Ye olde Op-Amp in Digital gear question again

[Fred Dieckmann]

Re: Re: Ye olde TIM misconception

Konnichiwa,



That maybe so. But what do you do if outside the audio range the noisefloor rises constantly with at least a first and often second order function from all the noiseshaping? Suddenly there is RF noise into the MHz at levels less than 60db below full scale. And your Op-Amp is actually trying to be an accurate lowpass filter and cable driver at that kind of signal.

It matters little what you call the resulting intermodulation products which invariably (eventually) fold back into the audio band. The best protection against such troubles is to use Op-Amp's with exceptionally wide bandwidth. They tend to treat the garbage that trips slower "Audio" Op-Amp's well up as simple signal and tend to amplify it in a quite linear fashion, so our lowpasses actually still work properly and the noise is contained.

Sayonara

The other way to deal with RF intermod is with low transconductance input devices, of which Jfets are a prime example. I have always like Jfet inputs for this reason. Using very fast op amps is likely to amplify and pass on this noise to the next stage as well as introducing more RF noise on the supply rails, where it can end up in other circuits due to lousy PSRR at these frequencies. If you are drop high speed amps into existing designs where power supply bypass is not optimum (and it seldom is) with film bypass caps located a few tenths of an inch from the op amp for example, and you have stability and overshoot issues. You also have conceivably more RFI in the circuit due to the amplification of the RF noise. The layout of the PCB traces to the feedback resistors are often long enough to have parasitics the make stability issues with high speed devices. I have been down this road with substitution of AC logic for HC in digital circuits. Sounded nasty for many of same reasons above.


Drop in very high speed op amps in to an existing layout with these potential headaches? Not on my watch!


Not fast but maybe half fast,
Fred

 

[Fred Dieckmann]

Some articles

high speed op amp ap notes

http://www.analog.com/UploadedFiles/Application_Notes/135208865AN-202.pdf

http://www.analog.com/UploadedFiles/Associated_Docs/556612917a.pdf

http://www.linear.com/pdf/an47fa.pdf

 

[ThorstenL]

Re: Re: Re: Ye olde TIM misconception

Konnichiwa,

The other way to deal with RF intermod is with low transconductance input devices, of which Jfets are a prime example.

This rarely helps that much, UNLESS combined with a sufficiently high GBWP, see OPA627 vs. OPA637 in an I/V position (tried both, 637 if you can fixeruppa the noisegain is much better). And that again means to lineary amplify the noise....

Using very fast op amps is likely to amplify and pass on this noise to the next stage as well as introducing more RF noise on the supply rails, where it can end up in other circuits due to lousy PSRR at these frequencies.

There is something to this, but it tends to be much less a problem than most people think, if you take care to do ONE specific thing....

If you are drop high speed amps into existing designs where power supply bypass is not optimum (and it seldom is) with film bypass caps located a few tenths of an inch from the op amp for example, and you have stability and overshoot issues.

The trick I like for that is to "piggyback" a Epcos (formerly Siemens) Stacked Film MKT Cap directly on top of the Op-Amp, connected between +V & -V. With DIL Cases I tend to use the 1uF unit, which with a residual parasitic inductance of 9nH is about as good as it gets. The short wires to connect the DIL Pins and the inductance of the DIL frame for the Op-Amp are higher than that....

The biggest advantage of this +V to -V bypass is that it does not louse up any groundding schemes, assuming the "designer" actually bothered to do a decent ground layout in the first place (I have been on occasion surprised, if rarely).

You also have conceivably more RFI in the circuit due to the amplification of the RF noise.

Well, if your Op-Amp insted slews and distorts you consider this to be better??!!

Sayonara

 

[ojg]

Re: Re: Re: Ye olde TIM misconception

Drop in very high speed op amps in to an existing layout with these potential headaches? Not on my watch!

Not on my watch either! Unless you can examine the layout of the exisiting board closely and make sure they followed all the rules.

Using these high-speed op-amps makes SMD and 2-4 layer boards absolutely necessary. Always put a capacitor in parallell with the feedback resistor to limit the bandwidth to something sensible above 20kHz. Although the amplifier can amplify RF signals, we don't need that. If long feedback traces are unavoidable, then an additional small cap directly from output to inverting input will help.

Good appnote-links, but the second one did not work for me...

-ojg

 

[Fred Dieckmann]

The missing link

http://www.analog.com/UploadedFiles/Associated_Docs/519453295054447556612917a.pdf

Other good stuff at Technical Support Seminar Materials
www.analog.com

I could not direct link to page.

 

[Fred Dieckmann]

I got timed out on edit so another post..........

"Well, if your Op-Amp insted slews and distorts you consider this to be better??!!"

If you have RF signals this large you probably have other problems that op amp substituiton alone are not going the solve. The discussion was about drop in replacements from my understanding and I have seen some very bad results for this practice. Changing the decoupling is not that easily done correctly by many who would make this drop in change. At at this point you are entering the realm of redesign and hence my references.
RF signals large enough to cause slew limiting in jfet op amps are going to cause even more problem after amplification by a very high speed op amp. Most of the popular jfet audio op amps today have gain bandwidths of 15 MHz or more.

OPA604 20MHz

OPA627 16MHz

AD825 44 MHz

AD823 16 MHz

AD8610 25MHz

I still stand by my original recommendation.