Waveform generator - amplifier stage

[miceuz]

Hi all,

I want to build a waveform generator based on AD9834 (http://www.analog.com/en/digital-to-analog-converters/direct-digital-synthesis-dds/ad9834/products/product.html) with specs as follows:

Voltage source will be +12V, so power rails for opamps will be +-6V.
Maximum frequency I want to acheive is 50MHz.
Output voltage from AD9834 is 0.6Vp-p, maximum output swing of generator - 10Vp-p

I want to acheive a gain of 17 in a single opamp stage. It this possible/feasible? I've seen a thread where a slew rate was mentioned, so I've looked for an opamp with this in mind. I'm thinking of THS3202 (http://www.ti.com/product/ths3202). Is this a proper choice?

[ptricks]

I want to acheive a gain of 17 in a single opamp stage. It this possible/feasible? I've seen a thread where a slew rate was mentioned, so I've looked for an opamp with this in mind. I'm thinking of THS3202 (http://www.ti.com/product/ths3202). Is this a proper choice?

gain of 17 is easily done. The part you linked though has fairly high noise which is what is expected when you have a part with such a wide bandwidth.
The MCP661 might be a better pick , depends on what comes after the opamp stage.
http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en543024

[miceuz]

The MCP661 might be a better pick , depends on what comes after the opamp stage.
http://www.microchip.com/wwwproducts/Devices.aspx?dDocName=en543024

But if I look at this figure - it's about 5MHz at 20dB. Am I looking at the wrong graph?

[grenert]

No offense intended, but if you are asking this question, I think you are likely to have difficulty designing a board that can handle a 2-GHz opamp.
At the very least, you should check out the Linear AN-47 app note by the legendary Jim Williams.

[Mechatrommer]

i havent look at the datasheet but...
1) dont use THS3202, because (i) its an 2GHz cfa way overkill for your 50MHz, its other limitation (noise, distortion etc will be unneccesarily greater). (ii) it cannot use ±12V rail in case you want to upgrade to ±10V FG output. use THS3091 or THS3092 instead if you are in cfa setup.
2) yes there is GHz range of vfa, but when we talking about 10dB of gain or above, thing can get tough for vfa. so you have two options... (i) go for cfa in one single stage of 20dB gain. (ii) if you really need vfa, then you need several stages of lowered gain.

[Mechatrommer]

well, based on TI recommendation, you dont put THS3202 or THS3091 as the first stage of your DDS, somehow for some reason i dont know havent study that maybe due to the differential current output nature of the DDS/DAC. here's the recomendation from THS3091 datasheet. they are paralleled for greater power, you can only use one 3091 (or THS3092 paralleled one chip) for reduced power FG.

[grenert]

1) dont use THS3202, because (i) its an 2GHz cfa way overkill for your 50MHz

There might be many reasons to not use this opamp for this application, but if you wanted to amplify a 50 MHz signal 17x, that comes out to a GBW of 850 MHz, so 2 GHz isn't totally out of whack.

[Mechatrommer]

1) dont use THS3202, because (i) its an 2GHz cfa way overkill for your 50MHz

There might be many reasons to not use this opamp for this application, but if you wanted to amplify a 50 MHz signal 17x, that comes out to a GBW of 850 MHz, so 2 GHz isn't totally out of whack.

cfa's opened loop gain profile is very very different from vfa's, by adjusting the Rf value you pretty much still get close to unity gain BW of cfa's. here's the THS3092 (cfa rated GBW 160MHz) and LM7121 (vfa rated at GBW 175MHz). at gain 10 (20dB), the LM7121 is only less than 10MHz BW but for THS3092, its still at > 100MHz with Rf 604 ohm. so cfa vs vfa is very different ballgame in term of GBW.

[grenert]

Thanks, Mechatrommer!  That seems amazing to me.  I need to do some reading on current feedback opamps...

[miceuz]

" An added benefit of the AD8130 is its industry-leading ac common-mode rejection (70 dB @ 10 MHz). This feature can be used to reject noise between the DAC digital ground plane and the receiver analog ground plane—a common problem in this type of mixed-signal application. "

See why here
http://www.analog.com/en/circuits-from-the-lab/CN0142/vc.html

Are you suggesting I use a differential opamp and use both outputs Iout and IoutB of AD9834?

[miceuz]

saying I believe. You will not be able to get 50MHz out of that chip Nyquist sampling theorem says the sampling speed must be at least twice the output frequency more is better. The chip is

I must say, I feel terribly embarrased  I've just made simple calculations using formula from datasheed regarding output frequency vs control word.

From what I've red today, I can expect to go as high as 30MHz with 75MHz clock (AD specifies 40% of clock as highest output rate). That changes the game a bit. Now my idea is to use an LC filter at the corner frequency of 30MHz and two stages amplification using http://www.ti.com/product/opa843

[grenert]

If you haven't seen this, check out the N3ZI DDS kit.  It has an AD9834, filtering, and a design for a discrete buffer stage (not included on the main board).
http://www.pongrance.com/super-dds.html

[nctnico]

1) dont use THS3202, because (i) its an 2GHz cfa way overkill for your 50MHz

There might be many reasons to not use this opamp for this application, but if you wanted to amplify a 50 MHz signal 17x, that comes out to a GBW of 850 MHz, so 2 GHz isn't totally out of whack.

The 2GHz bandwidth is only relevant for small signal gain. If you want large output swing then the slew rate is the dominating factor. For a 50MHz sine wave with 10Vpeak you'll need a slew rate of 500V/us.