Output Stage.
For the power output I thought about using MJE15030-31, which I used in my high-power amplifiers as drivers. I know they are not made for this, but do you think it will work for this little amp ? The transistors are rated for 8A and 50W. The peak output current in 4 ohms is 3A per transistor, and the average dissipated power per transistor si 6W max. Do you think they will hold ?
Another disadvantage would be that the gain is a bit nonlinear:
NPN hFE: 150 @0.7A / 100 @2A / 75 @3A
PNP hFE: 150 @0.7A / 120 @2A / 83 @3A
Will this matter much ?
I would be interested if you know of other more suitable output transistors, but only in the TO-220 package.
For better hFE linearity use D44H8/D45H8. They are not only flat past 3A, but they maintain full gain to 2 volts. Tailor made for low voltage supplies, when losing a miserable volt or not is a big deal. They have serious second breakdown issues at higher supply voltage, and above +/-18V or so the MJEs should be used. I think they are still available in DPAK, if you wanted to surface mount them too. Real world dissipation with music would be low enough (about half the 6 watt maximum). Can be driven with 2N3904/6 at these voltages too, so there would be a lot of SMD driver choices.
Use the H8 versions, not the H10 or VH versions, so that the gain will tend toward the higher bins. Unnecessary voltage or power handling always comes with a price.
The thread title is mistaken; it's not fully symmetrical thanks to Q9 (the bias spreader). Fortunately, fully symmetrical bias spreader circuits do exist, and work well. See Nelson Pass's US patent # 3,995,228 (November 1976) for a couple examples. Bob Cordell's power amp book exhibits an example too.
FJV1845FMTF and its comp
2N4403 / 1 type are also in SMD
In the spec graphs you'll see the noise measurement is for 1MHz bandwidth not 1Khz like the others. Check for typos, I'm informed the 1MHz measurement is correct. 7dB would in fact be of minimal impact.They are also available SMD. Over 3dB unfortunately. The match very tightly in my SPICE sim which I like about them.
2N4403 / 1 type are also in SMD
Last edited:
For better hFE linearity use D44H8/D45H8. They are not only flat past 3A, but they maintain full gain to 2 volts. Tailor made for low voltage supplies, when losing a miserable volt or not is a big deal. They have serious second breakdown issues at higher supply voltage, and above +/-18V or so the MJEs should be used. I think they are still available in DPAK, if you wanted to surface mount them too. Real world dissipation with music would be low enough (about half the 6 watt maximum). Can be driven with 2N3904/6 at these voltages too, so there would be a lot of SMD driver choices.
I don't really understand what you mean here... (maybe because I'm not a native English speaker).
You said that D44H8/D45H8 are a better option for output power transistors. Looks interesting, I will study their datasheet...
Then you said "They have serious second breakdown issues at higher supply voltage, and above +/-18V or so". Who are you talking about here ? My MJE15030/31 ? Or my drivers BCP56/53 ? I see that MJE340 (MJD340 SMD) is rated at only 10MHz...
Everything always matches perfectly in LTspice, because the same model is used for every device of the same type.They match very tightly in my SPICE sim which I like about them.
Unless you deliberately vary a parameter in the model for each device from the same type you place.
Jan
I don't really understand what you mean here... (maybe because I'm not a native English speaker).
You said that D44H8/D45H8 are a better option for output power transistors. Looks interesting, I will study their datasheet...
Then you said "They have serious second breakdown issues at higher supply voltage, and above +/-18V or so". Who are you talking about here ? My MJE15030/31 ? Or my drivers BCP56/53 ? I see that MJE340 (MJD340 SMD) is rated at only 10MHz...
It’s the D44/45’s that have a second breakdown problem at high voltages. Power handling falls off rapidly above 12 volts. There is a temptation to use the D44’s up at +/-25 or 30V, but but experience shows that they blow up. MJE1503x will actually stand up to it. On +/-18, either runs fine, but the D44’s will let you eke out a half watt more power because they are better in saturation.
The D44’s are actually easier on the drivers, because of that saturation behavior. When you go to turn them on all the way, a 200 mA transistor can supply the needed base current. Any reasonable choice of driver transistor works. I’ve used TO-92 2N3904/6. But if going SMD, I’d opt for a SOT323 over a 23.
MJD340 isn’t a good driver or output transistor here. Being a 300 volt transistor, something’s gotta give. Beta falls rapidly above 10V. So over half of your normal operating range, it’s not going to be very linear. On a 200 (or even 70) volt supply, sub-10-volt behavior doesn’t bother you as much. 10 MHz fT is the least of your worries.
As strange as it may seem, I haven't been able to find any decent low-noise transistors pairs in SMD package. The closest to what I want are the BCM847DS and BCM857DS. They say that the noise figure is 3dB typical. Will this be audible without music ? It is said that only the 10K volume potentiometer generate about 12.9 nV/√Hz, so it will ruin any low-noise transistor anyway... I don't know what to say... If you know a smd package with two NPN and two PNP matched transistors with lower noise than 3dB, that will not cost a fortune and they can be found easely, please let me know.
The noise figure indicates how much the thermal noise of the source impedance and the noise of the transistor together is above the thermal noise of the source impedance on its own, assuming the source has a temperature of 290 K. It's essentially meaningless when you don't specify what the source impedance is and what bias point the transistor is biased in.
In general, I wouldn't worry too much about the noise of the input stage of an audio power amplifier, unless the amplifier's gain is unusually high.
MJE1503x will actually stand up to it. On +/-18, either runs fine
So, after all, my output stage, with BCP56/53 and MJE15030/31 it's ok ?
From my simulations, the biggest V*I product (power dissipation) is at 8V and 1.8A.
For D44's, the DC SOA at 8V is 5A ! And for MJE15's 8A. They both should hold at +/-16V.
In general, I wouldn't worry too much about the noise of the input stage of an audio power amplifier, unless the amplifier's gain is unusually high.
The input stage transistors ar biased at 700uA each, and the gain of the amplifier is 9x (or 19dB). I dont't know what the source impedance is... maybe it is that 1K resistor from the imput.
The BCM's noise figure is specified in this conditions:
2.8dB: VCE = 5 V; IC = 0.2 mA; RS = 2 kΩ; f =10 Hz to 15.7 kHz; Tamb = 25 °C
3.3dB: VCE = 5 V; IC = 0.2 mA; RS = 2 kΩ;f = 1 kHz; B = 200 Hz; Tamb = 25 °C
So, after all, my output stage, with BCP56/53 and MJE15030/31 it's ok ?
From my simulations, the biggest V*I product (power dissipation) is at 8V and 1.8A.
For D44's, the DC SOA at 8V is 5A ! And for MJE15's 8A. They both should hold at +/-16V.
View attachment 1467248
Sure, that driver/output will work. Real world with a speaker load it will hit 1.8A at 16V (30 degrees reactive). You’re plenty safe at low voltages and you can see that as you raise the voltages, the D44 runs out of room sooner. Low voltage behavior is a little better - like when driving 2 ohms on a 12 volt supply. Start pushing 5 or 6 amps peak and that’s when you really see the limitations of 1503x at low volts. Even with 3.6A peak, you will probably be able to push a couple tenths of a volt more output swing with the D44’s. I guess the question is whether you want them badly enough to put together a Mouser order.
Real world with a speaker load it will hit 1.8A at 16V (30 degrees reactive).
How can I simulate someting like a real speaker, 30 degrees reactive, in LTSpice ?
For a single frequency analysis (for power dissipation, etc), make the R+jX (or R-jX) with a resistor and inductor (or capacitor). Set the R and L values to get the phase angle you want. To gauge what a speaker’s impedance curve does to frequency response, stability, or distortion you need a model for it. There is probably a way to import a .ZMA file and use it for the model. I know how to do it in the simulators I usually use (they are not free Spice derivatives, so procedures are different but the idea is the same).
30 degrees is often considered a practical worst case. Woofers will often hit as much as 60 or even 70 degrees at maximum phase deviation but the magnitude is always still high just on either side of resonance. When it gets down to about 30 degrees, the impedance is near the minimum, often only a few tenths of an ohm above that minimum. It usually happens in the upper bass/lower midrange where most of the power is.
30 degrees is often considered a practical worst case. Woofers will often hit as much as 60 or even 70 degrees at maximum phase deviation but the magnitude is always still high just on either side of resonance. When it gets down to about 30 degrees, the impedance is near the minimum, often only a few tenths of an ohm above that minimum. It usually happens in the upper bass/lower midrange where most of the power is.
The input stage transistors ar biased at 700uA each, and the gain of the amplifier is 9x (or 19dB). I dont't know what the source impedance is... maybe it is that 1K resistor from the imput.
The BCM's noise figure is specified in this conditions:
2.8dB: VCE = 5 V; IC = 0.2 mA; RS = 2 kΩ; f =10 Hz to 15.7 kHz; Tamb = 25 °C
3.3dB: VCE = 5 V; IC = 0.2 mA; RS = 2 kΩ;f = 1 kHz; B = 200 Hz; Tamb = 25 °C
As I don't feel like doing all the noise transformations and calculations, rough estimate only:
The noise figure spec is for a single transistor. I should really be doing noise transformations to be sure, but I think you effectively have two groups in AC parallel that each have two transistors in antiseries (differential pair), leading to the same noise figure you would have got with a single transistor.
That 200 uA from the noise figure spec is close to the theoretical collector current for minimum noise figure at 2 kohm source resistance with an hFE of about 290. Your collector current is about 3.5 times as high, but noise optima are so broad that this won't cause more than 1 or 2 dB of degradation.
When you just look at the thermal noise of the resistors when the amplifier is driven from a very low impedance:
Input filter about 1 kohm
Feedback network output impedance about 1 kohm
Emitter resistors: two of 180 ohm in series in one differential pair, but two pairs more or less in AC parallel -> effectively 180 ohm
Thermal noise of 2180 ohm at 25 degrees Celsius: 5.99125 nV/√Hz
Over a 13.5 kHz bandwidth, roughly the noise bandwidth of an A-weighting filter: 696.12 nV
Amplified 9 times: 6.26508 uV
A-weighted noise power into a 4 ohm load: 9.81281 pW, or -110.08207 dB with respect to 1 W
Noise at 1 metre distance in the free field when the loudspeaker has a sensitivity of 100 dB at 1 W, 1 m: -10.08207 dB(A)
Now adding 3 to 5 dB for transistor noise:
-7 dB(A) to -5 dB(A) at 1 metre using an unusually sensitive 4 ohm loudspeaker.
Chances are that in a normal living room, you will have a hard time hearing the noise even with your ear against the tweeter.
@MarcelvdG My god, you are really good at noise stuff ! 😀 Thanks for the calculations !
My speakers are 92dB (both the bass and the tweeter). This means that is even harder to hear the noise ? Or otherwise ?
My speakers are 92dB (both the bass and the tweeter). This means that is even harder to hear the noise ? Or otherwise ?
More sensitive speakers make it easier to hear amp noise. But 92dB, although high by modern standards is still pretty low. Compared to the 105dB of the horn loaded systems of yesteryear, or if driving a 112dB compression driver bi-amped. Most any modern transistor meant for audio small signal stages, if used with normal gain structure, is going to be quiet enough. Tubes were noisier, and that was when those 105dB/W systems were in fashion.
- Home
- Amplifiers
- Solid State
- The smallest audio amplifier for everyday listening (SMD, discrete, fully symmetrical, low power)