Saturating an amplifier demo

[pendragon]

Given recent discussions here, I thought a short report might be in order on how amplifiers can make slight improvements in received CNR for even a 'hot' LNB, but are also vulnerable to saturation. I recently finalized my toroid LNB configuration with a combination of Superdish FSS LNBs with DP feedhorns and Sadoun KBSL1 LNBFs. Both of these are bandstacked and mechanically fit very nicely together for 2 degree spacings on a T90. They have dramatically different internal gains.

There's not a lot of cable run involved for these dishes as the switches are in the attic on the other side of the wall from the T90s. But the LNBs have to drive a gauntlet of switches and taps to finally arrive at the wall plates for the receivers. The Superdish LNBs have moderate output levels that benefit from using amplifiers to maintain the CNRs of received signals. I chose the JVI Trunkline 20-TR13LA dual amps from Sadoun because they are quite inexpensive and suit this particular application well. The Sadoun LNBFs have a pretty hot output that needs little if any amplification for this setup. That makes it an excellent vehicle for this demonstration.

I chose 11771 V 14329 on 99W (Galaxy 16) for some quick tests with a Sadoun LNBF. I measured the CNR at the wall plate (after the full house signal distribution) for a variety of gains and two different amplifiers. Initially I used no amplifier and then inserted some high quality pads with 3, 6 and 10 dB of attenuation. I then made two separate runs with the same attenuators, but using the JVI amplifier and a Channel Master 5213IFD dual amplifier. The pads were applied at the amplifier inputs. This gave me a 23 dB range of output levels to consider. For grins I also took down the SQs reported on this signal by my Pansat 9200HD.

I've attached a couple of plots of the results. The amplifiers are in place for gains of 3 to 13 dB. As can be seen from the CNR plot, the JVI amplifier saturates very quickly as the input level to it increases. The CM does saturate, but nowhere near as quickly and with not with the same CNR deterioration. It's important to again mention the Sadoun LNBF has a lot of gain. Having done this test with the Superdish LNBs, but sadly not recording any data, I should add the CNR benefit is larger and there is no saturation with either amplifier.

It's worth noting that the CNR is improved right up to the saturation point of either amplifier. This is a strong statement that if you're not concerned about the absolute signal level at your receiver, you may be losing a detectable amount of CNR. For example if one compares a 10 dB attenuation to the best CM point, there is about a 1.4 dB difference in CNR. That's like dropping a 90 cm dish down to a 76 cm if your levels aren't high enough.

I added the Pansat 9200HD SQ plot because this receiver is known to be slightly deaf on the DVB-S side. It's hard to get consistent SQ results with this STB, as the numbers tend to bounce around. I logged the highest value seen over a couple of minutes as this has worked best for me in the past. Interestingly enough, the Pansat sees the saturation, but still enjoys improvements in SQ in spite of it. Sometimes saturation can be a lesser evil than insufficient input level.

 

[guapoharry]

Very interesting.

What grade of test equipment would suffice to make these sorts of measurements?

Would a Satlook spectrum analyzer be useful for this? Or is that the sort of thing that is OK to align LNBF's but not a unit that you send out for calibration....?

 

[zamar23]

Yeah, that's exactly what I was wondering about: how to find out a required pad attenuation value to prevent the amp's saturation while maxing out the line CNR.

Looking at your report, some questions come up:
How did you measure CNR?
Why choose this particular TP?
You varied the amp output by replacing pads. Is there a smoother way to fine tune the system?

 

[pendragon]

What grade of test equipment would suffice to make these sorts of measurements?

Measuring CNR directly is not easy. However many chip-level demods have all the statistical information necessary to do this. I'm sure many satellite meters and portable spectrum analyzers are doing exactly this when they provide a CNR measurement, but this is only conjecture on my part. STBs and PC tuners often display a massaged version of CNR as SQ.

Because one normally only cares about relative CNR variations, any of the above that are stable, monotonic and reliable would likely suffice for this purpose. I would prefer not to do it with my Pansat 9200HD because its SQ values jump back and forth between widely varying discrete values for a given signal. The PC tuners I've tested appear better in this regard.

If you're serious, I would look at using Linux. A number of the drivers pass the calculated CNR as a fraction of the tuner's maximum detectable CNR. There are several weaknesses involved, but they can be addressed because one can always change the code. The trick is to convince the Linux code gatekeepers. I now do many of my studies and alignments with custom Linux apps and drivers, backed and calibrated with my test equipment. It's a lot faster and generally more accurate than manual measurements.

I'm trying to figure out how to make such tools available to the FTA community, because for the most part all one needs is an appropriate PC tuner and an old beat up PC. It would be convenient to get the kernel level stuff into the Linux baseline, but I don't know if that will ever happen. The next problem is engineering this so it is reliable and easy to use. The other side of the coin is users have to know how to use it, because the best tool in the wrong hands can be dangerous.

 

[Cadsulfide]

Pendragon;

As always, it is a pleasure to see your posts. I have a couple of cheapie amps in my array, and they do improve SQ on weak signals, and as your graphs confirm can also wreck SQ if overloaded. Question on using attenuators to pad the amp inputs, most are internally configured as resistive "T" networks, will this not reduce the supply voltage seen by the lnb, and won't the DC I*2R drop in the pad cause it to smoke?

 

[pendragon]

Question on using attenuators to pad the amp inputs, most are internally configured as resistive "T" networks, will this not reduce the supply voltage seen by the lnb, and won't the DC I*2R drop in the pad cause it to smoke?

While not a common item, there are power passing attenuators that solve this problem by providing separate paths for the signal and power. Tech Tool Supply has a reasonable selection of high quality devices.