How do scope front ends automatically attenuate "dangerous" voltages for the ADC

[JoeN]

Most ADCs out there these days run at 3.3V or so, sometimes less.  Good scopes allow 400V input swings with automatic attenuation.  Obviously the input is scaled for the ADC, and it isn't simply scaled by a 1:300 divider all the time.  So how is it scaled into the ADC's range to get the best resolution on the actual input signal and also how is that done quickly enough to protect the ADC?  Any ideas?  I read this for fun (to a state of understanding what is going on perhaps 50%):

http://www.ti.com/lit/ug/tiduba4/tiduba4.pdf
http://www.ti.com/lit/df/tidrjw1/tidrjw1.pdf

TI's best idea for a "50-Ohm 2-GHz Oscilloscope Front-end Reference Design" is simply that the scope be "maximum input signal of ±3V".  Obviously a real scope front end has something in front of all this to scale the voltage, right?  10:1 probe isn't enough.

Help me think!  Everything I think of some HV gets through for at least a little white if the attenuation circuit sees high voltage after low and has to switch in more attenuation.  Thanks.

I wonder how my multi-meter does this so well also.  1000VDC/750VAC autoranging for that instrument.  I assume it is the same principle.

[EEVblog]

Scope (and multimeters) have input resistor dividers going into the ADC. The ADC (input amp actually ) has diode clamping on the input to prevent overload. The series protection resistance of the input ensures only limited current can flow, so no damage (up to the maximum specified, say 300V for scopes typically) regardless of the input range selected.

[EEVblog]

[ludzinc]

I also wanted to understand how to do this - here's my efforts in making a range switching meter. 

http://ludzinc.blogspot.com.au/2014/07/microcontrolled-analogue-gain.html

http://ludzinc.blogspot.com.au/2014/07/multi-channel-meter.html

[timb]

Calculate how much current can flow through 10Mohm at 400V. It's not a lot!

Here's an example of how protection is implemented.

First, we apply 25V to a 10x probe:



Now we'll up the voltage to 250V:



Notice how the input to the amp never goes above 5.5V? That's because the protection diode kicks in when the voltage goes above 5V (plus a diode drop). This works because the current is being limited through the impedance of the probe and the scope's attenuation network. So, only a few microamps are allowed to flow, which the protection diodes (and ultimately the 5V rail) absorb.

[JoeN]

I'll have to watch that vid to get more out of this.  Right now, I understand the protection angle.  Unfortunately, as far as i can see, a 200Vpp sine wave input going through this scheme now looks like a 10Vpp square wave input (with a bad rise time) when it hits the ADC.   Obviously there needs to be some scaling in front of the protection still.  I guess it will be a series of relay-selected attenuators and maybe a "worst case" divider to divide the input voltage down by 100 and let a MCU select which attenuator to switch in via its ADC.  Then despite whatever time it takes to kick in the attenuation, the protection will make sure the ADC pegs but doesn't break.  Basic idea correct?

I wonder why TI didn't just add this on the front of their design if it is so easy to do?  Their attenuation is limited.  Maybe it is because of the high cost of the relays they are using to preserve the high bandwidth of their circuit?  (the relays they use are about $50 each!  RF180-5 http://www.teledynerelays.com/pdf/electromechanical/RF180.pdf).

[timb]

I'll have to watch that vid to get more out of this.  Right now, I understand the protection angle.  Unfortunately, as far as i can see, a 200Vpp sine wave input going through this scheme now looks like a 10Vpp square wave input (with a bad rise time) when it hits the ADC.

No, you obviously don't understand the protection angle. Look at the second picture in my previous post. If you put a 200Vpp signal into it, you can't get more than a 5.5Vpp signal at the OP Amp (assuming you replace the ground connection to the bottom diode with -5V). Also, the +-5V rails can be replaced with 3V or 2.5V or whatever the input OP Amps run off of.

[JoeN]

I'll have to watch that vid to get more out of this.  Right now, I understand the protection angle.  Unfortunately, as far as i can see, a 200Vpp sine wave input going through this scheme now looks like a 10Vpp square wave input (with a bad rise time) when it hits the ADC.

No, you obviously don't understand the protection angle. Look at the second picture in my previous post. If you put a 200Vpp signal into it, you can't get more than a 5.5Vpp signal at the OP Amp (assuming you replace the ground connection to the bottom diode with -5V). Also, the +-5V rails can be replaced with 3V or 2.5V or whatever the input OP Amps run off of.

Yep, I get it.  You saved the op amp.  For sure.  No question.  You also destroyed the signal.  For sure.  No question.  97.5% of it is gone.  But, like I was saying, if the attenuation precedes that terminal protection then you get both the protection and the preservation of the signal, assuming the attenuator is done right.

[timb]

I'll have to watch that vid to get more out of this.  Right now, I understand the protection angle.  Unfortunately, as far as i can see, a 200Vpp sine wave input going through this scheme now looks like a 10Vpp square wave input (with a bad rise time) when it hits the ADC.

No, you obviously don't understand the protection angle. Look at the second picture in my previous post. If you put a 200Vpp signal into it, you can't get more than a 5.5Vpp signal at the OP Amp (assuming you replace the ground connection to the bottom diode with -5V). Also, the +-5V rails can be replaced with 3V or 2.5V or whatever the input OP Amps run off of.

Yep, I get it.  You saved the op amp.  For sure.  No question.  You also destroyed the signal.  For sure.  No question.  97.5% of it is gone.  But, like I was saying, if the attenuation precedes that terminal protection then you get both the protection and the preservation of the signal, assuming the attenuator is done right.

The attenuation would be between the section labeled "Scope" and "Input Amp" in my diagram.

Basically, you need to have the fixed 1M resistor there. After that you'd have switchable attenuation (with a "no attenuation" pass through as well) which feeds into an amplifier of some sort that has selectable gain. That is then fed into the ADC.

And yes, if you've got the scope in the 5mV/div range you'd "destroy the signal" but that goes without saying. What exactly would you except to happen? How would the signal be preserved? (Of course, if the scope was on the 50V/div setting the signal would be fine...)

[JoeN]

And yes, if you've got the scope in the 5mV/div range you'd "destroy the signal" but that goes without saying. What exactly would you except to happen? How would the signal be preserved? (Of course, if the scope was on the 50V/div setting the signal would be fine...)

Yes, I get it.  I was thinking all the way ahead to the dreaded "auto" button where the scope figures out the best attenuator to switch in.  In manual mode, if you have the wrong attenuator selected, I would totally expect the clipping to be a normal result.  I was also thinking from the perspective of my auto-ranging multimeter which does that by default.  I am thinking about this front end in a general way for all types of voltage measurement.  My initial question was to be able to have both that attenuation and not blow something out when it is in the wrong position.  The first answers answered the protection question and left the attenuation question hanging, especially with regard to auto-selection.  I had to think about it for a minute to see how to do both.   Now, how to preserve the bandwidth of the signal while running it through all of these components?  At least the TI schematic has some great ideas on that - use only really expensive parts. 

Thanks guys, I think I have a start here.

[matseng]

Detecting that the signal is "destroyed" is easily done by just scanning for values for hitting the limits of the A/D converters output range.

[tggzzz]

Calculate how much current can flow through 10Mohm at 400V. It's not a lot!

Here's an example of how protection is implemented....

I know you realise this, but the OP might not.

You are referring to 400V DC. If AC, then the parallel capacitor in the probe has to be considered. At 10kHz at 15pF capacitor's impedance is only 1Mohms, i.e. <<10Mohms.

Always check the scope and probe's frequency-dependent ratings.