estuart said:
OK Edmond, I'll take you up on your offer
. THD20 @ 1V below clipping (15W in this case) is a pretty stringent test. What do you get for THD20 at the rated power output of 12Wrms (19.6Vpp) into 4 ohms and at half power (13.8Vpp) into 4 ohms? Cheers,
Glen
EDIT: I think I see an error in your schematic - the junction of D1 and D2 seem to be connected to the input!
G.Kleinschmidt said:
OK Edmond, I'll take you up on your offer
Cheers,
Glen
EDIT: I think I see an error in your schematic - the junction of D1 and D2 seem to be connected to the input!
Hi Glen,
Here are the THD's. Notice that my previous figures were based on C5/6 tied to the emitter of the 1st pre-driver. If c5/6 tied to the VAS-output, THD is slightly higher. See below.
C5/6 -> e-Q30
Vo = 13.8V THD20 = 5.3ppm
Vo = 19.6V THD20 = 12.2ppm
C5/6 -> VAS-output
Vo = 13.8V THD20 = 5.7ppm
Vo = 19.6V THD20 = 12.8ppm
As for D1/2, you are right, but don't worry, this schematic was only a "cleaned up" version. For simulation I used a correct version.
I forgot to mention one important thing: without series base resistors for the O/P devices (R52/53), my sim reveals oscillations at 40MHz
We all know that sims are not always accurate, most of the time they produce flattered results. So, if a sim reveals a problem, this should be taken very seriously.
So, I'm afraid you have to modify your PCB lay-out. In that case you might consider to make also provision for the alternate Miller compensation (also add R32/33), see Bob's comments on this topic
http://www.diyaudio.com/forums/showthread.php?s=&threadid=94676&perpage=10&pagenumber=52 post # 518
(and of course provisions for TMC in case you changed your mind
)Cheers,
mikeks[/i][B] Just use good tone controls said:
I don't think Mikeks is joking, as he seems to be a fan of D. Self, who once wrote:
"I think tone controls are absolutely necessary, and it is a startling situation when, as frequently happens, anxious inquirers to hi-fi advice columns are advised to change their loudspeakers to correct excess or lack of bass or treble. This is a extremely expensive way of avoiding tone controls" (EW, sept., 1996, p709)
And I fully agree with this. Moreover, as some hi-end guru's (ahem) advocate heavily distorting amps, because of their superior sonic performance, I don't see what's wrong with manipulating the sound a little bit more and in a far less devastating fashion by means of a tone control.
, and the results were pretty close to the real thd figures. estuart said:
Hi Glen,
Here are the THD's. Notice that my previous figures were based on C5/6 tied to the emitter of the 1st pre-driver. If c5/6 tied to the VAS-output, THD is slightly higher. See below.
C5/6 -> e-Q30
Vo = 13.8V THD20 = 5.3ppm
Vo = 19.6V THD20 = 12.2ppm
C5/6 -> VAS-output
Vo = 13.8V THD20 = 5.7ppm
Vo = 19.6V THD20 = 12.8ppm
As for D1/2, you are right, but don't worry, this schematic was only a "cleaned up" version. For simulation I used a correct version.
I forgot to mention one important thing: without series base resistors for the O/P devices (R52/53), my sim reveals oscillations at 40MHz
We all know that sims are not always accurate, most of the time they produce flattered results. So, if a sim reveals a problem, this should be taken very seriously.
So, I'm afraid you have to modify your PCB lay-out. In that case you might consider to make also provision for the alternate Miller compensation (also add R32/33), see Bob's comments on this topic
http://www.diyaudio.com/forums/showthread.php?s=&threadid=94676&perpage=10&pagenumber=52 post # 518
(and of course provisions for TMC in case you changed your mind
Cheers,
Fantastic!
Edmond, you've earned yourself a Heineken 
At the rated power output, the THD figures are looking much better. I figured they would
I did notice the connection of the compensation capacitors being taken from the first emitter followers, but this didn’t really bother me because, as I think your second set of results confirm, the benefit from doing such is rather slim indeed - so the results are still quite accurate either way. This is a modification to the traditional miller compensation scheme that is more than bit overrated, IMHO.
The oscillation problem with the output transistors in you simulation is due, I think, to the fact that you are using models for 30MHz fT transistors. The MJL21193/4 transistors I specified only have an fT of 4MHz. In my hands, these suckers have been practically immune to oscillations in all but the most dodgily built prototype amplifier. I’ll be building my actual amplifiers with these devices, as I’ve got a couple of tubes of ‘em lying around. However, in order to make the design more flexible, I’ll update the schematic and the PCB for the base stopper resistors anyway. This is simple enough. That will make it compatible with transistors such as the 30MHz fT MJL3281A/MJL1302A pair, also available in TO-264.
Incidentally, a lot of people scoff at 4MHz transistors such as the MJL21193/4, but the benefits gained, in my experience, for using those fancy 30MHz or greater devices are far from proportional to the difference in fT. A lot of people are under the delusion that a 30MHz audio transistor will automatically switch 7.5 times faster than a 4MHz one. Not so! The improvement in linearity achieved by using the higher fT transistors is less pronounces in class A operation also, due to the absence of crossover switching.
As a matter of fact, this is a really good, on-topic thing to be discussing here, as it concerns the performance potential of high power audio BJT’s. Hmmmmm..........Now we have the simulated results of my design using 30MHz devices as a reference. Hmmmm........Now if only a particular someone with a rather fancy simulation package would be willing to go to the ON-semi website......Hmmm.....and then perhaps download the spice models for the MJL21193/4 transistors and then....................
Hint........Hint.........
Cheers!
Glen
Hi, Glen,
This is very interesting.
A question : what is the "better" result if we use 20-30mhz ft output transistor compared if we use 4-5mhz ft output transistors? In what POV the faster one is better? Xover distortion? Or what? Many well regarded (and robust) commercial amps are indeed using MJL21193/4 until this day in 2007 (while output transistors with 30mhz are easily available).
Are they really differ in sonics result?
Generally speaking, fast is good, slow is bad.
The very first practical complementary power transistors were introduced by Motorola, 40 years ago. These were the 2N3716-3792 series of TO-3 power transistors. They had an F(t) of 4MHz. This series of devices and their improved successors would give pretty good performance, but you had to use output coil buffering, or else overcompensate the circuit and reduce the slew rate to just a few volts/microsecond. I know this from experience.
Higher voltage devices that came from the same design series were even slower at about 2MHz. These devices could be driven to 50V/us, but no more, seemingly no matter what you drove them with.
When ring emitter transistors were introduced, we jumped up to 30MHz and more. High slew rate became easy and the same drive circuit could make 500V/us slew rate or more, with the faster devices. However, a small output coil was still necessary.
In order to remove the output coil, that was typically 2uH, but can be reduced to .5uH with effort in some cases, we had to slow down the amps from 500 to perhaps 150V/us.
This is a reasonable trade, I would think, as the coil can be problematic with some loads, as it will ring with a large cap load.
Today, we achieve over 100V/us slew rate without a coil in production power amps of many 100's of watts, using bipolar transistors with F(t) of 30MHz or more. I call that progress, but then that is just my personal opinion.
The very first practical complementary power transistors were introduced by Motorola, 40 years ago. These were the 2N3716-3792 series of TO-3 power transistors. They had an F(t) of 4MHz. This series of devices and their improved successors would give pretty good performance, but you had to use output coil buffering, or else overcompensate the circuit and reduce the slew rate to just a few volts/microsecond. I know this from experience.
Higher voltage devices that came from the same design series were even slower at about 2MHz. These devices could be driven to 50V/us, but no more, seemingly no matter what you drove them with.
When ring emitter transistors were introduced, we jumped up to 30MHz and more. High slew rate became easy and the same drive circuit could make 500V/us slew rate or more, with the faster devices. However, a small output coil was still necessary.
In order to remove the output coil, that was typically 2uH, but can be reduced to .5uH with effort in some cases, we had to slow down the amps from 500 to perhaps 150V/us.
This is a reasonable trade, I would think, as the coil can be problematic with some loads, as it will ring with a large cap load.
Today, we achieve over 100V/us slew rate without a coil in production power amps of many 100's of watts, using bipolar transistors with F(t) of 30MHz or more. I call that progress, but then that is just my personal opinion.
G.Kleinschmidt said:
Fantastic!
At the rated power output, the THD figures are looking much better. I figured they would
I did notice the connection of the compensation capacitors being taken from the first emitter followers, but this didn’t really bother me because, as I think your second set of results confirm, the benefit from doing such is rather slim indeed - so the results are still quite accurate either way. This is a modification to the traditional miller compensation scheme that is more than bit overrated, IMHO.
The oscillation problem with the output transistors in you simulation is due, I think, to the fact that you are using models for 30MHz fT transistors. The MJL21193/4 transistors I specified only have an fT of 4MHz. In my hands, these suckers have been practically immune to oscillations in all but the most dodgily built prototype amplifier. I’ll be building my actual amplifiers with these devices, as I’ve got a couple of tubes of ‘em lying around. However, in order to make the design more flexible, I’ll update the schematic and the PCB for the base stopper resistors anyway. This is simple enough. That will make it compatible with transistors such as the 30MHz fT MJL3281A/MJL1302A pair, also available in TO-264.
Incidentally, a lot of people scoff at 4MHz transistors such as the MJL21193/4, but the benefits gained, in my experience, for using those fancy 30MHz or greater devices are far from proportional to the difference in fT. A lot of people are under the delusion that a 30MHz audio transistor will automatically switch 7.5 times faster than a 4MHz one. Not so! The improvement in linearity achieved by using the higher fT transistors is less pronounces in class A operation also, due to the absence of crossover switching.
As a matter of fact, this is a really good, on-topic thing to be discussing here, as it concerns the performance potential of high power audio BJT’s. Hmmmmm..........Now we have the simulated results of my design using 30MHz devices as a reference. Hmmmm........Now if only a particular someone with a rather fancy simulation package would be willing to go to the ON-semi website......Hmmm.....and then perhaps download the spice models for the MJL21193/4 transistors and then....................
Hint........Hint.........
Cheers!
Glen
Glen,
I think you make a good point about a Class A design not needing as much a high ft output transistor because of the absence of crossover. This will be so as long as you are sure to provide enough turn-off current to the output transistor for the maximum rate of change of output current that will ever be needed. This needed amount of turn-off current will be greater with a 4 MHz ft transistor than with a 30 MHz ft transistor.
With regard to feedback loop stability, I believe that the substitution of a faster transistor in a stage should never cause instability. If it does, it is a sign that you don't have enough stability margin in your compensation, and that to some extent your compensation is depending on slowness of transistors, which can vary a lot, both from transistor to transistor and with current and voltage on the transistor.
Cheers,
Bob
Bob Cordell said:
Hi Bob,
I suppose you are talking about the global feedback stability. If that's the case, it has nothing to do with the kind instabilty that I have reported to Glen. They were oscillations around 40MHz, and they disapeared when base stopper resistors in the O/P stage were inserted (or when the series inductunces of the leads to these devices was reduced to zero, which is impossible in practice).
BTW the unity gain frequency of the global FB loop was 1.2MHz.
Anyhow, the instabillty occured inside the Miller loop or inside the output stage and, probably, this cannot be cured by chosing just an other Miller cap.
If you had an other mechanism in mind, I'm eager to learn about it.
Regards, Edmond.
To All,
To avoid unnecesarry discussions, allways specify exactly which feedback loop.
Bob Cordell said:
Completely agree, and as Mr. Curl pointed earlier, faster y better in a general sense.
Just to add a dime, note that faster transistors push the output stage pole forward, in turn allowing for a higher frequency dominant pole (or higher low frequency OL gain for the same 6 dB/oct. rolloff), meaning there is a larger high frequency feedback correction factor available. Of course this makes sense only for global negative feedback topologies.
Rodolfo
ingrast said:
Bob Cordell said:
Hi all. As Edmond has already pointed out, the 40MHz was parasitic oscillation of the output devices cured with base stopper resistors, not global loop instabliity.
Since I am publishing this design on the net, I will optimise the miller compensation of this design for the 4MHz devices and include base stopper resistors. That way, if anyone want's to copy the design, they can bung in 30MHz transistors without a problem for a modest improvment in high frequency lineaity. And if they are feeling adenturous, and want to get more out of the design with high fT transistors, they can then tweak the compensation themselves.
With regards to 30MHz transistors being better than 4MHz ones, I agree completely. But the performance differences when applied in typical audio circuits are not as great as some expect. D. Self has demonsatrated what can be achieved with lowly low-fT transistors with his 'Blameless' - A very basic topology that achieves in the vicinity of 0.01% THD+N at 20kHz and ~0.001% THD+N < 1kHz - with 2MHz fT output devices - definately "Hi-Fi", IMO.
Cheers,
Glen
lumanauw said:
Hi lumanauw.
30MHz is better than 4MHz for reasons others have already mentioned here, but the sonic performance of an amplifier ultimately depends on much more than the devices selected for the output stage.
I contend that one can make a perfectly adequate HiFi or PA amplifier with 4MHz devices and a horribly inadequate one with 30MHz devices, and vice versa.
Cheers,
Glen
While BJT's have non linear gain if you have a good class A driver for them it will work just fine. My drivers can supply 100ma of current to the darlington pair on the output. Even at full current output this is more that enough. MOSFETs at low currents are nonlinear and in my experiences are not to be used in audio because of the large current required to charge the gate at sufficient speeds. My current amp can drive a 2 ohm load to 240W without a problem. It uses BJT's and with the heatsink i have they don't even get hot.
This is my amp design. Let me know what ya think. I know it can use improvements.
http://byrne.isa-geek.com/brendan/amp.bmp
This is my amp design. Let me know what ya think. I know it can use improvements.
http://byrne.isa-geek.com/brendan/amp.bmp
IrishboyM4 said:
Its your own specific view.....The non-linearity could be solved by implementing/adopting some good techniques such as Transconductance Gain cell, Error-Correction..
For driving the mosfet you need sufficient gate current which is given by Ig=F X Qg, where F is operating frequency and Qg is total gate charge... and this value of Ig is much less when compared to the base current required for driving Bipolars at same frequency, which require much larger magnitude of base current in proportion to mosfets...