Differential ADC inputs

[Chris Holmes]

Hi,

I'm currently designing a simple digital oscilloscope as a hobby
project and am having some trouble with differential ADCs.

I picked out an AD9283 50 MSPS converter for the 'scope. It's a 3V
single supply part that takes a +-512mV differential input. I understand
that differential inputs are pretty much the standard on fast ADCs in
order to reduce noise.

What's confusing me is that the datasheet says the common mode
voltage can only be +-200mV maximum. I can program the common mode
voltage on the differential amplfier I'm using as a driver (AD8138), so
if I keep it simple and use a 0V common mode voltage, then one of the
differential inputs will go below ground. Is that normal with these
parts, even with a single supply? I need to have the converter DC
coupled in order to make a 'scope, but I'm unable to provide adequate
biasing to keep both inputs positive if the common mode voltage has to
be less than 200mV.

I looked around for dual supply ADCs at Analog, and they all seem to
be slow! What gives? I must be missing something obvious here (being
somewhat new to differential signaling).

Thanks for any tips!

Chris

 

[Pooh Bear]

Hi,

I'm currently designing a simple digital oscilloscope as a hobby
project and am having some trouble with differential ADCs.

I picked out an AD9283 50 MSPS converter for the 'scope. It's a 3V
single supply part that takes a +-512mV differential input. I understand
that differential inputs are pretty much the standard on fast ADCs in
order to reduce noise.

What's confusing me is that the datasheet says the common mode
voltage can only be +-200mV maximum. I can program the common mode
voltage on the differential amplfier I'm using as a driver (AD8138), so
if I keep it simple and use a 0V common mode voltage, then one of the
differential inputs will go below ground. Is that normal with these
parts, even with a single supply? I need to have the converter DC
coupled in order to make a 'scope, but I'm unable to provide adequate
biasing to keep both inputs positive if the common mode voltage has to
be less than 200mV.

I looked around for dual supply ADCs at Analog, and they all seem to
be slow! What gives? I must be missing something obvious here (being
somewhat new to differential signaling).

Without looking at the data sheets - something sounds wrong.

If it's single supply the common mode input clearly can't be negative
*unless* the'res on-chip negative volts generation !

Which it isn't.

Ok - you got me - the analog inputs are pre-biased to approx 1 V by on-chip
resistors.


Analog Input
The analog input to the ADC is fully differential and both inputs
are internally biased. This allows the most flexible use of ac or dc
and differential or single-ended input modes. For peak performance
the inputs are biased at 0.3 × VD. See the specification table for
allowable common-mode range when dc coupling the input.
The inputs are also buffered to reduce the load the user needs to
drive. For best dynamic performance, the impedances at AIN and
AIN should be matched. The importance of this increases with
sampling rate and analog input frequency. The nominal input
range is 1.024 V p-p.


The driver will need to do level-shifting. I assume that the AD8138/9 does
this.

Graham

 

[Mac]

Hi,

I'm currently designing a simple digital oscilloscope as a hobby
project and am having some trouble with differential ADCs.

I picked out an AD9283 50 MSPS converter for the 'scope. It's a 3V
single supply part that takes a +-512mV differential input. I understand
that differential inputs are pretty much the standard on fast ADCs in
order to reduce noise.

What's confusing me is that the datasheet says the common mode
voltage can only be +-200mV maximum. I can program the common mode
voltage on the differential amplfier I'm using as a driver (AD8138), so
if I keep it simple and use a 0V common mode voltage, then one of the
differential inputs will go below ground. Is that normal with these
parts, even with a single supply? I need to have the converter DC
coupled in order to make a 'scope, but I'm unable to provide adequate
biasing to keep both inputs positive if the common mode voltage has to
be less than 200mV.

I looked around for dual supply ADCs at Analog, and they all seem to
be slow! What gives? I must be missing something obvious here (being
somewhat new to differential signaling).

Thanks for any tips!

Chris

Chris,

I read through the datasheet for the ADC. Your analysis seems to be
correct. Since the maximum allowable common-mode voltage is 0.2 V, and the
maximum swing is +/- 0.512 V, one of the inputs MUST be driven below
ground to get a full range swing on the differential input.

There are a few things you could do that I can think of:

1) Find another differential ADC where the common mode range is
greater than or equal to 50% of the maximum differential input
swing. This would save you from having to drive below ground.

2) Provide +/- 5V supplies to the amplifier in front of the ADC so that
you can put the common mode voltage at, say, 0.15 V. (The supplies
don't have to be +/- 5 V. They could be +/- 3.3 or 2.7 or whatever.)

3) Create a false ground at some intermediate voltage just for the ADC.
This way, the amp in front won't have to drive below its ground.

4) Accept that you will not get a full range swing.

5) See if you can figure out how to reduce the voltage input range
so that the maximum code is produced with a smaller swing. Most
likely you could do this by tweaking VREF in. I didn't see anything
about this in the datasheet, however. Mabye it's not even be
possible.

Good luck!

--Mac

 

[Pooh Bear]

I read through the datasheet for the ADC. Your analysis seems to be
correct. Since the maximum allowable common-mode voltage is 0.2 V, and the
maximum swing is +/- 0.512 V, one of the inputs MUST be driven below
ground to get a full range swing on the differential input.

You read wrong.

The inputs are pre-biased at about 1 Volt above ground.


Graham

 

[Mac]

You read wrong.

The inputs are pre-biased at about 1 Volt above ground.


Graham

I didn't read wrong.

Now that I've gone back and looked at the datasheet again, I do see where
the inputs are pre-biased at 0.3 VDD, but I also see where it (still) says
that the common-mode voltage range is +/- 0.2 V.

In my opinion, the author of the data sheet has some explaining to do.

--Mac

 

[Pooh Bear]

I didn't read wrong.

Now that I've gone back and looked at the datasheet again, I do see where
the inputs are pre-biased at 0.3 VDD, but I also see where it (still) says
that the common-mode voltage range is +/- 0.2 V.

In my opinion, the author of the data sheet has some explaining to do.

Yeah - the common mode range is clearly specified in an ambiguous manner.

I wonder if they mean that the quiescent state of the Ain inputs can lie within
Vd/0.33 (+/- 200mV) ?


Graham

 

[Glenn Gundlach]

Hi,

I'm currently designing a simple digital oscilloscope as a hobby
project and am having some trouble with differential ADCs.

I picked out an AD9283 50 MSPS converter for the 'scope. It's a 3V
single supply part that takes a +-512mV differential input. I understand
that differential inputs are pretty much the standard on fast ADCs in
order to reduce noise.

What's confusing me is that the datasheet says the common mode
voltage can only be +-200mV maximum. I can program the common mode
voltage on the differential amplfier I'm using as a driver (AD8138), so
if I keep it simple and use a 0V common mode voltage, then one of the
differential inputs will go below ground. Is that normal with these
parts, even with a single supply? I need to have the converter DC
coupled in order to make a 'scope, but I'm unable to provide adequate
biasing to keep both inputs positive if the common mode voltage has to
be less than 200mV.

I looked around for dual supply ADCs at Analog, and they all seem to
be slow! What gives? I must be missing something obvious here (being
somewhat new to differential signaling).

Thanks for any tips!

Chris

Go back to that AD website and look up this chip.
http://www.analog.com/UploadedFiles/Data_Sheets/67526251AD8138_e.pdf
GG

 

[Mac]

Mac wrote:
[snip]

In my opinion, the author of the data sheet has some explaining

to do.

Yeah - the common mode range is clearly specified in an ambiguous
manner.

;-)

I wonder if they mean that the quiescent state of the Ain inputs can lie
within Vd/0.33 (+/- 200mV) ?


Graham

Yeah, that's about the only thing that makes sense. If I were doing a
DC-coupled design with this part, I would get an analog devices
applications engineer on the phone to clear it all up.

For an AC-coupled design, it wouldn't really matter, since the inputs
are biased to a place that makes them happy.

--Mac

 

[Chris Holmes]

Mac wrote:

[snip]

In my opinion, the author of the data sheet has some explaining

to do.

Yeah - the common mode range is clearly specified in an ambiguous
manner.


;-)


I wonder if they mean that the quiescent state of the Ain inputs can lie
within Vd/0.33 (+/- 200mV) ?


Graham

Yeah, that's about the only thing that makes sense. If I were doing a
DC-coupled design with this part, I would get an analog devices
applications engineer on the phone to clear it all up.

Yep, I checked with an applications engineer who confirmed that the
common mode range is Vd * 0.3 (+/- 200mV).

For an AC-coupled design, it wouldn't really matter, since the inputs
are biased to a place that makes them happy.

--Mac

Thanks for everyone's help on this!

Chris

 

[Ian Buckner]

Mac wrote:
[snip]

In my opinion, the author of the data sheet has some explaining

to do.

Yeah - the common mode range is clearly specified in an ambiguous
manner.

;-)

I wonder if they mean that the quiescent state of the Ain inputs can lie
within Vd/0.33 (+/- 200mV) ?


Graham

Yeah, that's about the only thing that makes sense. If I were doing a
DC-coupled design with this part, I would get an analog devices
applications engineer on the phone to clear it all up.

For an AC-coupled design, it wouldn't really matter, since the inputs
are biased to a place that makes them happy.

--Mac

Looks pretty clear to me. Figure 2 in the datasheet shows
that Ain and Ainbar are internally biassed to just over 30%
of the rail voltage, roughly 1 volt. The schematic for the
evaluation board on p11 shows the inputs as AC coupled,
so the bias is set by the internal resistors.
The +/-200mV CM range simply means that if you want
to DC couple, overriding the on-chip bias, you need to make
sure you finish up with 30% of the rail voltage +/-200 mV
on the inputs for no input signal.

Regards
Ian

 

[Mac]

Mac wrote:
[snip]

In my opinion, the author of the data sheet has some explaining to do.

Yeah - the common mode range is clearly specified in an ambiguous
manner.

;-)

I wonder if they mean that the quiescent state of the Ain inputs can lie
within Vd/0.33 (+/- 200mV) ?


Graham

Yeah, that's about the only thing that makes sense. If I were doing a
DC-coupled design with this part, I would get an analog devices
applications engineer on the phone to clear it all up.

For an AC-coupled design, it wouldn't really matter, since the inputs
are biased to a place that makes them happy.

--Mac

Looks pretty clear to me. Figure 2 in the datasheet shows
that Ain and Ainbar are internally biassed to just over 30%
of the rail voltage, roughly 1 volt.

That has already been established.

The schematic for the
evaluation board on p11 shows the inputs as AC coupled,
so the bias is set by the internal resistors.

I said, in the message you are replying to, that I wouldn't
worry about the AC-coupled case.

The +/-200mV CM range simply means that if you want
to DC couple, overriding the on-chip bias, you need to make
sure you finish up with 30% of the rail voltage +/-200 mV
on the inputs for no input signal.

While that is true, and has now been confirmed by the OP who contacted an
applications engineer at AD, it is NOT what the datasheet says.

When no other reference is specified, voltages are always with respect to
ground. In my first reading of the datasheet, I should have realized that
the +/- 200 mV spec was ridiculous, so I concede to being a fool.

But your interpretation was just a guess. Calling AD was the right thing
to do in this case, and in any case where the datasheet doesn't make sense.

Regards
Ian

--Mac

 

[Pooh Bear]

But your interpretation was just a guess. Calling AD was the right thing
to do in this case, and in any case where the datasheet doesn't make sense.

I agree wholeheartedly.

If you're not sure, you shouldn't ever be afraid to ask ! No-one should ever
think the worse of 'you' for so doing and 'you' could avoid some embarrassing
mistakes.

In this instance - had I been the designer, I would have phoned AD myself to
confirm my interpretation.

Graham

I only wish ( in a current project ) that the manufacturer could actually
*support* some of the more advanced functions of the device in question !