount lower down than the old ones.
> >
> > I have covered approximately half the work in your article. Where my work
> > diverged from yours is the point where the article seems to show that
> > non-leaking cylinders are added on if necessary, which may save on switches in
> > the short term but will cost you in the complexity of adding in leak proofing
> > later. I have always worked with the fundamental principle of no leaks in
> > steady-state, i.e. if a cylinder is not required to move the other way, pressure
> > is continuously applied to keep it where it is. Therefore all my logic building
> > blocks have been developed this way. For robots that have to apply permanent
> > forces to their leg cylinders, I would suggest a blanket no-leaks policy at the
> > start.
>
> I do not understand this point at all. Are you referring to the extra pistons
> added to circuit 7? I do not understand your concept of leaky cylinders, and
> leak proofing.
>
> If you always assume that a pressure *has* to be pressured, then your flip flops
> must be big. In the case where no outside force acts on a piston, you can
> create a flip flop out of one piston and one switch.
>
> Where ever possible I use a no-leak policy from the start. I wrestled with
> many, many prototypes of Quad242's circuitry before I settled on using two
> "leaky cylinders" to solve my timing problem.
>
> I'd be very interested seeing a Quad242 circuit that is completely no-leak. Are
> you game to give it a try?
Yes, I think it's circuit 7 - the penultimate one in your article where you say
you don't want 'A' to move twice in the cycle.
The irony is that a flip-flop uses fewer switches than anything else. It is by
nature a dual polarity system since all cylinders have both hoses connected into
the circuit.
I see that Quad242 has four legs that each have forward/backward and up/down
movement, that the F/B movement is all done together and that one set of legs
must be put down before the other is picked up, to avoid a gravity well. I'll
have a think about a circuit and post my ideas. Don't be surprised if it
initially uses more switches than your design!
The basic sequence looks very similar to a 3-variable octopus arm i.e. the
circuit goes 1-2-3-2-1-3, rather than 1-2-3-1-2-3 of an extended flip-flop.
Variable 3 is the F/B movement and 1 & 2 are the up/down of the sets of legs.
Watch this space!
> >
> > To add extra leak proofing to the AND gate, isolate the air supply by connecting
> > both the supply and the hose from the gate supply input to the middles of two
> > switches. Connect the right nozzles together and block both the left nozzles.
> > Only open the isolator (switches in the left position) when you are adding air
> > to the AND gate and isolate it when you don't want air to leak out if the output
> > cylinder is under load.
>
> I'll look at this.
Actually it can be done more simply, by using a single switch with the air
supply on the middle, the left nozzle blocked and the right nozzle being the
supply to the gate. Move the switch only between the middle and the left
position, by using a longer lever. The idea is that if another system hanging
off the air supply suddenly demands a lot of air, it would cause the pressure to
drop for all systems that were connected to the supply at the time. Isolating
the supply to each gate for the other systems prevents them losing pressure.
The pressure drop is a parasitic leak. I called it that because the system
demanding air is feeding off the pressure in the other systems.
> >
> > A half adder would use AND and XOR gates in electronic logic. Making a full
> > adder (two half adders and an OR gate) in pneumatics would use a lot of
> > switches. You would need two XOR gates, two AND gates and an OR gate, a total
> > of 28 switches with a leak-proof system. However, being a no-load system, you
> > don't need leak proofing, so you can save on the number of switches.
>
> As I said, I can get the full adder done in 17 switches, and three pistons.
> Using methods I still don't quite undersand, Mark Tarrabain has done it with two
> pistons, and a smaller number of switches.
I've not really gone in for logic larger than the basic gates. My aim is to
make robots that either move themselves or move something else, rather than
adding machines. However, if I knew the aim, I might have a go.
For any new design I'll work out what the sequence of the output variables is.
Then I'll see if an extended flip-flop will do it. For functions that all go
one way then all go the other way, this is the case. The other functions
require more logic that can be driven by the flip-flop. These other functions
can gate the flip-flop air supplies so that they are completed before the
flip-flop continues its cycle.
My pick and place robot uses an extended flip-flop, with three cylinders and
three switches. An exclusive-OR gate sits on the side of that by having its
switch banks added to the switches in the flip-flop. The air supplies to two of
the flip-flop control switches are gated by the output of the XOR gate, so that
the flip-flop waits for the XOR gate function to move. The three flip-flop
functions are elbow, wrist and grab and the XOR function is arm extension, which
moves in and out twice per cycle.
My hexapod circuit has a 7-way flip-flop for the main circuit and six lots of
two-way flip-flops for each leg movement. The leg flip-flops move a leg up
forward and down, and the backward movement is then gated by the main circuit.
The main circuit in turn is gated by each leg circuit. In this way the system
is made up of flip-flops, with all the legs moving backwards together at the end
of the main circuit cycle.
> >
> > My octopus arm logic has two banks of 4 switches, one bank of 3 and one bank of
> > 9! The first diagram had even more till I optimised the number of switches by
> > deciding which ones could handle two functions together without parasitic leaks.
>
> Again, can you please formalize the concept of parasitic leaks?
See above - my new comment explains this.
Mark
|
|
Message is in Reply To:
 | | Re: pneumatic cylinder: why not hydraulic ??
|
| (...) Hmmmm.... When talking about pneumatic gates in the past, I came to realize that different people defined gate inputs and outputs differently. Mark Tarrabain's single switch AND gate has two individual pressure inputs; one forces the piston to (...) (20 years ago, 11-Apr-04, to lugnet.technic)
|
38 Messages in This Thread:
 
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
