P Daniil

The motor construction and insulation should also be for use with an inverter . Non-inverter grade motor insulation and bearings will wear out quickly because of the long cable effects. Also, using a lower inverter switching frequency will not do much as it is the PWM waveform rise- and fall- time which cause the problem. Using a ruggedized drive (as BobLfoot suggests above) will also save you potential protective earth problems such as cable leakage causing fault trips and similar troublesome return currents.

The position drift you describe is typically caused by: 1. Loose mechanical coupling, backlash etc, and 2. Encoder dither. Our white paper looks at the problem and may be of interest.

A capacitive level sensor should do it. What kind/chemistry of "soapy" foam is it?

In addition to the excellent posts above, a DC motor driven by a four-quadrant SCR drive still has an advantage when braking: All the generated braking energy flows back to the supply line during normal SCR commutation offering practically limitless braking capacity (effectively driving the grid´s infinite inertia) as well as superior power efficiency (no energy having to be dissipated in a resistor as in the case with AC drives). So if the application involves significant braking energies, DC drives have still one thing going for them.

There are many ways false counting can arise. With the encoder still, noise can arise from: 1. Wiring ground loops (as kaiser_will above correctly pointed), 2. Common mode noise entering through the encoder power supply, 3. Faulty encoder output electronics, 4. Absent/wrong termination at the input card (with differential signalling), and 5. Faulty/latched-up input card. With the encoder moving, in addition to the above the most common cause is so-called dithering or a worn/damaged encoder optical disc. You may find our paper (pdf, 238K) of interest as it discusses these problems.

Most likely an external (SSRs rarely have them inside) RC snubber is needed to allow the SSR thyristor/triac to turn off graciously. The snubber RC values must be so calculated to limit the rate of voltage rising (dv/dt) at SSR turn-off for the given load (ie the relay coil) inductive current. SSR manufacturers usually provide typical values for typical loads and some of them offer ready, off-the-shelf snubber networks. Also keep in mind that an oversized snubber circuit will leak unnecessary current to the load which may prevent the relay/contactor to open.

Can't you use a small Voltage/Current Interface? (Potentiometer dividing the 24 VDC source to drive the interface to drive your 4-20 mA input). In this way you do not have to worry about linearity or potentiometer values.

Assuming this is an AC induction motor, as long as the V/f (volts to frequency) ratio is the same, you are OK. So as 480 x 50 / 60 equals 400 there is nothing to worry about the line supply. The motor however may not be able to develop the same power (ie increase the current to develop the extra torque needed) as this depends on the motor construction, power, slip and the load. The motors we get on this part of the world usually carry rating plates with 50 and 60 Hz sets of values. Is there a rating plate still on this old machine? If so, it may be of help. As a final check and to be on the safe side, I would drive the motor with an inverter at 400V/50Hz and see how the whole setup behaves. What is the power of the motor?

My approach would be first to have two universal, galvanically isolated, wired-OR connected signals, "Fault" and "Ready", driving all the devices (PLCs, supplies, I/O etc) so that when any supply/device has a problem (load fault or source problem) it will signal the others to shut-down/wait graciously (as opposed to chattering to death). This arrangement has the advantage of enhanced robustness as you have one signal that there is an operational fault and a signal that all is working OK and being able to handle the four combinations accordingly. In more complex set-ups (such as in a ship) where you have separate power and control supplies it may also be worth having respective "Ready" signals for each line. Following this you can then look for suitable surge/UV/OV devices/controls.

I have not seen or heard of a bad Linux distribution. We started switching to Mandriva on all our PCs a few years ago and have never looked back since then. We chose Mandriva then because we thought that our learning curve would be flatter with this distribution at the time. (In retrospect all effort went to learning the Linux fundamentals and way of thinking and not so much on the particulars of the distribution). Most of the machines are set-up as dual-boot to run existing Win32 applications (mainly CAD and accounting) which we have not yet managed to run under the Linux WINE environment. Blue screens are now distant memories to be told to grandchildren like black and white TV.

Edit: Good luck, please keep us posted.

Most probably they got partially damaged during the "power problems" and as a result could not supply their loads with the necessary current (and hence the 5-10 Hz power-up "hiccup"). Most chances are that their dc-bus electrolytics got "savaged" from the line fault (typically high voltage 300-400 V inductive spikes riding on the 230 VAC line) as they are the most vulnerable part in a power supply. With a partially working electrolytic the power supply can only store and convert minimum/unloaded power. My further guess is that the line neutral became unstable during the coupling/uncoupling of the shore power and, just as it is always good practice, I would regardless power the PLCs through a 400/230V transformer and stay away from the neutral. Most faults involve the neutral and in a ship the presense of the hull (also at neutral) makes matters worse by offering a catch-all connection to the line.

There is no black or white rule here as it all depends on the particular encoder construction (isolated body, earthed common etc) and the frequency of the signals you are trying to protect. With RF (radio frequency, above 1 MHz) signals "plane grounding" techniques must be used as connection inductance (to ground) comes to play. In effect you ground the shield at as many points as possible to ensure a minimum inductance connection to reference (the "ground"). At lower frequencies, wiring inductance is manageable and "single point grounding" techniques must be used. By grounding at a single point you ensure that no current flows through the cable/body shield (and in effect turning it to a noise generating/receiving antenna) and that no noise-generating "ground loops" are formed. Grounding should ideally be done at the receiver or weakest signal end to ensure minimum noise at the input processing stages. In industrial electronics most situations do not involve RF signals and single point grounding is preferably used as it does not interfere/mix with any protective/safety earth (unless intented). So the way to go is to take into consideration the manufacturer instructions (the encoder construction), the signal frequencies involved and the overall wiring/cabling layout to judge how noise is generated and conducted/transmitted. I would start with the manufacturer recommendation if their example/model is the same as the installation at hand (in theory they have checked it and works OK) or with "single point grounding" and make the final decision in the field. Regardless, a tidy and clean installation with good ohmic connections is a must before looking at any noise problems.

As it saves IP related hassles, I would call it something like "Beyond IPs" or the more tongue-in-cheek "Forget IPs".

It depends on the physical shape and layout of the capacitor system. For such a system I would suggest an insulated rod type hanging immersed in the monitored fluid. We have succesfully used such sensors in potable water as well as ship grey water tanks. Capacitive sensors are less sensitive/suited for oil detection due to the lower dielectric constant of oil.

What kind of fluids are stored in the tanks? There are many other types of sensor you can use. For such an application I would first look for a capacitive type sensor.

We have agonized (yes, agonized) on the issue of membrane fronts sometime ago and here is our experience from this part of the world: 1. What you pay is what you get. The faces are printed on the inside and to prevent inside wear/fading takes some extra steps and materials in their production. At a price you can also get mil-spec and vibration-resistant ones for >1 Mops (operations). Best value-for-money in our case proved to be four-color plus 100 Kops switch devices as this (anyway worst-case) figure outlives under normal industrial use other parts in the product such as the supply electrolytics. 2. As we produce in a high labour cost country we have also chosen this type of face as it saves design and production time and is less messy in assembly. 3. An additional plus is that they can be ordered/produced in small and large batches, ie their marginal cost sensitivity is manageable. Regarding price my non-expert view is that although nobody likes to pay more, your market is technical/engineering people who can understand/appreciate good craftmanship and materials so there should be no problem in charging the extra for the better quality materials and improved longetivity. Afterall this is a time saving device promoted in a high labour cost and technology intensive market. Good luck in your efforts.

If possible, I would detect current, to the VFD or the load, as a measure of the action taking place. Most probably this is already available in the VFD as part of current monitoring for overloads/faults and can be used to signal the PLC for good/bad operation via a VFD bit output. If such a facility is not available in the VFD then you will have to measure current to the VFD with an external current transformer at the PLC (and unavoidably use additonal bugdet/space resources). An additional benefit of current detection is that you also detect power train problems such as a broken/loose belt in your case. Edit: Another less accurate alternative is to pick fan-blade movement with a proximity switch to estimate fan speed.

My guess (I am completely unfamiliar with UL standards) is that this is to prevent primary-to-secondary fault current not flowing to protective earth but escaping outside the supply enclosure. On this part of the world, a double insulation transformer with physically separate (ie each winding on its own section, separated by a barrier) primary and secondary is usually specified for such a case. It is also a requirement in consumer 230 VAC devices such as small battery chargers and supplies. Depending on "duty", the specs require a minimum insulation strength and primary to secondary "creepage" (words in "" per IEC terminology).

Put Safety first! Here is a basic checklist: 1. Ensure that the crane will not travel past the track ends by using end limit switches which kill the power circuit directly. 2. If more than one crane is on the same tracks use an anti-collision system. 3. Alarms should be audible AND visual. 4. Ensure that the hoist drive is a dc-injecting type. 5. Do not allow lifting/hoisting of heavier-than-specified loads by using an overload alarm device. 6. Ensure that the left and right motion motors are connected in parallel and driven by the same inverter. This makes the crane stay on the tracks. When one side advances more than the other (remember that the wheels are not exactly the same and the tracks are not absolutely straight) one side slows down and the other speeds up to maintain phase synchronism.

As a lot depends on the surrounding machinery and wiring, there is no black and white rule here. The sometimes wrongly called protective "ground" (instead of "earth") is an additional source of confusion. This is really a safety net which should always be there and electrically have nothing to do with any machinery function. However as it surrounds all electrical equipment it is capacitively coupled to everything and so can be part of noise currents like the high frequency one generated by inverters. An additional problem is when the plant neutral and earth are used as one, typically in old factories. So, ideally one should not ground any isolated power supply potentials to maintain isolation and protection BUT by connecting to the earth one may divert noise currents from affecting susceptible devices. The most handy tool when investigating problems like this is a 100 - 1000 Ohm resistor. By connecting it between a supply potential and earth one can decide if it helps or not. The 100-1000 Ohms are relatively low for noise currents to flow and high enough to prevent destructive ground loops flowing via an earthing connection elsewhere in the plant such as from grounded equipment shields. Sometime ago such a resistor saved the day and a few red faces in a plastics factory where static was a big problem.

The MOV protects against over-voltages. It protects the coil insulation (when connected at the coil side) and the relay contacts and body (when connected at the contact side). In 24 VDC work the coil is usually more valnurable to over-voltages than the contact side. An anti-parallel diode across each DC-driven coil will conduct inductive current at turn-off and as vettedrivr suggested is the least you can do. An RC-snubber (a series connected resistor and capacitor circuit) across the contacts reduces the rate of voltage change (rise or fall) and so reduces/eliminates arcing when the contacts bounce at turn-on/off. It also allows for some leakage current to flow, but with 24 VDC circuits this usually is no problem. Ideally you should use all three methods. MOVs are selected according to clamping voltage and energy capacity. When selecting them check the data sheet DC clamping voltage (some manufacturers mark their MOVs for AC clamping voltage while others mark them for DC). The MOV diameter/size specifies the energy capacity and it is better to err on the larger side. The IN400x line is a typical choice for relay coils when the current is under 1 A. The RC-snubber values are calculated on the allowed maximum rate of voltage change, the driven coil current and the voltage (24 VDC in your case) and are not that critical for relay work. There are many ways (simple or more complex) to calculate them and a web search should provide a suitable method. You can either roll your own circuits or buy them ready. Most relay manufacturers offer sockets/bases/add-ons for their relays such as indicating LEDs, anti-parallel diodes, MOVs as well as RC-snubbers for various load currents.

PWM is generated by having a variable time interval (´pulse width´) triggered at constant frequency. So you need a constantly repeating event (say at T secs). Within this period you generate a pulse (with any suitable timer arrangement) which is proportional to the required power output (say t secs). The output power equals the ratio t/T times the full power of the heater. If the heater power is large and the system has large thermal inertia you should also consider integral cycle control, ie driving the heater with a number of whole line cycles (synchronized at the line zero crossings, at twice the line frequency 100/120 Hz) every fixed number of cycles. In this way minimum noise is generated in the line. Edit: To derive the required heater ouput you must first process the thermocouple measurement to generate an error value which must then be further processed with PID functions. You can find many and very good posts on this subject in MrPlc.

Ghostscript and its viewer GSview work well, run on Windows and Linux and handle many paper sizes. To use them, a postscript printer must first be declared (you can choose any from the available listed for your machine). You then print to a file for this printer and then convert the postscript format to pdf with GSview. They are available from here

The IBM Redbook "TCP/IP Tutorial and Technical Overview" may help you as reference material. You can download it (pdf format, 7.5 MB) from the IBM Redbook site here