Hi NUTTTR,
Your amplifier is indeed underbiased. If the amplifier designer could not keep the amp from going into thermal runaway, this is what they would do. I suspect the heatsink is sized a little small. When running at high power this amp may not be able to control it's bias.
Remove the resistor between the B-E of the Vbe multiplier and install the a lower resistance (say 250 ohms) in series with a 500 pot. Set the pot for maximium resistance. If you have a variac, increase the mains supply slowly while watching the voltage across the emitter resistor. Set the bias for 1 mV across the emitter resistor just to start. This should make the amp sound better but we want to see how stable the bias is. Let the amp idle for 1/2 hour and check the bias. Don't touch it unless there is a very big increase. Then, run music for 1/2 hour and check again. If the bias is stable, increase the music level and check again after 1/2 hour.
You can experiment with increasing bias levels until the circuit is unstable or you are happier with the sound. The is no point in a bias current much over 10mV (20mA ish). While you are at it, measure the voltage across all emitter resistor to check the match on your outputs.
-Chris
Your amplifier is indeed underbiased. If the amplifier designer could not keep the amp from going into thermal runaway, this is what they would do. I suspect the heatsink is sized a little small. When running at high power this amp may not be able to control it's bias.
Remove the resistor between the B-E of the Vbe multiplier and install the a lower resistance (say 250 ohms) in series with a 500 pot. Set the pot for maximium resistance. If you have a variac, increase the mains supply slowly while watching the voltage across the emitter resistor. Set the bias for 1 mV across the emitter resistor just to start. This should make the amp sound better but we want to see how stable the bias is. Let the amp idle for 1/2 hour and check the bias. Don't touch it unless there is a very big increase. Then, run music for 1/2 hour and check again. If the bias is stable, increase the music level and check again after 1/2 hour.
You can experiment with increasing bias levels until the circuit is unstable or you are happier with the sound. The is no point in a bias current much over 10mV (20mA ish). While you are at it, measure the voltage across all emitter resistor to check the match on your outputs.
-Chris
something is not quite right
Aaron,
The transistors that are getting 0.38 are probably off, IMHO that is too low, and would explain why you are getting no current through the output emitter resistors. Akmost all the idle current flows through both poles of the output, so if one half is turned off it will read very low.
If you have a pot of approx. the correct value, say 1k-5k you could install it in series with the 1k resistor, make sure it is at zero before you start (ie the total Rce is still 1k) then after install small fuses in your amp, very, very slowly increase the value. You should see the Vce for the d699 increase and following, the bias voltage for the output transistors.
The outputs get their bias from the drivers so there is a natural tendency for the drivers to bias up first. But once everything is beginning to conduct the Vbe should be fairly similar across the drivers, and all the outputs.
If the Vbe for the outputs doesn't increase as the output of the vbeM increases, or the disparity between the output poles doesnt diminish, something, probably the protection circuit, is dumping (shorting) the bias. If it does increase, beg borrow or steal at least 3 meters, then monitor the bias voltage for the drivers and (ideally both) output poles, plus the voltage across the emitter resistors. Slowly turn it up until you see approx. 0.55-0.6 Vbe for the outputs & drivers., or the voltage across the output emitter resistor pair reaches 0.02v. Once the d699 is generating something like 2.4v, you should see a fairly even spread of Vbe's for the drivers and outputs, if any are too high (>0.6) or low (<0.5) put it back to where you started and look for dead components in the protection circuits. If you are brave you could just remove the protection circuits, I'd want to draw out the schematic of this part to make sure I still had DC protection, but overcurrent is probably not useful unless you plan on bridging the amp.
If everything seems to work right, you will eventually have a pot with the correct series increment for the top resistor. Measure it, buy or make a resistor of the correct or slightly smaller value, then put it into the circuit. Be conservative here, while more bias is good, dead amps are not...This value is unique to this set of transistors, each channel will need the procedure repeated.
The d699 compensates for thermal changes in the output transistor Vbe, it's own vbe diminishes as it heats up, so the bias voltage drops as the temp on the heatsink increases. Tthere is a lag while the temp of everything equalizes, so you will always need to make a change, let everything warm up ,while monitoring, etc. Once you're close to done adjusting you will probably also want to use a hair drier or a heat gun to get the heasink up to, say, 50c, then retest all your bias points.
Good luck
Stuart
Aaron,
The transistors that are getting 0.38 are probably off, IMHO that is too low, and would explain why you are getting no current through the output emitter resistors. Akmost all the idle current flows through both poles of the output, so if one half is turned off it will read very low.
If you have a pot of approx. the correct value, say 1k-5k you could install it in series with the 1k resistor, make sure it is at zero before you start (ie the total Rce is still 1k) then after install small fuses in your amp, very, very slowly increase the value. You should see the Vce for the d699 increase and following, the bias voltage for the output transistors.
The outputs get their bias from the drivers so there is a natural tendency for the drivers to bias up first. But once everything is beginning to conduct the Vbe should be fairly similar across the drivers, and all the outputs.
If the Vbe for the outputs doesn't increase as the output of the vbeM increases, or the disparity between the output poles doesnt diminish, something, probably the protection circuit, is dumping (shorting) the bias. If it does increase, beg borrow or steal at least 3 meters, then monitor the bias voltage for the drivers and (ideally both) output poles, plus the voltage across the emitter resistors. Slowly turn it up until you see approx. 0.55-0.6 Vbe for the outputs & drivers., or the voltage across the output emitter resistor pair reaches 0.02v. Once the d699 is generating something like 2.4v, you should see a fairly even spread of Vbe's for the drivers and outputs, if any are too high (>0.6) or low (<0.5) put it back to where you started and look for dead components in the protection circuits. If you are brave you could just remove the protection circuits, I'd want to draw out the schematic of this part to make sure I still had DC protection, but overcurrent is probably not useful unless you plan on bridging the amp.
If everything seems to work right, you will eventually have a pot with the correct series increment for the top resistor. Measure it, buy or make a resistor of the correct or slightly smaller value, then put it into the circuit. Be conservative here, while more bias is good, dead amps are not...This value is unique to this set of transistors, each channel will need the procedure repeated.
The d699 compensates for thermal changes in the output transistor Vbe, it's own vbe diminishes as it heats up, so the bias voltage drops as the temp on the heatsink increases. Tthere is a lag while the temp of everything equalizes, so you will always need to make a change, let everything warm up ,while monitoring, etc. Once you're close to done adjusting you will probably also want to use a hair drier or a heat gun to get the heasink up to, say, 50c, then retest all your bias points.
Good luck
Stuart
Hi Stuart,
With the vari resistor, you said to add in in series with the 1k ohm one... from your diagrams, i would assume it is the "top" resistor - i am only worried about adjusting this obviously because if it does go to "0" ohm, wouldn't it make the amp go *pop* and that'd be it? I'm going to install a smaller fuse, i'm going to buy a large range and put the lowest one possible in!
If i was to adjust the "bottom" resistor how do i go about that?
Thanks
Aaron
With the vari resistor, you said to add in in series with the 1k ohm one... from your diagrams, i would assume it is the "top" resistor - i am only worried about adjusting this obviously because if it does go to "0" ohm, wouldn't it make the amp go *pop* and that'd be it? I'm going to install a smaller fuse, i'm going to buy a large range and put the lowest one possible in!
If i was to adjust the "bottom" resistor how do i go about that?
Thanks
Aaron
to replace bottom resistor
It is just as you would think, you use the variable resistor in the bottom position, adjust it to start at approx. the same resistance or a little more than the fixed resistor it is replacing, then slowly decrease the value.
Stuart
It is just as you would think, you use the variable resistor in the bottom position, adjust it to start at approx. the same resistance or a little more than the fixed resistor it is replacing, then slowly decrease the value.
Stuart
Great 🙂
I've got a 25 turn industrial vari resistor here that will suit it nicely 🙂
I just didn't want it to go pop 🙂 I thought better safe than sorry eh 🙂
I gather the "bottom" resistor would be 430ohm? I'd assume that to be the case anyway!
Aaron
I've got a 25 turn industrial vari resistor here that will suit it nicely 🙂
I just didn't want it to go pop 🙂 I thought better safe than sorry eh 🙂
I gather the "bottom" resistor would be 430ohm? I'd assume that to be the case anyway!
Aaron
better safe than sorry...
Aaron,
I'd go for safety and make it 500 to start with...it should drop the d699 Vce to approx. 1.8v, everything will be a little lower and you will be able to see if the driver Vbe changes and increases properly as you head towards the original value...
Keep up the good work,
Stuart
Aaron,
I'd go for safety and make it 500 to start with...it should drop the d699 Vce to approx. 1.8v, everything will be a little lower and you will be able to see if the driver Vbe changes and increases properly as you head towards the original value...
Keep up the good work,
Stuart
The heatsinking approach is long wind tunnels, and these easily show 15ºC temperature difference between one end and the other. To make things worse, the Vbe multiplier is placed at one end of the tunnel so it allways senses a extreme temperature [hotter or cooler than the rest of output transistors depending on the placement of the fan] instead of sensing the average temperature of the tunnel [middle placement]
So adjusting class AB bias for the hottest pair of output devices will force the rest of the [cooler] device pairs to operate still in class B, and adjusting class AB bias for the coolest transistor pair will cause thermal runaway in the rest of [hotter] transistor pairs [blown devices]
To make things even worse, the driver transistors are also placed on the main heatsink, reducing the negative thermal feedback introduced by the Vbe multiplier, or worse, creating positive thermal feedback [and thermal runaway]
It's obvious that the designer of this amplifier gave up in the task of making it class AB
Be careful when experimenting, don't measure bias on a single output device, measure on several devices and see the imbalance [disconnect the load momentaneously when measuring]
So adjusting class AB bias for the hottest pair of output devices will force the rest of the [cooler] device pairs to operate still in class B, and adjusting class AB bias for the coolest transistor pair will cause thermal runaway in the rest of [hotter] transistor pairs [blown devices]
To make things even worse, the driver transistors are also placed on the main heatsink, reducing the negative thermal feedback introduced by the Vbe multiplier, or worse, creating positive thermal feedback [and thermal runaway]
It's obvious that the designer of this amplifier gave up in the task of making it class AB
Be careful when experimenting, don't measure bias on a single output device, measure on several devices and see the imbalance [disconnect the load momentaneously when measuring]
Ok, so i have 2.2v across the Vce of the D669, bias is ~2.7mv or so once warmed up a little, been idle for about 20mins and the other channel (untouched) is still at room temp (21deg), the channel i am working on is about 27deg... So far so good, i am waiting for the amp to settle a bit, but it looks highly possible i can go straight to 5mv bias... not going to till this is tested more though!! Extra fans will be needed for cooling, i am thinking of drilling out the bottom of the case and have 2 fans UNDER each heatsink, that way it will keep the whole sink at an even temp, it does mean higher feet on the bottom of the amp though!! but it would solve my problems with potential heat issues...
Let me see if i get this bit right too, if i have about 5mv worth of bias, does that mean ~50ma of current is going through each device? If that's the case, does that make 500ma for the whole channel (10 outputs/5 pairs x 50ma)...?
Thanks for the help so far!! I've made a "pin base" for the pot to go into, allowed me to remove the resistor without taking the board out and attach the pot without doing it too... So seems like all is good so far.. I might keep the pot, considering it's a pretty high quality one, rather than put a resistor on it 🙂
Hopefully this goes well, but will test the sound shortly so i guess i'll know the sound difference (well, hopefully but may not through a sub!)
Thanks all
Aaron
Let me see if i get this bit right too, if i have about 5mv worth of bias, does that mean ~50ma of current is going through each device? If that's the case, does that make 500ma for the whole channel (10 outputs/5 pairs x 50ma)...?
Thanks for the help so far!! I've made a "pin base" for the pot to go into, allowed me to remove the resistor without taking the board out and attach the pot without doing it too... So seems like all is good so far.. I might keep the pot, considering it's a pretty high quality one, rather than put a resistor on it 🙂
Hopefully this goes well, but will test the sound shortly so i guess i'll know the sound difference (well, hopefully but may not through a sub!)
Thanks all
Aaron
Now i'm up to about 5.5mv bias across almost all the outputs... there's 1 that stands out quite a bit tho, about 6mv when all the others are at 5.5... however, once running for a while (@4ohm one channel only for one 3min song running at almost full power out of that channel) it is up to about 56deg (no fans at all on it at the moment) wtih about 9.1mv bias, while the others are more like 7.9mv-8.1mv... Why the difference? is it just the chip is a bit off? I can always order more to try and match them (same batch, etc) i guess... However it does make me wonder why this one is so far out! Adding fans in would help a MASSIVE amount at this point, obviously it runs rather warm without them! The thermal switch for the fans in on the other channel (yup, only switches on 1 channel!) so i'm gonna change that obviously! It idles at about, oooo, 40-45deg, which is pretty much the same as my peavey amp (runs motorola TO-3 outputs though), so not sure if this is right... The sub "seems" (to my ears anyway) quite a fair bit louder than before when running off one channel, i remember it sounded poor and didn't go that loud, but now it does sound better (again to my ears) and is very clear!
Aaron
Aaron
imbalance...
Aaron,
It is not unusual at all to find one transistor with higher bias, actually it's almost guraranteed. In this context it is about 10% high, which is not too bad, and might not matter, but it will definitely be the limiting factor as you approach the amps maximum output. As Eva points out the individual temp of the transistors will affect the current sharing, if the temp spread along the heatsink makes this guy a little hotter, that alone could explain the difference. The natural tendency of bipolar transistors is to hog the current, as they heat up the Vbe decreases and they take more current, so they heat up more etc. If you know you won't be using the full output of the amps you could change the emitter resistors to be, say 0.5ohm, or even 0.68, that would encourage slightly better sharing. Also the emitter resistors might only be +/-5% so you might find the one in question is lower than the others...
Sounds like you are nearly done. If you are seeing 5mv across the emitter resistors the bias current is still quite low. Ohms law tells you that the standing current is 5mv/0.3ohms, or approx 15 ma per output. Since you have 70v rails, your total idle dissipation is 5 * (70+70) * 0.015 or approx. 10w. This seems to induce a temp rise of about 6 degrees, so your heatsinking is good for about 0.6c/w, without fans, with the case open etc. This tells you just how hot the amp will get for different bias levels, but it also tells you that you can't remove the fans. At a maximum internal dissipation of approx. 400w your amp would be destroyed, the theoretical heatsink temp rise would be 400*0.6, or 240c...smokin'...I think had you continued to run the channel at full power it would have reached a much higher temperature...
Keep up the good work, it's great to hear the amp sounds better
Stuart
Aaron,
It is not unusual at all to find one transistor with higher bias, actually it's almost guraranteed. In this context it is about 10% high, which is not too bad, and might not matter, but it will definitely be the limiting factor as you approach the amps maximum output. As Eva points out the individual temp of the transistors will affect the current sharing, if the temp spread along the heatsink makes this guy a little hotter, that alone could explain the difference. The natural tendency of bipolar transistors is to hog the current, as they heat up the Vbe decreases and they take more current, so they heat up more etc. If you know you won't be using the full output of the amps you could change the emitter resistors to be, say 0.5ohm, or even 0.68, that would encourage slightly better sharing. Also the emitter resistors might only be +/-5% so you might find the one in question is lower than the others...
Sounds like you are nearly done. If you are seeing 5mv across the emitter resistors the bias current is still quite low. Ohms law tells you that the standing current is 5mv/0.3ohms, or approx 15 ma per output. Since you have 70v rails, your total idle dissipation is 5 * (70+70) * 0.015 or approx. 10w. This seems to induce a temp rise of about 6 degrees, so your heatsinking is good for about 0.6c/w, without fans, with the case open etc. This tells you just how hot the amp will get for different bias levels, but it also tells you that you can't remove the fans. At a maximum internal dissipation of approx. 400w your amp would be destroyed, the theoretical heatsink temp rise would be 400*0.6, or 240c...smokin'...I think had you continued to run the channel at full power it would have reached a much higher temperature...
Keep up the good work, it's great to hear the amp sounds better
Stuart
Do you mean 2.7V? If so, that's fairly close to the 2.86-2.89V cited by Self as optimal for EF topologies.
There is still the wind tunnel conundrum disccussed earler. If you have anyway of measuring it (a DMM with a temperatur function and a thermocouple, for instance), it would be good to know how big the temp difference is at opposite ends. It's always possible the difference isn't great enough to be a major concern -- at least you would know.
Thanks stuart!
As it gets to "running temp" which i'd be happy with at 50deg or so, hotter is always going to happen sometimes, but i'd prefer to avoid it! The bias is out at a level i'd be happy with at about 50deg, it was about 8.*mv on the high chip and about 7.*mv on the rest, which to me should be fine, if i turn it up much more it does get quite a bit warmer!!! the Vbe of the outputs gets much lower when i turn it up a fair way, (i.e. drops to 0.5v or so) so this is obviously the protection circuitry coming on a little (it does seem overly sensitive) but at full power this should also help keep temps down, and looks like it's biasing pretty much back to class b at clipping... Which is ok for me!!
I'll give the other channel a go tonight/tomorrow and see how it goes!
Thanks alot
Aaron
As it gets to "running temp" which i'd be happy with at 50deg or so, hotter is always going to happen sometimes, but i'd prefer to avoid it! The bias is out at a level i'd be happy with at about 50deg, it was about 8.*mv on the high chip and about 7.*mv on the rest, which to me should be fine, if i turn it up much more it does get quite a bit warmer!!! the Vbe of the outputs gets much lower when i turn it up a fair way, (i.e. drops to 0.5v or so) so this is obviously the protection circuitry coming on a little (it does seem overly sensitive) but at full power this should also help keep temps down, and looks like it's biasing pretty much back to class b at clipping... Which is ok for me!!
I'll give the other channel a go tonight/tomorrow and see how it goes!
Thanks alot
Aaron
The Vbe of bipolar transistors decreases aprox. 2mV for each ºC [or ºK] of temperature increase. It's easy to go below 500mV at high temperatures
The Vbe multiplier also suffers this Vbe effect and provides some compensation for Vbe changes of output devices with temperature
But this only works properly when the Vbe multiplier is at the same temperature than the output devices and has similar Vbe temperature coefficient
The Vbe multiplier also suffers this Vbe effect and provides some compensation for Vbe changes of output devices with temperature
But this only works properly when the Vbe multiplier is at the same temperature than the output devices and has similar Vbe temperature coefficient
Thanks Eva,
Does that mean even though i suspected it might be running class b at high volume it might not be? The temp difference between the hottest part of the heatsink and the vbe multiplier was about 8deg which is minimal (no fan at all) so i'm ok with that, i guess i'll have to see how it goes now!
Thanks
Aaron
Does that mean even though i suspected it might be running class b at high volume it might not be? The temp difference between the hottest part of the heatsink and the vbe multiplier was about 8deg which is minimal (no fan at all) so i'm ok with that, i guess i'll have to see how it goes now!
Thanks
Aaron
sam9 said:
With the bias set as is now, about 2.7v or so.... However the heatsink does get quite warm, when playing gets quite hot and bias climbs further (but will obviously stop at a certain point) the temp diff is about 8deg or so... nothing major and with a fan in the right place i might be able to alleviate this!
On a side note - what bias does class A run? Another side note - other than high voltage rails (153VDC on each channel (+-78)) what is stopping someone from making it nearly class a?
Some amplifiers are designed to get optimum bias only when they are hot
You have to consider that when the amplifier is playing loud, the silicon dies of the power transistors may be markedly hotter than the heatsink and this effectively increases biasing far above the idle value
The only way to know is to measure...
Leave a multimeter connected to one of the emitter resistors, play loud for some time and then quicky pause the music, disconnect the ouput and see what the multimeter has to say
You will se a bias value substantially higher than expected that progressively drops to the expected value as the dies of the transistors cool down to the same temperature of the heatsink
You have to consider that when the amplifier is playing loud, the silicon dies of the power transistors may be markedly hotter than the heatsink and this effectively increases biasing far above the idle value
The only way to know is to measure...
Leave a multimeter connected to one of the emitter resistors, play loud for some time and then quicky pause the music, disconnect the ouput and see what the multimeter has to say
You will se a bias value substantially higher than expected that progressively drops to the expected value as the dies of the transistors cool down to the same temperature of the heatsink
class A
Aaron,
Class A for rails of this voltage would need standing current of several amps, depending on the intended load, could be 10+ amps. This is a huge thermal load, well beyond the capabilities of your current output transistor set, heatinks etc. Even if you halved the rail voltage to +/-35 you'd still need to get rid of 300+ watts per channel continuously, not really practical, even if the amp design were capable of doing justice to the effort.
If class A is your goal, you'd probably do better to use the big components and rebuild the amps from scratch. You know, build some alephs in the cases or something. Your biggest task would be unwinding, tapping and rewinding the transformer secondaries to get 35-0-35, 35-0-35. Much better values to be working with, but still pretty high for class A. The unwinding and rewinding trick only really works for toroids...
If you like the way the amp sounds I'd button it all down and call it done. It can only get worse from here.
A job well done
Stuart
Aaron,
Class A for rails of this voltage would need standing current of several amps, depending on the intended load, could be 10+ amps. This is a huge thermal load, well beyond the capabilities of your current output transistor set, heatinks etc. Even if you halved the rail voltage to +/-35 you'd still need to get rid of 300+ watts per channel continuously, not really practical, even if the amp design were capable of doing justice to the effort.
If class A is your goal, you'd probably do better to use the big components and rebuild the amps from scratch. You know, build some alephs in the cases or something. Your biggest task would be unwinding, tapping and rewinding the transformer secondaries to get 35-0-35, 35-0-35. Much better values to be working with, but still pretty high for class A. The unwinding and rewinding trick only really works for toroids...
If you like the way the amp sounds I'd button it all down and call it done. It can only get worse from here.
A job well done
Stuart
My very limited understanding of BJT Class A is that the bias scheme for a push-pull (compementary output) is done by other means than just cranking up a Vbe multiplier. On the other hand, MOSFET Class A appearently does work this way more or less.
Righteo!! I was more wondering about class a than be game enough to actually try it!!!! Getting rid of 300+watts worth of heat is something i'm not willing to do at any stage, as it gets hot enough here without that!!!
Thanks for the help Stuart! I'm doing the other channel tomorrow it seems and i'll see how it goes and we should be right 🙂
Thanks
Aaron
Thanks for the help Stuart! I'm doing the other channel tomorrow it seems and i'll see how it goes and we should be right 🙂
Thanks
Aaron
You forgot to mention that a bridged amplifier dissipates at least 4 times the output in class A.
With vented tunnel heatsinks it would be possible to conservatively set the bias to a level where each SA/SC output transistor dissipates something like 40 watts at a junction temperature of 80 C/184 F above ambient.
That would be 800 watts of heat, with voltage at around 80V that comes down to 0.50 amps bias current per transistor.
800 watts of heat for far less than 100 watts Class A output in bridged mode.
At 100mA per output transistor you'll have 4 watt in class A,160 watts of heat, with heatsinks tunnelled and vented at some 4 degrees above ambient.
Without vents on both heatsinks with bias at 100mA for each transistor more like 25.
With vented tunnel heatsinks it would be possible to conservatively set the bias to a level where each SA/SC output transistor dissipates something like 40 watts at a junction temperature of 80 C/184 F above ambient.
That would be 800 watts of heat, with voltage at around 80V that comes down to 0.50 amps bias current per transistor.
800 watts of heat for far less than 100 watts Class A output in bridged mode.
At 100mA per output transistor you'll have 4 watt in class A,160 watts of heat, with heatsinks tunnelled and vented at some 4 degrees above ambient.
Without vents on both heatsinks with bias at 100mA for each transistor more like 25.
- Status
- Not open for further replies.