[arboristman]

this is from my buddy eqqmc2


c/p:



I am writing this post to address frequent problems of:

1) Bad diseqc switching
2) Interference Problems between multiple receivers when using diseqc switches.


These problems (in their majority) can be traced back to 3 possible causes:

1) Low or lack of isolation loss of 22khz signals by certain dual legacy lnbs: In this case, some lnbs do not isolate their lnb ports appropriately. When a 22kzh signal comes into one of its ports, the lnb fails to isolate the 22kzh signals and it allows leakage into the other lnb port, which then travels to say a diseqc switch. This additional noise produces problems with that 2nd switch and corrupts its switching table causing loss of signal on the associated receiver.

Solution: Remove or Isolate the offending lnb. To isolate the lnb, use another component with good isolation loss, such as a 3x4 multiswitch or 22kzh multiswitch that will prevent the unwanted leakage of 22kzh signal noise. There are also 22kzh filters that can be placed between lnb and diseqc switches to filter our any 22kzh signal from comming into the lnb.

2) Ground Loops: This situation happens when mutiple stbs have a different ground potential and a "floating ground" condition happens. This difference in ground potential can in fact affect the operation of diseqc switches causing unneccessary resets.

Solution: Install a grounding block between receiver and diseqc switch so that all receivers are grounded to a common point. The grounding block is connected to the electrical panel ground. See the excellent presentation by Satlight for more information on grounding.

3) Noisy Power Supplies in certain stbs: In this case the power supply of the receiver puts undue noise onto the 22kzh signal which then corrupts the behaviour of the diseqc switch (especially the cheap ones). In the forum of mytheather (thanks to revengineer and freelnb) the following solution and rational was posted:

Solution to the problem, by revengineer, is to modify the diseqc switch by adding a capacitor to serve as a filter of the noise :

Beg CP:


I've made some more progress. I found a 2 position diseqc design off of Freescale's website. It's application note AN2677 for whoever might be interested. It's not a very complicated circuit. It consists of a simple zener based power supply, a pair of transistors to switch each lnb, an input circuit to a microcontroller, and the microcontroller itself. After studying Freescale’s design, I was able to probe around one of the diseqc switches I have and found where the diseqc command enters the micro. Without any lnbs connected, the signal coming into the micro is clean. I can send a command to switch each lnb and I can watch the appropriate transistors switch on. Everything is ok until you hook up an lnb. Now if you’re switched to that lnb, a ton of noise shows up on the micro input pin. This explains why you can sometimes switch to an lnb, but then never be able to get off of it without disconnecting the cable. If you disconnect the lnb and send a new diseqc command, then switch then behaves. But what good is that if you can't have an lnb connected!?

I went a little further and tried suppressing the noise by dropping a small cap between the micro input and vcc. I went up to about 2200pF and the noise diminished enough so that the switch would work with lnbs connected. So far so good! It’s a little early to tell if this is it. It’s been a little flaky, but then I cranked up the retries to 3. I’m having trouble scanning without doing a “do now” command first and for some reason not all transponders are coming in either. I disconnected the switch and then scanned each lnb separately and got them all. I think my signal might be a little low. Now that they are scanned and diseqc switch is working perfectly! I’m done with it for tonight. I’ll spend some time later and see if I can find an optimal value capacitor.

The switch I’m working with is a DMS International. Inside it’s nearly identical to the GEOSATpro. The DMS appears to be a slightly newer revision of the GEOSAT. My feeling right now is that the common diseqc switches might be poor designs. The Freescale design uses a second transistor to the micro input – probably because of this noise issue. I can post some pictures later if someone wants to try this themselves. The parts are small and you’ll need to be very careful with a soldering iron to add the cap like I did.



From what I observed with the scope, it looked like this was a diseqc switch problem and not related to the twinhan. With an LNB connected to the switched port, a ton of noise shows up inside the diseqc which leads me to believe the source of the noise is the lnb and the circuit does not adequately suppress it. Can someone here confirm that a diseqc switch works fine with a set-top box and not with a twinhan? I'm not quite sure how the twinhan can be contributing to the noise since the signal looks fine without an lnb connected.

If the lnb is the source of the noise, an attenuator between the switch and the lnb might have the most potential of work. I think the signal strength coming out of the twinhan is fine and, as jerryt noted, is much stronger than the skystar.


Here's some more info:

I tried out a vp1020 and compared it to the 102g and found no difference seen with the scope.

I think this still could be mostly an lnb/diseqc problem. The signal from the dvb card to the diseqc looks no different with an lnb connected or disconnected. The diseqc command signal coming into the microcontroller looks very different depending on whether an lnb is connected.

Adding a small value capacitor from the microcontroller diseqc input pin and to vcc of the circuit seems to do an adequate job of squashing the noise. So far, I've found about 4700pF to look the best while values above 10,000pF starts to limit the entire signal. Most of my measurements have been on a DMS International switch. This switch is nearly identical to the GEOSATpro. I also modified a Sadoun S-DS41C and got it working too.

Some of these switches seem to be of very low build quality inside. I found the DMS International to have a physical broken transistor connected to lnb 4 - which only caused lnb 4 to not work. The GEOSATpro appeared to have a shorted transistor connected to lnb 1.

Any 0603 size 4700pF cap will work. Here's a part number from digikey.com 490-1571-1-ND. I don't know how one would find a local retailer of smt caps... I think you'll have to go mail order..

And some steps:

1. Carefully pry off the back cover. I don't think you can do it without bending the lid.

2. Carefully position the small cap with a tweezers and careful put down a dab of solder. It helps to have three hands, but us mortals have to do it with only two.

3. Repeat for the other cap.

Soldering small smt components takes some skill and a careful hand. It's easy to bridge the parts with solder. Maybe find a junk circuit board to practice on before you tackle the switch.
end CP

This is the rational nicely explained by freelnb:

Beg CP

Nice Work RevEngineer.

I believe your hunch about the twinhan putting noise onto the line under heavy load is correct(these switches work fine with other systems after all). This is what I have confirmed. Monitor the RF line under full load (LNB and switch on) and you will find a 200-400 mV peak to peak 350 - 400 kHz signal (noise) coming from the twinhan power supply. This is traceable all the way back to the LM317 high voltage regulator on the twinhan circuit card. This ripple noise coming from the twinhan under heavy load is then amplified at the switch by the transistor detector completely swamping the pin on the micro-controller with noise. In application note AN2667 as well as the DMS switch(I have one two) the 22k feedback resistor R4 over the 1K (DMS uses 1.2k) R8 provides the gain for the detector. Putting a 4700pf capacitor from the collector of Q3 to ground or VCC reduces the noise so the micro can decode the 22khz signal. I also found that putting a 4700pf cap to ground at the base of Q3 helps reduce the noise further. This creates a low pass filter of 33.9 KHz in conjunction with R8 1/(2(3.14)(1K)(4700pf))=33.9 KHz. This also has the advantage of reducing the 350Khz noise before it gets amplified. It works best with both caps in circuit.

Based on the above I thought that just filtering the twinhan power supply with more bypass caps would fix the problem WRONG. Here is why. The twinhan I found out uses the LM317 in conjunction with transistor switches and resistors to switch the LNB Supply Voltage between 13V and 18V. This is fine. This is not, it also uses this same method to generate the 22kHz signal,--- they are modulating the power supply at 22kHz to generate a 1.5volt square wave that rides on top the the 18 volt supply. This design has one inherent fault, you cannot filter the output of the LM317 or you also filter the 22kHz signal. Under heavy load the LM317 requires a fairly large capacitor to increase ripple rejection and improve transient response. I wondered why there was no capacitor at the output of the LM317 but now I understand why. This is not a very good design. They should have filtered the supply heavily then superimposed the 22KHZ on top of the DC using an inductor to create a low impedance at DC and a relatively high impedance at 22KHz.

The best solution is to modify the switch and leave the twinhan alone. I believe that the more expensive switches probably have filtering built in and therefore do not have this problem.

end CP.