Op Amp to drive a coax cable

[Galroc]

Hi,

New to this forum, and new to electronics in general. I am a software programmer but I needed and wanted to lean electronics to solve a problem at work. The arduino has made it extremely easy to learn and introduce myself to electronics. I am about halfway through Dave's youtube videos...

One part of my project is taking an incoming ±10 Volt signal on a coax cable and adding/subtracting ±1.6 V from it, and putting it back out on another coax.  <10Khz.

It is the output that I am having issues with, especially with longer coax cables. I guess it is the capacitance and/or inductance of the coax.

Initially I just used an op amp that we had available to combine the voltages and send the result down the coax. This did not work well at all with long coax cables (~10ft). I then added a voltage follower (2nd op amp) with a 220ohm resistor in the feedback loop. This worked better and is usable but I am not totally happy.

I saw that Dave used a 100 ohm resistor on the output of an op amp in his uCurrent, but he didn't put it inside the feed back loop which means it changes the voltage output by a very small amount. I am driving 100K ohm instrument and he was driving a voltmeter which would have a much higher resistance. His error would be much lower than mine and acceptable. 100 0hm/100K ohm would give me a 0.1% error at max voltage.

I am now redesigning the circuit to use a proper instrumentation op amp (AD8221).  I am not certain if it needs another voltage follower after it.

How do people get signals to propagate reliably down a coax post op amp?

Cheers

[Mechatrommer]

human dont usually send long distance info using coax, it will has large cumulative capacitive coupling along the distance and expensive. try using twisted pair. DC bias offset is not much of a problem over 10ft distance, only few tenth volt drop depending on cable quality/resistance, the problem is with digital (square signal), capacitive will make it flatter. though i'm maybe mistaken and not an expert in this, someone else may answer it more accurately. dont worry, you are not the only software guy coming here to play with an opamp and coax cable welcome to the forum and i hope you stick around.

edit: and some hi speed opamp will oscillate if driving capacitive load such as coax cable, usually 50ohm impedance (if coax is 50ohm type) is placed on the opamp's output before the coax to avoid this. better if you can afford another 50ohm ground termination, but you'll get half the amplitude, but max power transfer. YMMV

edit2: and i dont think changing to intrument amp will make any difference. the output is all that matters (albeit low noise level), not its intrumentation inputs.

[saturation]

With coax, its better to impedance match the op amp output to the receiver input, even at these low frequencies.



If you're using 50 ohm coax, then put a series resistor Zs and a terminating resistor Zl as 50 ohms as in the above photo.  Now adjust the output gain of Vs, which I presume is your sender op amp, to the voltage you require assuming a 50 ohm divider.

Is the receiver an inverting op amp too? If so, the gain resistors will be affected by the input Zl interacting with the gain resistors causing the output voltage to be not what you calculated.  It'll be easier to use the non-inverting mode or a general fix is put another voltage follower at the receiver input after Zl to buffer it before the 100k ohm instrument.

[Lightages]

Most op amps have problems driving a 50ohm load. A high current output buffer is usually required.

[jahonen]

I think that 50 ohm series resistor at the driving end is enough, no need to terminate the receiving end. If the resulting voltage divider causes too much error, perhaps you could add slight gain to the driving end to compensate for that?

Regards,
Janne

[Rufus]

Most op amps have problems driving a 50ohm load. A high current output buffer is usually required.

50 ohm coax isn't 50 ohms without 50 ohm termination. The OP's 10 feet of coax driving a 100k load at <10kHz is mostly just a capacitor.

The OP didn't define his 'issues' or his cable or what he is driving with. My guess is the op-amp isn't stable with that much capacitance on its output. Sticking a couple of hundred ohms in series with the output (outside the feedback loop) would likely fix it and give a small dc error and some frequency roll off.

[alm]

human dont usually send long distance info using coax,

Tell that to the cable TV companies. Back in the days, 10base2 had a limit of about 200 meter with standard network interface cards. For high frequency signals, distance is often primarily limited by skin effect losses, not by the output stage.

it will has large cumulative capacitive coupling along the distance and expensive. try using twisted pair.

A terminated transmission line is purely resistive, at least in theory. What's the point of using twisted pair for single ended signals?

Terminating the cable (either source or end) and/or a using high current buffer (for example a cable driver opamp like the LT1206) would also be my idea. 10ft is not particularly long, so the buffer may not be required.

[Galroc]

edit: and some hi speed opamp will oscillate if driving capacitive load such as coax cable, usually 50ohm impedance (if coax is 50ohm type) is placed on the opamp's output before the coax to avoid this. better if you can afford another 50ohm ground termination, but you'll get half the amplitude, but max power transfer. YMMV

edit2: and i dont think changing to intrument amp will make any difference. the output is all that matters (albeit low noise level), not its intrumentation inputs.

Thanks Mechatrommer

You are right, I don't really need an instrumentation op amp on the output, just on the input. I thought it might be simpler than using two separate op amps (second one being a voltage follower), which is what my current design uses. I can change that one to a dual op amp chip, with a voltage follower, and leave the other AD8221 for the input.

With coax, its better to impedance match the op amp output to the receiver input, even at these low frequencies.

If you're using 50 ohm coax, then put a series resistor Zs and a terminating resistor Zl as 50 ohms as in the above photo.  Now adjust the output gain of Vs, which I presume is your sender op amp, to the voltage you require assuming a 50 ohm divider.

Is the receiver an inverting op amp too? If so, the gain resistors will be affected by the input Zl interacting with the gain resistors causing the output voltage to be not what you calculated.  It'll be easier to use the non-inverting mode or a general fix is put another voltage follower at the receiver input after Zl to buffer it before the 100k ohm instrument.

Thanks Saturation.

I am trying to drive a scientific instrument which accepts -2 to 10V on BNC. It says it has a 100K impedance. I can't change that end, nor tell what it is using internally. I have other commercial devices that generate DA that drive that device fine over the same coax length.

From your above image, I can only manipulate Zs, not Zl.

It sounds like what I should do is what I am currently doing now, which is use a voltage follower (noninverting op amp) after a normal op amp (which does the voltage math).

But, should a resistor be used in series like Zs from your above image (which is what Dave did http://www.alternatezone.com/electronics/ucurrent/uCurrent%20Schematic.png ) , or should I use a low ohm resistor in the feed back loop of the voltage following op amp, which is what I am doing now, which won't change the voltage output?

follow up
should I use any capacitors on the output as well?
page 49: http://www.analog.com/static/imported-files/data_sheets/AD8221.pdf

I am using capacitors on the input of the other AD8221 to reduce EMI/RF.

Cheers

[Galroc]

The OP didn't define his 'issues' or his cable or what he is driving with. My guess is the op-amp isn't stable with that much capacitance on its output. Sticking a couple of hundred ohms in series with the output (outside the feedback loop) would likely fix it and give a small dc error and some frequency roll off.

Thanks everyone!

It is correct that there is no 50 Ohm termination. Only the termination inside the device.

I am looking for stability and an improved design. My current design works, but has noise in the 50-100mv range. I suspect it is noisy because I need to use a better Voltage reference for a DA that I have in the design. My second design includes this.

But, since I am redesigning everything, I wanted to make sure that I was designing my output the correct way. I want to avoid the DC offset error if possible

Here is what I am trying to drive: http://www.physikinstrumente.de/pdf/E_662.pdf

I am currently using a Op Amp as a voltage follower with a 220 ohm resistor in the feed back loop. This seems to work, but it is different than what Dave did in his uCurrent schematic, where he used a resistor in series, which causes a DC offset.

Are those my only two choices?

Cheers

[Galroc]

This is what I am currently doing, without the gain:

http://neurosciences.tripod.com/headstag.htm

Cheers

[Mechatrommer]

should I use any capacitors on the output as well?
page 49: http://www.analog.com/static/imported-files/data_sheets/AD8221.pdf

you are making thing worst if its directly infront of the opamp. put the separating 50ohm impedance then you can put any capacitance you like after that (refer to above post saying opamp instability driving capacitive load). capacitor on the input is common though to reject noise etc.
edit: there's no page 49 in AD8221.pdf, i got page 24 max only.

[Rufus]

But, since I am redesigning everything, I wanted to make sure that I was designing my output the correct way. I want to avoid the DC offset error if possible

A 200 ohm series resistor introduces a dc scale error not offset. The error is 0.2% and could be compensated for by increasing the amplifier gain by 0.2%. The high frequency roll off depends on the cable capacitance. I doubt it will be significant in your application.

A series resistor inside the feedback loop makes things worse. The output impedance of the opamp and capacitive loading is a low pass filter introducing frequency dependant phase shift in the feedback signal. The op-amp has internal low pass filters which also introduce frequency dependant phase shift. If the combined phase shifts add to 180 degrees the negative feedback becomes positive and the amplifier becomes an oscillator if there is still sufficient gain at that frequency.

Increasing the output impedance with a resistor increases the phase shift. A series resistor outside the feedback loop isolates the output from the capacitance reducing the phase shift.

[Galroc]

should I use any capacitors on the output as well?
page 49: http://www.analog.com/static/imported-files/data_sheets/AD8221.pdf

you are making thing worst if its directly infront of the opamp. put the separating 50ohm impedance then you can put any capacitance you like after that (refer to above post saying opamp instability driving capacitive load). capacitor on the input is common though to reject noise etc.
edit: there's no page 49 in AD8221.pdf, i got page 24 max only.

Sorry, Figure 49, not page 49.

[Galroc]

A 200 ohm series resistor introduces a dc scale error not offset. The error is 0.2% and could be compensated for by increasing the amplifier
gain by 0.2%.

mistyped...I meant scale rather than offset...

I didn't think about that....changing the gain to compensate for the DC scale.

I would think that the gain that small would be in the error of even the most precise resistors.

The high frequency roll off depends on the cable capacitance. I doubt it will be significant in your application.

A series resistor inside the feedback loop makes things worse. The output impedance of the opamp and capacitive loading is a low pass filter introducing frequency dependant phase shift in the feedback signal. The op-amp has internal low pass filters which also introduce frequency dependant phase shift. If the combined phase shifts add to 180 degrees the negative feedback becomes positive and the amplifier becomes an oscillator if there is still sufficient gain at that frequency.

Increasing the output impedance with a resistor increases the phase shift. A series resistor outside the feedback loop isolates the output from the capacitance reducing the phase shift.

Good to know and I might have actually seen that when I built my first prototype. Thanks.

I think I will go with a 100 Ohm series resistor, and then work out if changing the gain is reasonable considering the errors.

Cheers

[Galroc]

Good to know and I might have actually seen that when I built my first prototype. Thanks.

I just went back to look at my prototype schematics and confirmed with the actual hardware and I did end up putting in a small capacitor across the 220 ohm feed back resistor to eliminate oscillations after I built the circuit. As Dave says in his videos, you learn more when things don't work...

But now it works. I don't need very fast frequency response (<10Khz).
 
But it sounds like I should use a series resistor and accept the small DC scale error. We also do not really work at the 10V limit. Most of the voltage range that we use is actually 0.5 to 2V. Which works out to a maximum to 2mV error.

Really appreciate the advice. Analog electronics is something else...

[vk6zgo]

Analog  Video Distribution Amplifiers (VDAs) do most of the things you want to do,& present a low impedance (75 Ohms) to the coax cable,--these things are flat to better than 5MHz.
I'm not suggesting you obtain a VDA,but you can probably look up a common circuit configuration for one,& adapt that to your purpose.

VK6ZGO

[Galroc]

Thanks vk6zgo, I will look into those.

I have noticed there are Op Amps designed for unknown capacitiveloads like the AD817.


As of now, I will use a 100k Ohm resistor on the op amps output to match the impedance of the instrument I am trying to drive. This will halve the voltage, but then I can use a gain of 2 in the op amp. I will need to halve the input going into the op amp for one of the signals but I think I can do that.

Cheers

[Rufus]

As of now, I will use a 100k Ohm resistor on the op amps output to match the impedance of the instrument I am trying to drive.

No No, 100k source impedance will create a significant low pass filter with the capacitance of the cable.

[SeanB]

If the frequency is low and you are wanting high drive ( and have dual rails available to power it) how about using an audio power amp, like a TDA 2030. You might want to give it a fixed gain of 2 and then use another to provide the desired high impedance input. Not the best opamp around, but can drive large capacitive loads and has a high output current capacity. In this application it would probably not even need a heatsink. It would need an output zobel network though, and a low value inductor with a damping resistor would not hurt either.

[Galroc]

As of now, I will use a 100k Ohm resistor on the op amps output to match the impedance of the instrument I am trying to drive.

No No, 100k source impedance will create a significant low pass filter with the capacitance of the cable.

Ok...forgot about that, but I don't really need high frequency.

I will test on a breadboard I have on hand and a frequency generator next week.

If the frequency is low and you are wanting high drive ( and have dual rails available to power it) how about using an audio power amp, like a TDA 2030. You might want to give it a fixed gain of 2 and then use another to provide the desired high impedance input. Not the best opamp around, but can drive large capacitive loads and has a high output current capacity. In this application it would probably not even need a heatsink. It would need an output zobel network though, and a low value inductor with a damping resistor would not hurt either.

SeanD, I will look into that one....

I was reading about the AMP01 from Analog Devices today, even though it is expensive, it is designed to drive large capacitive loads.

[Galroc]

After bench testing an NA5534 op amp with a long coax cable, I settled on using Sallen-Key Low-pass filter to eliminate noise on the line. I ended up
with about a 225kHz cut off frequency. I could actually be more aggressive with the filter if I wanted since I only need about 10kHz.

It works really well and I see no issues on my oscilloscope.

Problem solved.

Cheers

[Bambur]

There is this video on MCHP Youtube channel related to the thread.

[Galroc]

Good video and I love to watch these videos since I am learning, but that op amp assumes a 50 Ohm load but I am driving a 100K Ohm impedance. But, the idea to match the 100k Ohm impedance with a 100k resistor and use a gain of 2 was proposed earlier, like shown in the video (except they use a 50 ohm). Except, it was pointed out that it would act like a low pass filter as well.

I just got in an evaluation board for use with an AD8221 instrumentation op amp. The AD8221 is unity gain stable and was able to drive a long coax just fine. No Sallen-Key filter needed to block the reflected signal.  There is a cost difference between an instrumentation op amp and a normal op amp.

My options appear to be to use a cheap op amp with a Sallen-Key filter or a more expensive instrumentation op amp.


Cheers

[SeanB]

NE5534 makes a very good RF transmitter if you connect it to a long cable, especially if the negative supply is not well decoupled. I had a mixer using them that was a pretty good AM transmitter, with the long cabling acting as an antenna. A free AM transmitter at around 1.6MHzwith the outputs running at close to toast levels.

[Bambur]

Good video and I love to watch these videos since I am learning, but that op amp assumes a 50 Ohm load but I am driving a 100K Ohm impedance. But, the idea to match the 100k Ohm impedance with a 100k resistor and use a gain of 2 was proposed earlier, like shown in the video (except they use a 50 ohm). Except, it was pointed out that it would act like a low pass filter as well.

I just got in an evaluation board for use with an AD8221 instrumentation op amp. The AD8221 is unity gain stable and was able to drive a long coax just fine. No Sallen-Key filter needed to block the reflected signal.  There is a cost difference between an instrumentation op amp and a normal op amp.

My options appear to be to use a cheap op amp with a Sallen-Key filter or a more expensive instrumentation op amp.


Cheers

Galroc, you need to think about matching impedances only when the physical length of transmission line is somewhat comparable to the electrical length or wavelength. There is no practical 100KOhm transmission lines (with very special few exceptions). You really should know the principles involved here.

[Galroc]

Galroc, you need to think about matching impedances only when the physical length of transmission line is somewhat comparable to the electrical length or wavelength.

I dropped that line of thought really fast. Look above.

You really should know the principles involved here.

Thus the reason why I asked....

I came up with using a Sallen-Key filter with an appropriate Fc and I bench tested it and I was very pleased. Do you have a better solution? Perhaps you have a favorite unity-gain stable op amp that would be usable in this situation?

Cheers

[jahonen]

I think that the filter is not necessary, and does not fix the basic problem (although it seems to work). 50 ohm series termination resistor at the driving opamp would have done just fine, and the loss could have been compensated using a gain of (50+100k)/100k = 1.0005 at driving end if desired (18 ohms from output to negative input and 36 kohms from negative input to ground).

It should be noted that although signal is not anywhere near that transmission line effects will affect, the driving amplifier might still have enough bandwidth to get upset from that. Thus it is always wise to put that series termination resistor in.

Regards,
Janne

[Rufus]

I think that the filter is not necessary, and does not fix the basic problem (although it seems to work). 50 ohm series termination resistor at the driving opamp would have done just fine.

Gees it isn't a series termination resistor. With 100k at the other end it isn't a transmission line.

The cable capacitance on the output creates another pole in the feedback loop which may cause instability. That is probably what was happening but the OP never really described what his problem was. A resistor in series with the output isolates the cable capacitance reducing the effect of the pole. 50 ohms might not be enough it depends on the opamp.

We don't know where the OP put his filter, presumably he is driving the cable with it. Not a good solution but the but it will work because filter poles will dominate any extra pole on the output.

[Galroc]

I put the Sallen Key filter around the outgoing op amp.

You can go to this link http://sim.okawa-denshi.jp/en/OPseikiLowkeisan.htm to see how it is designed.

R1  and R1 = 10K

C1 and C2 = 100pF

What I am driving with the op amp is a very slow microscope objective and its controller (100K Ohm impedance, not 50 Ohm). Speed isn't an issue, but accuracy is.

Rufus is right, I was getting capacitance from the cable interfering with the feedback signal of op amp causing all sorts of havoc. A 100ohm resistor could have isolated it, but I didn't want the DC scale error.

The op amp would oscillate easily. I need mVolt accuracy. 10 mVolts equals 100 nM at the objective. We use 100 nM steps.

Plus, I don't think the op am I was originally using was unity-gain stable. It was one we already had on hand. I will be using a TC7660. If anyone has any objections, let me know.

Last month I didn't even know the term "unity-gain stable", but I do now! I am a software guy and I am learning as fast as I can... I just sent off the PCB for manufacturing a few hours ago.

Appreciate all the advice..

Cheers!
 

[vk6zgo]

Perhaps I missed something,but you do realise that you can terminate the instrument end with 50 Ohms,& it won't bother the 100k bridged across it?
This is the standard method for using high impedance instruments in a 50 Ohm system.
The 100k doesn't affect the 50 Ohms much, either. :- 50 Ohms in parallel with 100k =49.975 Ohms!

All that should happen,is that you will get half the voltage amplitude that you had before,as you now have a  voltage divider across it.
Any adverse effects on the driver op amp should disappear in the terminated case.

[jahonen]

Gees it isn't a series termination resistor. With 100k at the other end it isn't a transmission line.

A transmission line does not stop being a transmission line if it is not terminated. It is the reflected signal from the mismatched (open) load which makes the line look like capacitive load from driver perspective at low frequencies. Series termination only terminates the line from the load perspective, assuming that the driver stands the impedance it sees. From the line driver perspective, the line is still mismatched. That's why RF stuff tends to be double terminated, it suppresses reflections to either direction, thus it gets best signal fidelity. Also, for power transmission matching is needed. But back to our subject.

Generally, behaviour of a transmission line is quite different from lumped capacitor in general. From interest, I made a small simulation using LTSpice, a comparison between a transmission line and corresponding lumped capacitor. As we see, there are some similarities and some differences.





One thing to note is that termination from the driver perspective could be done here in AC-fashion, just by adding RC at the load end. This makes the impedance essentially resistive for wide frequency range, without causing any DC error. One could infer from capacitor analogy that such thing might only make things worse since the impedance would only drop. This is however not what happens.

Regards,
Janne