Microvolts suffice on a bipolar process. RF gets rectified at the first BE junction and then the rectified signal attacks inside the loop. Meaning at the full open loop gain of the opamp. This is why one often hears rat-tat-tat-POP when placing a cell phone near a stereo.
Perhaps. A low signal levels you don't have much non-linearity. Once you get above about 50mV in bipolars you get effective half-wave rectification. Nothing is being "rectified" but the output looks as if it is.
You get rectification at the microvolt level, there is no 50mV threshold. A BE junction in the conductive region is always non-linear. For all practical purposes the stuff connected to the emitter in an opamp and other ICs can be considered a "heavy load" for RF in the GHz range. Then any ever so small RF signal entering the base will be rectified and the rectified signal drives the works inside the loop.
Hmm. Given that rectification is a quadratic process, it seems as though v ery small signals ought not to be such a big problem. If it takes 25 mV to give you an order-unity perturbation, then 1 uV ought to be nearly 100 dB down, not to mention the rolloff with frequency. N'est ce pas?
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
I've rarely seen RF rectification problems "internal" to a PC board, which is our current situation. The problems are usually in opamps that connect to connector pins, out to the world. All such inputs should be protected against RF and ESD.
Some stuff can't be protected without compromising performance; we warn the users about those.
We did have the problem with the LM3102 switcher, the step-recovery thing, where monster spikes from the synchronous switcher trashed bipolar opamps all over the board. Did you build your board, with the similar part?
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
We had a big problem on two board in my last job. They had a 2.4GHz TDMA transmitter module right next to the audio boards. At first we had no problems but then we the opamp supplier's process improved (our guess, anyway) and the 200Hz TDMA signal came right through the audio. it was a bear to fix.
They are usually in the same ballpark. GSM is the worst. For whatever reason they start the handshake at full power. On-off-on-off-on-off and it sounds like rat-tat-tat. Not sure what you mean with order-unity but the issue is that this goes in at full open loop gain. If we assume a modest 100dB of open loop gain then a 1uV rectified signal turns into
100mV at the output. 25mV would rail the amp hard. In a measurement setup or in audio gear that can seriously ruin the day.
OTOH this problem of bipolar process amplifiers has brought me lots of consulting work of the "Please make this go away" kind 8-)
This problem becomes hardcore in TIAs because there you often need the performance of bipolar RF processes, can't use CMOS. You also can't have any muffling capacitance to ground because then they go berserk. Especially CFBs since they do not like a lot of feedback capacitance to compensate. So you have to shield the heck out of the whole setup and make sure that no cell phones, Wifi devices and whatnot are being operated nearby. Not even an innocent looking Bluetooth mouse.
The point I'm making is that if it takes 25 mV to produce an order-1 change of forward current, which it does, then a 1 uV RF signal will produce only 1/(25000 ** 2)/2! = 7E-10 relative change.
If you're talking about a microvolt of rectified DC, i.e. 1 part in
25000, that makes more sense. But to get that much, you'd need around
sqrt(2*1/25000)*25mV = 2.2 mV, not counting the AC rolloff. That's a fair amount of RF.
Cheers
Phil Hobbs
In a measurement
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
I'm measuring about a 7-ns rise time on the LM3103. It has a pre-step like yours, but no SRD-style behaviour.
The two isolated supplies are quite a bit wimpier than the simulation. :(
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
I am still not quite following. If the BE diode path is operated in the very low current regions, which it usually is, then an excursion of 0.1% in voltage results in a 1% current change. Add in the usual parasitic capacitances and you can have a very efficient RF rectifier.
There is no AC roll-off, only the performance of the BE junction diode path at RF really matters and unfortunately that is often pretty good.
It's quadratic, though. For small swings the two half cycles almost cancel out when it comes to generating DC. To make DC out of sin(omega t) you need an even-order nonlinearity. For the linear term, superposition applies and so it doesn't cause interference with normal operation.
Well, that and the circuitry in front of it.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
How would they cancel out? The nonlinerity makes it favor one half-cycle over the other, thus moving the average substantially. That "hump" in average is what creates the noise into the first device.
There usually isn't much in front of it inside the opamp other than ESD diodes. Except for EMI-hardened versions but the remedies in there only go so far. I have had cases where moving to CMOS was the only option. I have a design right now where we must assume that personnel is on site and using their cell phones. Whether there are warning signs against it or not. They often think "Oh, sending a text really isn't like being on the phone ...".
If you look at the gross diode curve, it looks pretty nonlinear. But the more you zoom in on any region, the more it looks like a straight line, a resistor. The smaller the region, the straighter the line.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
Sure, but ... large absolute change in the current value in the direction up and very small change going down. That's what causes the EMI.
I had cases where the signal was truly miniscule yet caused the ADC result to go off the rocker and the client engineers looked at me as if I was proclaiming voodoo. The best was when I said "I bet someone came down the stairs and is texting on their smart phone while having a smoke outside right now". We could not see or hear the stair well because it was far away. So one of the guys when looking ... "Shazam!"
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.