It runs even hotter…
But it has a metal casing so it can be cooled much easier…
@wiedehopf @SoNic67
Interesting result: Thermal pad all over the stick to prevent short circuiting, inserted stick in Windows pc without antenna connected because that gives too much signal to see leaks, SDRSharp on screen, and then moving metal shaped objects over the thermal pad. In the graphs I am looking for increasing spikes and contrast.
Turns out…on top of the R820T2 and between R820T2 and that ‘gold box’ component in the same line of SMA data is the danger zone. Metal there will do something with the signal. I sticked a little heatsink (often used on the Pi LAN bridge chip) on top of R820T2 and again, a peak.
In addition there could be a zone below the center capacitor also, although impact is a lot less that the real ‘danger zone.’
Next step is to verify this by repeating the metal shaped objects over the thermal pad while looking at message rate and aircraft count.
You got to be kidding me…
Testing with USB cable…no more phenomenon.
Back to direct connection…no more phenomenon.
Yeah this indeed is voodoo.
1 Like
Hmm… in a sealed box that won’t make any difference. Been there done that.
Did hours of testing today.
It seems that metal around the red highlighted areas shown in the picture below attracts signal.
(Red+white spots difficult to confirm.)
Even a heatsink on the back side of the pcb is subject to becoming an antenna, albeit much less than on the front side.
My conclusion basically, which in the world of RF of course could change 5 seconds from now, is that installing a heatsink anywhere on the stick might degrade performance.
1 Like
Not so much that it becomes an antenna.
You change the capacitance on the board which can shift frequency response.
Interesting that the rtl-sdr v3 does have a thermal pad and still works fine.
Not sure on which side of the PCP it’s located though.
May well be that it’s on the other side.
Other side should be no problem because it’s a big ground plane.
The rtl-sdr dongles have the thermal pad on the bottom of the board. This is a v2:
It’s quite a thick pad:
1 Like
A heatsink on the back side of the pcb with a few millimeters of thermal pad in between might just be enough space to not be a problem anymore. As is confirmed to be their practice by the last picture. 
Starting from the beginning - how have you determined a heatsink is necessary?
No. It just would have saved you time and money.
What does “attract signal” mean?
If you spend some time learning the names and functions of the major components, it will become self-evident which areas should be treated carefully.
Taking the voltage regulator as an example (most dongles use a xx1117 chip*, so I assume this does too), the spot you have marked is the DC output. If placing metal near this spot changes your reception, you have bigger problems.
The area above the 820T2 is the L & C components making up the tracking filter.
One of the ‘gold box’ is a 1090MHz SAW filter, the other is the reference crystal. Hardly surprising these are sensitive areas.
What sort of thermal pad are you using and do you have a data sheet?
(the datasheet alone may tell you why you are not getting the results you expect)
** if the output of the reg starts vibrating, it may be due the choice of part numbers!
1 Like
@geckoVN
Chips get hot, cooler chips perform better, thus apply cooling.
It’s just that I am new to RF and did not see all the side effects coming that are not in play when installing custom cooling on for example a computer CPU.
‘Attract signal’ …I mean when a object starts doing things that makes an RF signal is being picked up by the stick, being either real or fake, unwanted both ways.
The test was done without antenna as noted in the post before the one with the image, but should have been noted in the one with the image also, my mistake. That means attaching an antenna might solve some of the spots… And that’s the RF problem…might. That’s why I also use ‘might’ in the conclusion.
While picking the large heatsink, my intention was never to cool all of the components. It was to get maximum cooling performance
Yes, I lack knowledge on the part of electro components and even more on RF. Studying these subjects will help, yet this is such a niche area and clearly some phenomena cannot be explained at all. So practical testing in my opinion has to happen any ways. There must be quite some people now happily running their sticks with big heatsinks, not knowing they gave themselves a performance penalty.
About that voltage regulator ‘LDO’… The main chips (RTL2832, R820T2) work at 3,3V and maybe even the whole board. When supplying 4,6V due to a longer USB cable, what are the consequences? Wiedehopf said a drop of performance. But that LDO is not so stupid to just take the input voltage, divide it by 5 and multiply with 3,3 I hope? I will measure today.
The thermal pad is Gelid Solutions Extreme Thermal Pad 1mm. I chose this one to level out unevenness with as little thermal pad as possible, because it’s conductivity is much lower than a metal heatsink one’s. Now after testing it becomes clear you need to have the heatsink some minimal distance away from the pcb. I am reasonabily sure (from testing) that covering the pcb with this thermal pad does not affect performance much, with a possible exception for the first centimeter from the SMA side.
Well, pretty much all of the effects can be explained. But i couldn’t explain it well because i haven’t studied Electrical Engineering 
Also most people couldn’t be the recipient of the explanation because they lack a basic understanding of the subject.
Understanding changing capacitance and how it affects for example an LC circuit is easier.
So you can get an appreciation for what is critical and what isn’t.
More than one chip on the board.
Have fun investigating the effects of reduced voltage.
Could well be that the LNA needs 4 V and still has an LDO which drops voltage by 0.8 V.
So in that scenario everything below 4.8 V, the LDO would still drop 0.8 V and the LNA would get insufficient voltage.
Really figuring out what the problem with low voltage is, that’s hard. For that you more or less need a schematic with all the datasheets for the components.
But you can sure test experimentally at which voltage the performance starts to drop, that would be interesting.
Sorry, what do you mean? Some are on 5V? Most are on 5V? Most are on 3,3V? All are on 3,3V?
If I would be able to investigate the voltage effects, I gladly would do so. Problem is 1) I am leaving the country next week for half a year, so a) I have to quickly get a reasonable stable setup and b) I cannot do ADS-B hardware testing in that time period. 2) I don’t have the tools to dynamically adjust input voltage.
Though, when changing between no USB cable, short USB (30-50 cm) cable and long (3 m) USB cable, no difference caught my attention concerning performance. Might be able to do a ‘real’ test on this.
How could a chip like 820T2 be designed let alone perform reliably if “some phenomena cannot be explained at all”?
Perhaps people do learn more studying a degree than you can teach yourself in a couple of weeks - who’d-a-thought?
To describe RF as “niche” is rather naive. You are saying mobile phones, television, wifi, gps, ethernet, satellite communication are all obscure or unusual.
If you read the datasheet I offered, perhaps you’ll be enlightened.
There is no datasheet offered for that product, but it is clearly intended for the OC community where no allowance for resistivity, dialectic constant, loss tangent and of course the tendency to absorb moisture (which will change all the other parameters).
A very brief trace of the tracks will answer that quesion.
An ‘LDO’ is a subset of linear voltage regulators. No linear regulator works by dividing and multiplying (that by definition is not a regulator).
Within its range of operation, a regulator puts out a constant voltage regardless of input.
Possible, but no one here has reported using this technique. Anyone adding cooling has described using fans or heatsinks either direct on the heat generating chips or the back of the board
Have you considered ceramic heatsinks?
The main gotcha with putting anything close to the PCB is that anything conductive will effectively turn into a transmission line (Transmission line - Wikipedia) or waveguide to some degree. At GHz frequencies this happens with even small conductors (the 1/10th-wavelength rule of thumb suggested on the wikipedia page is ~3cm at 1GHz) - e.g the PCB traces and any nearby metallic heatsink.
What they teach them today is just programming, and high level math. Nothing real useful in the trade, the professors don’t have a clue of what is required from an EE in real life.
They don’t know that there are electrical codes in effect. Half don’t know anything about the voltage distribution system in their country (they might answer sheepishly 120V or 230V, but nothing else).
SMD electronics RF design? There is not a teacher that knows how to do that, everyone that knows works for big companies and they keep it almost as a trade secret.
I say this after I mentored 10 young EE, fresh out of school. Talked to University Teachers about that and I was told that those are “trade” materials, not worthy of “academic” study.
University degrees aren’t really intended to be vocational training though. They are there to teach you how to learn, with some relevance to the subject at hand. The specific application of knowledge isn’t really their remit (especially at undergraduate level), since that is something you are likely to have to learn through professional development later on. A degree gives you a body of knowledge to build upon - it’s not an apprenticeship.
3 Likes
Talking about RF when sticking a big custom sawed heatsink on an FA stick and finding out the impact of that. Niche in practice.
It’s full with specs. Really nice for an electro noob.
In that case I would say, that nothing is working at 5V on the board, but I cannot be 100% sure, that’s why I am asking.
This seems to suggest something. So 4,75V might be the minimum input to get 3,3V output.
Yesterday I measured the stick on the computer, today on the Pi.
Pi throws out a higher voltage than the computer: 5,18V on the 3 meter cable without load. 4,85V on a load of 130 msg/s in Dump1090 and 4,82V with 250 msg/s. I cannot measure with a higher rate because the stick has to be in an inaccessible position then.
I also measured the LDO output… 3,32V. Further along the board…two or three times 3,3-3,32V. So it sounds unreasonable to get a performance penalty from a long USB cable because the LDO still puts out 3,32V…unless some other component is using 5V.
But now practice again… And a big surprise: Taking out the cable…boosted the msg/sec rate massively. So @wiedehopf is right once again.
(Yes gain is too high, it will be tuned when antenna is in final position.)
In theory it needs a 3 cm free radius? Damn.
All in all, once more the plans are ruined. 
At this moment I am strongly considering to replace the 3 meter USB cable with 3 meters of high quality coaxial cable. With the stick inside, active cooling would be possible then.
1 Like
They consider “vocational training” everything that they don’t understand. Including the National Electrical Codes, Lightning and Energy Codes, Lightning protection codes, production, transport and distribution of electrical energy, protection coordination, CAD.
It became a racket to provide Degree Paper for money.
No wonder that the required 5 years (when I graduated in 90’s) to get an EE degree became 4 year and now is just 3 years…
They “get” you with 3 year of Master on the back side though…
Well, not what I meant. See the wikipedia page:
A common rule of thumb is that the cable or wire should be treated as a transmission line if the length is greater than 1/10 of the wavelength. At this length the phase delay and the interference of any reflections on the line become important and can lead to unpredictable behaviour in systems which have not been carefully designed using transmission line theory.
How close it has to be to cause problems is going to depend on the geometry, the dielectric, etc. My point was more that this is probably why you’re getting unpredictable behaviour when you stick a big heatsink parallel to the PCB traces carrying GHz-frequency signals because you’ve just built a transmission line for those signals.
Those should all be on the top of the PCB, right?
Therefore a pad on the back side (ground plane) should be much less critical.