Jump to content

jokerman777

Returning Member
  • Posts

    68
  • Joined

  • Last visited

Everything posted by jokerman777

  1. ๐Ÿ˜ƒWas trying not to dump more flat-earth theories of T2 to the thread as I already did a bunch, but here are what (I think) my problems were: I didn't find the cause of the oscillations appeared at the output, but they are killed by upping the feedback caps from 5pf to 10pf. Some "AC components" I measured around output plate CCS might actually be oscillations excited by me sticking my handheld meter probes across floating nodes - I recall when I measure referencing ground it was different but better, but in the end when I probe with oscilloscope I see nothing there. When I trim both channels in place, turns out the 10M90S in -500V rail reached the boundary of current limiting so it's trying to sag and gives a sawtooth wave on that output rail voltage, hence appeared at the amp output. Lowering the current limiting resistors fixed it. When it comes to getting preamp tubes for T2, probably the most important and minimum requirement for the circuit to work is to have well-balanced triode sections for each individual tubes. Some permutations of the quad JJ tubes I got will give too much imbalance that is outside the correction range of the servo. Also I find my left channel having more tolerance on input tube mismatch for the servo to correct than the right channel - suppose the better matched the rest of the circuit is, the more tolerance on input tube mismatch you will get.
  2. Fixed this amp a while back and been running it day to day ever since. So far everything works and seems to be stable, I finally get to enjoy the music coming out of it ๐Ÿ™‚ It's definitely quite a thing building and troubleshooting it in the past 8 months or so, major lesson learned is that I could've gotten a scope on day 1 if I could turn back time. Many thanks to Kevin for providing such a design - the major heavy lifting of bringing back a great amp. Thanks to GeorgeP for supplying the hardware with all the building advices, or I would never have considered building one. And a special thank to Simmconn, who spent a lot of extra time not only helped me with diagnosing the issues, but also taught me a lot of the nuts and bolts about proper troubleshooting practices in general as well as demystifying the T2 circuit to me through patiently explaining. Of course there are a few more to mention, thanks to the folks out here who kindly helped me. It takes 10+ headcase veterans to help a newbie like me building a working T2 ๐Ÿ˜„
  3. @G600 I got my boards from GeorgeP and it should be whatever is the latest version of Kevin's design with original parts (as opposed to "modern version") Haven't got back to this thread for a while but I made some progress over the holiday time identifying the problem with a lot of help from simmconn as well as finally using a scope. -Balance: turns out by changing the permutation of my preamp tube locations in certain case I can get both channels balanced, likely some of my 6dj8 has better matched sections than others which were kicking the servo out of working range. -Offset: Replacing R73 with 10k pot I was able to trim out the remaining offset to near zero no problem. -Oscillation: quite possibly my last bit of problem, that I was observing low frequency (60-80Hz) sawtooth waves at all outputs. left-channel has seemingly matched waveforms on O+ and O- with hundreds of millivolts peak-to peak and right-channel has unmatched waveforms with one measures 1.42V pk-pk and the other much less and also reversed pattern along time axis. Slightly trimming the battery voltages doesn't seem to affect this behavior. When I got time again I will inspect things around the balance servo as well as breaking the loop to see if I can find out where it comes from. ๐Ÿ™‚
  4. Might have gotten a tiny bit more clue, been trying random things while waiting on a batch of K216/J79 replacement parts to arrive ($$$๐Ÿ˜ข). I first tried replacing Q6/Q7, Q11/Q12 with KSA1156s, nothing changed, then I tried putting both C12/C13 in there instead of either, nothing changed (thus removed one again). Then I tried removing 5pf feedback caps across R78, R79: -Q33 still measures weird, plate CCS still violates Kirchhoff's law and are imbalanced, one J79 still has positive Vgs -Output offset went straight to -80V -Balance servo seems to be working in this case, with O+/O- within 1V difference and C/D being 1/10 of that. -Setting meter to AC measure mode and probe between O+ and ground, 18V @ 400kHz... O- doesn't give any significant reading, near 0V at near 0 frequency. So first of all it's possible that by removing these caps I just introduced new/worse problems. But it also makes some sense if oscillation is a part (if not all) of my original problem? if O+/O- were swinging some big time AC that would kinda explain why DC measure around CCS voltage doesn't check out, and why Q33 is having invalid Vbe from DC measure to be turned on yet Q32 still seems to be conducting fine? I certainly have absolute 0 knowledge on how to track down and troubleshoot stability issues, probably next step for me is just mess around with anything that have a time constant in it - capacitors, right after my previous battles against all diodes and resistors. I'm planning on trying: 1. Replace C1 with mica, I don't know what this RC network does but I seen other people built working T2 having mica cap there (could be wrong, only eyeballed from shape), and I used a film. 2. put 5pf feedback cap back and then try slightly higher values for comparison 3. maybe recap balance servo just in case I messed up some .47u/.1u values, hard to tell what I had there from looking Not too hopeful if this will fix anything but at least cure some boredom while waiting on those rare MOSFETs..
  5. @simmconn Thanks for your explanation, once again ๐Ÿ™‚ Sorry for the confusion and yes I meant D6/D7 (or D10/D11). Specifically, taking one side of CCS as an example: 1. V(D4) + V(D5) = 3.6V, Vbe(Q6) = 0.45V, V(R17) = 3.2V, this is all good. 2. Probe across R18-R20, all gives around 2.4V 3. Probe across B-E of Q8-Q10, all gives 0.57V 4. Probe across D6: 1.8V 5. Probe across D7: 0.7V (and probe D6, D7 together gives about 2.6V) D6, D7 look as bright, first round of trouble shooting I already replaced all LEDs and elevated them from contacting the board. Amp powered off and I diode test them they all fine with around 1.83V of Vf. So 2-5 makes no sense only thing looks ok is Vbe across C3675s. I will go ahead and check board connectivity tho. (Everything looks wrong when one probe is stick to O+?)
  6. nvm, staring at the schematic a little bit more Q34-Q36 probably do this, referencing -560V supply negative battery voltage would like to be about 3x of V(R74) makes sense that it likes to stay somewhere between -540V and -550V
  7. I been focusing on the "working" channel in the past two week, with +/- offset of -5V/-10V (they will slowly drift up to around -9V/-15V or so over time) - I removed R65, R66, R69, R70, they measured fine, rest of the resistors in the string also measures fine. - I removed R88, R90, they measured fine, then R89 and R91 in circuit also measured fine. - Probing around offset servo referencing -560V supply, Q33 still in very weird state, I then removed Q33, it measured fine on transistor tester. All these parts I put new ones in anyway. I then tried running the amp with R88, R90 disconnected, this should take the balance servo out of the loop, well output measures around +30V/-40V.. So this is not really a working channel, the seemingly low offset is just because the original offsets are quite high but center around 0, and the balance servo is already working pretty hard to bring them together. And without balance servo the situation of Q33 is the same. C/D might be ok - they likes to stay at same potential and is usually around the middle point between 1/10 of +/- offset when servo is connected back. Moving on trying to find out the cause of the imbalance - voltage at the gate of Q24 and Q25 are about 5-6V apart, thus probably so are the voltages at cathode of the output tubes. Then here is another weird thing: one of the J79 has point something volts of positive Vgs. Then going to the plate load CCS of EL34, since this should have major impact on cathode voltage of the output tubes. R17/R27 measures fine, with around 3.2V drop across them, this checks out with 3.6V drop of D4, D5/D8, D9 and 0.45V Vbe of Q6/Q11. R18-20/R28-R30 however on one side all measured around 2.37V-2.4V, on the other side all measure around 2.2V. So there are 2 problems - first being they are not balanced, second being the voltage drop doesn't quite seem right - when I directly measure Vbe of Q8-Q10/Q13-Q15, they are all around 0.57V, which is fine. D6,D7/D8,D9 should have about 3.6V across them, so voltage across all these resistors should be around 3V instead? I then try measure voltage across D8 and D7, D8 measures fine with around 1.8V drop, D7 however measures only 0.7V... I don't think this is right as D7 is visually as bright as D8, also this should instead gives around 2V drop across all the 820ohm resistors, one more mystery.. For the imbalance of current I thought of thermal again, as one set of the CCS is closer to EL34 than the other. Had a friend came over help me elevating almost all resistors, then I re-soldered every joint on the component side of the board due to fear of cold joint after pulling - Power on again and nothing changes. So now it sums up to: 1. Balance servo might be ok or at least I will leave it alone for now 2. Weirdness around Q33 and I don't know what it's doing. 3. Positive Vgs on one J79 4. Things in plate load CCS measures weird and off. I don't know if I have a single problem or multiple, and not quite sure among them which is the cause and which is the effect. To my understanding, the majority of the current carried by plate CCS - thru R17-R20, this should be independent of any feedback stuff in rest of the amp? If that's the case I should try to fix this first, but I don't know what to do about it, all the resistor values measures fine when amp is off. And also an unrelated question - what is the factor in the circuit that prevents negative battery voltage from changing? When trying to adjust the trimmer closer to 740V, up to some point apparently negative voltage likes to stay where they are around -540V while positive voltage can still be trimmed down. (I also tried trimming battery voltages in opposite direction, up to 10V difference, this doesn't seem to affect output imbalance and they still likes to stay around 5V apart.)
  8. Thank you! I will take them out and check! This reminded me that, depending on how I set the trimmers, when turning the amp on with a bunch of things not conducting, O+ and O- can be as high as the positive 500V rail voltage, that certainly exceeds the 350V rating of R88, R90, R65, R66, R69, R70 I used, also voltage across C12/C13 can be a little above their 1kv rating. So I shall probably just replace them all anyway.
  9. I went back to right channel, wouldn't really call it working since balance is still a whack where offsets are a few volts apart. Battery JFET CCS is stable, no drift when trimming and I went back measured a couple times they sit at 6.55V fine. I thought because offset is low enough if I can figure out the balance stuff then I get a working channel at least from DC standpoint and I can use it as reference, turns out lotta things around output still measure weird. For left channel, output offset are both around -40V, point C and D both measures -3.5V. For right channel, output offset are about -4 and -10V, point C and D both measures -0.65V. So in both channels C and D are not exactly 1/10 of output offset and C and D seems always have the same voltage. Q37-Q40 are probably working fine, as current through R84 - R87 matches what the current mirror should do. For this, in both channels Vbe across Q32 is about 0.6V, which seems fine, Vbe across Q33 is however about -0.2V... that explains when I measure them together total voltage drop is about 0.4V. I don't know how much of these are affected by my 10Mohm meter impedance, I will try to see if I can borrow a high impedance volt meter to use. So far this has been pretty confusing.. But thanks again for your patience and explanation! Yeah overall the use of Q23 seems pretty marginal on paper, according to the datasheet it also seems it's operating on the boundary of SOA at 740V Vce with a few mA of Ic.
  10. Sorry that was a typo, I mean change RV1 after setting 6.55V across R42, fixed. ๐Ÿ™‚ I'm pretty ignorant would you mind explain how do you derive the theoretical value of 786V battery voltage in this setting? might help me understand how the trimmers actually affect the voltages. Also I assume the 6.55V reference voltage came from it being the center value of the range of which RV1 can set the voltage at base of Q17 (6-7V) when battery voltage is 740V, is that true? Thanks!
  11. Just tried parallel up R39 with 300K resistors, this should reduce the resistance to 51k and gives slightly more current available to drive the base of Q23, nothing has changed, so probably hfe is not my problem here.
  12. Replaced K246s in the right channel with a couple that I previously breadboard tested as CCS, everything is still the same. tbh I don't exactly know if I have problem with the CCS on it's own, right after trimming when I measure voltage across R42, it could be my meter throwing things off and some LED somewhere when off and voltage becomes around 6V, if I leave it alone for a bit came back and measure the voltage is back at around 6.5V, only slightly lower than 6.55V. The ones I took out also doesn't measure like anything suspicious on my transistor tester. I measured connectivity of things around Q32 and Q33 from both sides of the board, nothing suspicious. I also directly measured voltage drop from the base of Q33 to the emitter of Q32, it gives me 0.38V, would this make even a tiny bit of sense considering say the rise in temperature combine with the fact that the amount of base current going into Q33 is tiny? I feel there could be a relation between offset and my negative battery voltage? So I don't know if this is normal but when I set RV1 to R32 side, and trim RV2 to set voltage across R42 to 6.55V, my negative battery voltage will be reduced from -561V down to around -541V, positive voltages are somewhere above 240V. Then when I trim RV2 RV1, as positive voltage decrease, negative voltage likes to slowly kick back for a little bit. This happens in both channels, when positive voltage reaches 200V, right channel (lower output offset) kickbacks to -543V, and left channel (-40V offset) kickbacks to -547V. I don't know what's causing this, but as a consequence my negative voltage will never go below -540V in any case. (before replacing LEDs both will go back to -548V) Lastly I'm considering the hfe of Q23, as it was discussed in previous thread that we'd want highest hfe possible. Measuring them with my toy tester, at Ic = ~6mA, right channel has hfe of 44/47, left channel has hfe of 40/44. I have a small stack of back up 3675s, the highest I can get are 1 at 53 and 4 at 50, all the rest are between 40-50. I can try replace them but I'm unsure if that is related to my issue, as I currently have 4.8mA going through Q30 and Q31 which doesn't seem far off from 5mA on the surface and am also unsure if increase hfe by this small margin would help.
  13. Thanks, that makes sense! Strangely when I diode test BE junction on Q32 and Q33, they measured fine. Hook them up to transistor tester on board they also got recognized as NPN fine with correct hfe. Power on and measure again voltage drop across R73 and R64 are still the same with V(R73) only maybe 20mV higher. But voltage drop across R64 seems correct on its own with about 20mA running through Q32 so both devices should be on... I also measured actual Vgs for Q4 and Q5 both below 1V so not gonna worry about them now. I been also thinking about the diodes, since previously all my LEDs are de-rated by quite a bit (likely thermally) and replacing them made things better. Are D2 and D3 for protection purpose only and not meant to be driven into reverse breakdown under normal circumstances? least measuring voltage across them are near zero, so derating should not matter here? Tomorrow I'm finally gonna replace the K246s, since there is still something finicky about setting voltage across R42..
  14. Definitely regret saying the famous last words, all LEDs replaced in left channel and I'm still getting -30V ~ -40V of offset at output, and balance servo is not kicking in neither as +/- output likes to stay a few volts apart. I measured as much node as I can think of, after setting battery to -547V/200V. EL34 plate load CCS measures fine, current going thru R17-R20 adds up to 10mA Output tube voltages measures ok on one side, the other side measures worse with -453V at cathode and -457V at the gate of the 2sj79. Current going through Q26 and Q52 seems fine. input stage CCS voltage I'm not sure if I'm measuring the 207V correctly with my handheld meter with 10Mohm impedance, if this is indeed 207V then Q4 and Q5 are seeing a Vgs of about 7V? And lastly the bottom part point C and D are both around -3.5V. I also tried to measure a 2sa1413 with very low collector current on a scratchy breadboard circuit, at Ic = ~46uA hfe is somewhere above 90, which should be fine. Other than that I have set the battery voltage to -547V/200V, leaving the trimmers alone and turn off the amp, turning on again when high voltage kicks in for an initial period battery voltage likes to be -500V/245V, then all of a sudden it becomes -547V/200V again and stayed... Also isn't the current going through Q30 and Q31 coming from the battery via Q23? In that case I'd expect a few volts drop across R99 and R100? Which when I measured there isn't any. Is there anything looks obviously suspicious from my measurements? Would really appreciate any input, thanks!
  15. @simmconn Sorry about the confusion, for LED what I meant is, Vf = 1.8V @ If = 0.5mA is straight calculated from the schematic if I look at D1 and D23 and Kevin's notation on current through Q1, Q2, Q3, Q28, Q29, assuming base current are negligible and all Vbe are 0.6V, hence this is a theoretical reference value. Vf = 1.725 - 1.750V @ If = 0.5mA is when I measure LEDs on the breadboard, both from a new batch and from D1, D23 and another one in battery taken off of the board, as plotted, a little lower than the aforementioned ref value. Vf = 1.57 - 1.6xV is measured from D1, D23 on board when amp is powered on, I didn't measure individual ones from the string in battery but 10.7V drop across 7 of them falls in the same end - all are considerably lower than even the values measured from breadboarding. This made me assume all my on board LED Vf are degraded by a lot, hence so are the current through all the transistors referencing them, results in operating point of output tube to be off and hence the large offset. I just done replacing all the LEDs with the 3RDL-S and made sure they are elevated from the board in the right channel. Power on I can now relatively comfortably set voltage across R42 to around 6.55V. When setting battery voltages to -544V/200V and -543V/200V, output offset from +/- are around -3V/-8V. Checking voltage across D1 and it's now 1.79V. Now -3V/-8V offset values by themselves I believe is within sanity, the difference however made me wonder if there is something wrong with my balance servo.. But before worrying about that I'm gonna replace the LEDs in the left channel first. (I have not taken out the on board k246 for a measure yet just because I'm confident they are genuine so only after trying to fix the LEDs) Again many many thanks for your help and detailed explanation! I definitely started off not understanding a single thing about how this works but with your guidance I'm able to troubleshoot simple things little by little ๐Ÿ™‚ For the first time in a long time working on this amp that I have a good feeling of getting pretty close now (famous last word), ofc still some distance to go.
  16. I went on and measured more things outside battery while setting battery voltage to 750V: Seems like everything is a little off. I suppose the key to get correct low offset at output is to get the output tube operating point in place (without considering the feedback shenanigans that I don't quite understand๐Ÿ˜ต), with grid sitting on the -500V supply it should come down to getting the plate load CCS and cathode voltage in place. Other than the previously mentioned battery adjustment weirdness, measuring voltage across D1 and D23 I got 1.57V-1.67V, definitely on the low side similar to what I found in the battery. Also battery seems to be not outputing any current with little to no voltage drop across R99 and R100. LEDs With current noted on the schematic, at least for D1 and D23 they be running at about 0.5mA with a forward voltage drop of 1.8V (if all base current can be ignored). I took D1 and D23 and another one from the battery for a measure. LTL4213_old is taken off from the board, only plotting one as they all measured similarly, LTL4213_new is another batch I bought recently, along with the Cree C566D also a 2.1V red LED with nominal forward current of 20mA. These all measured pretty much the same, on the lower side of 1.8V but strangely never went down as low as around 1.6V as I measured in amp when operating. Only thing I can think of is derating due to the negative temperature coefficient? 3RDL-S is a 2.1V red LED with nominal forward current of 2mA, as expected the curve shifted to the right side. Seems to be a shot in the dark but I shall replace all LEDs with 3RDL-S and sufficiently elevate them from the board? (which I didn't do with the current LTL4213s on board..) 2SK246 I didn't take out the ones on board but gotten a few backup ones from ebay, hook them up on the breadboard and set voltage across 22K resistor to around 6.55V at 12V supply voltage from the bench. Then sweeping supply voltage nothing surprising and they regulates fine... Resistors I replaced R33, R34, R43, R44 with 500V rated resistors - as I just realized voltage across them will always be a little above 350V, and this doesn't change anything. So I think I'm going to replace all the LEDs this weekend, although it seems everyone else are using LTL4213 just fine, possibly still not the cause of my problem...
  17. This is really helpful, thank you so much! I'm way too ignorant on this so I wanna verify if I understand things correctly. More probing around the battery with Vgs on 2SK246 now, "/" is before and after trimming RV1 to bring down battery voltage to around 740V. So in theory assume base current of Q16 is negligible, while the regulation on the LED string is not ideal, Q22 shall work as a current source fine despite the Vds drop from 4V to ~2.65V, that means regardless the LED is not the source of my problem right? Now the current running thru R42 before and after are 298uA/284uA, current thru RV2 before and after are 303uA/287uA, while not super accurate but it seems the base current of Q16 is on order of magnitude of a few uA and had changed about 2uA, this can cause Vgs of Q22 to change about some tens of mV depending on RV2 value, enough to say that whatever Q16 is doing here, it's effect on the CCS is indeed negligible? That lead to the conclusion the first thing I should do is replace the 2sk246 right? And breadboard test them with a bench supply varying input voltage around 10-12V with R42 and RV2 in place. Is there a recommendation on readily available replacement of this part? Thanks again! ๐Ÿ™‚
  18. Many thanks @simmconn! Things I know so far: PSU voltages sits still as I tried to trim the batteries closer to 740V, none is loaded down. Also unlikely that I'm running on the boundary of current limiting because I trimmed one channel at a time while leaving another channel with batteries not conducting? monitoring total power consumption it is 30-40W lower than bringing both channels to be on. All resistors seems fine, no visible damage and ones below 100kohm I can measure on board and values are correct. Measured two C3675s in same channel on board with toy tester, hfe are both around 48 When I set RV1 to R32 side, I can comfortably set voltage across R42 to 6.55V, this is what I get: Then I leave RV2 alone and turn RV1 to reduce positive voltage to ~200V, I got: I remember reading somewhere that LEDs need to be 2.1V, mine with handheld meter all measured around 1.7V, but I guess around below 1ma bias current in circuit this sounds about right? I unfortunately don't have any spare k246, best I can do is pop one out and find replacement if they are malfunctioning. Gonna measure more nodes over the wknd
  19. Mine from Bdent looks like this. If anything my batch of 1413 are labelled as "K" grade and my batch of 1627 are labelled as "L" grade, on my toy transistor tester from Amazon the hfe measures are 100 and 60 respectively which kinda checks out. From my mouser list all resistors are rated 350V, 0.5 watt, most are Xicon except a few. I will double check on board resistor values from color code tonight though. ๐Ÿ™‚
  20. Q16-Q19 are 2sa1413, I also have a bunch of 2sa1627 as well, least both of their datasheet looks like a straight copy and paste from A1486 datasheet. Q23 is 2sc3675. I believe they are all genuine from reliable sources.
  21. ๐Ÿ˜‘Right channel is exactly the same story - negative voltage likes to bounce back from - 544V to -549V and stay when trimming RV2, with RV1 bringing positive voltage down output offset just keep increasing, and is -125V/-151V near -549V/200V I found one thing though - if I start off setting RV1 wiper to the upper R32 (9.1k) side, trimming RV2 I can easily set 6.55V across the R42 independently and they stay there, until I trim RV1 to bring down the positive voltage, they drift to something else between 6.2V-6.6V. This happens to all 4 batteries.
  22. thx, I will do that in the next few days (I remember last time I tried the behavior was similar). When I do that should I leave left channel in this current state?
  23. Thanks for the help! Yes I only used RV1 to adjust positive battery voltage after trying to set voltage across R42. And yes I can set overall battery voltages to 740V too: because the negative voltage is just staying there at -549V I just keep further trimming RV1 to bring down another 10V on positive voltage, that gives both battery voltage -549V/191V. Now output offset just increase to L- = -123V and L+ = -150V, measure R42 again it's 6.35V and 6.6V
  24. I just gave my best shot on left channel, don't know if it's normal at all but setting voltage across R42 isn't easy, when either voltage is below 7V trimming the pot don't make them go down continuously and they like to jump down a few hundred mV and hanging around there. Also trimming either pots affect both voltages across R42s and in general seems like they like to stay on different side of 6.55V. So the best I got is I got both stay between 6.5-6.6V, this is when battery voltages are๏ผš 216V/-542V and 231V/-547V Starting from there I use RV1 to bring down the positive voltage and it ended up at: 200V/-548V and 200V/-549V Remeasuring voltage across R42 they be what they wanted to be: 6.3V/6.7V Output offset is L+ = -103V and L- = -135V Also there is likely some oscillation going on in this state as I can hear the output tube squeaking. is setting voltage across R42 not suppose to be this hard and trimming RV2 can continuously bring it from 7V down to 6.55V and they will comfortably stay around there?
  25. Thanks! So to elaborate none of the things I measured are bouncing themselves all over the place when leaving the pots alone. Output DC offset simply follows battery voltage and goes + -> 0 -> - as I trim the pot but only likes -543V/245V with the lowest offset near 0. As for voltages across R42 - they start off well above 7V when battery is not conducting. Trimming RV2 will continuously reduce this voltage, until I get to around 6.8-7V ( which is also the -543V/245V; 0 offset point) then there seems to be a discontinuity where further trimming the pot will make it jump to around 6.3V. Only back and forth trimming both batteries I can sort of get 6.4V and 6.6V, it almost seems like V R42 really doesn't want to go to 6.55V specifically. Oscillation yes definitely happens as turning the pot makes all kinds of squeaking noise coming in and out from the tubes. I'm assuming batteries from the 2 channels are independent as they only share the PSU, so I only tried to bring batteries in one channel to conduction at a time. I guess another way to look at it is, is it normal that when both batteries are conducting, I can never trim the negative voltage below -543V? Passing that point it just likes to go back up towards -549V and stay there.
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use.