DAC output filter question

[Smillsey]

Hi guys

I am designing the output low pass image filter for a DAC that will be generating up to 75MHz sinusoids....

I needed DC coupling so i used this great app note to design the output circuit; https://www.ti.com/lit/an/slyt368/slyt368.pdf

There is a circuit with a filter and a circuit without a filter... but not much info about the filter design itself... I am waiting information from the FPGA engineer about the exact sample rate we will use in order to determine the roll off of the filter...

But my question is where would you put the filter?

See attached the circuits used thus far, I was thinking about just strapping on another op amp after the single ended conversion and implementing an active 2nd order low pass after the existing circuit........ Is there any downside to that?

PS. Ignore the AD8066, i wont be using that specific op amp...

The dac is a 0_20mA complimentary output IC, hence the current sources in the spice below...

 

[KlausST]

Hi,

The circuit - without showing the DAC - makes not much sense to me.
--> Please show the complete circuit.

What output voltage range do you expect?

Basically I recommend to use a (at least) first order low pass in front of the first OPAMP.
The DAC usually generate steps with high dI/dt causing high dV/dt, which may stress the OPAMP input and may cause regulation problems. The LPF reduces dV/dt.

Klaus

 

[Smillsey]

The two current sources represent DACOUTP and DACOUTN which are current output complimentary types….

i suppose I will have to find out how to add the filter between the DAC and the op amp then… it isn’t so simple with this type of DAC

--- Updated Oct 27, 2021 ---

UPDATE:

I just found a much simpler approach in the analog devices app note here : https://www.analog.com/media/en/training-seminars/tutorials/MT-019.pdf

I have then put together this circuit with 3dB @ 318MHz approx

Results:

Compliance voltage is 1V on the DAC outputs, so this looks ok...

I can see why the TI app note obsesses over the compliance voltage, but the above circuit is simpler and seems to do the job.

The DAC sample rate will be 800MSPS with 2x interpolation (actual 400MSPS clocked) so image frequencies will by 725MHz....

I could pull the roll off down to 150MHz and calibrate the generator (it will be calibrated anyway)...

Any thoughts?

 

[KlausST]

Hi,

Your current source drives (as far as I can see) symmetrical current against GND.
But it's impossible to drive current from GND to GND.

The DAC output is positive only against VCC.

*****
But the new schematic makes more sense.
You used differential LPF, but you could also do individual LPF with a C at each 50 Ohms resistor.

****
Still open the answers to post#2

Klaus

 

[Smillsey]

Hi,

Your current source drives (as far as I can see) symmetrical current against GND.
But it's impossible to drive current from GND to GND.

The DAC output is positive only against VCC.

*****
But the new schematic makes more sense.
You used differential LPF, but you could also do individual LPF with a C at each 50 Ohms resistor.

****
Still open the answers to post#2

Klaus

Thanks klaus

the output is 3V pk

the dac has complimentary outputs, 0~20mA

--- Updated Oct 27, 2021 ---

Thanks klaus

the output is 3V pk

the dac has complimentary outputs, 0~20mA

The dac is the AD9139

 

[FvM]

The filter specification is determined by the required image frequency supression. Your first order filter hasn't much attenuation at 725 MHz.
Did you see the 5th order LC Filter in AD9139 datasheet?

 

[Smillsey]

The filter specification is determined by the required image frequency supression. Your first order filter hasn't much attenuation at 725 MHz.
Did you see the 5th order LC Filter in AD9139 datasheet?

Thanks FvM

I did, but that won’t work with my DC coupled op amp circuit

 

[KlausST]

Hi,

3 V peak ... doesn't say much...

You simulate the Opamp, but don't use the current signals like the DAC has.
You use a different Opamp.
What information do you expect from this simulation?

Klaus

 

[Smillsey]

Hi,

3 V peak ... doesn't say much...

You simulate the Opamp, but don't use the current signals like the DAC has.
You use a different Opamp.
What information do you expect from this simulation?

Klaus

Good point

I was just pondering what type of filter I should employ…

I am not sure how much I need to attenuate the image frequencies (50dB, 100dB?)

I will edit the circuit when I find the model for the op amp I’m using.

sorry I should have been specific…

 

[FvM]

I did, but that won’t work with my DC coupled op amp circuit

Don't see why. LC low pass is perfectly DC capable.

 

[Smillsey]

To better clarify what I am trying to achieve :

1. Up to +/-3Vpk sine wave output
2. uHz to 75MHz output frequencies (sinusoids)
3. DC offset capability
4. AD9139 used as DAC (Cannot be changed now)
5. 800MHz DAC update rate
6. See current circuit attached

As you mention, it’s only a single pole filter with 3dB at around 300MHz, so I need much more attenuation but the circuit from the original TI app note I mentioned doesn’t detail the filters… are they just using passive high order butterworth do you think?

 

[FvM]

I am not sure how much I need to attenuate the image frequencies (50dB, 100dB?)

This solely depends on your application.

 

[Smillsey]

I suppose what it comes down to is I am not sure how to design a 5th order filter like the one in the datasheet and interface it to my op amp… looks like I know what I will be researching tomorrrow then

 

[KlausST]

Hi,

My problem is .. I don't know what's important for your application.

Some parameters...go through them and ask yourself what you need.
* absolute DC accuracy (do DAC_realistic simulation to find out if the result is good enough. I expect you become surprised ;-) )
* amplitude flatness over frequency
* phase shift over frequency
* common mode suppression
* waveform reconstruction
* impulse response
* overtones (especially check at high frequencies)
* sample frequency residuals
* what else...

Maybe you could give a short statement to each item.

Klaus

 

[Smillsey]

Hi,

My problem is .. I don't know what's important for your application.

Some parameters...go through them and ask yourself what you need.
* absolute DC accuracy (do DAC_realistic simulation to find out if the result is good enough. I expect you become surprised ;-) )
* amplitude flatness over frequency
* phase shift over frequency
* common mode suppression
* waveform reconstruction
* impulse response
* overtones (especially check at high frequencies)
* sample frequency residuals
* what else...

Maybe you could give a short statement to each item.

Klaus

Thanks Klaus

Please see below

* absolute DC accuracy : This is not an issue as its not critical to the application and can be calibrated, as long as its linear
* amplitude flatness over frequency : Not too critical as can calibrate, but 1dB to 75MHz would be ideal, 3dB acceptable
* phase shift over frequency : Not critical
* common mode suppression : Not critical
* waveform reconstruction : Removing image filter components is quite important.
* impulse response : Not critical
* overtones (especially check at high frequencies) : Not critical
* sample frequency residuals : Need to be removed

I suppose i just am not sure what type of filter to insert in the picture below;

I think i need a balanced LC filter like the design referenced here : https://ceworkbench.wordpress.com/2012/03/03/designing-balanced-lc-filters/

 

[Smillsey]

Hi all

I am now at this stage:

I need to insert a filter (passive) where the green circle is ....

Should I go for a balanced version like this?

Or two of these, one on each of the DAC outputs....

On each DAC output, the impedances are nominally 90Ohms on each line, looking "into" and "out of" the filters.

Which of course combine to 45Ohm

 

[FvM]

Your first filter isn't balanced because the capacitors are grounded. Both filters are double single-ended, just with different dimensioning.

The simulation schematic doesn't fit the DAC output configuration which is a current source. Furthermore you want an effective DC termination of 50 ohms for each DAC output to achieve optimal signal scaling. The circuit below is implementing it. The corresponding -3 dB frequency is 86 MHz.

 

[Smillsey]

Hi FvM

thanks for your help

What do you mean regarding it doesn’t fit the DAC output configuration? The output of my schematic is a current source … So how is it different?

--- Updated Nov 1, 2021 ---

I have to use a 45Ohm termination and lose some resolution because if I don’t I will exceed the 1V compliance voltage on the DAC

this is from the Ti app note mentioned previously