Craig Sawyers

Looks great Kevin - it is a bugger to work being so tough. But real Lignum Vitae has a wonderful perfume, doesn't it? Really looking forward to seeing the final thing. You definitely need pilot holes for the screws though, or drill and tap - just as you would for metal. As you know I tried 1/4UNC and M6 with perfect results.

That is my stock phrase to team members when things aren't happening quick enough: "C'mon guys - this isn't rocket science. Wait a minuite....". This instrument is a real pushing-the-envelope thing. There are two X-ray imaging spectrometers (to look at the chemical makeup of Mercury's crust) mounted on a 1 meter long optical bench. There is also an electronics box with a multi-output switched mode supply and a data processing unit on a 14-layer circuit board. Total weight is 8kg (17lbs). It has to be seriously radiation hard too. In order to do the science, the sun has to be spitting out x-rays, which only happen when the sun is in an active phase - when it also throws out massive amounts of radiation. One consequence is that FPGA chips, which are normally around $10, cost $19,000 each in the space qualified grade necessary.

The single large ground pin is specced at #16

Deliver the units to on time and to cost target. International team of around 80 involved and a budget of Euro14m. Stuctural and Thermal model delivered on schedule, Electrical Model due for delivery end November, Qualification Model due around Spring 2011, and Flight Model Jan 2012. Overall buck stops with me.

Due to start on Monday in a half-time employed role at Leicester University. I was actually doing the same thing for 2 1/2 years as an external consultant, but they could not cope with even longer as an external bod. So - and heaven knows why on earth I agreed to this - I said I'd take a huge pay cut to continue doing the job as an employee - and (UK) academic salaries suck big time. I'm not really salary-constrained in a major way, and I kind of have a good rapport with the team, and a sense of responsibility to see the thing through - and that swayed the decision. This is as Project Manager for a deep space instrument called MIXS (the Mercury Imaging X-ray Spectrometer) due for launch by ESA in late 2014. NASA have the second only spacecraft (Messenger) visiting Mercury about now; it is not in Mercury orbit yet, but doing fly-bys as it slows down enough to get into orbit. Ours will be the third mission (BepiColumbo), getting into polar orbit around the planet in 2020. All sorts of planetary gymnastics on its 6 year travel to catch up with Mercury - as has been the case with Messenger.

Oof.

The amazing thing is that the active, signal-route part of the circuit, is four tubes and four FETs. Everything else is either current sources or voltage sources. Plus the servo, of course.

Bored and twiddling my thumbs waiting for semiconductors to arrive to fix the zapped T2, I thought I'd try to get a grip on what the actual signal level circuit was doing. I've replaced all the current sources and 740V battery with ideal symbols. I've also left out compensation components, and some of the details of one or two current sources which I think are error handling, to prevent the thing blowing up if a tube goes down badly. I think the result shows the rather neat symmetry of the design, and the clever way in which DC voltage shifts are generated using the FETs and batteries to move the ground referenced input signal to -500V. File is here http://www.tech-enterprise.com/tekstuff/T2deconstruct.pdf

Interesting. I've just measured where my ESL57's are - 56" from the rear wall and a ceiling height of 90" - so a ratio of 0.62! I seem to have iterated to GC's golden ratio just based on moving them around and listening.

George Cardas is one of the audio business good guys. Fixated on golden ratios for sure, but that fixation has led to some exceptional results. His golden trapagon stuff is aimed at distributing the room resonances at the listening position in an acoustically acceptible way. But it seems to be suited to monopole speakers (ie ones in boxes) that are omnidirectional at low frequencies, and excite resonances in all three room dimensions. I'd be interested in learning what his thinking is on placement of dipole speakers (such as electrostatics, or Magneplanar panels) which excite resonances only in one room dimension.

I guess you were lucky you were only dressed as a wolfman - could've been a whole lot worse.

Reminds me of someone I met at a strange party long ago.

I vaguely recall slowing down some of the scenes to see the detail of what happened to the crew. Not pleasant. But you're right, a load of tosh.

Hokay - here's where I've got to. No progress on zapped channel. Waiting semiconductors. But some progress on the one with stuck battery voltage. Symptom was two LED's in one of the current source triples would blow. They would flash very brightly and then blow. One of the 3675's in that triple was open circuit too. So I spent a cheerless hour or two pulling semiconductors. Everything checked out fine - all MOSFET's absolutely fine, all junction transistors likewise fine (apart from that one 3675). Then I spotted it - it was either Q32 or Q33 - I had missed soldering a pin! Boy did I feel a dope - I must have been interrupted half way through soldering and skipped a joint. Anyway, whereas previously both batteries were stuck, but at significantly different voltages and LED's died, they are still stuck, but within a volt of each other at 570V. LED's are all OK (checked with a 9V battery in series with 6k8 resistor) so LED death is cured. Current triples and bottom current source are not turned on, but the battery LED chains are lit. So no anode current in the EL34's. About the only semiconductor I haven't checked are the JFETS in the batteries. If they were phut, that would explain lack of adjustablilty using RV2. Sounds like the next check. All other semis in the batteries are fine - checked in circuit just using diode test on the DVM.

I'd agree. Given Inu's comments on the fact that using old stock Amperex 6DJ8's killing the 2SK216's no matter what he did, I think that that one messes with this design with real caution.

Well, the two K216 left over from the build produce traces identical to those in the Hitachi data sheet - to the extent you could pretty much substitute the curves. My Tek 177 curve tracer won't generate enough voltage to trigger the gate protection zeners. But the drain-source diode seems to be present just using the diode function on the DVM. The thing I don't understand is the 2N3675 in one of the active batteries (the one connected to Q4) has suffered second breakdown - it is essentially short circuit between all three terminals. Q27 and Q28 (2SK316) are dead, as are D21 and D22 (100V zeners). Three dead LED's in the current source triples (D10, 11 and 7). So the sequence of events is a little puzzling, and I cannot think of a mechanism that would lead to dead LED's - other than infant mortality. Everything connected to them in the current sources is A-OK.

That is a dark and worrying though, Justin. I've got a couple of spares (they have the Hitachi logo, but that counts for nothing) that I'll throw at the curvetracer and see what they look like.

I do beg your pardon - you are quite right. I hadn't spotted the thermals that make the connection.

I've been peering at the photos of the original T2, and the only difference I can find after a couple of hours is that pin 9 (the shield between the two halves of the triode) is connected to ground in the original, but not in the clone. Also Stax use a grounded shield tube around each 6DJ8. I've also been searching the web for the significance of connecting pin 9 to ground, or not. Most audio applications ground it, but it is uncertain whether this has a grounding in theory or not. The changes in interelectrode capacitance seem trivial from the datasheet. But I also read that the 6DJ8 was originally introduced as a TV RF stage, and that with its high mu if it can burst into self oscillation it will. Most folks seem to add a 1k grid stopper to quite the things down. So maybe pin 9 to ground and a shield are important in taming the 6DJ8, particularly in an ultrahigh bandwidth cascode/constant current arrangement. Plus perhaps a grid stopper tight up to the pin, although that would need a track cut, and wasn't necessary in the orignal.

Inu, what adjustment protocol did you use? ie where to preset RV1 and RV2, and which order in which to adjust, whether they need to be adjusted simultaneously or not, etc. And how many meters, and connected where? Apart from KG, I seem to be number 2 to get to this stage (you were first), and we need to work out a foolproof method of setting this beast up to prevent auto-destruct setting of the battery pots, or there will be a further 17 learning curves down the line, and buckets of dead silicon.

That is very useful information Inu. I am using old stock Mullard ECC88/6DJ8, probably 1960's vintage. They measure fine on a tube tester, but maybe the T2 is sensitive regarding input tubes. As Kevin mentioned before, adjust Active Battery is tricky. While adjusting the active batteries, you will better to monitoring voltage between +/-Out to GND or +Out / -Out at same time. Not necessary to adjust it to exact 740V. My unit is balanced at 749 – 750V. If adjusted point of the RV1(2K) is out of the safe area, loud buzzing noise comes from the ear speakers. It sounds like oscillating and very loud. Turn both RV1 quickly to find the point that stop the loud noise.

Wierd is right. The channel that is nominally working still has a pair of dead LED's in one of the current triples. The more I think about it, it just has to be a tube issue. Everything that died on the deceased channel was associated with one of the EL34's. I'll do some more serious checks on those tubes and see if I can find out if that is the problem, although I can only get to 400V anode-cathode voltage on my tester (AVO CT160). JFET's are GR Idss band, which is what was called for. But even if the Idss band was an issue, it wouldn't explain mass silicon suicide. Anyway, I've just ordered a bucket full of spare semiconductors. Enough to fix the sucker, and plenty left over (which of course I'll now never need). After all, what is another hundred quid on this beast?

I'd like to think so. But I don't think that is the case - the only dead LED's that I have found are the ones in the current triples. I'm wondering whether it is something to do with the EL34's I'm using. Not minded to put in new tubes for the first run, I used a serviceable old pair - just running one channel at a time. Now those tubes measure pretty much OK - but they could be just off enough, or have some odd electrode short that doesn't show up on test, to trigger a problem. I'll fix the unit up, replacing the LED's as a matter of policy, and use new EL34's. It'll be a few days since I don't have spares for the FETs. Not the best of days....

Well this is interesting. The channel that self oscillated itself to death has for certain dead: Two dead 2SK216 (the ones in the cathode circuit) One 2SJ19 Two 100V zeners One 2SC3675 - the one in one of the 740V batteries Several dead LED's associated with the current sources. The last is interesting, because in one case only *one* of two series connected diodes is dead. This was established by using a 9V battery and 5.6k series resistor. In the channel with stuck battery voltage there is also at least a pair of dead LED's in one of the anode current sources (the pair associated with the current triple). Which would explain completely why I could not adjust the battery voltages. There are no other dead semiconductors in that channel (at least that I have thus found). All the power supply voltages are correct.

You know - that just happened to me. Uncertain what caused it in my case, but was merrily adjusting the active battery, both test points at very close to 740V, then a sound of oscillation (whistling noise). Never a good sign - then crack - certainly 2SK216 are fried. The other channel refused to adjust. Stuck at about 560V no matter what I do (-540, +20 WRT 0V).