AD8138 drive AD9434

Hi, 

The CML voltage of the AD9434 should be the same as the internal ~1.75V.  Since you are provising the common mode bias, you should disable the internal bias by writing 0X04 to register address 0x2C.  This will open the switch (which is closed by default for ac coupled signals) in figure 34.  The clock must be active for the internal bias voltage to be generated.  

Regards, 

David

Hi Khai,

I don't seem to find any violation on your application. What are you using to implement the circuit? Is it a PCB? could you attach that as well? WIth the current configuration, the AD8138 would work just perfectly fine.

Regards.

Hi Khai,

If you would check the specifications table, you will see that the 0.1 dB flatness is stated to be 30 MHz typical without any compensation capacitor on the feedback. Also, figures 5, 6, 8 and 9 would clearly show you the AC response of AD8138. These figures tell us that at frequencies above  30 MHz, the signal will peak up to approximately 1.5 dB at ~150 MHz.

As shown above, at 85 MHz, the amplifier peaks and the gain is not effectively 1 that's why you are getting a higher output than expected. What you can do is add a compensation capacitor on the feedback to remove the peaking. However, the bandwidth would dramatically decrease. See Figures 6 and 9. Or, pick another differential amplifier, perhaps a faster one to achieve a maximally flat response on your target bandwidth.

Regards. 

Hi Khai,

1. Yes, it's true but note that board capacitance could introduce instability as well, hence some peaking.

2. 0.5 pF will reduce the peaking, somewhere in between 0 pF and 1 pF but we are not certain about the exact value of the peaking since it wasn't tested at that CF.

3. There are other FDA in our portfolio that you might consider. Please see ADA4932 and ADA4937.

Regards,
jino

Hi,

Thank you JinoL for the reply and finally I found this is due to the wrong chipset from our suppliers, it suppose to be AD8138ARM but it happened to be AD8313ARM , it is a mistake of them which took me weeks, and now I change back to the right AD8138 and I could see the waveform swing and the correct common mode level.

The only problem now is that I used a Keysight N5182B to generate some 1Mhz to 100Mhz, but I discovered that the amplitude after the AD8138 increase significantly with the freqencies increases. the single end input Vpp is 640mV, and from 100Khz to about 30Mhz, the output voltage is flat and showed correctly at 320mV Vpp for single end, but when I tuned to around 85Mhz, the single end output voltage INCREASE to 500mV Vpp. could you please advise why?

Hi,

Thank you David for the reply and finally I found this is due to the wrong chipset from our suppliers, it suppose to be AD8138ARM but it happened to be AD8313ARM , it is a mistake of them which took me weeks, and now I change back to the right AD8138 and I could see the waveform swing and the correct common mode level.

The only problem now is that I used a Keysight N5182B to generate some 1Mhz to 100Mhz, but I discovered that the amplitude after the AD8138 increase significantly with the freqencies increases. the single end input Vpp is 640mV, and from 100Khz to about 30Mhz, the output voltage is flat and showed correctly at 320mV Vpp for single end, but when I tuned to around 85Mhz, the single end output voltage INCREASE to 500mV Vpp. could you please advise why?

Hi JinoL,

Really appreciate it and it is very clear to me, and I'm working on the 500Mhz sampling and using AD9434 as the digitizer, and could you please suggest:

[1] The figure 6 and 9 shows the max gain (Cf = 0) is just <1.5db, and when at 85Mhz, the gain seems less than 1db, this will not lead to the 320mV vs 500mV difference which ask for around 2db, do you think this is reasonable?

[2] The figure 6 and 9 showed the 1pF Cf curve, what if I add just a smaller Cf, such as 0.5pf, will it provide a little bit balance on the gain and bandwidth? Although it is not normal to use a < 1pf MLCC for me.

[3] would you suggest an easy to use device such as AD8138 which has a better flatness device and also a better -3db bandwidth (for example, 400Mhz to 500Mhz).

Thanks JinoL, very helpful!!