Allison Sim
Keantoken,
So, I've plugged the Allison into the simulation configuration you so kindly fixed for me. I think I've implemented it correctly. Please have a look. I also added a speaker simulation for the load. The THD numbers are quite remarkable. A question for you... the voltage source at the lower left it's -1.66540v how to you come up with this value? Just trying to learn...
I hope to have a board in about prototype board of this configuration driven by an LME49811 in about two weeks... I'll keep you posted.
Ken
Keantoken,
So, I've plugged the Allison into the simulation configuration you so kindly fixed for me. I think I've implemented it correctly. Please have a look. I also added a speaker simulation for the load. The THD numbers are quite remarkable. A question for you... the voltage source at the lower left it's -1.66540v how to you come up with this value? Just trying to learn...
I hope to have a board in about prototype board of this configuration driven by an LME49811 in about two weeks... I'll keep you posted.
Ken
In real life you probably won't get such incredible numbers, I'm sure you already know this. The simulator should be understood as a determiner of "theoretical best possible performance".
First, Q4 and co are a Vbe multiplier/bias circuit that is actually made obsolete by Q8 and Q9, which perform the same function. Here is what I recommend:
1: Use a Darlington output configuration instead of CFP. It worked for me with very low power output devices, but high power ones like the 1302a and 3281a have too much Cob, added to the fact that the collector is connected to the heatsink which increases Cob. With high power devices, my experience is that the CFP Allison is unstable (I tested it with the 1302a/3281a personally and it didn't work out, maybe with faster japanese devices). In my simulations distortion and general performance was not affected by this change.
2: The 2SD669/649 are good I think but they are obsolete and hard to get. I recommend you use the 2SC4793/A1837 (I provided you these models in the other thread). Or you can use the MJE150xx drivers more commonly used for the 1302A/3281A output pair.
3: Just to be clear, the schematic is how the LME49811 should be connected. You shouldn't use a bias device like a Vbe multiplier with the Allison because the Allison is self-biasing (your LME49811 model doesn't give the device justice in this situation because the Allison doesn't need a Vbe multiplier).
I tweaked this value to get the offset down. This voltage source corrects for the offset of the whole output stage.
We must live some ways away from each other, it seems we post in alternating day/night cycles.
- keantoken
First, Q4 and co are a Vbe multiplier/bias circuit that is actually made obsolete by Q8 and Q9, which perform the same function. Here is what I recommend:
1: Use a Darlington output configuration instead of CFP. It worked for me with very low power output devices, but high power ones like the 1302a and 3281a have too much Cob, added to the fact that the collector is connected to the heatsink which increases Cob. With high power devices, my experience is that the CFP Allison is unstable (I tested it with the 1302a/3281a personally and it didn't work out, maybe with faster japanese devices). In my simulations distortion and general performance was not affected by this change.
2: The 2SD669/649 are good I think but they are obsolete and hard to get. I recommend you use the 2SC4793/A1837 (I provided you these models in the other thread). Or you can use the MJE150xx drivers more commonly used for the 1302A/3281A output pair.
3: Just to be clear, the schematic is how the LME49811 should be connected. You shouldn't use a bias device like a Vbe multiplier with the Allison because the Allison is self-biasing (your LME49811 model doesn't give the device justice in this situation because the Allison doesn't need a Vbe multiplier).
I tweaked this value to get the offset down. This voltage source corrects for the offset of the whole output stage.
We must live some ways away from each other, it seems we post in alternating day/night cycles.
- keantoken
Attachments
Strange, Allison as a sliding bias only. I thought the emitter connection
in the center was somehow an integral part of the magic? Perhaps not...
Ah, but this one is not a rigid follower of the emitter, and needs a feedback
loop. No big deal if you were planning to have one anyway, but somewhat
less local than the emitter coupled version I'm more familiar with.
in the center was somehow an integral part of the magic? Perhaps not...
Ah, but this one is not a rigid follower of the emitter, and needs a feedback
loop. No big deal if you were planning to have one anyway, but somewhat
less local than the emitter coupled version I'm more familiar with.
The emitter coupled version works the best (by far if you can't provide balanced signals to the complimentary sides), but this one integrates with the LME49811 more smoothly. We could fix it up to drive the "normal" way, but that would require 4 more transistors and 4 diodes.
I have no reason to believe that this way will have problems, but it depends on how "balanced" the outputs of the LME4980 are. I think we will just have to see. There could be problems if the chip tries to bias adjust the Allison in some way, but I don't expect that either since the datasheet schematic shows an external Vbe multipler.
- keantoken
I have no reason to believe that this way will have problems, but it depends on how "balanced" the outputs of the LME4980 are. I think we will just have to see. There could be problems if the chip tries to bias adjust the Allison in some way, but I don't expect that either since the datasheet schematic shows an external Vbe multipler.
- keantoken
2: Yes, but to use those big devices for Q5/6 would be a waste of money. The 4124/4126 are actually quite excellent in this position (because they are very fast), they are my favorites next to the 2N5771/5769. But you have to make sure that they are within their limits current-wise. They should be fine with the LME49811
3: The Allison doesn't need external temperature compensation. The bias current is set solely by the Vbe of Q5 and Q6. If these transistors get hot, the bias current will go down. They can be mounted far away from the heatsink because there is no danger of thermal runaway. It's not practical to use the temp-NFB affect here either because you'd simply begin overdriving Q5/Q6 when the bias went down, which may damage them.
Thermal-Trak - interesting you mentioned this. We can't use these for temperature compensation, but we can use them to set the Allison up for emitter drive, as in the schematic. Although it could be a waste of money since we could simply use normal diodes here.
- keantoken
3: The Allison doesn't need external temperature compensation. The bias current is set solely by the Vbe of Q5 and Q6. If these transistors get hot, the bias current will go down. They can be mounted far away from the heatsink because there is no danger of thermal runaway. It's not practical to use the temp-NFB affect here either because you'd simply begin overdriving Q5/Q6 when the bias went down, which may damage them.
Thermal-Trak - interesting you mentioned this. We can't use these for temperature compensation, but we can use them to set the Allison up for emitter drive, as in the schematic. Although it could be a waste of money since we could simply use normal diodes here.
- keantoken
Attachments
so, do you prefer one configuration (the first schematic vs the second in this discussion) over the other?
Japanese transistors would be perfect, but let me explain why they are still not necessary:
If you look at Q1, Q2 and Q5, you see that Q5 will, at all times, only have at max 1.5V across it, which is require to bias the darlington outputs into operation. The same is true for Q6. These transistors never have to endure a large voltage swing, and actually have relatively low dissipation because of the low Vce.
I don't see it as necessary, but I would recommend the 2SC3423/A1360 pair as they have very low Cjo. I don't think it would improve performance significantly but you are welcome to see for yourself.
If I were to choose one schematic over the other I would choose the one I posted in post 270, since this method has worked in real life and is proven in simulation to work better than the earlier schematic. The performance question here is whether the two outputs of the LME49811 are equal, as we want to avoid common-mode signals through Q5 and Q6. The second schematic is more likely to work properly and predictably.
- keantoken
If you look at Q1, Q2 and Q5, you see that Q5 will, at all times, only have at max 1.5V across it, which is require to bias the darlington outputs into operation. The same is true for Q6. These transistors never have to endure a large voltage swing, and actually have relatively low dissipation because of the low Vce.
I don't see it as necessary, but I would recommend the 2SC3423/A1360 pair as they have very low Cjo. I don't think it would improve performance significantly but you are welcome to see for yourself.
If I were to choose one schematic over the other I would choose the one I posted in post 270, since this method has worked in real life and is proven in simulation to work better than the earlier schematic. The performance question here is whether the two outputs of the LME49811 are equal, as we want to avoid common-mode signals through Q5 and Q6. The second schematic is more likely to work properly and predictably.
- keantoken
Keantoken,
Well, it worked at least for a while. I think I somehow cooked the LME49811... It lived long enough to pump out a decent sine wave and play a little music. I didn't even think to ask, but, it seems to be a Class A bais scheme as it gets quite hot at idle, is this correct? The other configurations I' ve been working with are Class A/B. I have been running all these configurations on a heavy piece of aluminum plate that barely gets warm. With the allison configuration, it got down right hot, probably 160/180 deg F.
Ken
Well, it worked at least for a while. I think I somehow cooked the LME49811... It lived long enough to pump out a decent sine wave and play a little music. I didn't even think to ask, but, it seems to be a Class A bais scheme as it gets quite hot at idle, is this correct? The other configurations I' ve been working with are Class A/B. I have been running all these configurations on a heavy piece of aluminum plate that barely gets warm. With the allison configuration, it got down right hot, probably 160/180 deg F.
Ken
MJL21193,
Just read your Patchwork Reloaded thread, over a period of a couple of weeks, very nice work. So, what are the symptoms of occilliation?
Ken
I recall someone (maybe it was Pass?) posting something about a pair of offset
chip amps in an Allison like arrangement. The name Allison never came up in the
description, but the topology appeared similar...
Anyways, both chip amps are tricked to operate in Class A. As only half of the
totem in each gets cooked, I don't pretend to know the thermal implications???
chip amps in an Allison like arrangement. The name Allison never came up in the
description, but the topology appeared similar...
Anyways, both chip amps are tricked to operate in Class A. As only half of the
totem in each gets cooked, I don't pretend to know the thermal implications???
I learned much over the course of the Patchwork amp but still have a lot more to learn.
One of the signs of oscillation is the devices getting hotter than they should. Checking the output with an oscope will confirm this.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.