What makes a mosfet linear?

[lumanauw]

Mr.Pass writes that to get better performance out of a mosfet we can give more voltage and more current to it. This will result in more dissipation in a mosfet.
What really makes the operation linear and low distortion? Is it the high current or high voltage?
I dont have any simulator, could someone with simulator simulate this ?
1. Differential with mosfet (9610) working at +/-60V at 20mA CCS
2. The same differential as above working at +-20V at 60mA CCS
Both examples will dissipate the comparable heat, but one in high voltage, and one in high current.
Which gives more linear operation?
Which gives less distortion?
Which gives better audible performance?

[SeanPool]

Off hand, I am able to answer part of the question. Just my two cents: Increasing both will result in more linearity, but on a different scale.

If the voltage increases, the gate voltage of the mosfets increases. This results in a higher Vgs (in DC) since the source voltage stays the same because of the constant drain current. As a result, the transconductance gm increases, and does so linearly with Vgs.
If Ids increases, the transconductance also increases, but in the power of 0.5, i.e. g_m = (constant) * (Ids)^0.5.

When g_m increases, the overall gain of the amplifier increases. This gives us more linearity from the overall -tive feedback. (The overall negative feedback works better when the open-loop gain is high).

Please note that I did not talk about the linearity of the mosfets themselves when Ids increases. But generally, they tend to be more linear when Ids increases (in the sense that Vgs and Id curve is more straight).

-Sean

[akb1212]

Another question I'd really like to know more about is what does the working temperature have to say in all this.

The reason for asking this is Nelsson's saying: Run'em hot.

Now is this because the increased voltage/current gives better linearity, or is the temperature in itself also a contributing reason.

The reason for asking is that I'm in the starting process of building several AlephX' now. And I'm doing some considerations on how efficient cooling system to make (see my thread about it here ).

I'm able to make the transistors really cool if I want (by phase-change or water), but do I really want that? If one is able to control the temperature in a higher degree that normal, what temperature do I really want to aim for? Lower temps means the devices are able to handle more energy.
I know you can't think the same way as in computers where lower temps are better, no matter what temp.

Anyone with experiences on the sound on the same amps with different temps? I also know about the warm up time, and that the amp sounds better after warming up, but is it the increased temp that is the reason for this?

Anders

[sianturi]

Hello Anders,

As I remember, Nelson suggested running the amp so that the temperature of the heatsink (transistor) is about 50deg Celcius. I read it in another thread that Nelson wrote (don't ask me the exact link... I forgot). He also mentioned that Passlabs products were aimed at that temperature.

Correct me if I'm wrong....

[Nelson Pass]

Originally posted by SeanPool
If the voltage increases, the gate voltage of the mosfets increases. This results in a higher Vgs (in DC) since the source voltage stays the same because of the constant drain current. As a result, the transconductance gm increases, and does so linearly with Vgs.

Sean, while the rest of your comments are correct, I would have
to disagree with this part. Higher Vds does not typically result
in higher Vgs at the same current, rather the opposite is usually
true.

Higher Vds gives greater linearity, particularly if you are starting
out at low voltages, but you get decreasing returns above, say,
5 volts or so. Then the heat builds up.

Higher Ids similarly - when you look at the curves, you see that
the linearity quickly improves as you increase from 0, but at some
(usually high) current, improvement ceases.

What makes a Mosfet Linear? They are not all that linear, as
is true of the rest of the gain devices.

[azira]

Originally posted by lumanauw
Mr.Pass writes that to get better performance out of a mosfet we can give more voltage and more current to it. This will result in more dissipation in a mosfet.
What really makes the operation linear and low distortion? Is it the high current or high voltage?
I dont have any simulator, could someone with simulator simulate this ?
1. Differential with mosfet (9610) working at +/-60V at 20mA CCS
2. The same differential as above working at +-20V at 60mA CCS
Both examples will dissipate the comparable heat, but one in high voltage, and one in high current.
Which gives more linear operation?
Which gives less distortion?
Which gives better audible performance?

It's the current... but that's not the end of it.

FETS (and BJTs) have 3-regions of operation. cut-off, triode, saturation. For an amplifier, you need the FET to be in the saturation region. However, looking at the Vds/Ids graph, you'll see that it's smoothly curved. The operating region you want to be in is where the curve is 'flat' (saturation). This corresponds to a higher Ids. If you set your operating point so that you'll stay nicely in the saturation region for your input conditions, then it won't matter too much what current you choose.

Since FETS (and BJTs) aren't predictable when they are created, these curves can be slightly different for each FET. So, to be safer and assured that you're within the Saturation region, you'll want to run them with more current.

Lastly, gM the transconductance (or gain) of the FET changes with the current going through it. At higher currents, the gain is more constant, this leads to less distortion.

Refer to fig 1. and fig 6.
http://www.irf.com/product-info/data...ta/irf9610.pdf

[lumanauw]

Sianturi, apa kabar? Banyak dari Tanah Air yang ikut di website ini?

improvement ceases.

Mosfets comes in different sizes, TO220, TO247. For small signal stages (Differential and VAS), what is the general rule or indication up to how big standing current will give improvement? Will the body temperature gives a good indication (like said 50-55deg?). If this body temperature gives good indication that a certain mosfet is in the optimal working point, wouldn't it be that the current be different in case of we put heatsink or not to the mosfet?

The question rises after reading the article Balanced Zen Line Stage by Nelson Pass. In this article, Mr.Pass gives measurement result of the cct running in +/-30V, 40V, 60V.
This cct is working with R only, so if we rise the voltage, it automaticly will rise the standing current.

If the final answer is =ONLY HIGH CURRENT that gives improvement, wouldn't it be that we could make a mosfet works in relatively smaller voltage rails, but be forced by CCS to whatever current we want, to get max performance of this mosfet?

But if the final answer still includes factor of HIGH VOLTAGE to get good performance out of a mosfet, forcing the current by CCS wouldn't help so much. We just have to stick in high voltage rail, without nothing can be changed.

Why is this question rises? Making high voltage is more difficult and more expensive than making high current. Especially in the transformer and Elko's. Most of the Elkos I can find here, the good ones, only up to 50V. More than that like 63V rating or 80V rating or 100V rating are hard to find, especially if we look for good ratings (computer grades).

[SeanPool]

Higher Vds does not typically result in higher Vgs at the same current

- I guess this is what I don't quite understand (although irrelavant to the quote above): Let's say we have a CS, with the source connected to ground, and the gate gets the bias voltage from a voltage divider (R1&R2), which connects from the rail to the ground, tapped to the gate. In this setting, the gate will see the voltage dropped across R2. Since gate current is zero, this voltage drop may not be 4 volts. Actually, it may be a lot higher/lower. But you usually say that Vgs of a mosfet is 4 volts. This quite confuses me.

One explanation that comes into my mind is that the Vgs=4V is really because of the DC feedback from the Rs?

What makes a Mosfet Linear? They are not all that linear, as is true of the rest of the gain devices.

- This explains why different amps are sold at different prices: we are simply trying to make the most out of it. Actually, this is why Passlabs, Krell, and the likes, and this forum exist.

lumanauw

I forgot to mention: I remember from a thread (don't know exactly where) in which NP talked about how to improve the 30: Run the 30 with the output stage of the 60, with cranked up bias. Now the 30 has about 25 V rail, while the 60 has 34 V. So, *I think* he suggested that lower rail voltage with higher idle current should give you a better performance. (since the voltage is dropped to 24 V, you can run them at a higher current per mosfet for the same dissipated watts) -- please correct me if I'm wrong.

In addition to greater overall transconductance (means more linear in some frequency band, in general), what he suggested will result in a lower output impedance of the amp as well ---> you get more control over the load.

The catch: need to sacrifice the power as clippings will occur at lower power.

-Sean

[azira]

Originally posted by SeanPool


- I guess this is what I don't quite understand (although irrelavant to the quote above): Let's say we have a CS, with the source connected to ground, and the gate gets the bias voltage from a voltage divider (R1&R2), which connects from the rail to the ground, tapped to the gate. In this setting, the gate will see the voltage dropped across R2. Since gate current is zero, this voltage drop may not be 4 volts. Actually, it may be a lot higher/lower. But you usually say that Vgs of a mosfet is 4 volts. This quite confuses me.

In your scenario you'll have one of two things happen:
a) If the max current "allowed" to flow through the FET is larger than the CCS is providing, then the full current will go through the FET. Even though the FET would for the same Vgs allow more current to flow.

b) If the max current "allowed" to flow through the FET is less than the CCS is providing, then only the allowed current will flow, the rest will flow through another path or not at all. It is possible for a CCS to source less current than it was designed for if the loading doesn't allow for it.

- This explains why different amps are sold at different prices: we are simply trying to make the most out of it. Actually, this is why Passlabs, Krell, and the likes, and this forum exist

I suppose that is one thing. There are some BJTs and FETs out there that have more "linear" saturation regions. Basically that comes down to larger "early voltages" or the MOSFET equivelent of it. While these are more expensive, sometimes 3 or 4x more, it doesn't really account for the major difference in a lot of the prices.
I think that a lot more of it has to do with a well designed, maybe proprietary, topology and tighter tolerances in the other components as well. Well matched parts is a hands-on adventure that you're probably paying for too.

[sianturi]

Originally posted by azira


FETS (and BJTs) have 3-regions of operation. cut-off, triode, saturation. For an amplifier, you need the FET to be in the saturation region. However, looking at the Vds/Ids graph, you'll see that it's smoothly curved. The operating region you want to be in is where the curve is 'flat' (saturation). This corresponds to a higher Ids. If you set your operating point so that you'll stay nicely in the saturation region for your input conditions, then it won't matter too much what current you choose.

It's not just the 'flatness' of the curved, but also the distance between the Vds/Ids curve for each Vgs value that makes the mosfet linear. If the mosfet has uniform distance, then the linearity will be good. In term of uniformity of the distance between each curve, BJT is better.

Originally posted by lumanauw
Sianturi, apa kabar? Banyak dari Tanah Air yang ikut di website ini?

Hallo lumanauw....
Saya perhatikan ada beberapa dari tanah air yang ikutan forum ini. Tapi tidak terlalu banyak kayaknya.

I saw your website, it seems that you are really an audio addict. I went to college in Bandung, so maybe we could exchange some info on diyaudio parts source.... Jaya plaza is my favorite place though.

Cheers,