THS3217: THS3217 distortion when trying to get an O/P of 13dBm with frequencies over 100MHz.

Part Number: THS3217

Hello Engineers

THS3217 DC to 800-MHz, Differential-to-Single-Ended, DAC Amplifier.

I’m hoping to get some advice from anyone that has experience using THS3217EVM dev board.

The problem is, I am seeing is distortion (compression?) when trying to get 13dBm out of the device at higher frequencies. The part performs well up to around 100MHz. With increasing drive, the harmonics start to rise and eventually goes unstable and shows huge distortion with harmonics and sidebands.

Two plots show an input signal level of -4.0dBm and -2.0dBm.

The requirement is to output 13 spot frequencies between 1 and 250MHz at an output power of approximately 13dBm (2.8V p-p) into 50R.

Test setup.

Power supply ±5V, Input and output ground-centered.

RF signal generator, wideband balun (single ended to differential). Input impedance on dev board (R3 & R4) 50R to GND on each differential line. Gain of device is set to x10 (20dB). Output impedance (R34) 50R. Output power measured on a spectrum analyzer.

My interpretation of the data sheet suggests that the device should be capable of this level of output at these frequencies. Any suggestions would be welcome. Thanks Geoff.

  • Hi Geoff,

    The 13dBm spectrum on the right it looks like the device is clipping to me. Have you confirmed with an oscilloscope what the output looks like in that state and if it is clipping? From the left to right plot you increased the input power 2dB, but the output changed from 12.42dBm to 13dBm which isn't a 2dB increase, further telling me that it is clipping. Does the measured 13dBm look better at lower frequencies?

    With the 50ohm R34 and what I assume is a 50ohm spectrum analyzer, the measured 12.42dBm on the left would really be 18.42dBm on the output pin of the device or about 5.27Vpp. This would also mean that the gain isn't 20dB but more like 22dB, have you confirmed the gain? If you changed the gain on the EVM to 10V/V did you change the Rf resistor? 

    Thanks,

    Evan

  • Hi Evan

    Many thanks or your response. Here are the answers to your questions.

    The 13dBm spectrum on the right it looks like the device is clipping to me. Have you confirmed with an oscilloscope what the output looks like in that state and if it is clipping? 

    Yes I think it is clipping. Scope shows the onset of distortion. I can get a little more power if I up the rails to +/- 6V but device performs well at lower frequencies providing over 13dBm with +/-5V rails.

    From the left to right plot you increased the input power 2dB, but the output changed from 12.42dBm to 13dBm which isn't a 2dB increase, further telling me that it is clipping. Does the measured 13dBm look better at lower frequencies?

    Yes at frequencies below approximately 100MHz input and output track well with little distortion and power output of over 13dBm.

    With the 50ohm R34 and what I assume is a 50ohm spectrum analyzer, the measured 12.42dBm on the left would really be 18.42dBm on the output pin of the device or about 5.27Vpp. This would also mean that the gain isn't 20dB but more like 22dB, have you confirmed the gain? If you changed the gain on the EVM to 10V/V did you change the Rf resistor? 

    Yes 50ohm spectrum analyzer input. From memory the gain of 20dB looked right but will check again on the spectrum analyzer and scope. Have adjusted the output stage gain Rf = 140R, Rg = 34.8R (used 36R). These are the recommended values from page 38 of the data sheet. This should give input stage gain x2 and output stage x5.

    I am feeding the differential to single ended input stage directly into the output power stage and not going through the filter.

    The data sheet headlines states Full-Power Bandwidth: 500-MHz (5 VPP). It performs well at low frequencies, but I just can’t get the required power output at higher frequencies. Is this a limitation of the device or am I missing something?

    Thanks

    Geoff

  • Hi Geoff,

    Thanks for your responses. It is interesting that you aren't having the issue at lower frequencies, so it's not a pure resistive load/feedback issue. Do you continue to have the issue past 100MHz?

    Could be one of two things is happening: 

    1. At 100MHz and that power level there is some gain peaking or resonance that is causing the output to peak up at that particular point. Maybe why the gain is more like 22dB there. 

    2. You are reaching the full-power bandwidth of the device.

    Looking at the S-parameters with a network analyzer can tell you about #1. But like you said I find #2 unlikely, from the datasheet it should be able to handle that power at that frequency. I was even able to find some measured data that showed the OPS can do 8Vpp with 330MHz of bandwidth in that Gain of 5V/V with the correct Rf that you chose. Unless there is some other load we aren't accounting for. You said you aren't using the filter path, have you populated the filter section of the EVM? That section of the EVM can load the D2S even if you aren't using it. Is there any other load on pin 6? The reference buffer (VMID_IN and VMID_OUT) has some output limitations, are you using that at all?

    From Figure 4 of the EVM guide: 

    I found a working THS3217 board in our lab. I'm going to try that 13dBm at 100MHz with your gain to see if I can replicate it. 

    Thanks,

    Evan

  • Hi Evan

    Thanks for the quick response and your continued support.

    I had removed all the filter components from the board as the schematics suggests that they could have a loading effect.

    Pin 6 is connected to GND via a 200R as per page 47 of the data sheet.

    VMID_IN and VMID_OUT no connection

    My EVM is the same as the schematic on the data sheet page 47 with the exception of Rf and Rg giving 10V/V

    Other observations that may or may not be relevant.

    We originally had an OP stage gain of 1.5V/V (x3 device gain) but noticed on page 38 of the data sheet that this was not recommended. We upped the gain to 2.5V/V (x5 device gain) the gain flatness improved and I could get more power out at higher frequencies before distortion set in. Upping the gain to 5V/V (x10 device gain) gave marginal improvement.

    Curious observation. I been driving this from a DDS  AD9910 eval board sweeping up and down from 1MHz to 250MHz. On the up sweep, the harmonics could be seen rising at higher frequencies (as per my 1st plot) but only on the down sweep could I make it go completely unstable (as per my 2nd plot). I reproduced this using a bench RF signal generator. I have not been able to explain this.

    Kind regards

    Geoff

  • Hi Geoff,

    I think I was able to replicate this behavior on the bench. With an Rf=140 and the G=5V/V, 13dBm on the spectrum analyzer was seeing those spurs from about 110MHz on my setup up to about 200MHz, sweeping both directions. Increasing the supplies to +/-6V made it go away. Looking at the waveform with an oscilloscope it did look like a distorted sinewave not a clipped output limitation. The gain of the part also varies a bit over frequency by 1-2dB so the amplitude isn't totally flat, but that doesn't make it output limited. 

    I increased the Rf of the OPS from 140ohms to 200ohms, keeping the gain at 5V/V, and that improved the distortion. Instead of the 110-200 MHz range it shrunk to 140-160MHz, and 200-210MHz. Increasing the Rf further to 300ohms eliminated the problem.

    I think there might be some stability problems with the output amplifier in this specific configuration and in these specific frequencies. Increasing the Rf of the OPS will help with stability since it is a current-feedback amplifier. This will limit the bandwidth a bit so there are tradeoffs to that. Other things that might help with stability would be better decoupling.

    Hope this helps.

    -Evan 

  • Hi Evan

    That is a big help. I am reassured that you could recreate what I have seen. I will try adjusting the value of Rf  and look into the decoupling strategy.

    Many thanks

    Geoff