What's new
What's new

OT: Analog electronics help: buffer stage design

lathehand

Cast Iron
Joined
Jan 31, 2005
Location
San Francisco Bay Area
A few years ago, I built a small dynamic balancing machine and amplifier. The input to the amplifier is a switching network which provides plane separation and calibration of the meter to directly read the amount of unbalance. This network must remain double-ended – ungrounded – until the signal exits the network and is grounded in one leg to input to a 741 op amp amplifier.

The problem is that the switching network includes two potentiometers and the net effect is that the input resistance of the network changes during the process of separating the planes and calibrating the correction weights. The amplifier is my design but the network was reverse engineered from a professional amplifier. The original design used 10K potentiometers and I copied that. However, the internal DC resistance of the pickups is 6K and the potentiometers are loading the pickups. I determined that a good input resistance would be between 500K and 1M, and closer to 1M.

I thought the easy way would be to install a buffer stage between the pickups and the switching network so the pickups would always see the same impedance. The problem is that I am inputing a double-ended signal from the pickups and I need a double-ended output signal to the network. That has turned out to be more difficult than I had expected.

To summarize, I need a buffer with high impedance input which can be either single or double ended and a low impedance double ended output at about 30 HZ. I've thought of signal isolation transformers and I poked around the web and found audio driver chips such as TI DRV 134. http://www.ti.com/lit/ds/symlink/drv134.pdf I am not familiar with these chips so am looking for help and other circuits that might be suitable.
TIA
Carl
 
There are a lot of guys here who are also members there, but you might want to post this at the EEV blog forum, there are more electronics guys over there.
 
This may seem like an obvious suggestion, but unless your common mode voltage is something crazy why not just buffer each side of the differential signal separately?

A schematic would really help out here.
 
The simplest method would be to use an opamp on each input, with the gain set at unity gain, and then use the outputs of those, which will be quite low impedance, as the input to your network.

Since it does not seem as if you have any bandwidth issues, or issues with noise (as long as it is not too much) a 741 on each input should be fine, and will have a high input impedance.

If you need even higher impedance, a TLO72 will provide a very high imput impedance, and has both opamps in one package.

Any will need a reference to ground, but that can be a high value resistor. The TLO72 is fine with 1 megohm, although you do start to get into noise issues related just to the resistance value.

Although some of the specialty chips are good, I think they are un-needed and offer no advantage, given what you have mentioned so far.
 
I have built a lot of op amp circuits and this is exactly what I was thinking of. Modern op amps can have a very high input impedance and a very low output one. And to maintain the accuracy, op amp circuits have their gain set by the external resistors that are connected to them so you can either use high precision resistors for the input and feed back resistors or use a multi-turn pot for one of them and set them to the same gain for both lines of the balanced input circuit. Resistors are available in 1% and even 0.1% tolerances or, as an alternative, with a digital multimeter you can buy a batch of resistors and find two pairs that are within whatever tolerance that you need.

The rest is standard op amp design. This technique has been recommended by multiple sources, including the chip makers. Web sites like TI and others will have example circuits.

I would look for a dual, FET op amp chip. Places like Digi-Key, Mouser, and Newark are good places to look because they list all the specs and you can drill-down through a selection process to get the perfect chip for your circuit. AND they always have links to the data sheets for those chips, just a click away.

Note: On that 741, that is a great chip and still in production after more than 50 years. I have used them extensively since my 20s and I am 75 now. But there are far better chip designs available today. And many of them are pin-for-pin replacements for the 741, but with their improved specs. One of my favorite design tricks with op amps was to breadboard the circuit with a 741 and test it. If it worked, fine. If it was lacking in one way or another then I would just pick one of the improved designs with the same pin arrangement and plug it in. You may be able to replace the original 741 in your circuit with a better chip, perhaps with changes to the input and feedback resistors to get the circuit parameters you need. And avoid adding any additional buffer stages altogether. Again, the above mentioned web sites are your friend: the readily available parameters and data sheets will quicken your search.



The simplest method would be to use an opamp on each input, with the gain set at unity gain, and then use the outputs of those, which will be quite low impedance, as the input to your network.

Since it does not seem as if you have any bandwidth issues, or issues with noise (as long as it is not too much) a 741 on each input should be fine, and will have a high input impedance.

If you need even higher impedance, a TLO72 will provide a very high imput impedance, and has both opamps in one package.

Any will need a reference to ground, but that can be a high value resistor. The TLO72 is fine with 1 megohm, although you do start to get into noise issues related just to the resistance value.

Although some of the specialty chips are good, I think they are un-needed and offer no advantage, given what you have mentioned so far.
 
Thank you, gentlemen. I looked at the TI application report for inputs to ADC chips and there are a lot circuits which will take some time to digest. Conrad, will video amps work at my low frequency? Some of you recommended a unity gain voltage follower on each leg of the input signal and using the outputs from pin 6 as an ungrounded signal. I tried that, with inconclusive results. I think I have oscilloscope problems and I'll get on that and report back.
Carl
 
Re: video amps. Generally speaking, video amplifier chips are just amplifier chips that are a lot like op amps but they have a larger gain/bandwidth product. So they can operate at higher frequencies. Audio = 20 Hz to 20 KHz while Video = 15 Hz to 5 or more MHz. In actual ICs the low end usually extends all the way down to DC and it is only the high end that differs. Op amps have been used for both. And other designs that are not described as op amps are also used for both. Often those other designs are a lot like op amps but that description is not used on the data sheet. You would have to ask the chip designers for a reason.

But yes, most "video amp" chips will operate at low frequencies, all the way down to DC. In video circuits an external capacitor may be used to limit this at a higher value.

"Some of you recommended a unity gain voltage follower on each leg of the input signal and using the outputs from pin 6 as an ungrounded signal." Of course the output will not be grounded. You seem to be thinking about it not being referenced to ground. Apparently you have a source/sensor that has a differential output and that output is not referenced to ground. You call it "a switching network". That does not tell us much about it. Is it powered or is it a floating device that produces it's own signal, like a microphone or piezoelectric device? If it is powered, then it probably does have a ground reference. If it is a truly floating device, then things may be easier.

If it is powered, then there is probably a ground reference that will be, somehow, connected to your amplifier. Worst case would be through the AC power connections.

If it is truly floating, then the input stage(s) of your amplifier will provide a ground reference.

Op amp circuits are traditionally powered with a dual power supply; that is one that has a GROUND and equal positive and negative Voltage sources. For an inverting op amp circuit, the non inverting input will normally be referenced to that GROUND via a resistor that is calculated from the values of the other resistors in the circuit. That resistor to ground provides the ground reference to a floating device that is connected to the input of the circuit. And that floating device will be somewhere near that ground potential. With a differential input op amp circuit that same resistor to ground still exists and the floating device will still be approximately at that potential. And, due to the dual rail nature of the power supply, that ground will be half way between the two (+ and -) power rails.

If your "switching network" does have an actual ground reference, then that ground reference should be connected to the amplifier's power ground AND the output or both sides of a differential output, absolutely must be between the power rails of the op amp's power supply. In fact, depending on the gain in that initial, input stage op amp, those "switching network" output(s) will have to be fairly close to the ground potential of the op amp's power supply. Other wise, the output of that initial op amp will be driven VERY HARD to one supply rail or the other and remain there despite any variations in the input signal.

In short, you will have a ground reference weather you realize it or not and you must take it into account in your circuit. Without knowing the exact nature of your signal source device, it is hard to say how this should be done.



Thank you, gentlemen. I looked at the TI application report for inputs to ADC chips and there are a lot circuits which will take some time to digest. Conrad, will video amps work at my low frequency? Some of you recommended a unity gain voltage follower on each leg of the input signal and using the outputs from pin 6 as an ungrounded signal. I tried that, with inconclusive results. I think I have oscilloscope problems and I'll get on that and report back.
Carl
 
Some of you have pointed out that ‘a switching network’ does not provide enough information and there is no information on the input signal. So here goes: The input sensors are linear velocity transducers – two of them, one for each correction plane of the rotor – and they are simply magnets in a coil that generate a voltage proportional to the vibration of at that end of the rotor.

These ungrounded signals are added or subtracted in various combinations in a network comprised of three DPDT relays, two potentiometers and a load resistor. As I noted in the original post, the resistance of these pots and load resistor are too low for my pickups. I could simply swap them out for higher values but that would not totally eliminate the loading as the pots may have to be adjusted to a low resistance. So, I thought to insert an impedance buffer between the pickups and the relay network and my first thought was to use some chip.

I woke up this morning thinking that there a problem with a chip solution: the signals must remain ungrounded throughout the switching stage. At its output a relay connects one side or the other of the signal to ground and the other side to a 741 op amp connected as an amplifier. I got to thinking that none of the chips which have been suggested so far would allow grounding of either side of their output signal. I suspect they would go kaput rather quickly.

That leaves transformer isolation and a quick look found this: TTC-105: Arndt : Telecommunication Coupling Transformer 1:1 : Passive Components I have ordered two of these to play with and they should be here in about a week.

OK, guys, did I do a better job describing the circuit this time?

BTW, what I thought was problem with the scope turned out to be two dead 741's - the ones that I was using to experiment with using a voltage follower connected to each leg of the signal. I do not know what killed them as I did not ground their outputs. I'll leave that mystery alone for now.
 
It sounds as if the coil type sensors are ungrounded because they are being connected in the "various ways" that would not work right as op-amp outputs.

In that case, can you "break into" the network AFTER the relays, and before the stuff that is low impedance? That would be the place to put the buffers, and is what I was thinking you meant. That may not be possible, as the network may be set to load the coils to calibrate. But in that case, the loading and attenuation is known, and intended.....

And, I know you mentioned the impedance issue, but is it really an issue? Are you sure there is a bad effect from it? The 6k is a noticeable attenuation, but may not be a killer since the absolute amplitude may not be that important, and the gain can be (presumably) adjusted later in the circuit if it affects a readout of weight to be added, etc. it seems like a known loading that is wanted.......

Maybe you are overthinking the matter?

More details? I am not sure we are being asked the right question here.
 
I would consider redesigning this circuit in the following manner.

You need to switch (the relays) the various sensors into combinations that either add or subtract their signals. Your existing circuit uses an op amp as it's input stage.

One great thing about op amp circuits is that a single op amp can add or subtract any number of analog signals simply by using separate input resistors. The values of these input resistors can be selected/adjusted to provide individual scaling factors for each of these input signals. So your relays can send the signals from each of these sensors to the appropriate input resistor of that input stage for each of the configurations that you want. In other words, instead of adding/subtracting their signals before they get there, add/subtract them in that input stage. This design would also allow a single sensor to be used with different scaling factors by directing it to different input resistors for different configurations. For instance, in one measurement sensor A could be added with a scale factor of 2.3 and in another one it could be subtracted with a scale factor of 1.7. Likewise for each of the other sensors.

Since separate inputs could be used for adding or subtracting the signal from a given sensor, the sensors could have one side grounded and an additional op amp or other type of amplifier could be added to them to handle any impedance or signal level problems that you have. These additional amplifiers could also have adjustable gains to allow calibration of each sensor without effecting any of the others. Adjustable DC offsets could also be incorporated.

You might also consider using electronic switching instead of those relays. There are CMOS switching chips that make this easy. They can be controlled with logic level signals. The 4066 comes to mind, but I am somewhat out of date and there are probably better ones available.
 
That is quite a decent idea. And doing that would potentially allow using buffers on the input if that is still needed.

This assumes the input network is reasonably simple and can be changed in that manner without too much work and re-calculation.
 
JST: You are right in one sense that I am overthinking the project. My background is in power - up to 115 KV and 22,500 Hp motors. I worked in building construction, not engineering. No tubes, transistors or ICs. What I know about electronics I picked up on the side and one thing that I learned is the desirability of matching impedances between outputs of sensors and inputs of processing electronics.

So I got out my HP voltmeter which will measure down to 0.1 mV AC and a decade resistance box and started experimenting to determining the loading of the pickups by various input resistances. It is low frequency so I did not consider reactance. I was quite surprised by the difference in the sensor output voltage between 10K and 1 M input resistance.

At some point in the balancing process, the amplifier is no longer able to detect the zero crossing point of the sine wave and the strobe "tumbles" - flashes randomly. I figured that I could improve the sensitivity of the amplifier by increasing the available signal level to the edge detector module. To do this, I want to insert an impedance buffer stage between the sensor and the varying and rather low resistance of the plane separation switching circuitry.

This was and still is a learning and research project. Never done an electronics circuit before and this one has 12 op amps, a opto-coupler, and a FET. I've done a lot of learning and this little detail of input impedance matching is part of it. I wire circuits, turn on the power and see what happens. I do not have the electronics background to fully understand what happens, but that has not held me back so far.

Enough about me: back to circuits. The 4066 may not work. As I noted in my previous post, I do not think that electronic solutions will allow a double-ended output to be grounded after the plane separation and calibration network. I stand ready to be corrected in this assumption. Grounding a leg makes no difference to a transformer so I borrowed one yesterday and will test it later today.


And yes, JST, I could simply increase the amplification as necessary and the amplifier has terminal blocks to make changing the feedback resistor easy. Comparing the sensitivity of the system with increased amplification vs sensor impedance matching is the next thing to do.

EPAII: Paul, I think. A summing amplifier. Didn't think of that even as I wired various op-amp circuits out of my textbook. The problem is that the summing is done while the signal is double-ended and the ground is applied after the summing junction. But it is an attractive idea that eliminates the impedance problem. And another but: but implementing a summing amplifier will require quite a revision of the plane separating circuit. I may give this some thought and experimentation for a learning experience, but not actually implement it in the finished circuitry.

I thought I was asking a relatively easy question and have been surprised by the discussion it has generated. Thank you for taking the time to input your knowledge.
 
The thing about that summing circuit is that you can do all the summing from unbalanced inputs (one side grounded), because sum, difference, etc can be done in the circuit instead of via connections of the balanced (not grounded) coils. You still need switching, but less of it.

And, the coil signals could be amplified first, and then summed etc, rather than needing to do it before amplification

Probably the existing circuit would be helpful, as we are all making slightly different guesses as to what exists now.
 
I breadboarded the 1:1 signal isolation transformer and it works perfectly. I'll build the circuit into the balancing amplifier and test it in actual operation, but I think I have arrived at the solution to my question. Thanks for the help.
Carl
 
If it works, it works. Great.

And yes, it is Paul. Not hiding anything: that's what I go by, my middle name. Perhaps I should incorporate it in my tag line. I picked the user name many years ago when I was new to this internet stuff.
 
It is surprising what does and does not stick in my brain. Somewhere either you signed a message or someone else knew it and for that unknown reason I remembered it. Thanks again for the help.
 








 
Back
Top