I am using the HMC575 evaluation board and I am trying to validate HMC575 for our design. I have a 3 Ghz signal coming in (from a signal generator) and I put that signal through a band pass filter (2.7GHz to 3.1GHz) to clean any extra spurs. I have compensated for the cable loss in my calculations, and I have connected the output of the HMC575 eval board directly to a phase noise analyzer using SMA header. but when I compare my empirical results with the datasheet, I can see that there is up to 12dBm difference in some cases with the output power measured and the power reported in the datasheet.
I can observe that a 3Ghz component is fed through from the input to the output and in some cases, the amplitude of the 3Ghz on the output is a lot more than the 6Ghz component. The following is our empirical results:
Input power (dBm), 3Ghz sine wave
HMC575 output @ 3Ghz (dBm) (Feed-through from the input on the output)
HMC575 output @ 6Ghz (dBm) (desired doubled frequency on the output)
Delta (6Ghz - 3Ghz)
Besides the 3Ghz input feedthrough, you can observe that the output power vs. input power is significantly different from what the datasheet specifies; for instance, according to the datasheet, a -8 dBm input power at 3Ghz, should result in a -4 dBm output at 6GHz. However we observe in our experiments that the 6Ghz output is -16.1 dBm!
For reference, I have included the output signal which is connected to the phase noise analyzer for an input signal at 3Ghz with -2 dBm power:
Can you please tell me whether the datasheet is incorrect or I am doing something wrong?
The datasheet is correct.
One thing I notice (doesn't really change the result) is that your column for delta should be either taking the abs value or referencing output power (3 GHz ) - output power (6GHz) so that all values are either negative or positive. The formula you're using makes it look worse than it really is (there isn't a 24 dB delta in isolation from a -10 dBm input to +9 dBm input).
I'm assuming the input power level shown in your chart also accounts for the loss of your filter (the HMC575 isn't overly sensitive to drive level until the power drops below 0 dBm for a 6 GHz output (see Output Power vs Drive Level plot in DS).
Your plot here shows 11.34 dB output using a -2 dBm input after any cable, connector and filter losses. Datasheet shows a nominal 15 dB output (upper right plot) for these conditions; 3.7 dB of losses for the cable, filter and connectors doesn't seem unreasonable. Output power is about right at higher drive levels but is low at low power levels and isolation is off.
If you truly believe the part is defective please return the part for failure analysis. Please note that these parts are ESD Class 0 so best ESD handling practices are essential to prevent damage.
Thanks for the prompt reply. Unfortunately, I think something may have gotten a bit lost here.
For clarity, the column named, "HMC575 output @ 3Ghz (dBm)", refers to the feed through from the 3GHz input signal. The "HMC575 output @ 6GHz (dBm)" refers to the measured output power of the desired doubled frequency.
Also note that the current test set up was calibrated and we accounted for cable and filter losses; the input power measurement was measured at the output of the SMA (input to the HMC575 eval board).
So a couple of observations;
1) The delta measurement that I listed was intended to be used to help highlight that the difference between the desired (doubled) output frequency and the feed through from the original 3GHz signal. In particular, based on my observations, an input power level of less than -4dBm results in the 6GHz output frequency to have less power than the feed through from the 3GHz.
This is a problem because we intend to use this device to generate a 6GHz Dev Clock signal to feed an AD9162. Right now, it looks like the DLL on the device is locking on to the 3GHz feed through instead of the desired 6GHz output.
2) It looks like the output gain from the device appears to be less than what was listed on the datasheet. The datasheet indicates the following:
My observations are inconsistent with this graph; in particular, with a 3GHz input signal, at a measured power level of -4dBm, I should expect to see a 10dBm output signal. Instead, I only see 4.67dBm (As noted in the "HMC575 output @ 6Ghz (dBm)" column).
What could account for the variance here?
I agree that the output power is low.
Do you have another EVB that you can try? If not maybe on of our local FAE's can help get one to you.
The internal DC blocking cap may be damaged and reducing the output power. These are Class 0 (HBM) parts that pass 100V fail 150V so easily damaged.
I understand that you calibrated out the losses of the external system components but you might try testing the EVB without the filter, cabling, etc (connect directly to the SA or through a broadband DC block) and see what you get. Perhaps there's an issue with the cable or the filter is loading the part in someway that it's causing an issue.
Other thoughts that you've likely already considered:
Is Idd ~90 mA?
Are you getting 5.0V on Vdd1 and Vdd2? If these are shared by a common supply maybe separate them and see if this makes a difference (normally not a problem as we typically connect both). If you can, let me know the current drawn by each supply.
If supply voltage is being pulled low or current is higher than expected, perhaps C1, C2, C3 or C4 have been damaged. Could see if this makes any difference (if supply voltage and current is o.k. these are likely fine).
There is no "thru" path on the EVB but I would expect this to only make a dB or so difference. Perhaps there's an issue with the connector or the soldering of the part to the EVB?
RL at 3GHz input & 6 GHz output is 15 dB respectively so unless there's a problem with the board, part or assembly this should be o.k.