Re: Re: Re: Re: weak inversion
Of course you can have a look at it, but not right now. I rather prefer to wait for the final results of the PCB version to make sure the design is bug free.
Also, I need Ovidio's permission to show his part of the job: the ouput stage + error correction (I did the front-end: input stage, NDFL stage, VAS and common mode control loop)
So, please have some patient for a while.
I'm sorry for your block, I hope it'll be repaired soon.
Cheers, Edmond.
G.Kleinschmidt said:Ok Edmond, sounds good. Do you guys plan to show the finished circuit? I’d like to see it!
I’ll be stripping my engine down tomorrow morning, but it looks like I’m going to have a weekend without much to do (can’t really go anywhere until my motor is back together, and the machine shop isn't picking it up the block till Wednesday).
I hope to get a major update on my web site for my 12W and 1kW amp projects done.
I’ve got a 48-transistor preamp and tone control circuit, along with a 8 transistor power supply to add to the 12W amp page for a start!
Cheers,
Glen
Of course you can have a look at it, but not right now. I rather prefer to wait for the final results of the PCB version to make sure the design is bug free.
Also, I need Ovidio's permission to show his part of the job: the ouput stage + error correction (I did the front-end: input stage, NDFL stage, VAS and common mode control loop)
So, please have some patient for a while.
I'm sorry for your block, I hope it'll be repaired soon.
Cheers, Edmond.
Regarding several posts before, I have a couple of questions for Mr. Hanson and Mr. Curl.
Both of you use low value base stopper resistors on your BJT outputs. (~1 to 5 ohms) I am sure you have good reasons why you do. Why these values? What are the benefits and drawbacks to values above and below these values. I see some designs with no base stoppers?
Whatever the reasonings that apply to the question above probably applies the the choice of base stopper on a BJT driver also. If so what range of values give you the best solution.
Regarding BJT's in the driver stage, do you like to tie the emitters to the output node or just tie them together? What are the different tradeoffs. Also is there a way to determine the best value/s for the emitter resistor/s?
Thanks, DonS
Both of you use low value base stopper resistors on your BJT outputs. (~1 to 5 ohms) I am sure you have good reasons why you do. Why these values? What are the benefits and drawbacks to values above and below these values. I see some designs with no base stoppers?
Whatever the reasonings that apply to the question above probably applies the the choice of base stopper on a BJT driver also. If so what range of values give you the best solution.
Regarding BJT's in the driver stage, do you like to tie the emitters to the output node or just tie them together? What are the different tradeoffs. Also is there a way to determine the best value/s for the emitter resistor/s?
Thanks, DonS
Base stopper resistors give a REAL resistance to counter the NEGATIVE resistance caused by the follower phase shift and a cap load. Negative resistance means oscillation. How much R? It should be as small as practical. I use 10 ohms, but I hope to use less in future.
I never needed this resistor in my earlier, low F(t) output devices (4MHz).
I never needed this resistor in my earlier, low F(t) output devices (4MHz).
estuart said:
Experimentally it was concluded that the MOSFET subthreshold current follows an exponential law, with a slope (defined as the voltage required to increase the current 10x) S=n*kT/q*ln(10) where n is a factor >1. If n=1 then S=60mV/decade which is the theoretical minimum subthreshold slope.
It was concluded after mucho debate in the 70's and early 80's that the MOSFET subthreshold current is essentially a diffusion curent (there are two more components that for the purpose of this discussion can be neglected). A MOSFET in subthreshold conduction has an energy band diagram that looks like the base of bipolar transistor, so it is natural that the MOSFET currents look and smell like BJT.
A very basic subthreshold model shows that for Vds>>kT/q the subthreshold current is proportional to W/L (the transistor geometry), the electron (or hole) mobility, the squared thermal voltage kT/q and the equivalent bulk capacitance Cb (which in turn depends on the transistor substrate donor/acceptor concentration). The exponential factor n (which ultimately explains the slope variation) is n=1+Cb/Cox. As mentioned, Cb depends on the substrate donor/acceptor concentration (Na or Nd), while Cox dependes (to a geometry factor) only to the gate oxide thickness. Therefore:
Xox decrease -> Cox increase -> n decrease -> sharper subthreshold
Na or Nd increase -> Cb increase -> n increase -> softer subthreshold.
also:
T increase -> softer subthreshold.
The subhtreshold current is also importand because it defines the Ioff current of the MOSFET device (that is, the current at Vgs=0). To minimize Ioff, we need to compromise:
L increase -> performance decrease (transconductance)
Vt increase -> performance decrease (high threshold voltage)
n decrease -> Na or Nd decrease -> ”shortchannel” effects increase -> needs Xox decrease -> field in oxide increase -> reability issues
To summarize:
The subthreshold slope is S=(1+Cb/Cox)*(kT/q)*ln(10) so it essentially depends on the transistor design and substrate properties. Therefore, differences between n and p channel are natural. Also, very important for power applications, S depends linearly with the temperature, therefore one would expect a significat dependence of subthreshold conduction with the thermal regime of the device. This is one of the reasons why Nelson's Pass answer to the question "how do we set the optimum bias" was correctly "set it, let is warm up, then adjust it again".
Or you could try the ThermalTraks. Once you get your design dialed in, you only have to set it once cold, and then fine-tune one time only when it's warm.
The trick is to get the compensation just right. With an emitter follower triple, there are three different junction temperatures. Two of them basically just follow ambient (assuming that you designed the thing right, like Bart Locanthi showed us in 1966, and the pre-drivers and drivers are running class-A). The output junctions are affected by the ambient and the drive level and the load impedance.
And of course it gets trickier because the diodes inside the ThermalTraks don't have quite the same tempco as the transistor. But if it were easy, *everyone* would be doing it. And what's the fun in that?
The trick is to get the compensation just right. With an emitter follower triple, there are three different junction temperatures. Two of them basically just follow ambient (assuming that you designed the thing right, like Bart Locanthi showed us in 1966, and the pre-drivers and drivers are running class-A). The output junctions are affected by the ambient and the drive level and the load impedance.
And of course it gets trickier because the diodes inside the ThermalTraks don't have quite the same tempco as the transistor. But if it were easy, *everyone* would be doing it. And what's the fun in that?
janneman said:
For my designs the Sankens are useless. First of all, they are only doubles and not triples. A double creates far too much loading on the front end and increases the distortion about two orders of magnitude. I suppose that this could be corrected with feedback, but even with feedback it is best to start with a linear design.
The second problem with them is that the drivers are running in class AB. Locanthi showed us 40 years ago the best connection for a bipolar output stage (class A pre-drivers and drivers), and that is not possible with the Sanken parts.
Charles Hansen said:
Yes that's a bummer. I wish they had left out the Re and instead provided access to the driver for this. Interestingly, they now changed the type number but it appears to be the same transistor but without Re, and the same package except with the former Re pin snipped off close to the package. What's going on there??
Anyway, I am building an error-correction only amp (no global feedback) with them, I'll see how that turns out.
Jan Didden
Re: emitter follower triple
Hi, I'm not sure that I understand your question. My philosophy is to build the most linear circuit possible without using negative feedback. In my experience, adding feedback changes the sound of the amplifier in a negative manner.
As far as a correlation between distortion and "transparency", I can't really say. Part of the problem is that when we say "distortion", we normally mean "THD + N as measured with steady-state signals and a purely resistive load". But in my experience, there are many other types of distortions also. Some of them have not yet been measured, and perhaps may never be.
(Think about the state of modern physics. At one time, we thought that protons, electrons, and neutrons comprised the fundamental particles of which all matter is made. But now they have discovered well over a hundred "elementary particles"! Do you think that we will ever be able to identify all of the true fundamental particles?)
http://en.wikipedia.org/wiki/Fundamental_particles
I would have to say that there is no direct correlation between "distortion" (meaning "THD + N as measured with steady-state signals and a purely resistive load") and "transparency". Otherwise the Halcro would be the most transparent amplifier in the world and everyone would simply copy their design approach.
maxpou said:i read the document of "Tcircuit" and i have a question
Hi, I'm not sure that I understand your question. My philosophy is to build the most linear circuit possible without using negative feedback. In my experience, adding feedback changes the sound of the amplifier in a negative manner.
As far as a correlation between distortion and "transparency", I can't really say. Part of the problem is that when we say "distortion", we normally mean "THD + N as measured with steady-state signals and a purely resistive load". But in my experience, there are many other types of distortions also. Some of them have not yet been measured, and perhaps may never be.
(Think about the state of modern physics. At one time, we thought that protons, electrons, and neutrons comprised the fundamental particles of which all matter is made. But now they have discovered well over a hundred "elementary particles"! Do you think that we will ever be able to identify all of the true fundamental particles?)
http://en.wikipedia.org/wiki/Fundamental_particles
I would have to say that there is no direct correlation between "distortion" (meaning "THD + N as measured with steady-state signals and a purely resistive load") and "transparency". Otherwise the Halcro would be the most transparent amplifier in the world and everyone would simply copy their design approach.
