|Pass Labs This forum is dedicated to Pass Labs discussion.|
Please consider donating to help us continue to serve you.
Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
||Thread Tools||Search this Thread|
|31st July 2011, 11:04 AM||#11|
Join Date: Feb 2011
Hello again. Thank you Analog S.A. for the 40mA, it's close to the 50mA I had it at. Thanks James, I'm doing it now....
Shorted Input Jacks
Using Oscilloscope (Tek 7603), since I have no milivolt meter...
Right Channel at speaker side of protection relay across Zobel Network. +11mV D.C, with about 2.5mV R.M.S. broadband A.C. noise with a bandwidth exceeding 100MHz superimposed.
Left Channel at output binding posts, +13.5mV D.C. Offset with about 800uV RMS broadband noise superimposed with a bandwith exceeding 100MHz. With Nacamichi off, broadband noise is still present indicating is is probably being generated in the Oscilloscope's Y amp or being picked up by the probe.
This model, the PA-7E II, is a different circuit from the PA-7E, there is no D.C. offset adjustment trimpot. The front end is a diffirential cascode stage consisting of a 2SK240 dual fet, (two thermally coupled fets in a aluminium can), to the drains of this going up to the emitters of two 2SC2240s in a grounded base arangment with their bases 10.2 volts (zener) above the sources of the 2SK240. The collectors of these 2SC2240s provide the differential drive to the bases of Class-A cascode voltage gain stage, a pair of 2SA970s with a pair of 2SA1370s beneath them in a similar, zener biased grounded base arrangment, another pair of 2SA970s sit above the first grounded base pair of 2SC2240s as a current mirror. A 2SC2705 sits beneath the sources of the 2SK240 as a 4.3 mA current sink and 5th 2SC970 above the bases of the grounded base 2SC2240s and zener to provide a 650uA current source. The current source and sink share an 82K resistor to bias up their diode stacks. The input, 50K down, 220 ohm 330pF R.F. LPF goes into one fet gate and the feedback network into the other. 50K from output shunted by 2k5 in series with 220uF 16v Nichicon "MUSE" bipolar electrolytic. This describes the front end of this unit.
After the time it has taken me to single finger type this, I measured the Right Channel, Thermally equlibriated bias across the 1 ohm emitter resistors of the output stage,....
Positive PNP side, measuring from the end most device inwards, 26mV, 29mV, 30mV, 31mV, 31mV, 31mV, 36mV.
Negative NPN side, from end inwards, 29mV, 27mV, 28mV, 29mV, 31mV, 30mV, 23mV. They don't seem to match up, but they varied by up to +-2mV even during the time it took to measure them. Even a cool draught on the heatsink seemed to make them fall slightly.
The number of 200 watt incandescent lamps (in parallel), in series with the amplifier supply was decreased from four to two, so the glow of the filaments was visible. The bias pot was touched about 5 degrees counter clockwise and the filament glow increased in intensity indicating that this direction was increasing the bias. After this adjustment the other two lamps were switched back in to decrease the source impedance of the supply and the thing left to equibrate for a further 15 minutes or so. Periodically the back of the forfinger was run along the row of output devices to check to see if any were running hotter or cooler. It was noticed that the four driver devices, emitters inwards rather than the collector inward output devices, but of the same type numbers, viz 2SC3856 and 2SA1492, which are positioned in the centre of the row of 18 transistors on the heatsink, were running noticeably warmer.
The second lot of measurements, oddly enough, were almost the same as the first, so the bias pot was touched another 3-5 degrees CCW again...
The oscolloscope was monitoring the output offset the whole time, no appreciable (>+-5mV), change was noticed.
Third Set of measurements..., PNPs 36, 39, 36, 37, 38, 37, 41 (264mA in total)
NPNs 39, 37, 38, 40, 42, 40, 35 (271mA in total)
Some of the devices are starting to creep up to and over the mark, so we will proceed gingerly......just a slight tweak of 1-2 degrees...
Fourth Set.............................., PNPs 41, 45, 44, 45, 46, 46, 51
NPNs 45, 43, 44, 46, 49, 47, 42
The transistor drawing 51mA was noticeably hotter than the others and the drivers were getting hot enough to warrant a second careful feel....
It has been decided to back the bias down a touch so the bias pot was "backed up" 1-2 degrees to where it was before.
Fifth set............................................... Not Much Change
Back it down further and wait....
What is really important when adjusting the bias of any amplifier with bipolar output transistors, is not the actual value of the measurements, but which way they drift with time,
because bipolar transistors possess a negative temperature coefficient. So with a constant base current, there should be a corresponding larger collector current and the relationship between them is called the beta, or current gain of the device. Most small transistors have a beta of around 100, but power devices, like these in this amp, it is lower, between about 30 and 50.
So for 1mA into the base of an NPN device, 30 to 50mA is pulled into the collector regardless of the collector voltage, (as long as the transistor is not saturated and its collector voltage is low).
So the collector voltage multiplied by the collector current is a good indication of the power dissipation of the device. In this particular unit, the rail voltage is +-75v, so that's roughly the collector voltage. At 40mA per device each is dissipating 75 x 0.04 watts or,3 watts, enough to make 'em quite hot if not firmly fastened to a heatsink. Now the problem is that the beta value increases as they get hotter, so for a constant base current, the collector current rises and this heats the device more and the collector current rises....and so on until,....poof something melts, usually part of a device die.
To counter this effect, a transistor, called a vbe multiplier, in the bias network, but is fastened to the heatsink, "feels" the heat and its collector current rises too, but as it does th reduces the bias to the output devices. In some designs this thermal sensing is not enough and a thermistor is placed in the base circuit of the vbe multiplier, this is attached to the heatsink too and increases the retarding of the bias when things get hot. The only drag in all of this is the thermal inertia of the heatsink which slows the communication of heat from a hot output device to the vbe multiplier and thermistor. So if an output device has really started to "run away", the vbe multiplier will not feel it soon enough to stop it, and, again, poof, another stuffed output device. It's sort of loke balancing a ball on a hill, if you notice it start to roll down and stop it, all is O.K., but of it has rolled too far and is going too fast, you will never catch it. As the bias pot of an amp like this is adjusted, at low settings the crossover distortion is bad, but the "hill" has a dent in the top into which the "ball" will always roll, and hot output devices will always tend to cool down. However, as the bias is increased, "the dent in the top of the hill" gets shallower and at some point it is flat. Further increases make the "flat area" into into an "ordinary hill" and the "roll down of the ball" is inevitable. The trick is to retain that "dimple", a state of unconditional stability, and the device never gets the chance to run away.
Here in Tasmania it is winter and the heatsinks of this amp are freezing cold when it is off. In the winter cold, the devices have no trouble getting heat away, but when summer comes and the ambient temperature is high, that's like turning the bias up and removing the "dimple on the hilltop", and things could get nasty, so it's always a good idea the back the bias down by a little safety margin, particularly if adjusting in a cold room in winter. On a very hot summer day, one could tweak the bias right up to the mark, knowing cooler weather would simply back it off a touch. So this is the rationale as to why one is so careful with bias adjustments of equipment of this sort. After you adjust it, you want to see it back off slightly as it comes to thermal equlibrium, not keep gradually advancing with gathering pace into eventual runaway. This is another reason for the incandescent lamps in series with the supply, of the output devices do tip over and run away, the increased collector current will "pull against" the higher impedance of the lamps and the power supply/collector voltage will fall, the lamps will start to glow as a warning and the whole lot will come to a second equlibrium, like a valley around the hill, before any real damage is done.
So, what is it doing now?
Sixth Measurement......................PNPs 43, 46, 44, 43, 44, 43, 47
Won't bother to do the NPNs, this is still a bit high, but its settling into the "dimple in the hill". The really hot device, drawing 51mA before, has backed down to 47mA. So now we need to back off the bias as a safety margin for hot summer days to come....
Seventh Measurement..................Back down around 31-36mA
Tweak up again
Eighth Measurement....................Back up around 34-41mA
Tweak down again
Ninth measurement.....................Back down to 31-38mA Range
Try blasting heatsink with neice's hair dryer to simulate a hot day
Tenth Measurement...................30 down to 24, 37 down to 31, so the vbe multiplier is working!
Tweak it up again
Eleventh measurement...............30-38 while the heatsink is still hot from the hairdryer
Twelfth measurement..................36-46 as it cools (42 at TP)
Thirteenth Measurement..............40-50 as it continues to cool (43 at TP)
Fourteenth measurement..............35-43 after cooling with damp sponge, (37 at TP)
Notice how it peaks in the middle and falls off at either end, this is exactly what is wanted and the range of values obtained over this temperature range, (estimated to be from 45C* down to around 20C*) cluster around 40mA. The "falling off at each end" characteristic is obviously from the vbe multiplier and its associated NTC thermistor between its collector and base, in series with fixed 750 ohms. This is where I will leave it, now I have the other channel.....!
Final warm measurement, 37-42, 39 at the test point. Turn off for a couple of hours...
Stone Cold Heatsink, 31-41 and rising rapidly as dies heat up. (27 at test Point).
After a few minutes before heatsink has had a chance to be warmed by transistor dies the bias current range for the PNP devices from endmost to innermost, next to drivers was
61-69mA, (64mA at test point), at this stage enough current is flowing to cause all four 200w lamps to glow at a low level. It seems to hold at 61-70mA until the vbe multiplier starts to feel the wamth.
After 5 minutes or so, the lamps start to dim slowly and range drops to 56-65mA, (59mA at test point), Aafter about 10 minutes, the glow from the lamps is barely visible in a lit room and the current range is down to the 47-57mA range and 51mA at the test point. After 15 minutes or so, the glow from the lamps has faded from view and the range is 41-51mA and 45mA at the test point. It is expected to take about an hour to stabilize at the 37-42mA range.
Half hour later, 38-48mA, 43mA at test point, so its creeping down...
Hope this is of some use to you all, not only in a specific sense, but a general one too in understanding the nature of the bias in bipolar output stage amplifiers.
|Thread Tools||Search this Thread|
|Thread||Thread Starter||Forum||Replies||Last Post|
|Help With Amp For B&W Matrix 3's||led_zeppelin||Solid State||0||9th October 2006 02:14 AM|
|Matrix Amp?||Mike L.||Tubes / Valves||1||6th October 2004 05:57 PM|
|B&W Matrix 3's||led_zeppelin||Swap Meet||4||13th July 2004 08:37 PM|
|B&W Matrix 3's||led_zeppelin||Multi-Way||0||15th December 2003 01:58 PM|
|To matrix or to not matrix?||Trist||Subwoofers||8||15th July 2001 09:48 AM|
|New To Site?||Need Help?|