So again it’s a front end design problem. The problems going to be knowing it’s happening. Try turning off a few kHz and treating it like either a frequency offset problem or mix it down once it’s digital.
Thats generally a better option than an audio DSP unit that sits outside the AGC loop and can only operate on the recovered audio out of the rig. Sounds like you are messing with an AD9361 worrying about split gain tables and DC trackers taking a shit. If you can work out the connector clearance issues the DSP-100 would be a good solution as its an IF DSP unit that sits within the receivers AGC loop. I don’t really get this part, you are talking about a homodyne getting scrogged because it’s got an intermittent signal at DC? I guess it depends on the architecture. How do you feel with signals that dwell around DC for a bit in light of DC tracking loops? Having a big in band blocker generally comes down to front end design so you can do something to knock him out before conversion. If they are blacking our your front end you are fucked.
Excision if you have enough dynamic range. Add complexity slowly by controlling how and when you move.ĭepends. Some algorithms go into the log domain because it makes fast math easier and compresses the range of signals. They will have obsessed endlessly over the noise figure of the RX chain, so get them to fill you in and create a model. You need to talk with your RF guys and understand noise and response time for the real system. Butt voor linux, apple en Windows, gratis maar eenvoudig en maar 1 stream tegelijkertijd mogelijk om aan te sturen. Then you have some bearings on the extent of your gain control. DSP plugin met winamp speler (kies sound input) gratis, werkt alleen met SHOUTcast server versie 1 en versie 2 Nicecast voor apple werkt met SHOUTcast en Icecast (geen eigenschappen bekend) 40,- dollar. How many bits, or how much SNR do I need to receive this. Pick a max input level, something like 10 dBm and then think about your modulation sensitivity. Most systems have about 120-140 dB of range. Sometimes floating point models can mess with you because you are modeling way more dynamic range than any real system would have. You need to be sitting at max gain and cutting out as signal arrives. The point of AGC is to maximize sensitivity. Too much clever math only obfuscates the knobs you have to turn.Īlso it looks like you are sitting at the wrong quiescent gain. It’s about comprehensability and adjusting for known signal chartaristics. Having done a metric fuckload of AGC honestly, keep it simple. Essentially don’t let it jump around like a nut, just correct if you know you are very wrong. Go there, but now stay there until you have a good reason to leave. So now the task is understand where you need to be as quickly as possible. Single pole IIR, average power, square law detector. Run an envelope detection scheme of whatever your desired complexity. Each has trade offs.įor dealing with a bursty signal though I would recommend trying something simple first. We can add time constants to delay the rate of adaptation in a few spots. They are using an adaptive filter approach so we are optimizing against our desired gain state. Once we’ve made a decision we need to keep from making another decision to abruptly.
0 kommentar(er)
