Power amplifier output depending on input cable length?

[isso]

Hi, I have a modulated RF signal source in 1200-1300 MHz range and a power amplifier. Those are connected with a piece of coax cable. I noticed that the length of that cable affects the power amplifier output power a lot, it can vary by up to 6-7 dB depending on the cable length. Besides, if I just hold that coax with my hand the power output decreases by about 2 dB. Wondering why would this happen? Tried with 2 different cables with similar effect. Thanks

Oh and just to be clear it's not the cable losses issue as a shorter cable actually may result in a lower output power.

[HB9EVI]

looks to me like the input impedance of the amp is not properly matched;  if possible plan a small pad on the input side of the amp to match 50 impedance as close as possible in the interconnection

[thundertronics.com]

For narrow band operation you can use cable length multiple of half wavelength, which work as impedance repeater (impedance repeats each λ/2, also may be visualized making full circle on a Smith chart). This trick is sometimes used in microstrip circuits when line with required impedance is too thin or too thick, and can be replaced with λ/2 impedance repeater. Not sure about 2dB decrease when holding that coax cable, It may indicate poor decoupling of amplifier, variation is small though.

[rfclown]

looks to me like the input impedance of the amp is not properly matched;  if possible plan a small pad on the input side of the amp to match 50 impedance as close as possible in the interconnection

I agree, but the impdance mistmatch can be on either side (or both). If you put a pad on the amp input, and then don't get much change with cable length (within reason of cable loss), it is the amp input impedance. If you put a pad on the source output, and then don't get much change with cable length, it is the source output impedance. If it varies either way, throw up your hand and go home (just kidding).

[isso]

Thanks for your responses, I was working on troubleshooting this issue and the following came up:

1. The main problem was the coax cables, I bought those off Aliexpress and I would never realize that such a simple thing as a coax may be so low quality. After replacing it the issue was mostly gone, however some random power output fluctuations remained.
2. After spending more time on this it turned out that the RF attenuator that I was placing between the PA output and the power meter was bad, i.e. the 20 dB rated attenuation would fluctuate between 21 and 25 randomly in time. This was another item that I bought from Aliexpress mistakenly thinking that attenuator is a simple enough part to warrant buying from there. BTW I had 2 attenuators that I bought from different Aliexpress stores, one of them failed shortly after applying 10W power (although it was rated at 50W) and another developed attenuation inconsistency that took me a week to troubleshoot replacing pretty much everything in the PA thinking it had a problem until it dawned on me that my measurements might be wrong...

After replacing the cable and the attenuator with brand models the performance is acceptable now. There is still some small variation of output power vs cable length, so looks to be impedance issues, however it's not critical.

[HB9EVI]

1. the coax on Ali can be both: good and bad quality; sometimes it's try & error; when you find a good supplier, stuck with it; at least for me that's working well. For critical applications I prefer coax from reliable sources; they cost more, but you're not going on a wild goose chase if something doesn't work.

2. the same goes more or less for attenuators from such sources. your one seems to have a thermic problem. I experienced already the same issue - burned resistors loaded with powers far below the specified rating. the good thing is, that most of those attenuators can be opened and the resistors can be replaced by reliable ones. Just take care with the removal of the old one, you never really know if it contains BeO

[vu2tve]

rf cable performance can also significantly degrade if they get bent or twisted. I've had problems in the past with RG58, and thinner wires, where an accidental twisting or bending of the cable made the performance very unpredictable.

It was a lesson for me when I was new in RF design, and have since learnt to be more careful.

If ever, visual inspection shows a sharp bend in a cable, and I am facing problems, I replace the cable before any further debug.

[jonpaul]

We use American made Belden cables and have great results. Their 4.5 GHz rated coax is the best but needs special connectors and Belden crimp tools.

We avoid anything  Chinese, all junk.

Jon

[m3vuv]

seems to me chineese is another word for crap!,best avoided,lost count of the amount of shit counterfiet products/components i have got from china,the latest tip 121 transistors,supposedly darlington with an Hfe of 20,go figure!!

[jonpaul]

For SMPTE 2014 paper, We tested and used Belden 1694A coax  RG6/U 4.5 GHz  rated.

75 Ohm SDI, 3 db attn at 20 MHz for 100m length.

https://catalog.belden.com/index.cfm?event=pd&p=PF_1694A&tab=downloads

I am sure the Belden cat has similar low loss wideband for 50 Ohm.


Jon

[isso]

On a related note: I ordered BLF642 RF MOSFETs from Ali a while ago (because Digikey hiked up the price from $40 to $160 per piece in one day!).
Ali was still selling for the original price, and I thought those MOSFETs are too niche to be faked. Boy I was wrong!!
Ordered from 2 different suppliers, and both batches showed about 100 times less power output than expected.
I decapped the transistors to check the die, and those looked like LDMOS MOSFETs, but absolutely different from the original, and twice smaller in sizse. Also missing any marks. Compare the attached pictures, first is the original die, and the second is fake.
One seller agreed to refund, and another said that my pictures don't prove anything...