LMX2594: How to match the PLL with a pull-Up inductor

Sofiane Aloui

Part Number: LMX2594
Other Parts Discussed in Thread: TIDA-01410

Dear,

I did perform power measurments (LMX2594 EVM). The maximum output power doesn't reach my desired power level~ 6 to 7 dBm @ [9 to 11GHz]

According to what I red in this topic, many users got 1 to 2 dBm with a resistive load. I got the same level. The only way for me to figure out, is to use a 1nH inductor Pull-up. The issue is that I do not have a tuner to determine the output impedance in order to optimize the power transfert.

Do you have a file that provides the output impedance of the PLL (or the impedance that you find thanks to the tuner) ? If not, do you have any realized or simulated design (an output matching network) that has been used at these frequency range.

Thank you,

over 7 years ago

0 Simon Damphousse over 7 years ago

TI__Expert 5370 points

Dear sofiane,

We will have recommendations on output matching and power soon. The best thing to do for good operation while trying to maximize power is to use a pull up inductor and a pad to have a broadband match for the device.

I sent a request for someone to provide more details. thank you for your patience.

Regards, Simon.

0 Dean Banerjee over 7 years ago

TI__Mastermind 28021 points

Sofiane,

I think inductor followed by resistive pad is a better approach as it has better control of the impedance matching and also there is no DC drop accross the resistor. The impedance of the inductor should be "High" at the frequency of interest relative to whatever load is after the pull-up and 1 nH seems a good value for 9-11 GHz.

Here are some measurements I have for absolute power level for 1 nH inductor.

8 GHz: 9.9 dBm

9 GHz: 9 dBm

10 GHz: 8.2 dBm

11 GHz: 7.6 dBm

Now this assumes no board losses and perfect matching. In reality, you might lose something to board losses and matching; assuming these to both be zero is not realistic. If you put a 3 dB pad, you might not get as much power as you want, but it will be better matched and higher power than the resistor.

Now the other approach you were likely thinking was to do some sort of tuned match and then you are asking for the output impedance. I have tried to measure this with different approaches and the results do not correlate very closely. I do believe the output impedance is lower near 5 GHz and that it is not high next to 50 ohms. Here is the approach from the method I trust the most ...

See how I estimated the output impedance ...

At higher frequencies, things get complicated. I am pasting a table below of my estimation of output impedance. To get this, I did the following:

1. Measure the output impedance without the device, but with 50 ohm pull up and called it "Pull-Up Structure"

2. Attach to spectrum analyzer and use stub tuner to find setting to maximize output power. Then I measured stub tuner impedance and calculated complex conjugate. Called this "Measured Impedance". Then I solved x || y = z to try to get the true buffer impedance without pull-up. This is not high confidence measurement, but it implies that the magnitude of the output buffer changes and is not high next to 50 ohms.

Fout	OUTxPWR	Measured Impedance		Pull-Up Structure		True Buffer Impedance
Fout	OUTxPWR	R	jX	R	jX	R	jX	Magnitude
2000	20	40.9	21.5	49.1	-4.3	-7.5	83.8	84.1
2000	50	38.2	15.2	49.1	-4.3	19.6	88.6	90.7
5000	20	58.3	13.6	34.4	3.2	-79.0	7.0	79.4
5000	50	60.4	30.7	34.4	3.2	-59.9	15.3	61.9
10000	20	88.7	-32.3	59.7	-28.1	-169.8	127.8	212.6
10000	50	91.3	-8.3	59.7	-28.1	-75.0	143.8	162.2
15000	20	29.9	5.3	27.1	52.4	35.5	-13.4	38.0
15000	50	28.2	3.5	27.1	52.4	31.9	-12.6	34.3

Regards,
Dean

0 Sofiane Aloui over 7 years ago in reply to Dean Banerjee

Dean,
Thank yo for your answer. THis will help.

The first approach looks easier to manage. To verify before designing, Can I just replace the Resistive load of the demo board by 1nH, add an extern 3dB SMA Attenuator and observe in the spectrum the power? Am I rigth ?

The second approach is also interesting. I will try to put some equations to better understand how you get (R & Jx) . Am not confident when I see R<0.

NB: your OUTxPwr is set @20 & 50. Is that equivalent to 31mA (just above OUTA_PD) ?
Thank you
Regards,
Sofiane

0 Dean Banerjee over 7 years ago in reply to Sofiane Aloui

TI__Mastermind 28021 points

Sofiane,

Yes, for the first approach, you can do as you say. Realize that the board may have some losses due to the trace and SMA, but the inductor pull-up followed by the resistive pad is a good way to go. If you do the resistive pad with resistors, it gives the options to short it out, use resistors for resitive pad, or potentially tinker with inductors/capacitors to get a better match.

As for trying to match the tuned load, this might be largely impacted by the board impedance and the pull-up component. As I calculate theoretically R<0, for the device itself, it's hard to take these numbers if you try to match the impedance with a tuned circuit. If you try this approach, then it should involve some manual tinkering on the board to be used itself so it accounts for the board as well.

Also realize that our EVM has some losses; you could get higher single-ended power if you route it single ended as we do in reference design (TIDA-01410).

Regards,
Dean

Clock & timing

Clock & timing forum

LMX2594: How to match the PLL with a pull-Up inductor