Highly Linear ADC Driver Amplifier - need suggestions

[Darol Klawetter]

I'm currently researching the amplifier options for the output stage of a RF receiver I'm designing. This stage will have a single-ended output and will drive an ADC that has a SFDR of 100 dB. Of course, I need a highly linear amplifier to prevent washing out my ADC performance, but I don't want to use one with differential I/O, which is typically used to drive high-performance ADCs. I would prefer something simple, like a gain block that would reduce my parts count. The frequency range of my output is 2 MHz to 50 MHz, which is too low for most RF gain blocks' optimal frequency range. My goal is to drive 2.5 vpp, while preventing distortion that kills the SFDR of my ADC. A gain of 15 dB to 30 dB is acceptable. Any suggestions?

Thanks,

Darol Klawetter

 

[Joerg]

I'm currently researching the amplifier options for the output stage
of a RF receiver I'm designing. This stage will have a single-ended
output and will drive an ADC that has a SFDR of 100 dB. Of course, I
need a highly linear amplifier to prevent washing out my ADC
performance, but I don't want to use one with differential I/O, which
is typically used to drive high-performance ADCs. I would prefer
something simple, like a gain block that would reduce my parts count.
The frequency range of my output is 2 MHz to 50 MHz, which is too low
for most RF gain blocks' optimal frequency range. My goal is to drive
2.5 vpp, while preventing distortion that kills the SFDR of my ADC. A
gain of 15 dB to 30 dB is acceptable. Any suggestions?

AD603?

http://www.analog.com/static/imported-files/data_sheets/AD603.pdf

Then you can set the gain via a control voltage. These things are very
good, I used them in ultrasound systems where signal integrity is very
critical. Unfortunately they have become a bit expensive for my taste so
I haven't used one in new designs in a while.

 

[Darol Klawetter]

I'm currently researching the amplifier options for the output stage of aRF receiver I'm designing. This stage will have a single-ended output and will drive an ADC that has a SFDR of 100 dB. Of course, I need a highly linear amplifier to prevent washing out my ADC performance, but I don't want to use one with differential I/O, which is typically used to drive high-performance ADCs. I would prefer something simple, like a gain block that wouldreduce my parts count. The frequency range of my output is 2 MHz to 50 MHz, which is too low for most RF gain blocks' optimal frequency range. My goal is to drive 2.5 vpp, while preventing distortion that kills the SFDR of my ADC. A gain of 15 dB to 30 dB is acceptable. Any suggestions?



Thanks,



Darol Klawetter

Thanks for the suggestions. My ADC card (which is separate from my RF receiver card) already has a balun that is used to drive the diff input of the ADC from a single-ended SMA input. I want to drive this single ended input without having to use a diff output driver into another balun. I'll do it ifnecessary to get the performance I need, but I'm trying keep my parts count down on receiver's output stage.

Joerg, the AD603 has 60 dBc harmonic distortion.

 

[Tauno Voipio]

Thanks for the suggestions. My ADC card (which is separate from my RF receiver card) already has a balun that is used to drive the diff input of the ADC from a single-ended SMA input. I want to drive this single ended input without having to use a diff output driver into another balun. I'll do it if necessary to get the performance I need, but I'm trying keep my parts count down on receiver's output stage.

Joerg, the AD603 has 60 dBc harmonic distortion.

It seems to me that you're looking for a high dynamic-range HF
front end amplifier. Whole volumes have been written about
the subject. You should start with the ARRL Handbook and
follow the leads (if you have not done so).

I strongly suspect that you need to go to a discrete
component design. The power levels needed to handle most
of the HF range will be much for integrated designs.

 

[Darol Klawetter]

No, I'm just looking for an acceptable single-ended driver for my IF output stage.

 

[Darol Klawetter]

There are lots of opamps that would work.





--



John Larkin Highland Technology, Inc



jlarkin at highlandtechnology dot com

http://www.highlandtechnology.com



Precision electronic instrumentation

Picosecond-resolution Digital Delay and Pulse generators

Custom laser drivers and controllers

Photonics and fiberoptic TTL data links

VME thermocouple, LVDT, synchro acquisition and simulation

Hmmm...maybe so, but I haven't found one with distortion products that are lower than my ADC noise floor. Can I live with that? Probably, but I would prefer to use an amp that doesn't degrade my ADC performance.

Darol Klawetter

 

[josephkk]

I'm currently researching the amplifier options for the output stage of a RF receiver I'm designing. This stage will have a single-ended output and will drive an ADC that has a SFDR of 100 dB. Of course, I need a highly linear amplifier to prevent washing out my ADC performance, but I don't want to use one with differential I/O, which is typically used to drive high-performance ADCs. I would prefer something simple, like a gainblock that would reduce my parts count. The frequency range of my outputis 2 MHz to 50 MHz, which is too low for most RF gain blocks' optimal frequency range. My goal is to drive 2.5 vpp, while preventing distortionthat kills the SFDR of my ADC. A gain of 15 dB to 30 dB is acceptable. Any suggestions?

Thanks,

Darol Klawetter

Read the thread. Listen up, the high performance ADCs use differential
for good reasons. Quit goofing around and go fully differential and dump
the baluns. 100 dB is one heck of a range (about 26 bits), so whatever
the converter cost figure on at least 20% of that for the input amplifier,
maybe even over 300% as much. If it is differential input then give it
what it wants and is designed for. Then use really good layout or you will
waste all that capability.

?-)

 

[Robert Macy]

Read the thread.  Listen up, the high performance ADCs use differential
for good reasons.  Quit goofing around and go fully differential and dump
the baluns.  100 dB is one heck of a range (about 26 bits), so whatever
the converter cost figure on at least 20% of that for the input amplifier,
maybe even over 300% as much. If it is differential input then give it
what it wants and is designed for. Then use really good layout or you will
waste all that capability.

?-)

26 bits?, why not more like 18 bits?

 

[josephkk]

26 bits?, why not more like 18 bits?

100 dB SFDR. Linear at well better than 1 part in 10^5 in voltage,
including noise contribution. Then again 1 part in 10^6 is only 20 bits,
but spur free dynamic range at some given sample rate needs some bits of
excess to average out the inaccuracies. Maybe only 20 bits, but at well
over 100 MS/s what ADCs do you know of that are really linear for that
many bits. It usually takes 3 more bits minimum to control the
quantitization noise to a worthwhile level.

?-)

 

[Robert Macy]

100 dB SFDR.  Linear at well better than 1 part in 10^5 in voltage,
including noise contribution.  Then again 1 part in 10^6 is only 20 bits,
but spur free dynamic range at some given sample rate needs some bits of
excess to average out the inaccuracies.  Maybe only 20 bits, but at well
over 100 MS/s what ADCs do you know of that are really linear for that
many bits.  It usually takes 3 more bits minimum to control the
quantitization noise to a worthwhile level.

?-)

thank you for the explanation, didn't pay attention to that "SFDR"
ignored it entirely. at least 20+3 is less than 26

It's just that I'm used to routinely using my 24 bit system out to 22
bits and almost all of that damage comes form the supporting
electronics, didn't know at MS/s would need so much 'extra' overhead.

 

[Darol Klawetter]

Read the thread. Listen up, the high performance ADCs use differential

for good reasons. Quit goofing around and go fully differential and dump

the baluns. 100 dB is one heck of a range (about 26 bits), so whatever

the converter cost figure on at least 20% of that for the input amplifier,

maybe even over 300% as much. If it is differential input then give it

what it wants and is designed for. Then use really good layout or you will

waste all that capability.



?-)

As I began researching my amp options, I did decide to go with a differential amplifier. I was hoping I could meet my goal using some single ended options. Anyway, I'll be using a diff amp that was designed to drive my ADC, which is a LTC2217 from Linear Tech. I'll be using the LTC6401-26 diff amp.

About the 100 dB SFDR: For ADCs, this measurement is taken by capturing data at the max ADC clock rate and doing a frequency spectrum calculation (typically with a FFT). You will see that 100 dB SFDR can be attained with a 16-bit ADC, even if the ENOB is less than 16-bits.

Darol Klawetter

 

[Spehro Pefhany]

I used the LTC6402 and didn't like it. It tended to oscillate at high signal
swings.

ADA4950-1 is spiffy. ADA4960-1 looks even better.

Nice HF performance. They do spec the DC performance but it's pretty
ugly- 50uV/°C typical.


Best regards,
Spehro Pefhany

 

[josephkk]

As I began researching my amp options, I did decide to go with a differential amplifier. I was hoping I could meet my goal using some single ended options. Anyway, I'll be using a diff amp that was designed to drive my ADC, which is a LTC2217 from Linear Tech. I'll be using the LTC6401-26 diff amp.

About the 100 dB SFDR: For ADCs, this measurement is taken by capturing data at the max ADC clock rate and doing a frequency spectrum calculation(typically with a FFT). You will see that 100 dB SFDR can be attained with a 16-bit ADC, even if the ENOB is less than 16-bits.

Darol Klawetter

And that is the other tradeoff. Your ENOB is likely no more than 12 maybe
13 bits with a 16 bit converter setup at max speed. Still sounds like a
pretty cool design.

?-)

 

[josephkk]

thank you for the explanation, didn't pay attention to that "SFDR"
ignored it entirely. at least 20+3 is less than 26

It's just that I'm used to routinely using my 24 bit system out to 22
bits and almost all of that damage comes form the supporting
electronics, didn't know at MS/s would need so much 'extra' overhead.

Actually i overreacted a bit. The situation is not really all that dire,
you can trade off ENOB as well, and this is running at 100s of MHz. It
starts getting difficult and fighting the components and their intended
usage usually gets you much more trouble than it seems to save.

?-)

 

[[email protected]]

As I began researching my amp options, I did decide to go with a differential amplifier. I was hoping I could meet my goal using some single ended options. Anyway, I'll be using a diff amp that was designed to drive my ADC,which is a LTC2217 from Linear Tech. I'll be using the LTC6401-26 diff amp..



About the 100 dB SFDR: For ADCs, this measurement is taken by capturing data at the max ADC clock rate and doing a frequency spectrum calculation (typically with a FFT). You will see that 100 dB SFDR can be attained with a 16-bit ADC, even if the ENOB is less than 16-bits.



Darol Klawetter

Suggestion, read the spec's on those parts you listed.

If you're trying to get 100dB out of them @ 50MHz,
with your state Vo, the spec's say otherwise.

OIP of the amp and SFDR of the A/D won't cut it at 50MHz.

But you are correct re the theoretical calculation of SFDR.

gl