Craig Sawyers

A really wierd Family Guy in which Quagmire verbally lays into Brian. Really uncomfortable watching from start to finish.

Around a long time ago, I flew into Philli domestic, which was like a regular shopping mall - really nice. I then transferred to International, which in those days was the most depressing building on God's green earth. Steps leading up to it was populated by hookers and dealers. In the hangar-like building there was something called the Oasis Bar. This was a plastic palm tree screwed to the wall, and two nailed down tables with fixed chairs. Beside it was a booze vending machine. I swear. Then they cancelled my flight - so I had to get back to domestic and catch a flight to Boston, and thence back to the UK. My first and last experience of Philli International.

That is truly awful. Like you I'd like to kick nine bells out of people like that. A dose of their own medicine meted out by a rugby front row would make them think twice about pulling a similar stunt. When my son was at University, he went through a goth phase - which meant that he was a target when out for a few beers. On the way back one night three early 20's guys (who should have known better) stood in his path and blocked it. After Rob - at 6'6" and a rock climber - advised them he'd like to move peacably home, and getting the expected "what are you going to do about is fucker", he said "OK, I'm going to count to three, and then your friends will have to take you to hospital" - One, two (missed out three) and then nutted the guy. Went down like a pack of cards with broken nose. "Anyone next?" says my son as he steps over their fallen comrade. No takers...they headed off at some speed leaving their friend where he lay. But three brave lads taking on a woman - they deserve no such considerate treatment.

Ah - those are interesting questions. Because there is a little slop in the holes in the heatsinks, I loosely attached the heatsink bracket to the circuit board with a couple of transistors with washers etc. I then assembled the case with the brackets/pcb in place, and temporarily screwed the brackets to the heatsinks and snugged up the screws. I then tightened the transistors down to lock the bracket to the PCB in the correct place. Then off with one end panel and slide the PCB and attached brackets out. Now install all the transistors, tighten down and solder. You can then take the heatsinks off the case, smear them with heatsink grease, and bolt the brackets on and tighten down. You should be able to assemble the case without difficulty after this. And yes - it was a bit of a tricky process - but keep at it with two temporarily attached transistors per heatsink, until you are happy that there is no significant gaps between the brackets and heatsinks. On the umbilicals, I twisted the heater wires and put them in a separate sleeve (I just happened to have some high temperature glass braid of the right diameter). Twisted so that the AC current would not couple to sensitive things like the bias, and in a sleeve because they are connected to -500V; possibly overkill, but I had the sleeving anyway, so what the hey.

The 10m90s is a three terminal current regulator, used here as a protective current limiting device in series with each power supply rail. The parts you are suggesting as replacements are rectifier diodes. There is no substitute to the 10m90s.

Like I said - Digikey is showing zero stock.

That thought was going around in my mind, as a discrete option (but would be messy to implement). I had no idea that such high voltage single devices were available. The gotcha is that I couldn't find stock anywhere. Digikey list them at around $6 each, but have zero stock. Mouser also list them, zero stock with a 20 week lead time with MOQ of 605 pieces.

Heh! When my son was applying for university eight or nine years ago, he needed to fill in my occupation. "Intellectual prostitute" I shot back. "what!!!", "Yeah - intellectual prostitute. I can do any job you want me to, in return for money". I let him off the hook - eventually.

That sir looks like an interesting transistor. Do you have a source?

Well, that is interesting. Because I had a model already part developed for the battery, I thought I'd press on. The precise transistor type for Q16 to Q20 is not so important, but the current drawn into the base of Q23 is. Now I do not have sensible models for the actual transistors, I simulated the base current into Q23 by using a much higher hFE part, and connecting a resistor from base to emitter to pull more current. The behaviour of RV2 is to set the constant current through fet Q22, so I just used a constant current source there instead. Likewise I have used a 5mA constant current source from the bottom of the battery to -560V (arbitrary ground in the simulation) and a 10k resistor from the top of the battery to +810V (= 250 + 560) to simulate the load of Q5 (K216). Decreasing the current through Q22 increases the base current into Q23 and reduces the battery voltage. It has the effect of turning Q16 off and shunting current into Q17 and hence into the base of Q23. Likewise turning RV1 in the direction of R33 increases the current though Q17, and hence also reduces battery voltage. This seems to work fine, and gives a voltage of between 14 and 20V above -560V when the battery voltage is 740V. For any setting of RV1, the position of RV2 shifts to maintain the same operating point voltages. This works fine with RV1 throughout its travel provided that the current into the base of Q23 is not too high, ie 50 - 70uA. But once the base current of Q23 becomes big enough, approximately more than 100uA, it becomes impossible to turn it on. That corresponds to the specified hFE of a 3675. Now Kevin measures 50uA base current, which implies that the real-world transistors used in this position are better than the data sheet implies by a factor of two. The main point of all this calculation and modelling is because I *know* that my batch of 3675's are marginal for Q23. Swapping to the 4686 should be fine, but the hFE is lower than the 3675 by a datasheet factor of 2 - so I have to find more base current. The model suggests that decreasing R39 from 62k to 47k should have no other effect than increase the current drive to Q23. So what is the purpose of RV1? It looks as if it does the same job as RV2. I think (?) that its purpose is to give more adjustability in the case of (a) high Idss of Q22, ( high Vgs(off) of Q22 and © low hfe of Q23. Thoughts anyone? I'd like to examine all angles of this part substitution before firing the beast back up.

That is a sound idea - interview practice makes it easier when you go for something you really want to do.

Neat! Megavolt has a physics doctorate from Berkeley, but found that this sort of scary activity was his true calling - watch the video here Dr. MegaVolt and his Amazing Tesla Coil . Check out him cooking a sausage in the discharge from his 8 1/2 foot high machine - it uses a 15kW generator to power it. He usually has someone standing by with a fire extinguisher because his faraday suit sometimes catches fire. The two person performance on the vid is him and his wife! Thank heavens for nutters like this.

Is that Dr Megavolt? Dr. MegaVolt: Man Meets Lightning

Nice! Here's one I prepared earlier (2002) http://www.tech-enterprise.com/tekstuff/DSCN0622.JPG

Hope it was nothing exotic: read ultra-expensive.

WOOF!

I think I've been the only one with 2SC3675 troubles - and that is restricted to the one in the voltage source batteries. Formally that part is sailing very close to the wind - the spec sheet says that at 5mA collector current second breakdown happens for Vce greater than 600V. Depending on the LED in the current source Q30 the current is closer to 4mA or less. At 3mA it should just about do 750V. Now spec sheets are always conservative, but the statistical spread of characteristics must have put my batch just about where the spec sheet says. They went short in both channels, which is classic second breakdown failure. This happened spontaneously once in my class A Krell - that generated a hell of a lot of smoke! Failure of the 3675 took out two LED's in the anode current sources, a pair of 100V zeners (they glowed briefly!), and all four 2SK216's. That it happened on both channels identically indicates that it is not a stuffing issue (I checked that first), dry joint, or probing mistake - which points the finger at the 2SC3675. So once I get the hot stick out, I'm going non-authentic 2SC4686A (KG's suggestion). They should be good for 1200V at up to 8mA. We'll see if that cures the problem.

Well, I guess you could go for high reliability versions, such as JANTS space grade. But at $50 for a 1N4148 diode you'd better buckle up for remortgaging the house. And they are ceramic bodied surface mount, so they wouldn't fit anyway. Just spent well over a hundred grand (UK pounds) on the components for six circuit boards about A5 size, and only three of those have space grade components. The real killers were the ACTEL FPGA's, which were a cool $19k per chip (three off needed), and A/D convertors at $6k each (15 off needed). So I'd leave well alone with the semiconductors. Lots of games to be played with passives to tune the sound quality.

Yes please. I see you're using Linear's LTSpiceIV. I downloaded that wonderful freeby quite a while ago, but have not used it in anger yet. This sounds like the perfect excuse to do so. Where did you get the models for the transistors?

Don't you pity pro tour cyclists who spend 6 hours a day in the saddle. They suffer from something called "saddle boils", which are open weeping sores around the crotch area. They lather that particular area with a mixture of nappy cream and vaseline before setting off each day to fend off the problem as best they can.

Thanks Kevin. Kind of confirms my ball park numbers. I guess if I were designing this from scratch I'd put quite a bit more current down the long tailed pair to ensure enough margin (power dissipation in the 2SA1486's is not a problem), or make the output transistor a darlington to reduce base current.

OK, having been listening to the BH for a month or so, I'm girding my loins for the repair of the T2. Since the death mechanism for me seems to be second breakdown in Q23 (the 2SC3675 in the battery) I was trying to get to grips with the design of the battery. Now no-one else has had problems with Q23 in this location, so it is probably a batch issue - the batch that I bought was just too marginal on SOA. I realise however that I have an imperfect understanding of the operation. I'll go through what I've worked out, so anyone can chime in. First consider Q16/17, which is a long tailed pair. Nominally the base voltages should be very close to each other. The base of Q16 is fed by constant current source Q22 and resistor R42. The base of Q17 is fed via a resistor chain with RV1 in the middle. Looking at the resistor values, with RV1 set at the R32 end, Q17 base voltage (WRT the top rail) is 6.2V. Similarly, with RV2 set to mid track, the equation for an FET constant current source with Idss about 6mA is a drain current of 0.28mA. Flowing through 22k resistor R42 this produces 6.2V at the base of Q16 - precisely the same as that at Q17. So all well so far. The current through R39 is (6.2V - Vbe)/62k = 5.6V/62k = 90uA, which is shared between Q16 and Q17 - so 45uA through each transistor. Q18 and Q19 are simply voltage sharing - the bases are connected to the halfway point of the battery, so Q16/17 and Q18/19 only see half the battery voltage - important since these are 600V parts. Q20 is a perfectly standard Widlar current mirror load, which boosts the gain of the long tailed pair Q16/17. Problem is that Q23, the 2SC3675 has an hFE of around 50. Its collector current is determined by 5mA constant current source Q30. So the base current of Q23 is 100uA. And here is nub - there does not appear to be enough current in the collectors of Q16/17 (45uA through each) to drive 100uA into the base of Q23. So what am I missing?

Yeah - cheap those resistors aren't. But very, very good. And the 5mm pitch means you need to lay out the board to suit.