Help with selection of opamp please!

[random77]

Hi all,

I've been trying to determine which opamp will be most suitable for my project. I'm using opamps for amplification and for a bandpass filter. At the moment the amplification is broken up into 3 stages (limited to a gain of 100 for each stage), but I can change it to two stages if I have to. These are the characteristics of the input signal:

- The input signal to the first stage of amplification is sinsusoidal at a frequency of 5kHz and the DC offset is at 2.5V. The actual amplitude of the signal is variable, but for the first stage it is gauranteed to within the supply range which is 0-5V.
- For my application, voltage clipping is not a concern at the output of any stage in amplification.
- Note that for the second and third stages, it's possible that a clipped sinusoid could be going into the input terminals of the opamp.

As a result, I've come up with the following requirements for the amplifier:
- Supply voltage = 0-5V
- Slew rate ≥ 2*pi*5000*5 = 157079V/s = 0.157V/µs
- Gain bandwidth product = 100*5000 = 0.5 MHz
- Common-mode voltage = 2.5V
- Differential input voltage = Supply voltage (not 100% sure about this one though)
- Input voltage range = Supply voltage
- Low noise
- Preferably rail to rail operation

Up until just a few days ago, I've been using the LM833 (datasheet attached) and generally speaking, it's been performing relatively well. However, during the testing I found out that if the input voltage to the second or third stage is not restricted to something less than that of a clipped sinusoid, the output becomes distorted. (i.e. the output is not representative of a clipped sinusoid, but still has SOME characteristics of the right frequency) even though the amplitude going into the input to the opamp is always going to be restricted to the supply voltage)

Sorry, I don't have pics for the distorted output at the moment.

So my question is does the input signal at any stage have to be sinusoidal? Or is there some other specification that I have to consider to select the right opamp?

I also tried the TL972 opamp (datasheet attached) but the output of this is very distorted and does not represent the signal at all - I haven't had much time to check if it was connected properly, but I suspect that it didn't work because the differential input voltage has to be < ± 1V (which I assume does not satisfy my signal)

Please help - I don’t know what to consider!

 

[MrAl]

Hi,


Try a higher voltage supply voltage for the LM833 and see if that helps. It would also help if you would post your entire circuit so we can see how you have it set up and biased.
The higher supply voltage will increase the input common mode range so the amp might behave better.

 

[random77]

Hi,

Here are the schematics for the first stage (first attachment) and second & third stages (2nd attachment) The overall gain according to the schematics is a million but I'll be reducing that by changing the feedback resistor in the second and 3rd stages.

Cheers

 

[CafeLogic]

Those are some pretty tame requirements. Personally, for 5V stuff, I like the microchip line. http://www.microchip.com/ParamChartSearch/chart.aspx?branchID=11015&mid=11〈=en&pageId=79. I have found them to perform well, they have low offset voltage, ridiculous input impedance, and most importantly they are cheap.

Since you've got a gain of 100, I am going to assume you are working with signals well under 50mV. Anytime your doing that in a differential configuration, you've got to watch out for at least two things. 1 in input offset, 2 is common mode error caused by resistor mismatch. Just to give you an example, lets say input 1 is 2.5V and input 2 is 2.55V. In that situation, worst case 1% mismatch would cause a common mode error of 0.1V which is twice your input differential!

So anyway, if I were you, and I wasn't starving for BOM, I would grab myself a MAX4209 100V/V fixed gain amp with a typical offset of 3uV and a typical error of 0.03% for the low-low price of $2.41 USD (no they aren't paying me).

Digi-Key - MAX4209HAUA+-ND (Manufacturer - MAX4209HAUA+)

 

[audioguru]

The datasheet for the LM833 shows a minimum supply of 10V but yours is only 5V so of course it doesn't work properly.
Your second and third opamp stages have no DC input voltage reference (at half the supply voltage) so of course they don't work.

 

[MrAl]

Hi,

Here are the schematics for the first stage (first attachment) and second & third stages (2nd attachment) The overall gain according to the schematics is a million but I'll be reducing that by changing the feedback resistor in the second and 3rd stages.

Cheers

Hi again,


It looks like your circuit would work but you have to raise the supply voltage to more than 5v, like at least 10v or even 15v to get it to work right. The reason is because the op amp you chose does not work correctly at that low voltage. If you really need to use a supply of only 5v then you do need to select another op amp that can work down that low, and if you need rail to rail output then you need to make sure the op amp can do that too.
Try raising the voltage to 10v and see if it works better.

 

[random77]

Hi,

I'm a bit confused. Doesn't it say on the LM833 data sheet I attached that the supply voltage can be ±2.5 to ±15V under "operating conditions"? So as long as the difference between the two supplies is atleast 5V and no more than 30V it should operate correctly?

Also the reason why there's no DC bias required for the second and 3rd stages, is because the output of the first stage is already biased at Vcc/2 (i.e. there's no decoupling cap between each stage)

 

[audioguru]

Hi,

I'm a bit confused. Doesn't it say on the LM833 data sheet I attached that the supply voltage can be ±2.5 to ±15V under "operating conditions"? So as long as the difference between the two supplies is atleast 5V and no more than 30V it should operate correctly?

You are looking at the datasheet from ST Micro. They copy National's LM833, they did not invent it. Their written copy is WRONG because they are Italian and they No Speeky Zee Engrish. None of their graphs show plus and minus 5V like the graphs from National Semi show.
Here is from National Semi:

 

[MrAl]

Hi,

I'm a bit confused. Doesn't it say on the LM833 data sheet I attached that the supply voltage can be ±2.5 to ±15V under "operating conditions"? So as long as the difference between the two supplies is atleast 5V and no more than 30V it should operate correctly?

Also the reason why there's no DC bias required for the second and 3rd stages, is because the output of the first stage is already biased at Vcc/2 (i.e. there's no decoupling cap between each stage)

Hi again,


Actually the common mode input requires as a rule of thumb that the input never gets closer to either supply than 3v. Thus, with a supply voltage of plus and minus 3v (6v total) you could only input 0v (chuckle). With a supply voltage of plus and minus 5 volts (or 10 volts total) you should be able to input up to plus and minus 2 volts (4 volts total range). If you use a 10 volt supply and bias the input to 1/2 Vcc which is 5v then that gives you an input range from 3v up to 7v to work with.
Will you try the darn 10v supply already? (Chuckle)
Use a 9v battery if you have to, or even two 9v batteries in series for 18v.

 

[random77]

Hi again,

Unfortunately, we're not able to use a 10V supply. I'll test it out tomorrow anyway to see if you're right, but for the purposes of our project we're restricted to a single 5V supply.

I actually found that datasheet for the LM833 (national semiconductor) a little while back, and when I saw first saw it I was thinking "Awesome, I can get more gain out of this amplifier than the one from ST micro!" But then I realised it required a higher voltage supply which we can't do. So I was just under the impression that even though they both had the same NAME, they're still different amplifiers..

Anyway, it seems like the LM833 is not suitable in any case. So if I find a rail-to-rail input op amp satisfying all the requirements (I don't really care about the output being rail-to-rail, more just the input being able to cope with rail-to-rail voltages I laid out before) that's not from ST micro, will it be ok?

I think I have found a suitable opamp - the AD8692 (datasheet attached). There is only one concern that I have: they say that the maximum rating for the input voltage is -Vcc -0.3V and +Vcc +0.3V (This is fine)

HOWEVER, under the test conditions for a 5V supply, they say that the maximum input voltage should only be 3.9V (rather than 5V which is what I want) Would it be ok to use or should I try something else?

Cheers

 

[MRCecil]

Hi,

I'm a bit confused. Doesn't it say on the LM833 data sheet I attached that the supply voltage can be ±2.5 to ±15V under "operating conditions"? So as long as the difference between the two supplies is atleast 5V and no more than 30V it should operate correctly?

Also the reason why there's no DC bias required for the second and 3rd stages, is because the output of the first stage is already biased at Vcc/2 (i.e. there's no decoupling cap between each stage)

Don't be confused. The ST datasheet is correct. OnSemi & National's datasheets do not list the range of Vcc. To double check, I ran a sim with the LM833 (Cadence motolib model) with Vcc at both 5V & 10V with 1mv peak input and both came out with a gain of 93.71 in the same configuration you posted for the first amp with a gain of 100. You are correct in your thought that +/-2.5V is equilivant to Vcc=5V with Vee at ground and a pseudo-ground of 2.5V.

Your second statement is also true as you can see in the attachments posted below. The only thing is, if you must have three stages, I would suggest you distribute the gain over all three stages rather than trying to get the lion's share up front. With Vcc limited to 5V, you are set at around a gain of 12.6 per stage before clipping begins at the third stage with an input of 1mv peak. Perhaps it would be cheaper to go to two stages with a gain of ~44.7 each to reach the max gain of 2k at 1mv peak input. Or what ever you need as you left certain parameters out of your description of what it is you are trying to accomplish.

Anyway, and to answer your initial question, I would go with the LM833, which is a very close match to the MC33078 I have often used in designs and a proven stable amp. it may not be a total rail-to-rail amp but it does work fine if not overdriven.

 

[MrAl]

Hello again,


Yes, the input common mode range at Vcc=5v for that device is -0.3 to 3.9v, so you have to stay within that range unless this amp doesnt have a problem with an overdriven input of which data on that i could not find. Couldnt you just use maybe a 50k resistor in series with the input cap and that would limit an input signal of plus and minus 2.5v (5v total) to plus and minus approximately 1.25v, and that added to 2.5 comes up to 3.75v which is below the required 3.9v. That could very well work out ok since you've got plenty of gain.
The amp should work with the extra 50k resistor so you could try that. Looks like the output swings rail to rail so the second stage would have to be limited in a similar manner.
It may even work without the extra 50k resistor as the input can go rail to rail without destruction, so it's just a matter of whether or not the output inverts when the input goes too high. Some amps do some dont. If this does then you have to limit the input, if it doesnt then you dont have to limit the input.
BTW the rating of -0.3 to +0.3 is the max rating which only describes the max input it can take before it gets damaged but that doesnt mean it works as an amplifier above 3.9v as the input range spec states. This means that the ic chip will not get damaged if you go up to 4v input, but it may no longer function properly as a well-behaved amplifier.

 

[random77]

Hi again,

I was planning on distributing the gain evenly over each stage, once I had worked out a suitable gain value. I think it's definitely worthwhile spending the extra couple of dollars on the more reliable opamps in our case, as we have submitted the final PCB design for our project and can't make any significant alterations to the tracks or the number of components.

I have found one that copes with rail to rail inputs/outputs and based on the datasheet, it seems I shouldn't have any problems with it. It's a bit more expensive, (about $5 each) but I definitely want a reliable working project that's slightly more expensive than one that's unreliable FYI the datasheet is attached.

Thanks for all your help and information everyone - I really appreciate it! (way more useful than learning about charge control models..)

 

[MrAl]

Hello again,

Oh yes, that looks like a nice find there. If i needed 0 to 5v input range and have a 5v power supply i would like to see exactly what they have on that data sheet, 0 to 5v input range for a 5v supply! That's nice.
Yes anything from AD is going to be higher priced but they do make some good stuff. Im sure there is a cheaper way to do this too if you ever get around to looking at it again in the future.

 

[random77]

Yes and they brag about how it's the industry's lowest noise precision CMOS amplifier in the description. How much truth there is to that is another story though.

But it is a fair bit more expensive - approximately 5 times more than the LM833. If I was going to redesign the circuit, it would definitely be worthwhile limiting the input voltage so that there is a wider (and cheaper) range of opamps to use.

Btw, I tried using a 10V supply for the LM833, but I think in the process of testing it before, I might have damaged the opamp as the output was different again.

Thanks again