Hi, I have to measure the voltages of hundreds of batteries and display them on a PC. For the start the number is 200. 20 batteries in series form a group and we have 10 groups in total. I need 400 differential Inputs. The old technology we are replacing uses Relays. If the user selects the Relay Group, the analog voltages of this group can be read out. I want to replace the Relays with ICs. Any suggestions? Should I use multiplexers? Thanks
Because of the high voltages which are probably involved, it might be better to stick with the old technology. Reliability of most relay multiplexers of this type can be improved with simple inexpensive additions to the ATE, and setting up a preventative maintenance schedule to replace the relays on a regular basis.
Hi. I'm assuming you're not satisfied with the reliability of your relay multiplexer, and you're looking at replacing it with a solid state system (based on analog switches) in order to improve reliability.
Before you walk away from the relay concept, you should be aware of its advantages. Open is really open (100s of megohms or gigohms between contacts). If you've got a good relay contact, closed is really closed (milliohms). You're not going to get that ratio between open and closed with any analog multiplexer IC. Not only that, but an analog switch also has leakage currents between switches on the same IC, which isn't a problem at all with relays. And as long as the relay contacts are rated to withstand voltages greater than that seen in your battery system, you don't have to worry about finding high voltage multiplexer ICs. You didn't mention the voltages you're measuring, but that could be a serious problem.
Another big advantage is that relay contacts are immune to ESD (electrostatic discharge). People who switch to solid state multiplexers have to deal with the fact that they're now working in an anti-static environment. Static can easily kill many analog multiplexers. Of course, many analog switches are made with ESD protection, but you always pay for that with a fairly dramatic increase in leakage current and usually with a decrease in the ratio between off resistance and on resistance.
If reliability is your only issue, you might be able to improve that quite easily and keep a working system running without the hassle of redesign.
First, you should look at the type of relay you're using. Since you're switching into a multimeter, you should choose a relay that's made for "dry switching", meaning that there's not enough current or voltage to wet the contacts on make or break (the "cut" on the data sheet is usually 5 V at less than either 1 mA or 5mA). Standard relays won't work well for test. You didn't mention the type of relay you're using, so that's one thing you might want to look at. If the minimum switching current isn't specified on the data sheet, you can call the relay manufacturer to find out. Reed relays are generally OK for dry switching, but you should check. If you don't have reed relays, ones with bifurcated contacts are frequently capable of doing this job.
If you have reed relays, you should also check the withstand voltage specified on the data sheet. Many reed relays are only specified to switch 100VDC or less. Again, something to check, since you didn't mention which relays you're using or the voltage you're switching.
Another issue which might be a problem is the battery environment, which sometimes has a very acidic and corrosive atmosphere. This affects relay contacts as well as any other metal. If this is an issue, it also affects electrical connections and such, and should be addressed first. This will be a killer no matter what you do. Sealed relays might give you better results, but it's better to improve the environment. Another option would be to place the test equipment in a remote location with good ventilation. These are basically DC measurements, so a couple of hundred feet of cable shouldn't affect your measurements too much unless you're measuring individual voltages in a millisecond time frame.
Every reed relay manufacturer specifies an open contact capacitance, as well as a capacitance between contact and coil. Usually this is only several pF, which isn't much of a problem if you're just using just one. But if you're using hundreds of relays in parallel, 4pF can be multiplied to the point that it's a big issue. You haven't described the layout of your test setup, but your description implies that you've got 10 banks of 40 relay contacts, each of which go to another set of
20 DPST contacts. You might want to look at the schematic of your test layout, and imagine a several pF cap in place of each N.O. contact. You'll see that, in fact, you're switching hundreds of pF of capacitance when you're switching voltages. That changes things quite a bit, because now you're talking about surge currents that are limited only be the resistance of the wire and the relay contacts. That can easily reduce the life of the relay contacts and make a very unreliable system. This is valid no matter what the input impedance of the meter, which is generally a DC current and a capacitance added to the above.
Looking again at the data sheet of the relay, you'll see a maximum specified current, along with a specified number of electrical operations at that current. Generally, you'll also see a specified number of mechanical operations (at essentially no electrical current). The first number for reed relays can be as low as 50,000 to 100,000 operations, and the second is usually in the millions. At a bare-bones minimum, you should add enough series resistance to the test line so the peak switching current for the capacitive load is less than the specified maximum relay current. Usually that means adding a small series resistor between each voltage and its relay contact. The less current you switch, the better. I like to start with a series resistor that limits current to 1/10th of specified maximum or less, if I can. I'll then somewhat arbitarily call the expected life at 1/2 to 1/5 of mechanical maximum (depending on balance between maintenance costs and cost of bad readings).
You have to take a good look at the load impedance of the meter here. If you've got a very high (e.g. 10 Megohm) input impedance meter, adding a 2K series resistor (0.5 amp max. reed relay contact current, set resistor for 50mA pk. at 100V) shouldn't cause you any problem. Your voltage readings will only change by 0.2%, which isn't a problem for most ATE applications. If it is, you can compensate with software. But if you've got a load resistor at the meter (which may affect the above calculations of switching current a bit), or have a low input impedance meter (many DVMs have different input resistances for different ranges), you may have a voltage divider between the series resistor and the meter. Your best bet here is to obtain 0.1% resistors (they're cheaper by the hundred) to minimize the resistance uncertainty, and compensate in software for the voltage divider.
Now it's time for a little math. Take a realistic look at how many cycles you're putting the relays through per unit of time, and extrapolate that to find out how frequently you have to replace your relays as preventative maintenance. Your inside set of 20 relays will cycle 20 times for every time the outside 400 cycle, so you'll have to replace the inside set of relays 20 times as frequently. This number will give you a good idea of the maintenance costs of your relay multiplexer system per test cycle and per year.
I found it useful in systems like this to get a small Omron counter with LCD display, voltage input and internal 5-year battery, and place the increment input of the counter across one of the outer ring coils (they're on longer, so there's no problem with activation time). I would then document that the counter be periodically checked, and the relays be replaced on a regular schedule.
If the existing relay system is older or has been well used, the contact resistance of the relay sockets may be a bit of a problem. This is especially true if people have been swapping out relays at random in frustration. By putting together a test program to measure resistance of closed contacts with new relays, you can get a good handle on the status of the relay sockets with several insertion/extraction cycles (reseat relay, measure short circuit resistance, repeat). If some show up bad, it might be best to replace all of them. The good thing is that, if you have a rational preventative maintenance cycle, you limit the number of socket insertion/extraction cycles to a minimum.
By this point, you've got a good handle on how to make your automated test setup very reliable, as well as getting a handle on how much maintenance of the wear parts (relays) will cost to keep it reliable. Having that, you can make a judgment on whether to go with analog multiplexer ICs and the failure mechanisms associated with them.
If you want more follow-up information, please include the following:
Maximum voltages switched
Floating or grounded batteries or meter
Type of meter used
Load resistance at meter and load resistor (if any)
Type of relay used
Measurement frequency (once per millisecond, second, minute, hour?)
Moral of the story -- old ain't necessarily bad. ;-)
You will undoubtedly want to look at multiplexers. But one thing to be careful about is the common mode range of your differential inputs. You don't say what voltage each of the 20 batteries in series is, but even if they are single cells you are talking about more than 30 volts. Make sure your multiplexers can handle the entire series voltage. If each battery is more than one cell, you may want to think twice before you get rid of the relays.
Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis
I haven't thought this all the way thru...you probably need different parts to get accuracy, but here's the concept. Configure a LM3900 op-amp as an integrator. Use HV fets to connect even numbered batteries to the positive input. Use HV fets to connect odd numbered batteries to the negative input. Use series resistors to scale the currents appropriately. two threshold detectors on the ouput. Use shift registers or a processor to leapfrog the fet switches up the chain generating a triangle (sort-of) as you go. Measure the triangle half periods to get the voltages.
I'm sure it's more complicated than that, but it's a place to start thinking. mike
Return address is VALID but some sites block emails
with links. Delete this sig when replying.
Hi, common mode voltages are allways going to be a headache. An alternative is to use a small micro to measure a small group of cells powered by the cells it's measuring. Each processor is connected to the next with an opto coupler in a long daisy chain forming a serial link which connects to the PC. The micro just measures the cells its connected to coverts it into an ascii value adds a cell number and sends it down the link.
If you are using conventional relays, it will also improve their life if they are physically mounted so that the gap between the contacts is vertical. This allows any dirt or particles worn off the contacts to drop away after a few operating cycles.
In electromechanical telephone exchanges, this was always good design practice.
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
Hi, the batteries can be connected in series which makes 2.0V * 200=
400V max. The max. differential voltage will be 2.0V. I don't know if the INA117 solution is possible for these voltages . If I use the relays, The voltages of 10 series batteries will be switched to a micro for conversion. An Optocoupler has to be used to send the data to the PC.
If this is a one-off job and cost is an object, your best bet is to obtain a commercial DAQ + relay scanner on eBay. You should be able to get what you need for under $500, and start measurements pronto.
Since you can connect any cell to your A/D with relays I assume the battery stack is floating. If this is true you can reference the center of the battery stack to ground. That would give you +-200V to the inputs of the INA which is within the spec range.