Hello all,
About a month ago I completed my TubelabSE build and it has brought me great listening pleasure. I constructed it as per build instructions with no modifications whatsoever. I wanted to build it this way to establish a baseline for future modifications.
Yesterday I left the room to make something to eat and when I returned (not sure how much time elapsed) I smelled that awful smell....the electronic death smell. I immediately raced over to the amp and noticed the left 45 with a nuclear orange glow on the plate and heat damage around R5. As fast as I could I cut the power and waited for the unit to cool down.
I think I have narrowed down the cause of the failure, but I wanted to cover all my bases and have someone double check my thinking so this doesn't happen again.
What I think happened was simply thermal runaway. I have had meters monitoring the bias of both 45's ever since it was constructed and noticed one of the tubes was always 1 to 2 ma above or below the other and would always have to adjust the bias point to match. I chalked it up to the tubes being over 70 years old and to be honest I wasn't quite sure how stable this should be since the power supply isn't regulated and the mains power fluctuates between 117 and 122.
I checked for a shorted bias pot, but I didn't think this would cause such thing unless it shorted wide open allowing no bias voltage at all, but the pot tested fine. I then checked the mosfet for shorts as this is all i can do as I don't have a transistor checker. The mosfet tested fine. All other resistors, capacitors, and semiconductors tested fine.
I think the tube went into thermal runaway (yes, they were run above spec at 320V and between 26 and 28mA although George hasn't had a problem with any tubes at this operating point) and the massive current drawn by the tube overwhelmed R5 causing it to heat up dramatically. Amazingly it did not fail and still tests ok (of course it will be replaced though) and did some damage to the board since it was mounted on the bottom of the board.
Let me know if my logic here sounds correct or if you have any better ideas I'm game. Thanks!
About a month ago I completed my TubelabSE build and it has brought me great listening pleasure. I constructed it as per build instructions with no modifications whatsoever. I wanted to build it this way to establish a baseline for future modifications.
Yesterday I left the room to make something to eat and when I returned (not sure how much time elapsed) I smelled that awful smell....the electronic death smell. I immediately raced over to the amp and noticed the left 45 with a nuclear orange glow on the plate and heat damage around R5. As fast as I could I cut the power and waited for the unit to cool down.
I think I have narrowed down the cause of the failure, but I wanted to cover all my bases and have someone double check my thinking so this doesn't happen again.
What I think happened was simply thermal runaway. I have had meters monitoring the bias of both 45's ever since it was constructed and noticed one of the tubes was always 1 to 2 ma above or below the other and would always have to adjust the bias point to match. I chalked it up to the tubes being over 70 years old and to be honest I wasn't quite sure how stable this should be since the power supply isn't regulated and the mains power fluctuates between 117 and 122.
I checked for a shorted bias pot, but I didn't think this would cause such thing unless it shorted wide open allowing no bias voltage at all, but the pot tested fine. I then checked the mosfet for shorts as this is all i can do as I don't have a transistor checker. The mosfet tested fine. All other resistors, capacitors, and semiconductors tested fine.
I think the tube went into thermal runaway (yes, they were run above spec at 320V and between 26 and 28mA although George hasn't had a problem with any tubes at this operating point) and the massive current drawn by the tube overwhelmed R5 causing it to heat up dramatically. Amazingly it did not fail and still tests ok (of course it will be replaced though) and did some damage to the board since it was mounted on the bottom of the board.
Let me know if my logic here sounds correct or if you have any better ideas I'm game. Thanks!
I don't know the circuit so I can't make an intelligent assessment. However, if it's resistance coupled, I would replace the grid coupling capacitor. What happens there is a bit of leakage as it warms up tends to raise the grid voltage a bit. That makes the tube draw more and warms up the area, which increases the leakage in the capacitor, and there goes runaway.
This also may be why you have seen variations in bias, maybe temperature related.
This also may be why you have seen variations in bias, maybe temperature related.
Binaural,
Did you mount all of you components on the bottom of the board? If so make sure that your FETS have plenty of heat sink. Replace them anyways even if the measure fine. I had mine do a similar thing as a result of the heat sink being loose. FET quit working and bias went through the roof.
The amp uses a FET follower which results in a very low grid impedance (assuming it is not fried) so runaway would not be my first focus. Most likely the FET failed or bias was lost do to a bad connection or other component. bob91343's comments are also another possibility regarding the cap if it is leaking. They are close to the FET heat sinks and could get quite toasty. Also check very carefully your solder connections.
FYI, I have abandoned my bottom mounted Tubelab SE board, not because it didn;t sound good or its design, but the thermal issues in having the components flipped made it to unreliable. George does not recomend doing this.
Did you mount all of you components on the bottom of the board? If so make sure that your FETS have plenty of heat sink. Replace them anyways even if the measure fine. I had mine do a similar thing as a result of the heat sink being loose. FET quit working and bias went through the roof.
The amp uses a FET follower which results in a very low grid impedance (assuming it is not fried) so runaway would not be my first focus. Most likely the FET failed or bias was lost do to a bad connection or other component. bob91343's comments are also another possibility regarding the cap if it is leaking. They are close to the FET heat sinks and could get quite toasty. Also check very carefully your solder connections.
FYI, I have abandoned my bottom mounted Tubelab SE board, not because it didn;t sound good or its design, but the thermal issues in having the components flipped made it to unreliable. George does not recomend doing this.
+1 to the FETs. Also, FETs can go bad without "blowing" like BJTs and diodes often do. I wouldn't trust them. Bias drift may be indicative of the FETs getting really hot. However I wouldn't rule out the tube either. I have a couple of persnickety 45s whose filaments kink slightly when operating (not enough tension from the spring hook?) and come very close to touching the grid wires.
Another consideration are those trimmers. They are cermet pots, which really aren't meant to endure a lot of tweaking...usually in the 100s of operations. I've replaced mine with carbon trimmers for this reason.
Another consideration are those trimmers. They are cermet pots, which really aren't meant to endure a lot of tweaking...usually in the 100s of operations. I've replaced mine with carbon trimmers for this reason.
SGregory,
My second choice would've been the mosfet. Other than the tubes, they are the only thing mounted on the top of the board, so in this case heat wouldn't have been an issue. The coupling caps are mounted on the bottom mainly because they are so big (Audio Note .47uf copper in oil caps.) I haven't checked the tube yet to see if it's ok. It has lost some of it's slivering but I assume that's because of all the emission inside the tube from the glowing plate. If it hasn't killed it I know it has at least knocked some life out of it, so I'll probably have to go shopping for a new tube as well.
Like I said, I built it with no mods for a baseline and was planning on upgrading R5 with a choke, so I guess now would be the time to do it! I was also planning on replacing the B+ power supply caps with motor runs, so they will no longer be on the board. I think the final amp will have the tubes on some kind of extensions to mount them on the top plate of the chassis so this should allow everything to be mounted on top of the board. I'll probably have to mount the 5842 grid stoppers off of the board since the extra space between them might cause oscillations since this tube will oscillate if given the chance....
Thanks for the suggestions. I was planning on replacing the fets just in case they were the culprit. They're cheap and will hopefully save me much head scratching and pain in the process.
Joe
My second choice would've been the mosfet. Other than the tubes, they are the only thing mounted on the top of the board, so in this case heat wouldn't have been an issue. The coupling caps are mounted on the bottom mainly because they are so big (Audio Note .47uf copper in oil caps.) I haven't checked the tube yet to see if it's ok. It has lost some of it's slivering but I assume that's because of all the emission inside the tube from the glowing plate. If it hasn't killed it I know it has at least knocked some life out of it, so I'll probably have to go shopping for a new tube as well.
Like I said, I built it with no mods for a baseline and was planning on upgrading R5 with a choke, so I guess now would be the time to do it! I was also planning on replacing the B+ power supply caps with motor runs, so they will no longer be on the board. I think the final amp will have the tubes on some kind of extensions to mount them on the top plate of the chassis so this should allow everything to be mounted on top of the board. I'll probably have to mount the 5842 grid stoppers off of the board since the extra space between them might cause oscillations since this tube will oscillate if given the chance....
Thanks for the suggestions. I was planning on replacing the fets just in case they were the culprit. They're cheap and will hopefully save me much head scratching and pain in the process.
Joe
Just take the PA tubes out and measure the voltages at the socket to see what is going on.
When I lost a FET, my bias voltage went so negative the tube was completely cut off.
Sounds like you might have had a runaway, particularly in light of exceeding the published spec. That's always great fun, but sometimes not without consequences.
Win W5JAG
When I lost a FET, my bias voltage went so negative the tube was completely cut off.
Sounds like you might have had a runaway, particularly in light of exceeding the published spec. That's always great fun, but sometimes not without consequences.
Win W5JAG
I am having a similar problem. After months of operation today I heard a bop, and bad smell. What is happening is that, if I measure the voltage on the grid of the 300B, it starts to go down, and then suddenly it gets all the B+. I hope I was fast enough to switch it off and the tubes are still safe.
I though the problem was only on the left channel, so I replaced only the mosfet on the left. Nothing.
So I run out of FET, and I replaced the Toshiba K3563 with two IRF840. I am not sure if they can work there, but I had the same behavior.
After I removed the coupling caps, but still the same.
Any hint ? I am out of ideas.
I think the first failure was caused by a lose screw on the heatsink.
D.
P.S.: Any chance that the tubes are still OK ?
I though the problem was only on the left channel, so I replaced only the mosfet on the left. Nothing.
So I run out of FET, and I replaced the Toshiba K3563 with two IRF840. I am not sure if they can work there, but I had the same behavior.
After I removed the coupling caps, but still the same.
Any hint ? I am out of ideas.
I think the first failure was caused by a lose screw on the heatsink.
D.
P.S.: Any chance that the tubes are still OK ?
Now I am really puzzled. The right channel is OK.
On the left I do not manage to put the bias below -15 V. I replaced all the resistors, the trimmer, the coupling cap and the mosfet.
I am doing the measurements without the rectifier. I really do not know what to think any more. Please help.
D.
On the left I do not manage to put the bias below -15 V. I replaced all the resistors, the trimmer, the coupling cap and the mosfet.
I am doing the measurements without the rectifier. I really do not know what to think any more. Please help.
D.
On the right channel I have D: 418 V S: - 70 V G: -70 V Voltages are not precise, as I wanted to be quick.
On the left D:418 V and S & G: 357 V.
On the left channel the coupling cap is not connected, but this should not make any difference.
I am really puzzled, I changed the FET many times, It's a bit difficult, as I am running high in voltage, and without tubes I can go well over the max voltage of the FET.
Thanks,
D.
On the left D:418 V and S & G: 357 V.
On the left channel the coupling cap is not connected, but this should not make any difference.
I am really puzzled, I changed the FET many times, It's a bit difficult, as I am running high in voltage, and without tubes I can go well over the max voltage of the FET.
Thanks,
D.
You need to check and report back the voltage of the bias supply. Check for both DC and AC. I have had the diodes (d2-d3) fail for the supply as a result of the FET failure before. Then check/replace the dropping resistors (r5/r6).
Looking at the results listed I would think that the left may have a problem in the FET bias voltage divider and/or R25 is either open or smoken hot. (FET may be bad as well.)
What are you using as a FET? The 2700 is good to 900V and the recomended replacement of 2sk3563 is good to 500V. If you are running that high of a voltage you may wish to switch to a 2sk3564.
Looking at the results listed I would think that the left may have a problem in the FET bias voltage divider and/or R25 is either open or smoken hot. (FET may be bad as well.)
What are you using as a FET? The 2700 is good to 900V and the recomended replacement of 2sk3563 is good to 500V. If you are running that high of a voltage you may wish to switch to a 2sk3564.
Thank you for help.
All the resistor of the voltage divider have been replaced. R5 looks fine, I have a big one there.
The problem I have, and I do not understand is that if the B+ is not connected to the power drive both the voltage of the grid go negative, very negative, up to -300 V. It was always like this.
The right channel is complete and shows normal behavior.
On the left channel I have:
1) No coupling cap
2) No MOSFET
If the jumper of the power drive is closed I get all the B+ on the grid, and I do not understand where it comes from.
Note that with the jumper open, the voltage stays negative.
I have to say that replacing the components I damaged the surface of the board. Is it possible that is shortening somewhere ? If it is, it's not obvious, with no voltage, I measure 0.5 Mohm between grid and power drive B+
for the test I lowered the B+ tp 330 V with no load.
Thanks,
Davide
All the resistor of the voltage divider have been replaced. R5 looks fine, I have a big one there.
The problem I have, and I do not understand is that if the B+ is not connected to the power drive both the voltage of the grid go negative, very negative, up to -300 V. It was always like this.
The right channel is complete and shows normal behavior.
On the left channel I have:
1) No coupling cap
2) No MOSFET
If the jumper of the power drive is closed I get all the B+ on the grid, and I do not understand where it comes from.
Note that with the jumper open, the voltage stays negative.
I have to say that replacing the components I damaged the surface of the board. Is it possible that is shortening somewhere ? If it is, it's not obvious, with no voltage, I measure 0.5 Mohm between grid and power drive B+
for the test I lowered the B+ tp 330 V with no load.
Thanks,
Davide
Without the B+ hooked up to the Power Drive, it does not conduct any current. Only the bias voltage divider will conduct a little. Without the current there is no voltage drop across R5 and R6 so the voltage will be very negative. The grid voltage will then be pulled down. When you have it in this condition you are also creating a VERY large Vgs condition. The gate will be held at the potential divider voltage while the source will be at the negative rail voltage. There will be no voltage drop across R14 and R25 as there will be no current there as well. Probably enough to toast the FET again.
You will probably need to get both FETS installed and connected properly to see the negative rail in the range of -150V and not exceed the Vds of the FET.
For reference, I measured a good 2sk3563 and resistance from drain to source is >10Mohm in the reverse direction it is 2.8M ohm. iirc my blown fets measured about .5M
I have no idea what your jumper is.
You will probably need to get both FETS installed and connected properly to see the negative rail in the range of -150V and not exceed the Vds of the FET.
For reference, I measured a good 2sk3563 and resistance from drain to source is >10Mohm in the reverse direction it is 2.8M ohm. iirc my blown fets measured about .5M
I have no idea what your jumper is.
Thanks,
I will make some more tests following your hint.
For the jumper, I am referring to the contact that is soldered close to the 10 ohm resistors to feed the power drive.
I still do not understand, how can I get the B+ on the grid with:
1) No coupling cap
2) No tube
3) Ne FET.
Thanks,
Davide
I will make some more tests following your hint.
For the jumper, I am referring to the contact that is soldered close to the 10 ohm resistors to feed the power drive.
I still do not understand, how can I get the B+ on the grid with:
1) No coupling cap
2) No tube
3) Ne FET.
Thanks,
Davide
Let me make sure I understand what you are saying.
On the left channel, with no coupling cap, tube, or FET you get B+ on the grid. The resistance between the grid and B+ is 0.5Mohm, in the above configuration.
I guess the question is what is the resistance between the Grid and C-. If the above is correct then it does sound like you have a carbon track on the board, possibly from changing parts. Most likely place is right were the FET solders in. Clean both sides VERY well and see if the resistance improves. DO the source resistors measure okay. If so then I would suspect that the short between B+ and the grid has a much lower resistance.
I will have to pull my tubelabSE board out tonight and look for the jumper next to the 10 ohm resistor you are talking about. I can't get my head around where it is.
P.S. I found it.
On the left channel, with no coupling cap, tube, or FET you get B+ on the grid. The resistance between the grid and B+ is 0.5Mohm, in the above configuration.
I guess the question is what is the resistance between the Grid and C-. If the above is correct then it does sound like you have a carbon track on the board, possibly from changing parts. Most likely place is right were the FET solders in. Clean both sides VERY well and see if the resistance improves. DO the source resistors measure okay. If so then I would suspect that the short between B+ and the grid has a much lower resistance.
I will have to pull my tubelabSE board out tonight and look for the jumper next to the 10 ohm resistor you are talking about. I can't get my head around where it is.
P.S. I found it.
Last edited:
At the end, it was a short on the board, I cleaned it, but the only way to fix it was to cut the path of the B+ to the fet, and connect it directly.
I can do good work when I assemble things, but for me, to do reworking is always a mess, and things look like battlefield after.
Thanks for help,
D.
I can do good work when I assemble things, but for me, to do reworking is always a mess, and things look like battlefield after.
Thanks for help,
D.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.