OT: Analog Output Multiplexer

We have a possible application coming up where we will be controlling the pneumatic brake on 33 unwinds. We are going to do a simple diameter based brake control as the tension regulation specs are very open. The potential customer is asking us for the lowest cost idea we have. However, they want to be able to adjust all the tensions from a single location.

We are still toying with the idea of using an ultrasonic sensor measuring roll diameter through a pot to the analog input of an E/P. However, the thought of a panel with 33 pots staring at an operator doesn't sound too friendly.

One of the other guys here came up with idea to add a plc and HMI. Roll diameter would be measured using a prox on each unwind spindle relative to a prox on a known diameter positively driven roll. the time ratio between pulses would give you the diameter ratio between the two. We could do this very cheaply for 33 unwinds.

However, now we need to get 33 analog outputs to the E/Ps. this will get kind of expensive.

Has anyone ever come across a 1 to N analog output multiplexer, with N being any number between 4 and 32? We would like to have 12 bit resolution on this. It can be unipolar. As an alternative we could use a freestanding 1 to N DAC multiplexer board if it would accept higher voltage digital inputs. In this case we would limit it to 16 channels at 12 bit or 64 channels at 10 bits so we could fit all the interface on a single output module.

Any other ideas would be appreciated.

Thanks,
Keith

Use Profibus E/P's ?

Or how about using fail freeze e/p's and relays to switch the supply to each e/p and have one analogue output feeding all the e/p's. (maybe i/p's might be better here)

Originally posted by L D[AR2,P#0.0]:

Or how about using fail freeze e/p's...

Click to expand...

You've got me there. I've never heard of such a thing. That would work, though. Can you give me a supplier?

Thanks,
Keith

Has anyone ever come across a 1 to N analog output multiplexer, with N being any number between 4 and 32?

Click to expand...

Yes, we built a motor tester that had only 1 analog input. However the input was switched to (up to) 64,000 possible combinations, best I remember, of the 9 different motor leads plus ground. You might say it was a 1 to 64000 multiplexer, using the PLC relay outputs to switch the 1 analog input to all the possible combinations--sort of a harebrained idea to use PLC outputs to control where an analog input is connected. It used an Automation Direct DL-205 PLC with five Relay 8-Output modules and 1 analog input module, and worked very well. It only needed about 40 seconds to figure out the AC connections for all leads and diagnose any open, shorted, or grounded leads.

I don't know why you couldn't use the same method to switch an analog output (using enough discrete relay outputs) to different destinations. The method is to do the relay switching, wait a short time, apply the analog output to the new connection, then use the PLC to look at the results.

1 analog and 33 discrete outputs will be cheaper than 33 analog outputs.

Watson smith do a fail freeze i/p, I'm sure there must be others

http://www.norgren.com/ws/resources/F16099+422+Website.pdf

Lancie1-

We need to have the E/P or I/P maintain pressure when we are not refreshing it. So I actually don't want just a 1 to N multiplexer. I need a 1 to N multiplexer with a sample and hold on each output.

However, L D[AR2,P#0.0] idea to use fail freeze I/Ps may do the trick. I need to look a little closer at how long it takes to gain control of the output when the signal goes away. The Norgren unit simply closes both regulating valves on loss of signal. That will determine my refresh rate as the I/P needs to reach close to it's command level before I switch away from it.

Thanks guys.
Keith

Yes, I see that you need some way to lock the signal at the last level, or lock the signal converter at it's last level until the next refresh. That could be done with an extra output to an open/close valve downstream of the converter to only allow it to change during its turn at the trough. The fail-freeze converters would probably be the best hope, if they do not add too much more to the cost.

When a signal current is first applied to the instrument, there may be a short delay (a few seconds) whilst the capacitor charges, before control of pressure is achieved.

Click to expand...

Therefore you would have to leave the ouput switched to each I/P fail-freeze convertor for 3 or 4 seconds. It would take about 3 x 33 = 99 seconds to update all of them. If that is too long, then add 1 more analog output to cut the time in half.

However, they want to be able to adjust all the tensions from a single location.

Click to expand...

Will any tension levels be the same? If so, that would reduce the complexity. You could send the same signal to all units that need the same tension, reducing the total multiplex time.

Another possibility to reduce the update time would be to step the signal, starting out at 20 mA to charge up the capacitor in the fail-freeze, then once it is nearly charged, reduce the signal to the actual required level, wait for the tension adjustment, then switch to the next unwinder.

The possibilities for optimising system update rates and performance are interesting to say the least. For example, if two (or more) signals need to be reduced, but one by more than the other, you could reduce them both, turn one off and then reduce the other without having to wait for the unit to power up.

Will there be a signal to let you know that a new roll has been loaded into each unwind ? The last pressure will be the setting for a core and when a new roll starts up you would want to force an update to that unit.

Originally posted by Lancie1:

Will any tension levels be the same?

Click to expand...

We intend to design this to be completely freeform. This is a sandwich type product and the tensions in the various layers are different. Even if the tensions were the same we don't have any control over the diameters of the material they will put on each unwind. So identical tension settings wouldn't necessarily equate to identical brake torque.

Originally posted by L D[AR2,P#0.0]:

Will there be a signal to let you know that a new roll has been loaded into each unwind ?

Click to expand...

We will have them reset the roll diameter at the HMI. If they choose not to we will know the correct diameter within three roll revolutions anyway, which is probably fast enough for what they are doing.

However, I just got to thinking about how these E/Ps operate Maybe I'm better off going the other way. I use a P/E to read the actual pressure in the brake circuit. I use a small valve with an orifice to supply air to the brake. I just open the valve until the P/E is satisfied. If I really want to get fancy I use sliding mode control. If I don't like the cost of all the analog inputs, N to 1 multiplexers are much easier to find than 1 to N multiplexers.

Thanks for the ideas guys.
Keith

You know, E/P indicates a voltage signal. You can parallel a whole bunch of thes off a single signal if that is really what you want to do. If the total impedence of the parallel circuit got too low you could add some signal conditioners.

However, with an 8 channel analog output under $300 I'm not sure you coulddo all of the engineering etc. for $1,500.

Originally posted by Tom Jenkins:

You can parallel a whole bunch of thes off a single signal if that is really what you want to do.

Click to expand...

But I need to control them all independently. But as you indicate, the trick may be to look for cheaper hardware. With our 'vendor of choice' 33 channels is $5000, not $1500.

I think ultimately I need to pick my poison and drink. Go with the cheaper hardware that I don't know and take the hit on learning to work with it (as well as software costs) or use a platform I know and design a solution that fits my price range. It'll probably end up pretty close to a wash in the end, I suspect.

Keith