Take the 2-minute tour ×
Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. It's 100% free, no registration required.

I have a rather strange problem. I have a breadboard circuit on a double sided PCB with plenty of ground plane and a good well decoupled power supply. I am amplifying a small medium frequency signal (mV, hundreds of kHz) and bandpass filtering the region of interest.

Initially, I thought to use the AD829 - very low noise, plenty of bandwidth. The problem appears to be the it is too good! I was getting lots of noise especially at high frequencies leaking through. So, changed the AD829s with the cheap TL051 (10x worse noise spec, 30x lower bandwidth). Problem solved - noise gone!

Now obviously the problem has disappeared because the TL051 cannot handle much in the way of high frequency, which is fine. What worries me is two things. First, how legitimate is it to use low spec components to "hide" a problem like this. And second, do manufacturers ever "upgrade" the spec of their amplifiers? One day, will I put in a TL051 that has been improved and discover the product now longer works?

This is the bandpass section which really looks bad with AD829s

enter image description here

share|improve this question
    
The AD829 'likes' to see 50 ohm's. Your relatively larger impedances will mean that all the (not shown) capacitances can have more of an effect. (for the invertering configuration, input C at the inverting input can cause gain peaking.) A spectrum of you noise might tell you something. –  George Herold Sep 2 at 2:16

3 Answers 3

The question is "is the extra noise you saw relevant to what you are trying to achieve"? In other words is the noise in the bandwidth of interest of your signal? If it isn't, can it still cause you problems? Also, are you sacrificing anything by using the TL051 - if not then use it but there are other considerations.

Op-amps may specify so many nano volts per root hertz but, there is also the L.F. noise (usually specified as so many micro-volts peak-to-peak between 0.1Hz and 10Hz). Was the "so-called" good device actually worse than the TL051 in this area. BTW I haven't looked so I don't know but you shouldn't make assumptions.

The other equally important noise source is due to input bias currents and these are usually in the pico-amps per root hertz range. If you are using large input/feedback resistors this noise can dominate and, you might find that the TL051 is significantly better than the AD829 in this area. Again, I haven't looked but it's always best not to make assumptions.

EDIT - the TL051 is 150x better than the AD829 for input current noise. TL051 is 0.01pA per sqrt(Hz) whilst the AD829 is 1.5pA per sqrt(Hz). This will make a significant difference if the size of resistors on inputs and feedback are 1kohm or above. At 1kohm, the AD829's input current noise becomes about the same level as its input voltage noise and at 10kohm it becomes comparable with the TL051's voltage noise.

Power supply rejection ratio is another thing to consider - if there is noise on the power supply (as there inevitably is), which device is better at rejecting it?

Common mode rejection - if the amp configuration is differential, you would naturally assume common-mode noise would be always near zero - not so, check the data sheets and you will see that C.M.R. ratio gets far worse at higher frequencies.

I'm just trying to say that it's always a more complex picture!!!

share|improve this answer
    
Well, I suppose my real question is how solid is a design going to be if it relies on using cheapo parts because they work better than the "real" ones? If I had started by using TL051 then this question here would never have arisen, because I would not have realized there could be a problem in the first place. –  user32885 Jun 25 at 9:39
    
I checked the two devices and the input bias current noise of the TL051 is 150 times lower than the AD829 - this may, in your circuit make it the preferred device. If you post the relevant part of the circuit it might become clearer. –  Andy aka Jun 25 at 9:44
    
Quick question - how do I do that? Screenshot jpg? –  user32885 Jun 25 at 9:53
    
Yes - there is a button during edit you can use for embedding a pic –  Andy aka Jun 25 at 10:21

From your schematic, it looks like you're not using the compensation pin for the AD829. The datasheet recommends using up to as much as 68pF on this guy when running with low gain. Since this chip is so fast, you can often have unintended positive feedback at high frequencies due to the layout.

I would recommend examining whether or not you should be using the pin 5 compensation in this application.

share|improve this answer
    
Good point. Elsewhere the 829 is used with gain > 5 –  user32885 Sep 2 at 7:18

You can generally count on the "typical" parameters not varying too much from the published specifications, but that might be 3:1, so it's unwise to depend on them too much. If that sounds vague, it is. They're not going to make a 3MHz amplifier like the TL051 work at 100MHz, otherwise many applications would experience undesirable behavior such as (unwanted) oscillation. Unity gain bandwidth is probably one of the better controlled, yet unspecified, characteristics- I would expect maybe +/-30% to 3 sigmas but that's just a guess, and worth no more than what you paid for it. It typically varies 10% with temperature on top of unit-to-unit variations.

Comparing noise from a bipolar video amplifier with a JFET audio amplifier is going to depend a lot on your circuit. It sounds like there is noise that is either self-generated by other parts of the circuit (maybe switchmode power supply) or external sources that are making it through your bandpass filter, rather than intrinsic amplifier noise.

One pet peeve of mine is that they often don't state the 1/f corner frequency for current noise, which means that you can't really predict how much the total RMS noise will be if it's less than the stated test frequency, They also often truncate the noise voltage curve to below the frequencies where multistage amplifiers have an enormous hump in the noise. For example, the excellent LT1028 has noise about 7:1 higher at 300kHz than at lower frequencies in the white noise region.

Another gotcha is nonlinear behavior in the presence of strong noise- bipolar amplifiers tend to be worse than JFET amplifiers.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.