|
Hello all, My backstory is that I blew the BS170 finals on my high bands QDX rev 5 (12V). I had just completed a QSO and I had recently messed with my antenna setup, perhaps resulting in SWR between 1.5 and 2.0. Symptoms were that QDX started drawing too much power, which caused the USB PD cable I was using to panic and drop from 12V to 5V, which caused QDX to go into disk drive mode. I then?unplugged it; by then the BS170s were pretty hot. IC2 also seemed hot if I recall correctly. I removed the BS170s and confirmed all voltage regulator outputs looked normal, receive filter tests looked normal, and I could receive FT8. Then I proceeded to replace the BS170s with TN0110s as suggested in some threads here. I also added 56V 1W Zener diodes as overvoltage protection. Since the TN0110s have reversed pinout compared to BS170, and since I wanted to keep the leads short for better heat transfer, I kept them standing up straight. I put a raspi heatsink on the open metal on the board between the?pairs, which seems to help draw some heat away. I have an order for some of the cute TO-92 individual hat heat sinks that I also plan to add. It seems like this fix/mod is working well. Power output on all bands looks fine and comparable to before I blew the BS170s. I've only tried operating 20m FT8/JS8 so far, and the finals stay below 100F for that. My PSKReporter spotting seems fine, and my QSO rate seems okay; not?amazing but I'm having fun. However, I have noticed that two of my TN0110s (Q9 and Q11)?seem to heat up significantly more than the other pair once I start transmitting. Could this suggest an issue with my T1 winding? Or an imbalance in the components (I didn't check them thoroughly before installing)? Thanks! Peter |
|
Thanks! I tried checking IC5 using my DC voltmeter, running QDX at 6.3V and using WSJT-X "tune" mode to transmit into a dummy load.
The Tx input to all AND gates is 3V. The Clk inputs are close, but not quite balanced: Clk0 which drives Q8/10 reads ~1.46V while Clk1 which drives Q9/11 reads ~1.325V. (Of course this is a smoothed reading of the oscillating signal.) The IC5 outputs are more imbalanced, reading about 1.7V to Q9/11 and 2V to Q8/10. Does this sound like a problem with IC5? I can try replacing it, but I don't relish the prospect. I hope there's not an issue further upstream. I have a scope somewhere but not sure it's operational. I guess a spectrum analyzer could help confirm a functional imbalance in the output? |
|
Hello Peter,
The first thing to do is to change the input voltage to 7.5 volts.? This is just above the minimum, where the 5-volt regulator can regulate the voltage.? With the higher voltage, measure the gates of Q8, Q9, Q10, and Q11.? They should be close to 2.5 volts each when transmitting and close to zero when receiving.? If not, then IC5 is most likely bad. With the heating imbalance, I agree with Allison that IC5 is most likely the problem.? The above test is to confirm. I, too, would be curious to see how you have installed the TN0110s. 73 Evan AC9TU |
|
Thanks! Rechecking at 7.5V, finals and IC5 look okay during Rx. On Tx it's imbalanced: 2.5V on the Q8/10 gates, 2.25V on the Q9/11 (hotter) gates. I guess I should?try to replace IC5?
I'm a bit concerned that I see imbalance on the Clk0/1 inputs to IC5 as well: 1.6V driving Q8/10, 1.45V driving Q9/11. As a proportion of 3.3V, that seems similar to the imbalance seen at the output? (Incidentally, it would be handy if there were a "tune" mode in FT8CN Android app, but I didn't find one.) |
|
Hi Dean and Evan, here's a picture of my TN0110 finals installed:?
(The Zeners are on the back side as suggested by?LA3ZA.) At first I thought I'd bend the finals down and put a heatsink across their flat tops, but laying them down would have lengthened the leads. Shorter leads seemed more important, so I just put them in standing up and dropped the heatsink between them. It seems to at least draw some heat up from the board that way. The heatsink I used was suggested by Barb for the other PA mod:? I also ordered some of these cute hat heatsinks: Not sure they'll really help, but keeping things cooler gives a little peace of mind after blowing one set of BS170s. |
|
I guess it's a fair question why this part of the board should need heatsinking. It did seem that my heatsink tracked the temperature of the finals, but this wasn't at all a careful plan or a careful observation. Once I have my imbalanced heating figured out I will revisit this. The real question is what's the right way to put in TN0110 finals if you want to try them.
|
|
Thanks Bojan! Dean, interested to hear your experience. Please check the other threads on TN0110, in particular it was mentioned that the higher switching speed could mean bigger kickback if used in QDX. I took this to mean you should plan to use Zeners also.
Some thoughts on the heatsink... If most of the cooling of the finals is through the leads, especially the drain, then keeping leads short is good. I then thought putting a heatsink on the board near the leads was reasonable (plus honestly, I had a heatsink in my hand and a big flat space open on the board and just said why not). On the other hand, most of the area I put the heatsink on was already designed to be a heatsink for the BS170s (but in the other direction!) so that makes it look pretty dumb. What I observed in this setup is that the heatsink does warm when the finals warm (but more slowly). Given the above, this seems interesting to me, but since my QDX has other heating issues I won't get more into it until the other issues are fixed. I also observed that?my finals stay below 100F while chasing QSOs at 5W FT8 (so say 25-40% duty cycle). If you have the hat heatsinks also, you could consider tying them to the middle heatsink mechanically. If the fins form a solid interface this might start to make some sense, although still quite a hack to fit the not-designed-for mechanical situation. There are probably simpler and better options. Apologies for sharing half-baked parts of my setup; I tend to err on giving more information but I should have included more disclaimers. |
|
I believe that less than half the 'likely' cooling goes through the wires and that would be in the unlikeliest of cold board + cool air scenarios? (the Difference between BS170 and standard TO92 cooling model has never been fully explained by the maker) ? Keeping the plastic package cool (say 40? - 45C)? rather than floating around 70C when under load ) will reduce the junction temperature by? around 20 C at 0.5W Pdiss? as it helps both the traditional Tjc path and the wires that first contact and then exit through the epoxy package and so can get cooled more/sooner at that stage too? |
|
Hi Peter, I like what you’ve done so far. Keeping the TN0110 leads as short as possible is in line with many comments that the leads conduct most of the heat away from plastic TO-92 packages. The Raspi black-finned heat sink is also a good idea to help improve heat transfer from the board itself. Finally, the metal transistor hats will be a plus but likely marginal. Still, I’d do this too. There are copper finned heat sinks in the Raspi form factor that would be better than black aluminum so you might want to look into these as well. Using black paint will not help at all and will probably decrease heat transfer. Not a good idea, I’d scrape off that paint. The TN0110 will produce a larger voltage kick at turn off than the BS170 but TN0110 is a 100V device compared to 60V for the BS170 so you have considerably more safety margin. However, the lower gate threshold of the TN0110 will produce a higher drain current which is why you get a higher voltage kick at turn off. I’d go with one of the diode bypass methods offered in this forum for a better safety margin even though it’s probably not necessary. Some here have experimented with cranking the gate driver AND’s up to 6V. While this will help drive the BS170 to be a better switch, I doubt this would help at all with the TN0110 so don’t bother with that mod. The TN0110 is already a better switch since its gate threshold is lower. Finally, the imbalance you’re seeing could be due to mismatched transistors with different thresholds. I intend to match my TN0110’s with a DC threshold measurement using a superstrip breadboard. But investigate the other possibilities suggested here first. A scope on those gates would be very useful for troubleshooting this if you can get your hands on one. Overall, you’re doing great with these mods and keep up the feedback. We’re rooting for you! Tony - AC9QY On Sat, Dec 30, 2023 at 2:42?PM Peter Li <chinasaurli@...> wrote: Thanks Bruce! Once I have the finals imbalance sorted out I will try adding the hats and report back if I have anything useful to add. |
Bruce Akhurst