HI Glen
Last time I calculated the input capacitance of a bipolar gave a very interesting comparison. Every bipolar is different, and MOSFETS have a vast range of amps/volts so the comparison I made was just a snapshot and no way exhaustive.
The net result of my comparison was that the charge was similar between the two. In the case of a MOSFET the charge is a few nF spread out with a couple of volts, in the case of a bipolar the base charge was several tens of nF - but at about 1/10 the voltage swing.
I'm sure different transistors will give different charges but this shows that on one hand MOSFETS need large gate drives which some early circuits seem to imply isn't the case.
I don't know enough about the latest high fT transistor constructions to develop a suitable model. Anyone know the build of these? As a guess the emitters would be "hollow" centred types with perhaps a little current-spreading resistor loading; the bases must be relatively thin and the collector design probably is the key to the second-breakdown performance. Not the base, then, that RCA always said ...?
Cheers
John
Last time I calculated the input capacitance of a bipolar gave a very interesting comparison. Every bipolar is different, and MOSFETS have a vast range of amps/volts so the comparison I made was just a snapshot and no way exhaustive.
The net result of my comparison was that the charge was similar between the two. In the case of a MOSFET the charge is a few nF spread out with a couple of volts, in the case of a bipolar the base charge was several tens of nF - but at about 1/10 the voltage swing.
I'm sure different transistors will give different charges but this shows that on one hand MOSFETS need large gate drives which some early circuits seem to imply isn't the case.
I don't know enough about the latest high fT transistor constructions to develop a suitable model. Anyone know the build of these? As a guess the emitters would be "hollow" centred types with perhaps a little current-spreading resistor loading; the bases must be relatively thin and the collector design probably is the key to the second-breakdown performance. Not the base, then, that RCA always said ...?
Cheers
John
john_ellis said:
Hi John,
I tend to agree. On the average, the Cgd or Cbc between MOSFETs and Bipolars is not grossly different.
When I calculate the hybrid pi capacitance for a bipolar from ft, I see a large capacitance but, as you mentioned, a small voltage swing. Indeed, the difference in gm from bipolar to MOSFET tends to take this voltage swing difference into account. In very rough terms, the gm of a MOSFET is about 1/10 that of the gm of a bipolar at the same operating current.
Take a 30 Mhz ft bipolar at 1 amp, where gm is 40 and wt = 1.9e6. The hybrid pi capacitance comes out to 0.22 uF, or 220,000 nf (did I get that right?).
Now for comparison, calculate the effective ft of a MOSFET at 1 amp, where its gm is 4.0 and its Cgs is 5000 nF. It is 800 Mhz. Wow! Even more than I thought. Did I get that right? I was expecting a couple hundred MHz from my recollections.
Anyway, we know this is not the whole story, but it is one way of understanding one way in which MOSFETs are fast.
Bob
Bob Cordell said:
Yes, that looks right, and for a 4MHz transistor the value goes well above a uF.
Plugging in the values, I’ve been surprised at just how big Cib is, but I’ve also been surprised just how much it’s effective value is reduced in the emitter follower configuration, particularly at low Ic.
When building complementary BJT output stages, one method of improving large signal linearity is to use more parallel pairs than is necessary to avoid SOA limitations, so as to avoid beta droop at high peak currents by sharing the load current between lots of devices.
This makes for an interesting comparison between such a stage constructed with 4MHz devices and with high fT devices. When you parallel up lots of pairs, the proportion of the output stage’s complete input capacitance provided by each BJT’s Cib reduces, but the contribution of Cob for each BJT remains the same.
One thing I found interesting is that Cob has not been effectively reduced in the high fT devices. Under the same test conditions, a 60MHz NPN Sanken has the same Cob spec as a 4MHz MJL21194 of 250pF, and its PNP complement matches the MJL21193 at 500pF.
The more pairs connected in parallel to share the load current, the more the contribution of Cib goes down in proportion to Cob, and since the high fT devices don’t have an advantage over the low fT devices with respect to Cob, the reduction in net input capacitance obtained by using the high fT devices can be less than one may expect.
Cheers,
Glen
G.Kleinschmidt said:
I agree completely, and many of these observations map to MOSFETs as well.
Paralleling lots of output devices has many benefits, but one does have to watch out for the increased load capacitance placed on the drivers. And, yes, for the bipolars, the Cib is not only bootstrapped to a large degree, but also it does not tend to multiply as much as devices are paralleled, since each of the paralleled devices is operating at a reduced current and hence exhibits less Cib.
Cheers,
Bob
Hi Bob, Glen
In a so-called emitter-follower output, most amplifier circuits use a high impedance voltage amplifier stage to the driver/output pairs. In this configuration the output transistors cannot really be said to be operating in EF mode. However, to the input, I agree that the current swing needed in a BJT stage is "stretched" out over the higher voltage swing so it appears to be a lower capacitance, but the current swing remains the same.
Glen is also correct in that the base charge depends on the current flow, so at lower currents the base charge falls off very quickly. The question then is with parallelled transistors whether the effective capacitance is any different from a single pair. Although each transistor runs at a lower current, the individual capacitance is lower, but this has to be multiplied by the number of transistors in parallel to get the total. Probably the net charge for parallelled transistors remains more or less constant but Cob increases.
Ironic perhaps that the high fT transistors have flatter gain characteristics than the MJ21193/4 (which aren't bad) but once they pass the peak, fall quickly, so parallelled devices are more necessary than for MJ21193/4.
cheers
John
In a so-called emitter-follower output, most amplifier circuits use a high impedance voltage amplifier stage to the driver/output pairs. In this configuration the output transistors cannot really be said to be operating in EF mode. However, to the input, I agree that the current swing needed in a BJT stage is "stretched" out over the higher voltage swing so it appears to be a lower capacitance, but the current swing remains the same.
Glen is also correct in that the base charge depends on the current flow, so at lower currents the base charge falls off very quickly. The question then is with parallelled transistors whether the effective capacitance is any different from a single pair. Although each transistor runs at a lower current, the individual capacitance is lower, but this has to be multiplied by the number of transistors in parallel to get the total. Probably the net charge for parallelled transistors remains more or less constant but Cob increases.
Ironic perhaps that the high fT transistors have flatter gain characteristics than the MJ21193/4 (which aren't bad) but once they pass the peak, fall quickly, so parallelled devices are more necessary than for MJ21193/4.
cheers
John
john_ellis said:
Here's something to keep in mind. In a bipolar design with a triple Darlinton output and fairly high-ft output devices, there is a lot of current gain between the output of the VAS and the load - maybe on the order of 50 X 50 X 50 out to a MHz. At the same time, the output impedance of the VAS falls to quite a low value at high frequencies as a result of the shunt feedback action of the Miller compensation. This, combined with the beta-squared of the two Darlinton-connected driver transistors, can result in the output transistor actually having a decently low driving impedance. Thus, in many cases of good design, they probably are acting like emitter followers.
Cheers,
Bob
Bob Cordell said:
In the high power output bipolar stages I was considering, with lots of output pairs in parallel, Darligton connected driver pairs (or similar) are pretty much mandatory, unless you want to run a really hot VAS.
Driver transistors with 30MHz fT's are common, so there is no reason why a high power output stage using 4MHz BJT's cannot be provided with a sufficently low drive impedance.
Cheers,
Glen
G.Kleinschmidt said:
In the high power output bipolar stages I was considering, with lots of output pairs in parallel, Darligton connected driver pairs (or similar) are pretty much mandatory, unless you want to run a really hot VAS.
Driver transistors with 30MHz fT's are common, so there is no reason why a high power output stage using 4MHz BJT's cannot be provided with a sufficently low drive impedance.
Cheers,
Glen
Yes, that makes sense to me. In this regard, it might be interesting to contrast the use of BJTs with MOSFETs for use as the drivers for the output BJTs.
But here is another question: Given the availability of 30-60 MHz ring emitter BJTs from Sanken, ONSEMI, etc, why not use them as opposed to the 4 MHz BJTs? It it because of price? Or ruggedness?
Cheers,
Bob
Bob Cordell said:
Please define ruggedness.
About a year ago i tested the R from case to sink of a Sanken BJT in the big MT200 package, used aluminium oxide insulators from Aavid Thermalloy which i adapted to the size of the Japanese device.
Measure the effective cooling surface of a TO-3 and an MT200, they're close to identical. However, the geometry and construction of the Japan package is superior in keeping the contact surface flat. Additionally, TO-3 casing material quality is rather poor. Picture the warping of a metal surface if you raise its temperature by 100F.
I've got a transmission oil cooler hooked up to a TH700 automatic tranny. Many would be surprised how many lbs-ft torque an auto transmission handles with proper cooling. I can post a picture of what a clutch plate assembly looks like after running without cooling for a brief period.
A lot of gents post calculations on SOA curves (peak measurements made on a device that's not attached to a heatsink), but cooling issues seem less interesting.
Attachments
Bob Cordell said:
G’day Bob.
Greater freedom from parasitic oscillation? Price – definitely. I wouldn’t agree with ruggedness though. There are some beefy 4MHz devices, but the SOA of the most expensive Sanken BJT’s aren’t lacking over the lower fT devices.
I stocked up on MJL21193/MJL21194's way back when Motorola was still making them, and I'll continue using them until my stock is depleted (and probably afterwards too). At the time they were readily available and were less than half the price of the 30MHz devices.
The MJL21193/MJL21194 is still a pretty popular output pair, both for HiFi amps and PA amps. I guess this may be due, to a degree, to price. From another thread I’ve learnt that Bryston uses these BJT’s exclusively in their top-of-the-line ‘SST’ series amplifiers. From the few reviews of these amps I’ve looked at, they look to measure quite well, especially in terms of THD out to 20kHz.
Considering the amount of money Bryston charges for these amps though, one would think that they could afford the fancier BJT’s. I think that it may just boil down to the fact that 4MHz devices are good enough for the application.
Cheers,
Glen
john_ellis said:
I also try not to use Miller compensation on the VAS, but many other forms of compensation also effectively reduce the output impedance of the VAS at high frequencies. The form of compensation that I use in my MOSFET power amplifier with error correction also reduces the output impedance of the VAS at high frequencies, for example.
Cheers,
Bob
Bob Cordell said:
But why .....Some Hexfets such as IRF240/920 are also available in TO-3 Metal Package...I have used IRF250 alot in metal Packages
Hi Bob
my development 100W amp uses TO-3 MJ21193/4 - they have better thermal ratings than the plastic (250W instead of 200W). They also have reasonably good performance. (10 kHz THD <0.003%)
But last time I looked MOSFETS in TO-3 were priced for going out of fashion...
cheers
John
my development 100W amp uses TO-3 MJ21193/4 - they have better thermal ratings than the plastic (250W instead of 200W). They also have reasonably good performance. (10 kHz THD <0.003%)
But last time I looked MOSFETS in TO-3 were priced for going out of fashion...
cheers
John
Workhorse said:
They are very expensive and offer little in return for the extra price. You can buy more than two TO247 devices for the price of a TO3, and get twice the SOA in the bargain.
Bob
Bob Cordell said:
If we ignore the price tag, then TO-3 has a "STEEL" casing while TO-247 is just a mix of copper n plastic...BOB, TO-3 is a LUXURY item in Mosfet reign and there are only few amps on this planet who afford this piece of Steel. Their Heat Dissipation capability is much greater and they are very easier to mount on heat sink provided no mica insulators used.
The Steel Mosfet is a Mosfet with Muscle !😀 😀 😉
The "Shining Chrome Thing" claiming its superiority over dull black plastic.