Just been doing some simulations of the P101 amp, and wondered why the MJE350 is used for the VAS? It is only running at 5mA, so the dissipation will be between 200 and 300mW for 40 to 60V rails, which is well within the spec of most TO92 transistors.
The simulations indicate that using a TO92 device such as the MPSA92 or 2N5401 give around 10dB lower distortion, due to the lower Ccb.
The simulations indicate that using a TO92 device such as the MPSA92 or 2N5401 give around 10dB lower distortion, due to the lower Ccb.
According to this site, Project 101 - High Power, High Fidelity Lateral MOSFET power amplifier , the high-power version is designed to survive +/- 70 V rails. When the amplifier is driven into clipping, the collector could go negative due to the bootstrapping, so the voltage across the PNP can momentarily exceed 140 V.
During positive clipping with a low-impedance load, you might get some extra current flowing straight from the PNP through the Zener diode to the output.
Anyway, since you have a simulator, you can just simulate what happens during hard clipping at maximum supply voltage with the output open and with the output connected to a low-impedance load. If everything remains within the voltage, current, power and second breakdown (if specified) specs of your favourite TO-92 PNP under all conditions, you can try the TO-92 device.
By the way, the ZTX558 could be interesting. It can handle 1 W at 25 degrees C, its maximum emitter-collector voltage is 400 V and it has a reverse capacitance of at most 5 pF at 20 V. It's in an unusual package that's a bit flatter than a TO-92.
During positive clipping with a low-impedance load, you might get some extra current flowing straight from the PNP through the Zener diode to the output.
Anyway, since you have a simulator, you can just simulate what happens during hard clipping at maximum supply voltage with the output open and with the output connected to a low-impedance load. If everything remains within the voltage, current, power and second breakdown (if specified) specs of your favourite TO-92 PNP under all conditions, you can try the TO-92 device.
By the way, the ZTX558 could be interesting. It can handle 1 W at 25 degrees C, its maximum emitter-collector voltage is 400 V and it has a reverse capacitance of at most 5 pF at 20 V. It's in an unusual package that's a bit flatter than a TO-92.
The background to this is that I am considering the P101 circuit as a candidate for an active speaker application. The plan is to make a single board with two power amps, a PSU and an active crossover on it. It will be run off +/-42V supplies and only ever into 8 ohm drivers, giving around 50W linear per driver (and ~70W at clipping).
I'd like to stick to mosfet output devices, owing to their ease of biasing and ruggedness, so the P101 or the classic Hitachi/Maplin designs are both options. I noted that the Symasym uses TO92 transistors, which prompted this enquiry.
I'd like to stick to mosfet output devices, owing to their ease of biasing and ruggedness, so the P101 or the classic Hitachi/Maplin designs are both options. I noted that the Symasym uses TO92 transistors, which prompted this enquiry.
premium stuff, sajti, very nice suggestion! always better than mje350. I believe you meant KSA1381.
Sorry, I missed that it needs PNP. 2SA1381 of course!
Sajti
For that matter, you could use any of the old CRT video drivers which are typically high linearity, high voltage transistors but the OP wants to know if he can use TO92 transistors which may be cheaper and easier to source.
One point made was that dissipation was well within most TO92 specs. but saying there is only 200-300 mW dissipation is too close to safe limits for my liking and only considers quiescent current. Don't forget that at 10kHz or in a fault condition, peak gate switching current can also be significant.
TO92L (larger package) CRT and audio drivers are also available from China. The pinout should be compatible with TO126 format too, though with most of the sellers, it's doubtful that they are genuine, full spec. components.
One point made was that dissipation was well within most TO92 specs. but saying there is only 200-300 mW dissipation is too close to safe limits for my liking and only considers quiescent current. Don't forget that at 10kHz or in a fault condition, peak gate switching current can also be significant.
TO92L (larger package) CRT and audio drivers are also available from China. The pinout should be compatible with TO126 format too, though with most of the sellers, it's doubtful that they are genuine, full spec. components.
That issue seems to be OK.
I've been playing around with the simulator, making various tweaks and refinements to the circuit, and with the VAS run at 6mA, the current variation at 50W/10kHz out is just +/-0.3mA.
That was using MPSA92 VAS and a MPSA42 current source load. They are spec'd at 625mW dissipation at 25deg.
Clipping is the real concern, where normal operation becomes a fault condition just by overdriving. This and other faults like a shorted output stage or oscillation are inevitable when you tinker with amplifiers. There are more sophisticated protection schemes but the simplest way to deal with them is to accommodate with better safety margins for the earlier stages.
The 25 deg. spec. condition is the sticking point. You can't hold a BJT die at 25C as long as it is dissipating some amount of power. Basic datasheet specifications that use 25C as a refererence temperature may be true for a chip directly mounted on an infinitely large heatsink but it's impractical for TO92 package devices in typical ambient operating conditions. As the device continues to warm up from cold, the margins look narrower when the proper thermal derating is applied. This is true for all transistors and applications but is usually accommodated by referring to the derating graph or using advanced sim. models and programs that do the calculations for you, if you plug in the relevant figures. Perhaps you already have these sim features in use but I haven't assumed so.
As a matter of interest, Mooly uses TO92 transistors for the VAS and gate drivers in his popular Mosfet design. These are 625 mW, 600 mA max. rated 2N5401/2N5551: My MOSFET amplifier designed for music. This is fine but there are differences, starting with the obvious one, that P101 VAS drives the output devices directly with just one transistor and the aid of a bootstrap cap.
I don't think there will be too much harm done if you test small transistors at +/- 43V and full power to a dummy load but I suggest that you allow for fitting a T0126 or TO92L transistor in your board design all the same.
The 25 deg. spec. condition is the sticking point. You can't hold a BJT die at 25C as long as it is dissipating some amount of power. Basic datasheet specifications that use 25C as a refererence temperature may be true for a chip directly mounted on an infinitely large heatsink but it's impractical for TO92 package devices in typical ambient operating conditions. As the device continues to warm up from cold, the margins look narrower when the proper thermal derating is applied. This is true for all transistors and applications but is usually accommodated by referring to the derating graph or using advanced sim. models and programs that do the calculations for you, if you plug in the relevant figures. Perhaps you already have these sim features in use but I haven't assumed so.
As a matter of interest, Mooly uses TO92 transistors for the VAS and gate drivers in his popular Mosfet design. These are 625 mW, 600 mA max. rated 2N5401/2N5551: My MOSFET amplifier designed for music. This is fine but there are differences, starting with the obvious one, that P101 VAS drives the output devices directly with just one transistor and the aid of a bootstrap cap.
I don't think there will be too much harm done if you test small transistors at +/- 43V and full power to a dummy load but I suggest that you allow for fitting a T0126 or TO92L transistor in your board design all the same.
Just checked on the thermal specs for the MPSA42/92, and it looks bad. The junction to ambient spec is 200deg/W, which would give a 50deg rise at 250mW, and therefore a 75deg operating temp. The derating curve shows that the dissipation spec drops to 375mW at 75deg, leaving barely any margin.
A TO126 packaged device would be a wise move! In my active speaker application, the most likely failure mode would be a shorted driver, so there could be some high VAS currents via the gate protection zeners in that situation (at least until the output fuse blows).
The KSA1381/KSC3503 combo looks good, although not very widely available.
I think you are mixing up junction and ambient temperatures and are implicitly adding the self-heating twice. Ian Finch doesn't make a clear distinction between ambient and junction temperatures either.
Suppose the worst-case ambient temperature is 50 degrees C. With 50 degrees of self-heating, the junction temperature then becomes 100 degrees C. As long as that's less than the transistor's specified maximum junction temperature, everything is OK - that is, the transistor will then meet the manufacturer's reliability targets.
Equivalently you can look up the allowable dissipation at 50 degrees ambient temperature by looking at the derating curve. The sloped part of that curve is simply the difference between maximum junction temperature and ambient temperature divided by thermal resistance.
If you want to improve reliability beyond the manufacturer's targets, you can derate the maximum junction temperature a bit. The manufacturers usually don't specify what their reliability targets are and what the activation energy of the worst failure mechanism is, so it is difficult to say whether this is needed and how much it helps.
Suppose the worst-case ambient temperature is 50 degrees C. With 50 degrees of self-heating, the junction temperature then becomes 100 degrees C. As long as that's less than the transistor's specified maximum junction temperature, everything is OK - that is, the transistor will then meet the manufacturer's reliability targets.
Equivalently you can look up the allowable dissipation at 50 degrees ambient temperature by looking at the derating curve. The sloped part of that curve is simply the difference between maximum junction temperature and ambient temperature divided by thermal resistance.
If you want to improve reliability beyond the manufacturer's targets, you can derate the maximum junction temperature a bit. The manufacturers usually don't specify what their reliability targets are and what the activation energy of the worst failure mechanism is, so it is difficult to say whether this is needed and how much it helps.
Ah yes, you are right. The derating curve is against ambient not junction temp, so the thermal resistance is already included. The dissipation spec would then be around 500mW for 40-50deg ambient (inside my speaker cabinets), more than twice my expected quiescent dissipation.
Since you live in UK why not buy decent transistor like 2SB649A from Profusion Plc? Even Hitachi in their application note for Lateral fet amps recommed it for high power versions. 2SA1381 would be much more difficult to find. 2SA1209 would be impossible to find but it's probably idel for the job since it's purpose made for audio frequency amps predriver applications.
All TO92L and most TO126 are long gone. A couple KSA/KSC Fairchild types remain, but don’t expect them much longer. The MJE and BD13x’s still seem to be all over the place but they just aren’t as good. Most MJE 350 are way better than 10 MHz, but 50 seems to be where they top out and Hfe is lower and more variable than real VAS transistors. Still a good go-to device for an EF3 predriver and ok-ish for VAS if you’re not trying for single digit ppm THD20k.
It is kind of a shame about the TO92L going away. Relatively small dies with a 1 watt dissipation was very useful for a lot of things. But it’s a sign of the times. And not even very many of the 92L and 126 devices were converted to surface mount equivalents, either. I guess you’re just supposed to use 3886 chip amps for stuff below 60 watts or so and an IRS 2092 and a couple hexfets for anything above that.
It is kind of a shame about the TO92L going away. Relatively small dies with a 1 watt dissipation was very useful for a lot of things. But it’s a sign of the times. And not even very many of the 92L and 126 devices were converted to surface mount equivalents, either. I guess you’re just supposed to use 3886 chip amps for stuff below 60 watts or so and an IRS 2092 and a couple hexfets for anything above that.
Profusion's general purpose transistors are now copies but usually quite good ones from Unisonic, who at least identify their products and publish datasheets. A similar full-pack part is H649A and there are plenty of B647A generic copies around, which are a TO92L version that might suit you better. However, check the Cob of some of these identifiable brand BJT drivers - 27pF and even more, which isn't considered ideal for VAS applications. They are essentially drivers for a single pair of BJT power transistors.
Much of the sound character of amplifiers is determined by the VAS design and transistor characteristics. Many popular designs that are still considered very good sounding amplifiers, don't even compare technically with today's topline consumer products but If you like the P101's sound quality or at least the idea of it, I would think twice about changing from MJE350 or similar, without a compelling reason. For an active speaker system, I would instead, be trying to avoid sound character issues. Lower Cob, better linearity devices could well become necessary after carefully listening to the end results using different amplifiers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.