VBE multiplier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am trying to get the Luxman M-117 amp to bias properly. Attached to this post is the amp schematic. When I try to bias the amp higher to reduce cross over distortion I can't get it to thermally compenstate correctly. When monitoring the heat sink termperature and the emitter resistors the voltage stays constant but there is a very slow creep up in termperature. I think this means that the VBE multiplier is undercompensating at higher temperatures. I am trying to keep the amp at 45C at idle so that it won't overheat when driven. I am told it is better to slighly overcompensate. Right now I think it is slighly undercompensated at idle, which I don't think is safe.

I have mounted the hear sensor directly on top of the middle plastic sanken 2sc2922 output transsitor to reduce thermal lag. This is still a plastic to plastic interface so I don't know how effective it is. The original design simply mounted the sensor in the middle of the heat sink but the response was even slower.

I have read Self's book but he only talks about increased tempco vbe multipliers for CFP output stages not EF, like this is.

I tried the 2 transsitor Szikial pari ala Self which is supposed to have increased tempco over the 1 transistor version. It is stable and seems to have slighly higher tempco but the termperature still creeeps about .1 degree centrigrade per 30 minutes. After 2 hours it is up .5--.7 C. It does not seem to stop.

Some of you said on other threads that you could vary the tempco of the Szikail par by changing the base emitter resistor of the 2nd transistor. But I don't think this is true as the tempco does not vary much when I vary the value of this resistor.

Now I think I am left with 3 options:

1.Do a better job encasing the sensor in insulating material to avoid the slow heat loss?

2. Try using transistors in the Sziakel pari VBE with lower voltage (faster) or higher hfe to increase compensation? Will this work?

3. Try to come up with practical value of Self's increased tempco vbe multiplier based on the amp schematic I have attached. I am not smart enough to do this but I was thinking of using the one with a diode voltage reference? Can anyone suggest practical value to start.

4. I tried creating a 4 diode string a la Leach but it undercompensated. Someone suggested using Shotkcy diodes due to the lower Vf will have a higher compensation factor. Is this a worthwhile approach?


Bear in mind that the Vq of the emiiter resistor is held fairly firm. It seems to heat up, then the Vq goes slightly down to compenstate and then it heats up again, and then the Vq starts to go down again, which is all well and good. But eventually this pattern of going slighly down and then back up starts to drift upward. It takes at least a half hour to an hour before the slow upward drift starts.

I have spent almost 6 months trying to figure this out so I throw myself at the mercy of the forum.
 
If I understand your post correctly you’re worried about thermal runaway with extended operation (instead of reaching some stable working point the output bias current slowly creeps up over time and the amp temperature steadily rises accordingly)?

That schematic’s far too hosed up and incomplete to be of much help. One thing that you can do is give the 2nd stage current source a negative temperature coefficient. To do this mount the diodes that are used for reference in the VAS source (just above and to the left of the Vbe multiplier in your diagram, if I’m seeing it properly) on the heat sink with some epoxy. As the temperature of the heat sink goes up the voltage across the diodes will go down and the current through the Vbe multiplier will go down accordingly. This will cause a drop in voltage across the Vbe multiplier that’s added to it’s own TC. Seems to work fairly well, drawbacks are a decline in second stage slew rate as temperature goes up, and an increase in offset if a current mirror isn’t used to interface the differential to the 2nd stage. The slew rate issue can be addressed simply by starting out with ample 2nd stage current to begin with. Offset probably won’t be a serious issue if the DC gain is unity (a capacitor is used in series with the gain setting resistor, or an effective servo is in use).

I know, probably not the answer you were looking for. Discrete Vbe multipliers are limited in what they can do though.
 
I would try doubling the output transistor emitter resistors to
0.47 ohm * 6 after setting the bias current for low distortion with the
.22 ohm resistors in place. This keeps the current down while maintaining more class A drive on the output. It also reduces the
effects of heating on VBE as part of the drop in now across the output
resistors.
speaker output will not be affected much as .47 * 3 = .153 ohms
 
That makes perfect sense.

The original amp was biased at 8mv across the emitter resistors where Self recomends 46.2 mv for .22 emitters.

So I tried to get it up to about 25mv, which is still suboptimal.

So I either need to get a bigger heatsink-- not a viable option

Fan cooling makes no sense if it needs a fan even when idling.

Or I could eliminate one pair of the output transistors.


Now there are more facts I have not disclosed because I could not put it all in one e-mail.

The amp was a stereo amp I converted into a monblock because the thermal protection circuity kept kicking in. So now instead of six 17 amp sankens (2sc2922) on one heatsink there are now 12 sankens on 2 heatsinks. If I eliminate one pair from each heatsink I would go down to 8 transistors on two heatsinks instead of 6 tranistors on one heatsink. I think this would permit me to bias the transistors more optimally without thermal runaway.

It drives a 2.5 ohm woofer tower. So now I am concerned about using only 8 transistors because of beta droop.

It did 200 watts into 8 ohms but only 250 into 4. But with the transformer now only supplying one channel and the filter capacitance going from two 15,000 uf (30,000 uf total) to 60,000 uf total and with plus/minus 80 vots on the rails, the amp may try to sink alot more current into 2.5 ohms.

With the low impedance of the woofer tower I am not sure whether it is better to bias more optimally with 8 transistors or suboptimally with 12.

By the way the amp sounds fanstastic.

Any thoughts?
 
Hi,
what does the schematic show immediately before the output device bases? It has three rectangles with a letter inside each.
It appears to connect the Vbe multiplier to the supply rail and to the output base resistor. Is there a resistor in the connection to the supply rail?
It also looks like there is a pair of resistors to the output feedback tapping. Can you confirm?
 
I cannot digest all of your great suggestions.


WGski: I am trying to observe carefully the long term temperature creep. This is moving up only .1c per half hour. The Vq drops as temperature goes up but then it can't fully compenstate and goes back up to approximately the same value. It would appear that to keep the temperature from creeping up the VBE would have to respond faster to the increase so that the temperature does not cumulate. This is a very long term phenomena but it would appear that vq would have to dcrease more or faster to prevent the slow thermal run away. I am not sure whether this is a heat sink deficiency or a delayed response that causes cumulative error. Does it make sense that this problem would get worse when I try to bias at a higher vq?

Kaos, you are like a graduate school professor talking to a student in college. Which LEDs do you suggest I attach to the heatsink? There are 2 attached to the bottom of the resistor chain and one on the top. All 3 of them are Led LTZ MR15 (RED). Do I try mounting them all on the heat sink. I could put them in the same holes I made in the heat sink for the leach diode string for the VBE that I never implemented which are centrally located on the heatsink. Does anyone else agree with this approach because it is very creative? Does this make sense with an EF output stage?

Stocktrader. YOur idea is over my head. I know if I change the emitter resistor values it will change the current limited protection circuit. It will also improve current sharing which helps with 12 output transistors mounted on 2 heatsinks. I like the fact that it will lower the heat sink temperature because more current will be dissipated by the resistors and less by the transistors but I am afraid of this significant of a change. Does anyone else agree with this approach.

I need more help but thanks so much for all of your wonderful ideas.
 
Andrew, at least I can answer this question.

the rectangular box with the BCE designations refers down to the bottom edge of the schematic where you see the 4 small square boxes each of which are darlingtgon pairs.

There is one for the positive and negative drive circuit.

I think Self would refer to this as an output triple or beta enhancer , since it has a predriver and driver transistor in each box.

There are two 270 ohm resistors (540 ohm total) between the plus minus predrivers

Self uses one 330 ohm resistor for faster switch off. Thought about listening and changing this value to 330 ohms, make sense????

The drivers have two 68 ohm resistors for 144 total, which is very close to Self's circuit with 150 ohm so I was going to leave it alone. But I did add the suck out capacitor across the resistors

I think this is a EF type II output stage according to Self.
 
fglabach said:
I am not sure whether this is a heat sink deficiency or a delayed response that causes cumulative error. Does it make sense that this problem would get worse when I try to bias at a higher vq?


It is hard to seperate the two phenomena. In a deayed response situation, the bias may increase temporarily (or decrease) and again settle out after a change in dissipation. It's only really "thermal runaway" when the steady-state bias current creeps up as the temp increases. If the bias current is stable, the temp will eventually stop increasing - but it may be at a temp that is too high for comfort. The only remedy in that case is either reducing the bias (dissipation) or getting a fan. But if you overcompensate such that the bias goes down with temp, it will no longer be optimally biased per Self as the temp goes up. In situations where the heat sink isn't really big enough, it is the only acceptable answer. The optimal bias is quite a bit higher than is typically designed for when manufacturers select the heat sink.
 
The diode that I’m referring to is circled in green. If this were mounted to the heat sink to sense temperature while the associated transistor is allowed to remain at ambient, then the current source would develop a negative TC related to the heat sink temperature. The diode would need to be a red LED or a combination of series diodes (not a blue type). If you do choose to try this just be careful that you don’t accidentally hook the diode up backwards or damage it with ESD. Whether a single red LED will be enough to give you the TC that you want will remain to be seen.

diode_example.jpg
 
Kaos:

Thank you for the helpful diagram.

Yes, that is a red LEd for the current source.

It should be easy to move that off the board to the main heatsink

I will finish this project during the weekend. Sorry for the delaly and then I will find out whether there is enough additional negative co-efficient from the one red LED.

Some more information. The temperature rise is very slow. This is a very slow long term thermal delay. The VBE multipier actually lowers the vq in response to higher termperature.

What I don't understand is what is the difference between increasoing the tempco of the VBE multiplier as opposed to your idea of adding negative tempco from the VAS which may current limit the driver stage. Why have you shifted your attention to the VAs? This stuff is over my head but I still will pursue your idea.

By the way I live in LA as well.
 
A single Vbe multiplier can only give so much in terms of a practical temperature slope. In a discrete design it will always have a lower temperature rise than the outputs in an EF stage due to the mechanical nature of the situation. After that one has to start looking elsewhere for options to tweak the temperature curve. In your case it sounds like there may be a general heating of the circuit over time causing drift. The VAS current is a simple option to explore to counter this longer term effect. The red LED can be replaced with several discrete diodes to increase the TC if needed, though the value of the associated resistor in the current source would need to be adjusted to maintain proper VAS current if that were done.
 
wg ski:


Here are my readings starting one hour after warm up:

41.7 C 21.3mv at 9:30 a.m.
42C 21.41mv at 9:45 a.m.

42.5C 21.35mv at 10:00 a.m.

42.6C 21.6mv at 10:05 a.m.

42.7C 21.7mv at 10:10 a.m.

42.8C 22nv at 10:12

42.9C 21.54 mv at 10:13

42.7 21.26 at 10:14 am

42.8 21.23 at 10:14

42.8 21.17 at 10.14 a.m.

42.9 21.76 at 10:15

43.5 21.85 at 10:30

43.4 21.5 at 10:38

43.5 21.6 at 10:45.


Is this thermal run away? Insufficient negative tempco? or inadequate heat sink?
 
That doesn’t look too bad, though it’s difficult to tell from just a few degrees change. Ideally a slight negative TC is better than a slight positive one, but it’s hard to get a discrete design to be right on the mark. Looks like the current will slowly rise with temperature, but not destructively so. I guess you have to ask yourself “what’s my goal?” If you want the amp just to not blowup you’re probably there (though cooler running is better than hot, especially with a heavy load like you have). If you want a closer tracking or a negative TC then you still have some work to do.
 
Kaos

Since the amp is driving a woofer tower I think my goal has to be to get a slight negative tempco.

Do you think there would be any benefit from putting it in a lower voltage faster VBE multiplier such as a BD139.

By the way the readings I provided are with a Szikai two transistor vbe multiplier that I installed which I believe has slightly higher negative tempco then the one transistor used in the circuit.

If your idea works I may go back to the one transistor VBE because it is less prone to problems.

Thanks again.
 
Doesn't look like it's going anywhere. Those mV fluctuations could even be due to the line voltage varying. To be sure, either run it for a week straight and monitor it, or crank the music up for half an hour, or whatever it takes to get the heatsink up to about 60-70C. Then back off the music and take readings. It should settle out (with the HS temp still elevated) within a few minutes if it's stable. If it's still in the 22mV range I'd say it's pretty good.

If you're running over 40C just idling, the heat sink really isn't big enough for a class AB/B amp.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.