lumanauw said:
David,
Mr Curl's JC-1 has 18 Sanken MT200 output devices per channel, Re's are 0.1 ohm, that gives over 29 watts in class A for 15mV.
More interesting is the economy switch on the Parasound JC-1.
The output stage bias dropping to a level of 10 watts in class A should translate to less than 10mV across Re.
lumanauw said:
I agree. The use of a good number of output transistors makes sense. The only caveat being that the design needs to be able to properly drive the capacitance that inevitably accumulates with the addition of more and more output transistors.
The way I look at it, the addition of more output transistors helps in three ways:
Increased available power dissipation and less junction temp rise
Increased available Safe Operating Area
Decreased Beta drop-off at high currents (BJTs) (less ft drop as well)
These attributes also allow one to reduce significantly the effect of gm doubling without unduly hurting bias stability. Reduced gm doubling will happen with reduced total effective ballast resistance. But putting a really small ballast resistance on a single BJT is asking for trouble. But putting, say, 8 0.22 ohm ballast resistors on each side of eight output pairs effectively parallels them to get an effective one-sided ballast resistance of only about 0.03 ohms.
This would mean that the worst possible gm variation due to one or both sets of ballast resistors being involved in signal conduction, would be the equivalant of an impedance change of only about 0.015 ohm. This, up against a 4 ohm load, would only create a 0.4% change in static gain as we exit the Class A region into the Class B region. (for this simple example I'm assuming the gm of the transistor devices is significantly larger than 1/0.22 ohm, meaning more than about 100 mA per transistor; I know, this is fairly toasty Class AAB operation, but it makes the numbers easy for the illustration).
Cheers,
Bob
speed vs beta (bjt's)
i used to work for an equipment manufacturer, and we used mj15022's and '23's for the output devices. we had a deal with MOT to preselect our transistors for our amp. apparently one of the parameters that was part of the selection process was that the beta MUST be 75 or less. well, a few years later, we had a new parts buyer, who found transistors from Newark or Allied for about half of what we were paying MOT, so he bought a lot of them. i was the final test tech as well as the customer service tech at the time, and one of the tests was to run the amp up to about 10 watts at 100khz (to simulate somebody hitting the FF button on a casette deck with their volume up). well, a bunch of the production amps began failing this test, blowing fuses and output devices. the problem was common mode conduction (both output devices turned full on at the same time, effectively shorting the + and - rails together). so i began to investigate what was causing this. i checked date codes on the offending output devices against the ones that passed the test. date codes themselves weren't really any help, but the fact that our company name was on the ones that passed, and not on the failing ones was a clue. so i began testing batches of the transistors to see what was different. all of the preselected ones directly from MOT had betas between 50 and 75, while the ones that failed were between 80 and 150. i went to the parts room and pulled the last bin of preselected transistors, and tested them....... all between 50 and 75. so i pulled a bin of the transistors from the new supplier..... about 2/3 of them were above 75. so i went to the buyer and told him that we were buying twice as many transistors for the same price, but only a third of them were useable, and he would have to pay someone to do go-nogo testing on them, definitely not cost effective...... apparently the beta and the junction capacitance are proportional on bjt's, and it does make a difference.
i used to work for an equipment manufacturer, and we used mj15022's and '23's for the output devices. we had a deal with MOT to preselect our transistors for our amp. apparently one of the parameters that was part of the selection process was that the beta MUST be 75 or less. well, a few years later, we had a new parts buyer, who found transistors from Newark or Allied for about half of what we were paying MOT, so he bought a lot of them. i was the final test tech as well as the customer service tech at the time, and one of the tests was to run the amp up to about 10 watts at 100khz (to simulate somebody hitting the FF button on a casette deck with their volume up). well, a bunch of the production amps began failing this test, blowing fuses and output devices. the problem was common mode conduction (both output devices turned full on at the same time, effectively shorting the + and - rails together). so i began to investigate what was causing this. i checked date codes on the offending output devices against the ones that passed the test. date codes themselves weren't really any help, but the fact that our company name was on the ones that passed, and not on the failing ones was a clue. so i began testing batches of the transistors to see what was different. all of the preselected ones directly from MOT had betas between 50 and 75, while the ones that failed were between 80 and 150. i went to the parts room and pulled the last bin of preselected transistors, and tested them....... all between 50 and 75. so i pulled a bin of the transistors from the new supplier..... about 2/3 of them were above 75. so i went to the buyer and told him that we were buying twice as many transistors for the same price, but only a third of them were useable, and he would have to pay someone to do go-nogo testing on them, definitely not cost effective...... apparently the beta and the junction capacitance are proportional on bjt's, and it does make a difference.
Re: speed vs beta (bjt's)
This is very interesting. We often don't think of higher beta in a BJT causing a problem. I'm guessing that it may have been something else that happened to be correlated with the higher beta. Perhaps the ft on these types of BJTs is greater on devices with less beta. Just guessing. Its also conceivable that the new devices were counterfeit. Of course, that doesn't explain the original spec'ing requirement of beta less than 75. It would be interesting to know what led the original designers to that.
If the circuit problem was common mode conduction, a circuit design with a better/higher output transistor turn-off current capability might have been the better approach. Long before common mode conduction causes destruction, it causes a very high level of secondary crossover distortion because of the inadequate turn-off of the output transistors. This would typically show up as fairly high THD-20 at high power.
Bob
unclejed613 said:
This is very interesting. We often don't think of higher beta in a BJT causing a problem. I'm guessing that it may have been something else that happened to be correlated with the higher beta. Perhaps the ft on these types of BJTs is greater on devices with less beta. Just guessing. Its also conceivable that the new devices were counterfeit. Of course, that doesn't explain the original spec'ing requirement of beta less than 75. It would be interesting to know what led the original designers to that.
If the circuit problem was common mode conduction, a circuit design with a better/higher output transistor turn-off current capability might have been the better approach. Long before common mode conduction causes destruction, it causes a very high level of secondary crossover distortion because of the inadequate turn-off of the output transistors. This would typically show up as fairly high THD-20 at high power.
Bob
Hi beta usually makes a better amp, not an unstable one. Low beta should be available at a cheaper price than normal beta devices.
However, it is difficult to make a stable amp with very low value emitter resistors. In the old days (1967), I used 1 ohm. Did I screw up! Then I shifted to .05 ohm, but I had to design a better bias tracking network to do this.
However, it is difficult to make a stable amp with very low value emitter resistors. In the old days (1967), I used 1 ohm. Did I screw up! Then I shifted to .05 ohm, but I had to design a better bias tracking network to do this.
Re: Re: speed vs beta (bjt's)
Hi Bob,
A year or so ago I had a similar experience from purchasing dubious MJ15003/4 transistors over the counter.
I found this was caused by capacitance causing the transistors to start conducting similtaneously at high frequency due to phase shift.
Bob Cordell said:
Hi Bob,
A year or so ago I had a similar experience from purchasing dubious MJ15003/4 transistors over the counter.
I found this was caused by capacitance causing the transistors to start conducting similtaneously at high frequency due to phase shift.
lumanauw said:
While it is true that the voltage drop across RE is the important parameter for optimizing static crossover distortion, this is so ONLY in the context of the given values of RE in the circuit. In other words, this may be the best you can do FOR THAT PARTICULAR RE CHOICE. As a very rough rule of thumb, if you then cut the RE value in half, and re-bias appropriately, you will cut static crossover distortion (and any gm doubling effects) in half.
Also, think about what amount of idle bias would correspond to 25 mV across 1 ohm - only 25 mA. Not a lot going through crossover in the presence of signal swings in the ampere range.
Another thing that gives cause for pause with this kind of biasing is when you realize that 22 mV of bias corresponds to a delta T of junction temperature of only 10C. Without good tracking, this 22 mV could be an elusive value to maintiain in practice under varying conditions of signal swing.
Cheers,
Bob
Re: Re: Re: speed vs beta (bjt's)
Although this may not have been a case of counterfeiting, it is the case that counterfeit parts are becoming a very serious problem in the industry. The matter is now being exacerbated by the European ROHS requirements, incentivising unscrupulous organizations the re-mark parts as being ROHS compliant when they are not. It is also a very serious problem in the memory business, especially in the red-hot FLASH memory business. I know of a case where counterfeit "Toshiba" 2 Gb FLASH memory parts had Samsung die inside them (FLASH memories have a manufacturer's code that can be read from the device, that is how one can tell). No one is immune from the surge of counterfeiting in the semiconductor business. The safest thing to do is to buy only from the few distributors that the legitimate semiconductor manufacturer specifically authorizes.
Cheers,
Bob
Nico Ras said:
Although this may not have been a case of counterfeiting, it is the case that counterfeit parts are becoming a very serious problem in the industry. The matter is now being exacerbated by the European ROHS requirements, incentivising unscrupulous organizations the re-mark parts as being ROHS compliant when they are not. It is also a very serious problem in the memory business, especially in the red-hot FLASH memory business. I know of a case where counterfeit "Toshiba" 2 Gb FLASH memory parts had Samsung die inside them (FLASH memories have a manufacturer's code that can be read from the device, that is how one can tell). No one is immune from the surge of counterfeiting in the semiconductor business. The safest thing to do is to buy only from the few distributors that the legitimate semiconductor manufacturer specifically authorizes.
Cheers,
Bob
john curl said:
John
It seems to me that uncomplementarity of output devices plays some role.
If the follower of one polarity is faster than the other, than one could observe 'dead' crossover zone on rising slope and cross-conduction on lowering slope or vice versa when signal includes fast zero-crossing transients.
One usually assumes slight cross-conduction less harmfull than dead class B transition in highest-fi audio, so over-biasing is sometimes desirable.
I feel myself happy with something like 35mV drop on emitter resistor as gm doubling isn't so terrible at this range.
regards
Adam
Emitor resistor RE has conflicting interest. If we use some output transistors in parrarel, RE is needed for current sharing. Bigger is better. Also for thermal stability bigger is better.
But from GM doubling POV, smaller is better.
A member here (JP) don't use RE, he show that adding RE (any size) makes high order distortion rises.
But how to parrarel some output transistors?
Anyone has use anything besides resistor for RE? For example, uses 8A or 16A schottky (drop about 0V3) in place of RE? Like in this cct (MBR735)
But from GM doubling POV, smaller is better.
A member here (JP) don't use RE, he show that adding RE (any size) makes high order distortion rises.
But how to parrarel some output transistors?
Anyone has use anything besides resistor for RE? For example, uses 8A or 16A schottky (drop about 0V3) in place of RE? Like in this cct (MBR735)