Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Opamp capacitive load?

Status
Not open for further replies.

TQFP

Junior Member level 3
Joined
Aug 11, 2007
Messages
30
Helped
1
Reputation
2
Reaction score
1
Trophy points
1,288
Location
Florida
Activity points
1,666
Is there a somewhat simple way to determine if an opamp load is capacitive? How do you know, exactly? I am trying to use an opamp as a simple unity gain source follower, but all the datasheets have huge sections on special requirements for capacitive loads.

I can not control the load of the opamp, it is a pre-existing circuit, and not all of them are the same. They are all common in that there is a load resistor of 390-ohm to 590-ohm (varies), plus various capacitors and / or inductors used to develop the signal and couple it to another transistor stage. The signal is composite video, if it makes any difference.

Since all the circuits have at least the 1 resistor (tied to ground), are the capacitors and inductors (which I assume are making a filter) part of the load?

Thanks,
Matthew
 

Yes, there is a very simple way to identify if the load is capacitive by visual inspection of the circuits.

1. Check if a resistor is placed between op-amp output and the load.
This resistor is usually called the isolating or current-limiting resistor.
If there is no resistor, this load is very likely to be capacitive.
If there is a resistor, this load is very likely to be inductive.

2. Check that this load has a Zener diode or Transient Voltage Suppressor, and Bypass Capacitors (in parallel) to the load. If there are any of these, load is very likely to be capacitive.

BTW, unity-gain buffer (more commonly and correctly called, instead of source follower that refers more of MOS transistor configuration) offers medium output impedance around 150k to 200k Ohm. This is sufficient for capacitive load.
 
Thanks for the info. Here are three actual output stages that I will need to be able to drive:

63_1308149516.jpg

85_1308149516.jpg

65_1308149516.jpg


Two of them look capacitive, the other resistive, based on the criteria above. Yes? The "Y" signal is where I would input in the last one, sorry it is backwards...

I was having great fun with a SPICE simulator, testing my basic driver circuit, until I thought about needing an opamp designed for the video signal I'll be dealing with (composite video, about 5MHz). When I did a search for "video opamp", everything got a lot more complicated (and expensive.)

The initial design was simply my signal to the opamp on the + input, the output tied back to the - input, +5V and ground. But the "video" opamps start warning about capacitive loads, output series resistors, feedback resistors based on load, etc. I started glazing over.

Also, these "video" opamps are all in the 80MHz to 300MHz range. Do I really need something like that for composite video?? Composite is only 4.5MHz or so, so why do I need an opamp rated at 80MHz?
 

Image 1 is intended for resistive load as well as inductive load. Not capacitive load.
You will need to modify the circuit by adding a series capacitor (e.g. 1uF which can be polarised or non-polarised) at MONVID to drive capacitive load.

Image 2 is a 2nd order passive bandpass filter. This circuit is intended to filter noises from the signal, thus it has nothing to do with capacitive load.

Image 3 is a single-transistor inverting amplifier with output DC decoupling capacitor. It is intended primarily for signal (like audio, but can also be used for video).
It is still OK for capacitive load.

You mentioned 4.5MHz, I believe you are refering to PAL system. If you include chromas and audio, you should be looking at 8MHz bandwidth.
Then you will need 16MHz or more.
I suggest you look at circuits that can operate around 25MHz.

If possible, check out IC and reference design from MAXIM and Intersil and Philips.
 
Last edited:
You will need to modify the circuit by adding a series capacitor (e.g. 1uF which can be polarised or non-polarised) at MONVID to drive capacitive load.

I'll be generating the signal coming in at COMVID. The circuits shown are all pre-existing in host computers and I have to have a single circuit that can drive them all equally well (or as well as I can get anyway.) I'm replacing an original device (TMS9918A VDP), and these are examples from computers that used that VDP.

Image 2 is a 2nd order passive bandpass filter. This circuit is intended to filter noises from the signal, thus it has nothing to do with capacitive load.

So if I am feeding the signal in at Vin, what kind of load am I looking at?

You mentioned 4.5MHz, I believe you are refering to PAL system. If you include chromas and audio, you should be looking at 8MHz bandwidth.
Then you will need 16MHz or more.

NTSC actually, I was just being vague and over-stating the frequency for a little bit of a buffer. I will not be generating audio. Actually, I'm only going to generate 16 shades of gray, so no color either.

I found some examples on a few websites where composite video can be generated with just two resistors (which I have replicated), however you have to know the load resistance to select the resistor values. Since the load varies in my case, and I can't control that aspect, I started looking at using an opamp. Everything was going nicely until I read the datasheets about capacitive loads and such, for the video opamps.

I suggest you look at circuits that can operate around 25MHz.

If possible, check out IC and reference design from MAXIM and Intersil and Philips.

I've been looking around, but as soon as you search on "video opamp", the frequency specs go to the 80MHz+ ranges. So you suggest I just find an opamp in the 25MHz range without specifically looking for one designed for video?

Thanks for the info, it is much appreciated!
 
BTW, unity-gain buffer (more commonly and correctly called, instead of source follower that refers more of MOS transistor configuration) offers medium output impedance around 150k to 200k Ohm. This is sufficient for capacitive load.
It's very unlikely and even more unwanted to have kohm output impedances for a buffer. I really wonder which circuits or devices you are referring to.

Video equipment is most using 75 ohm impedance matching, preferably at both source and load side. It's the only way to connect cables of certain length without causing double edges or echo images. The fact, that the discusses circuits are designed far from impedance matching súggests, that they are intended for internal short range signal connections only. Connecting a mostly reactive load to an amplifier, either discrete transistor or opamp circuits, brings up resonance peaking or even oscillations.
 
...............
Image 2 is a 2nd order passive bandpass filter. This circuit is intended to filter noises from the signal, thus it has nothing to do with capacitive load.
..............
SkyHigh, second order with 4 reactive elements?
 

When you drive pure capacitive load (typical example is piezo-actuator) will required medium impedance at several k-Ohm.

For many op-amp, you got to choose if you are driving data or video signal.
For video, op-amp for this purpose will have Rout at 75-Ohm by default.
For data, op-amp for this purpose will have Rout at 50-Ohm by default.
The datasheets of various op-amp will tell you these.

For video, 75-Ohm is line-matching, but note this is single-ended unbalanced mode only.
Coaxial is usually done like this.

For longhaul transmission, the approach will be very different. Balanced mode will be adopted for higher SNR, you will need more than 75-Ohm. You need a centre-tapped passive balum usually a T-network to ground with two 75-Ohm on the lines (which can be differential).

It's very unlikely and even more unwanted to have kohm output impedances for a buffer. I really wonder which circuits or devices you are referring to.

Video equipment is most using 75 ohm impedance matching, preferably at both source and load side. It's the only way to connect cables of certain length without causing double edges or echo images. The fact, that the discusses circuits are designed far from impedance matching súggests, that they are intended for internal short range signal connections only. Connecting a mostly reactive load to an amplifier, either discrete transistor or opamp circuits, brings up resonance peaking or even oscillations.
 

Theoretically, 4 reactive elements should give you 4 Poles, but not always USUALLY 4th order.
Look at where the inductors are placed. That's right, in series.
1st node has a shunt C and series L, this is the 1st order that acts as low-pass filter.
2nd node has a series L and shunt R, but this node only servez as Zero, no Pole. In other words, only increasing Z at higher frequency compared to no or little voltage drop at DC.
The 2nd order kicks in with the series C after 2nd node. This will act as the high-pass filter.
With one 1st order high and one 1st order low-pass filter, you get a 2nd order band-pass.

SkyHigh, second order with 4 reactive elements?
 

I suggest you first examine the reference circuits of these IC that operate at 80MHz.
I think it has to do with noise-shaping by oversampling the input signal to achieve higher SNR, especially on the color called chronas (which is very sensitive, compared to Black/White).

Although you chose to design for NTSC (2MHz bandwidth-saving than PAL), the operating principle for composite video is the same.


I'll be generating the signal coming in at COMVID. The circuits shown are all pre-existing in host computers and I have to have a single circuit that can drive them all equally well (or as well as I can get anyway.) I'm replacing an original device (TMS9918A VDP), and these are examples from computers that used that VDP.



So if I am feeding the signal in at Vin, what kind of load am I looking at?



NTSC actually, I was just being vague and over-stating the frequency for a little bit of a buffer. I will not be generating audio. Actually, I'm only going to generate 16 shades of gray, so no color either.

I found some examples on a few websites where composite video can be generated with just two resistors (which I have replicated), however you have to know the load resistance to select the resistor values. Since the load varies in my case, and I can't control that aspect, I started looking at using an opamp. Everything was going nicely until I read the datasheets about capacitive loads and such, for the video opamps.



I've been looking around, but as soon as you search on "video opamp", the frequency specs go to the 80MHz+ ranges. So you suggest I just find an opamp in the 25MHz range without specifically looking for one designed for video?

Thanks for the info, it is much appreciated!
 

Theoretically, 4 reactive elements should give you 4 Poles, but not always USUALLY 4th order.
Look at where the inductors are placed. That's right, in series.
1st node has a shunt C and series L, this is the 1st order that acts as low-pass filter.
2nd node has a series L and shunt R, but this node only servez as Zero, no Pole. In other words, only increasing Z at higher frequency compared to no or little voltage drop at DC.
The 2nd order kicks in with the series C after 2nd node. This will act as the high-pass filter.
With one 1st order high and one 1st order low-pass filter, you get a 2nd order band-pass.

Hi, Skyhigh.
May I give you the 4th order transfer function for the circuit under discussion?
Note that R2 is the input resistance of the BJT.

Numerator (nominator, sorry) : (C2 R1 R2)*s

Denumerator (Denominator):

R1 +(C2 R1 R2 + L2 + L1)*s

+(C1 L1 R1 + C2 L2 R2 + C2 L2 R1 + C2 L1 R2 + C2 L1 R1)* s^2

+ (C2 C1 L1 R1 R2 + C1 L1 L2)*s^3

+ (C2 C1 L1 L2 R2 + C2 C1 L1 L2 R1)*s^4

Regards
LvW
 
Last edited:

Nice calculation. Good for exam. Still... An engineer is trained to be practical and apply knowledge. Just theoretical isn't enough.

Re-look at your calculation again. In practice, capacitors and inductors used in filters are very small values, except for resistors usually in kOhm range.
For example, I can use C1 100nF, C2 300nF, L1 50nH, L2 220nH.
Put these into your calculation.

Voila! Immediately s^3 and s^4 terms are too insignificant compared to s^1 and s^2 terms. Therefore 2nd order stands! :cool:

I hope you were taught this in university.


Hi, Skyhigh.
May I give you the 4th order transfer function for the circuit under discussion?
Note that R2 is the input resistance of the BJT.

Numerator (nominator, sorry) : (C2 R1 R2)*s

Denumerator (Denominator):

R1 +(C2 R1 R2 + L2 + L1)*s

+(C1 L1 R1 + C2 L2 R2 + C2 L2 R1 + C2 L1 R2 + C2 L1 R1)* s^2

+ (C2 C1 L1 R1 R2 + C1 L1 L2)*s^3

+ (C2 C1 L1 L2 R2 + C2 C1 L1 L2 R1)*s^4

Regards
LvW
 

Nice calculation. Good for exam. Still... An engineer is trained to be practical and apply knowledge. Just theoretical isn't enough.

Agreed. (By the way: I didn't calculate by myself; it was a symbolic analysis program).

Re-look at your calculation again. In practice, capacitors and inductors used in filters are very small values, except for resistors usually in kOhm range.

So you compare the values for capacitors/inductors with resistors? Independent on frequency and on their impedances?
Interesting approach! Is this the way of "practical" engineering thinking you were taught?
I am very sorry, but you see that I cannot agree to this. For my understanding you are going to argue that in filters capacitors/inductors are small enough so that their influence can be ignored? In this case, where comes the frequency dependence from?
Did your hear about filters in the MHz range with values in the pF and nH range?

For example, I can use C1 100nF, C2 300nF, L1 50nH, L2 220nH.
Put these into your calculation.


Yes, I did. And the high frequency characteristics is determined - of course (!) -by these reactive values only.

Voila! Immediately s^3 and s^4 terms are too insignificant compared to s^1 and s^2 terms. Therefore 2nd order stands!
I hope you were taught this in university.


I hope you were NOT taught this in university. This would be a bad reputation for your university.
Each relevant textbook tells you that higher order frequency terms come into play always with rising frequencies - even in case of small coefficients.
And you certainly will agree that - in particular for filters - the high-frequency behaviour is of certain interest.

With regards
LvW
 

When you drive pure capacitive load (typical example is piezo-actuator) will required medium impedance at several k-Ohm.

I don't see, how this statement and the succeeding considerations should be related to unity gain buffers. My comment was referring to exactly this statement:
BTW, unity-gain buffer (more commonly and correctly called, instead of source follower that refers more of MOS transistor configuration) offers medium output impedance around 150k to 200k Ohm. This is sufficient for capacitive load.
I don't doubt, that "medium impedance" drivers, and particularly current sources play a role in driving capacitive loads. But they obviously aren't unity-gain buffers, and 150k to 200k is just an arbitrary number.

My impression so far is, that you are telling a lot, but it's neither exactly related to the problem nor well-founded.
 

No, I don't agree.
Let's be subjective, your argument doesn't prove anything, beside theorectical calculation. You have to come back to reality about practical engineering.
The fact is you do not understand that only significant values are considered in any engineering design because the impact is the strongest.

Now, do yourself a favour, run a simulation and you will see that your 3rd and 4th order figures have almost no impact. Your 2nd order figure practically shapes the frequency response.

To make it simple for you to understand, does 5th, 7th and 9th harmonics of a Fourier Series matter, when you have 1st, 2rd and 3rd harmonics?
I can tell you in every college, unless yours don't, you simply accept 1st (2nd if any) and 3rd, ignore 5th onwards.

Back to the topic on filters, 4th and higher orders only considered when you signal really operate at the verge of the -3dB cut-off and beyond. If that happens, I bet your professor will tell you to PRACTICALLY re-select your C and L values to widen the bandwidth, instead of getting you to re-calculate your 3rd and 4th order.

I will end here. Hope this is something for you to re-think about engineering in practical world. Cheers!

Sometimes I wonder if you actually ask someone for a change of $1.21 when he can give you $1.20
What difference does it make for losing 1 cent?
 

I have the values for this circuit, I guess I didn't think it would matter in determining the kind of load...

L202 22uH
L203 8.2uH
C200 82pF
C201 10uF
R200 560-ohms
Q200 2N3906

All the circuits shown above are follow-on stages from the device creating the composite video, which is originally a TM9918A VDP, and now my FPGA-based replacement. I'm just trying to generate composite video that will be compatible with the existing circuitry in all these various computers. The original 9918A has an internal MOS transistor driver with an open source, and expects a 490-ohm external resistor to develop the signal. I could not get a MOS transistor circuit to work for me, so I turned to an opamp since the loads are different in every computer the chip is used in.

Throw normal 75-ohm loads out the window, my circuit is not the end driver going to a monitor. I simply need to develop a linear 0V to 1.92V signal across a variable load, which seems to range from about 390 to 560 ohms *resistive*, but I don't know what the effect the inductors and capacitors will have on the opamp (which is why I posted.)

The first circuit seems to be closest to what the original 9918A datasheet specifies, and looks to be totally resistive, and should not be a problem for the opamp. The other two circuits add the extra inductors and capacitors which complicated the load determination for me.

A basic unity-gain opamp circuit seemed to the be thing I needed, but I made the mistake of reading the datasheet and all the warnings about capacitive loads.
 

Sometimes I really wonder if you are LvW.

Well... I can only tell you to do more analog frontend design to learn and appreciate.
There are more to learn in practical enginering than arguing alot.
Go and read up and DO more on driving piezo actuators to know what's all about. If

Please relate to the topic to help the thread originator TQFP and stop disrupting this thread by asking your questions to please your queries that are not helping TQFP.




I don't see, how this statement and the succeeding considerations should be related to unity gain buffers. My comment was referring to exactly this statement:

I don't doubt, that "medium impedance" drivers, and particularly current sources play a role in driving capacitive loads. But they obviously aren't unity-gain buffers, and 150k to 200k is just an arbitrary number.

My impression so far is, that you are telling a lot, but it's neither exactly related to the problem nor well-founded.
 

TQFP,

Some questions. Did you try out the recommended reference design given in the datasheet?
You said you couldn't get the MOS transistor cct to work, which one do you mean? The recommended circuit or one of your posted images?
 

Please relate to the topic to help the thread originator TQFP and stop disrupting this thread by asking your questions to please your queries that are not helping TQFP.
Honestly, I won't post questions related to your contributions, if I don't doubt their technical correctness and relevance for the original problem as well. But I agree, that commenting your suggestions also doesn't help the original poster and I will stop it now, unless you'll shake electrical theory to the very foundations.

You'll easily see that I'm not LvW. But obviously, we agree about emphazing theoretical correctness.
 

SkyHigh, thank you for some practical hints.
Now I have learned that
(a) A 4th order denominator can be simplified to 2nd order,
(b) small capacitor/inductor values can be - independent on frequency - neglected against resistors,
(c) the influence of frequency terms of higher order can be compared with the influence of higher harmonics,
(d) ... sorry, but your "example" of $1.21 vs. $1.20 is not worth mentioning it here.

.............
To make it simple for you to understand, does 5th, 7th and 9th harmonics of a Fourier Series matter, when you have 1st, 2rd and 3rd harmonics?
................
Sometimes I wonder if you actually ask someone for a change of $1.21 when he can give you $1.20

PS: It was never my intention to "fight" with you.
Instead, in my first comment (posting #7) I only wanted to point out that your simple and lapidar statement in posting#4 was not correct "Image 2 is a 2nd order passive bandpass filter". In fact, it is of 4th order. That's all.
If you think that a simplified view is appropriate - feel free to do it. But it is good engineering practice to be correct as possible - and to explain and verify (!) some simplifications. Otherwise, other readers (beginners!) are confused.
Regards
LvW
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top