High impedance vs frequency

[0xDEADBEEF]

I'd like to measure a Wheatstone bridge with a lock in. The bridge has something like 20kOhms resistance. I would like to measure with some fairly high frequency like 50kHz but my cable capacitance is causing trouble and low passing at 10kHz or so.

Two questions:

1) What would you recommend cable wise? The normal 6mm BNC cables have a lot of metal in their shield. Would twisted pair wire lower the total capacitance? Would shielding around the twisted pair influence the capacitance? Are thin coax cables better?

2) I am also experimenting with a pulse transformer, and I am thinking about buying a new one. The manufacturers specify "impedances" of 600Ohms for example. I imagine that is for an old 600Ohms wave impedance cable standard. I measured my transformer and it only has about 40 Ohms DC. What do the manufacturers specify? Impedance is frequency dependent isn't it? Where can I get pulse transformers that work at ultrasound frequencies (audio ones seem to be fairly easy to get)?

 

[sophiecentaur]

What quantity do you want to measure about the Wheatstone Bridge?

 

[0xDEADBEEF]

A varying resistance.

 

[sophiecentaur]

Surely you measure it WITH a Wheatstone Bridge?
I'm still not sure what you actually want to do.

 

[0xDEADBEEF]

Is it that badly explained? I have a variable resistance sensor in a harsh environment connected with noisy cables. But I need fast readings. So I put it in a Wheatstone bridge because I am only interested in relative changes and supply the bridge with an oscillating voltage from a lock in. The output will oscillate at the same frequency, for the lock into detect. For the noise cancellation the higher the frequency the better.
Unfortunately the whole Wheatstone bridge needs resistances of around 20kOhms so it has fairly high impedance. Hence it is not very stiff and will not work when the cable capacity or the frequency are too high.

My question was basically what cable configuration has the lowest capacity. My thinking is the less metal the better. But I am unsure how a shield around a cable would effect capacity.

The pulse transformer is supposed to decouple input and output, because there are some grounding issues. But the data sheets I see don't seem to make sense, because impedance should be frequency dependent.

 

[Adjuster]

Is it that badly explained? I have a variable resistance sensor in a harsh environment connected with noisy cables. But I need fast readings. So I put it in a Wheatstone bridge because I am only interested in relative changes and supply the bridge with an oscillating voltage from a lock in. The output will oscillate at the same frequency, for the lock into detect. For the noise cancellation the higher the frequency the better.
Unfortunately the whole Wheatstone bridge needs resistances of around 20kOhms so it has fairly high impedance. Hence it is not very stiff and will not work when the cable capacity or the frequency are too high.

My question was basically what cable configuration has the lowest capacity. My thinking is the less metal the better. But I am unsure how a shield around a cable would effect capacity.

The pulse transformer is supposed to decouple input and output, because there are some grounding issues. But the data sheets I see don't seem to make sense, because impedance should be frequency dependent.

The transformer manufacturers probably know what they are talking about, don't you think? It is true that a transformer's primary, secondary and leakage inductances will all be frequency - dependant: what the data sheets are telling you is that the transformer will give a certain specified performance when driven from and / or terminated in certain impedances.

Many small pulse transformers will be designed for applications where the impedance will be dictated by some sort of transmission line, typically tens to hundreds of ohms. You would not get the maker's specified performance in your application because of the larger impedance. There are a few ways round this, but which would be accessible to you would depend on your level of electronics.

The easiest way would be to find a transformer designed for the application. There are for instance microphone transformers designed for about 50Kohms input, but of course these won't get you much past 20kHz.

The next possibility would be to try to find a lower designed impedance transformer designed to work down to a much lower frequency than the lowest you need, in the same ratio as of your working impedance to the working impedance.

Finally you might consider making a battery operated preamplifier to drive a transformer from a controlled lower impedance - I assume that what you are trying to do is make a properly balanced pick-up of signal from the bridge, to feed to your lock-in?

 

[sophiecentaur]

I may be dumb but I don't see where 'Stiffness' or "lock in" necessary relate to a Wheatstone Bridge. I think the problem may be that this is specific application of a Wheatstone Bridge, which is a very general term describing a basic bridge circuit. This is why I need some more description before I can hazard a guess as to your specific problem.
If you are bothered by noise, perhaps you could introduce some heavy filtering - or does this "lock in" refer to some phase locking?

 

[0xDEADBEEF]

From http://en.wikipedia.org/wiki/Lock-in_amplifier :

A lock-in amplifier (also known as a phase-sensitive detector) is a type of amplifier that can extract a signal with a known carrier wave from extremely noisy environment (S/N ratio can be -60 dB or even less.


I hope you know more about cables...

 

[berkeman]

I hope you know more about cables...

Take it easy, DEADBEEF. Have you considered bootstrapping the shield to eliminate the cable capacitance? Not sure if that's compatible with the lock-in connectorization, though.

 

[Bob S]

My question was basically what cable configuration has the lowest capacity. My thinking is the less metal the better. But I am unsure how a shield around a cable would effect capacity.

Use a double-shielded coax cable with the largest diameter ratio of the inner shield to the center conductor (or select the highest cable characteristic impedance), and the highest propagation velocity (over 90% c if possible), and use Berkeman's suggestion to bootstrap the [STRIKE]outer [/STRIKE] inner shield. Avoid cables like double-shielded RG-8 with Z=50 ohms and v = 0.66 c.

Bob S

 

[sophiecentaur]

From http://en.wikipedia.org/wiki/Lock-in_amplifier :

I hope you know more about cables...

I know that it is possible to be measuring several things with 'cables'. Is it, perhaps, temperature or strain that this cable would be registering? As I implied earlier, it is good not to assume that a shorthand question will be understandable by readers who may actually be able to contribute to your question. This is not a specialist forum.

Do you know a lot about Wheatstone Bridges in general - or is this the only application which you have used one for? When you say that the Bridge "needs 20k ohms" do you mean that you have no choice of resistances for the other three arms of the bridge? Bridges work on ratios so why not choose an appropriate resistance in the reference arm?

Rather than trying to eliminate the cable capacitance, it might be an idea just to balance it out, using similar cable in the reference leg. Bridges work on that principle. In any case, the capacity of a coax cable need not be much more than a few tens of pF per m - and that would be pretty much independent of how much metal is used in the braid - particularly at such a low frequency. (I do know that much about cables!) Would you be using many metres of cable?

 

[0xDEADBEEF]

Thank you for all the advice. It's a temperature sensor going to about 15KOhms. The other resistances won't get around that.

I tapped a few more sources and got some answers... I think that I have a plan now how to solve my problem. The cable is simply too long, so I will build my own amplification stage with some filtering.

For the cables it seems that changing to twisted pair is seldom worth it, because coax has less interference and thinner coax will not change the capacity too much. So first choice is trying to shorten the cables.

btw the impedance of pulse transformers is given at some common reference frequency like 1kHz. but with a fully differential amplification stage I will not need the transformer anymore.

I might annoy you some more when the design is still not working. ;)

 

[f95toli]

Now I am curious. Why are you trying to measure a temperature sensor at 50 kHz?

A standard temperature bridge (say a Picowatt AVS-47B) that uses square wave pulses can easily measure resistances up to at least a couple of hundred kOHms even with long cables and a lot of capacitive filtering, although as far as I remember this is done by balancing.

It might be worth looking in the AVS-47B manual (www.picowatt.fi), you might find something useful in there (the manual is actually worth a read anyway, there is a LOT of useful info in there).

 

[0xDEADBEEF]

I did remember that bridge I guess it was you who recommended it earlier and I looked at the specs recently. I would love to use it. My problem is speed. Depending on the accuracy I need 50 samples/second or so and I am really only interested in relative changes so it doesn't matter if the first 3 digits are right if the last ones flicker.

 

[sophiecentaur]

From your remarks you seem to be ignoring the whole principle of the Wheatstone Bridge. You can choose whatever resistance (impedance. actually) you want for the 'other' arm of the bridge. In this case, why can't you use 15kOhms in series with an identical piecs of cable - to balance out any capacity problems. If you connect the screens / inners appropriately and route the cables in more or less the same way, a lot of the interference pick up would also tend to cancel. This is standard for measurement with long cables. I believe.
I'm, sorry if this is all obvious to you but you haven't made it clear that you get these points and I am anxious that you should get a good solution to the problem.

 

[0xDEADBEEF]

In a way you are right. I do ignore a lot of the Wheatstone bridge principles. I moved the reference resistances away from the resistance to be measured and it is not very balanced at all. But it seems I am not the first person to build a 3/4 Wheatstone bridge, it is pretty much impossible do get around that (thermal instability, space, commercial resistance decade...) but I am still thinking about more balance.

The capacity problem is a mayor issue, and it cannot be solved by balancing. If the leads to the Wheatstone bridge (in the middle) have too much capacity they don't charge up fast enough to transmit the oscillating bridge signal.

 

[sophiecentaur]

Well at least we're talking the same language now.
I don't see why the length of coax (at less than 100pF per m) should pose such a problem at a mere 50kHz. If you also have the same length in the other arm, any frequency tilt / phase shift (which is what it is really) will not matter as it will balance out. I don't understand your reluctance in that direction because it's the only way you're going to deal with the capacity problem - unless you tune it out with an inductance.
Your idea of using a transformer (at the far end, I presume) is a good one but, unless you do a similar thing in the other leg with a dummy resistor of similar value to your sensor, along with a compensating length of cable, there will be similar phase and frequency response problems which could well affect the balance of the bridge and give false resistance readings.
If you don't intend to use the bridge as a bridge then why not just measure I and V straight?

 

[Bob S]

Comment on bootstrapping mentioned in post #10. The basic purpose is to reduce the effect of the capacitance of the input cable by reducing the voltage between the center conductor and the inner shield. In order to bootstrap the inner shield of a multiple shield coax, the input signal V_in is amplified in a voltage follower configuration with gain, and a fraction of the output voltage (< 1 x V_in) is tied back to the inner shield. If the feedback signal is too high, the circuit could become unstable. But if V_feedback ~½ V_in, the effect of the cable capacitance will be cut in half.

Bob S

 

[berkeman]

I fixed it. Thanks Bob!

 

[0xDEADBEEF]

Due to a forum bug or equally likely due to temporary insanity on my part I didn't see Adjustor's post until today. I didn't know it was called bootstrapping, but I knew it as guarding. I gave it some thought, but I don't think it would have worked because my signal is comparable to the output offset voltage of a buffer amp.

I didn't know that you could choose cables by signal speed, and always thought it was much lower. I heard somewhere that metals have an index of refraction of 8 or so, so I assumed the signal speed would be limited by that. Interesting...

 

[0xDEADBEEF]

Well at least we're talking the same language now.
I don't see why the length of coax (at less than 100pF per m) should pose such a problem at a mere 50kHz. If you also have the same length in the other arm, any frequency tilt / phase shift (which is what it is really) [...]

It is not a phase shift, but the cable is forming a low pass: Take 4m of cable. 15kOhm input and you have a 25kHz low pass. The signal is just charging the cable before the voltage rises noticeably.

 

[sophiecentaur]

It is not a phase shift, but the cable is forming a low pass: Take 4m of cable. 15kOhm input and you have a 25kHz low pass. The signal is just charging the cable before the voltage rises noticeably.

Well, if that ain't a phase shift I don't know what is. (Look up 'phase characteristics of an RC filter', anywhere). The actual transit time of a few metres of cable is vanishingly small wrt a cycle at your frequency. You can consider the cable as a lumped component at that wavelength (6km in air).
When you are so assertive about matters like that I can't be sure what to take seriously and what not to take seriously.

"I hope you know more about cables" Now where did I read that?

 

[0xDEADBEEF]

Sorry I was grumpy because I realized I was set back by at least three month. The low pass will shift the phase, but it will kill the amplitude, because the cable is a capacitor in parallel to my measurement device.

 

[sophiecentaur]

If you only have 4m of cable, the capacity shouldn't be much more than 400pF. The reactance at 50kHz is not all that low (about 8k). This, in parallel with your 20kR, will have an impedance of magnitude 7K - not very different from the original 20kR and will have only a small effect in detecting a null on the bridge.

The very first sentence in your OP is not clear. I am now assuming that you mean that you want to measure a resistance of about 20k, using a Wheatstone Bridge. I don't understand how you seem to have no choice of the impedance that you put in the reference arm. Another length of cable (or even a capacitor of the same value) can be used to eliminate the effects of the reactive component. You seem to be rejecting the one great strength of using a bridge for this sort of measurement. Is this bridge, in fact a piece of equipment with just one pair of terminals on it?

What can be causing such a high level of noise (narrow band around your 50kHz probe frequency)? Using a phase locking amp should be giving you a bandwidth of, perhaps, a few tens of Hz - you could do as well if you used a simple home made filter!.

You seem to be asking for three digit accuracy with sampling at 50Hz. This suggests that you will need an SNR of about 60dB. Just what level of noise / interference are you expecting in this measuring environment (in a 100Hz bandwidth)? Is the noise spectrum continuous?

It is always difficult to make suggestions 'by remote control' when there are so many unknowns at my end but I get the feeling that you are committed to a certain measurement technique that may not be optimal. You do not make it clear why you don't want to modify your bridge method to gain an improvement - is this because you are stuck with certain equipment and can't modify the setup?

And, finally, please tell me what 'stiffness' has got to do with this? I can't think how this can be relevant to the electronics of the problem.

 

[0xDEADBEEF]

Sorry for picking this up again. My amplifier is in the making and I simply removed all the cable that I possibly can...

Just to be sure, I checked your suggestion in SPICE and played with the resistances. If I make the reference arm with less ohms I can double my signal but not much more. The frequency characteristics stay qualitatively the same, and the bridge is not balanced anymore.

About the noise: The bridge can only be driven with 10mV or less. The signal will be around 1uV. This is why I have so much noise. Filtering is important, and I will do a lot more of it when the amplifier proves insufficient.

The noise is a good bastard mix: Slowly varying dc offset, line noise, radio pickup, voltage spikes. I am mostly worried about the total level. When the input AD of the lock in starts clipping it gets useless very fast.

I mean by stiffness the ability of an output to keep up a voltage when you draw current. The lack of this is basically my problem. If the voltage across the output of the bridge should be 1uV the cable transporting this signal has to charge up first, drawing current through the bridges resistances. The speed of the charging is most strongly limited by the largest resistor in the bridge (I expected this and was confirmed by the simulation). If the bridge was more "stiff" i.e. if the bridge had lower internal resistance, it would charge the cable faster.

I hope this clarifies some issues. I am quite hopeful, that my new amplifier will solve my problems, so I won't need help before is is ready

 

[sophiecentaur]

Why are you bothered about DC and RF noise if your bridge receiver is so well filtered?
Did you not take my point about SNR only being relevant once you have specified the bandwidth of the system and the noise components around your probe frequency?
If you are concerned by the 80dB loss then you need to do some tuning of the cable impedance. But a bridge looks for a null, doesn't it? So you would expect to be looking for a point at which the unbalance signal is (near) zero.

You never said why your bridge resistance needs to be what it is and have not commented on the notion of balancing out (or tuning out) the unwanted reactance apart from dismissing it as not being an option; this tuning would also have the effect of providing some selectivity against out of band noise. But, in fact you don't seem to want to use the bridge as a bridge at all. I can't understand that either.

As for the use of the term "stiff", why not just refer to source impedance? Stiffness is normally used as a mechanical term, I thought.
I hope the new amplifier helps but I am still sure that performance could be improved in other ways.

 

[0xDEADBEEF]

Why are you bothered about DC and RF noise if your bridge receiver is so well filtered?

The lock in is a filter, pretty much a bandpass as sharp as I you can make it, if the AD chips don't clip.

Did you not take my point about SNR only being relevant once you have specified the bandwidth of the system and the noise components around your probe frequency?

If the input doesn't clip, everything is fine. The only purpose of the filtering is to prevent the clipping.

If you are concerned by the 80dB loss then you need to do some tuning of the cable impedance. But a bridge looks for a null, doesn't it?

In my case the bridge doesn't look for a null, it just shifts the signal so you measure deviations instead of absolute value. This is also commonly done for strain gauges.

You never said why your bridge resistance needs to be what it is and have not commented on the notion of balancing out (or tuning out) the unwanted reactance apart from dismissing it as not being an option; this tuning would also have the effect of providing some selectivity against out of band noise. But, in fact you don't seem to want to use the bridge as a bridge at all. I can't understand that either.

OK, so enlighten me. I attached a sketch of my bridge, including the parasitic capacitance of the signal cable. What would you put into "balance" it, so it does not low pass. I need to measure the small deviation "e". The bridge is driven with at least 50KHz.

As for the use of the term "stiff", why not just refer to source impedance? Stiffness is normally used as a mechanical term, I thought.

It is also used in electronics for sources that don't change under load.

 

[Adjuster]

Due to a forum bug or equally likely due to temporary insanity on my part I didn't see Adjustor's post until today. I didn't know it was called bootstrapping, but I knew it as guarding. I gave it some thought, but I don't think it would have worked because my signal is comparable to the output offset voltage of a buffer amp.

I didn't know that you could choose cables by signal speed, and always thought it was much lower. I heard somewhere that metals have an index of refraction of 8 or so, so I assumed the signal speed would be limited by that. Interesting...

Propagation velocities in cables are more typically two-thirds that in free space. They depend on the construction, and particularly on the dielectric used, because that is where the fields (mostly) are. The speed is therefore more linked to the refractive index of the dielectric than that of the conductors.

 

[sophiecentaur]

0xDEADBEEF
This is interesting. I have two observations.
Firstly,it would have helped if you had actually stated the layout more explicitly. The bridge, according to your diagram, is mounted remotely from the detector (not clear uless your setup was in view). I was assuming that the only remote component would be the variable resistor and you were assuming that I knew what you meant. A diagram, much earlier on, would have been very useful. From the position of the 400pF, in your diagram, it appears that is the capacitative loading between the bridge and the detector amplifier. It makes more sense now. I was assuming that you would have a remote resistor and would compensate for the lead length by having a similar lead in the reference arm. I mentioned this several times but, of course, I must have appeared 'loony' in the context of the circuit which was quite clear to you.
Secondly it strikes me there is a' generation gap', here. When I say "filter" I imply a passive filter with linear components - when you read "filter" you assume a digitally implemented filter.

Digital implementation of any analogue quantity always requires an analogue anti aliasing filter, at the very least (explicitly or implicitly); you don't want to sample any signals that you don't want actually to measure or which will embarrass your DSP. To get the best out of your system you will need a suitable analogue filter. A fairly narrow band pass filter (b/w suitable for the rate of change of R that you want to monitor, centred on your probe frequency. If you design / select a filter design which includes the (estimated) 400pF in it and which will reject DC and out of band components and not attenuate the 50kHz signal then you will improve matters considerably. The simplest filter would involve a large C in series (to reject DC and low frequencies and a shunt L, to resonate at 50kHz. You just don't need to be losing all those dB's which your present system introduces. Play with Spice and put in some suitable values - which you can easily work out for yourself. Good luck.