I hope I don't come after the battle...
I didn't yet read all the thread but I think one parameter for BJT to MOS choice is cost.
For a given output power (let's say 100-200W/8ohms, 200-400W/4ohms), which of those is lower price?
We can include the fact that BJT may need paralleling, stronger drivers... we may include the fact that power supply voltage may be slightly different...
I think it is always possible to find a solution for cost-no-object designs. I don't care for them, I know they will find.
But I think the real test is affordability for mass market products.
And why Yamaha, Denon, Pionner still use BJT for their up to 200W amp???
Hope you didn't answer those questions yet, I still have a few pages to read... going back to that work.
Regards
Laurent
I didn't yet read all the thread but I think one parameter for BJT to MOS choice is cost.
For a given output power (let's say 100-200W/8ohms, 200-400W/4ohms), which of those is lower price?
We can include the fact that BJT may need paralleling, stronger drivers... we may include the fact that power supply voltage may be slightly different...
I think it is always possible to find a solution for cost-no-object designs. I don't care for them, I know they will find.
But I think the real test is affordability for mass market products.
And why Yamaha, Denon, Pionner still use BJT for their up to 200W amp???
Hope you didn't answer those questions yet, I still have a few pages to read... going back to that work.
Regards
Laurent
Hi Laurent,
You bring up a very good point: cost. While we as high-end people often don't like to talk about cost, there are usually tradeoffs in anything but a truly cost-no-object device. On the average, it is fair to say that BJTs are a bit less costly than MOSFETs, but one has to be careful about generalizations and how much of associated collateral costs are involved, like perhaps the need for more devices, or more complex protection circuitry, or beefier drivers.
It really depends on the quality target and the cost of the output devices must be put in perspective with the other costs of the amplifier, such as power supply, heat sinking, exotic or ultra-high quality passive components, etc. The cost of good power transistors of either BJT or MOSFET varieties has also come down somewhat over the years. Very good devices of each type can be had for well under $5.
Among MOSFETs, the IRFP240/9240 pair is very cost-effective and capable of yielding exceptional performance in the right (designer) hands. They are relatively inexpensive as a result of high-volume manufacture for non-audio applications. The Toshiba 2sk1530/2sj201 are better MOSFETs, but they cost quite a bit more - just under $10 in some cases.
Then there are the lateral MOSFETs. They are quite expensive and less readily available. They are a relatively obsolete technology pretty much reserved for audio, and that makes them expensive. This is not unlike the situation with tubes. Some people prefer the sound of the laterals, but they will pay a premium for it. I personally do not consider lateral power MOSFETs to be superior to verticals in the sound department, again given good amplifier designers for each. Under fault conditions, they are often more robust than verticals, however.
In general, I don't let the cost of the transistors play a big role in my designs. For BJT, I like the ThermalTrak Onsemi Devices. For MOSFET amplifiers, I like the Toshibas, but have very often used the IRFs.
Cheers,
Bob
As long as the cost of active parts is still a small portion of the complete amplifier (heatsink, cases, transformers are higher value), you are right.
I will have a look to the price of some devices, as it is for DIYing some monobloc amp for my home theater (lots of speaker to active filter), and as maybe I will have to do some others for friends, I assume I will face cost sooner or later. DIY cost is almost higher given lower quantities.
Regards
Laurent
I will have a look to the price of some devices, as it is for DIYing some monobloc amp for my home theater (lots of speaker to active filter), and as maybe I will have to do some others for friends, I assume I will face cost sooner or later. DIY cost is almost higher given lower quantities.
Regards
Laurent
Hello Bob,
In fact I found this thread as I was hunting some information about MOSFET, your name comes close to it. I read with great interests the various papers you did concerning MOSFET use in audio amp.
As I am on the phase of learning everything about power amp, I am reading many books about this. The last I read was the one from Douglas Self, and as I could find almost no information about MOSFET in his book (guess why (Doug, if you are on this web site, your book is great!)), I supposed you wrote one.
Pfff... couldn't find one. What is your plan for next week end?
Can you or anyone recommand me a good reading about MOSFET for audio power amp?
I think some different opinions are required to promote MOSFET. Up to now, I have more data on bjt and decision is hard to take.
The battle looks funny, sometimes like Tube/Bipolar...
(I work on RF, we had the same battle many decades ago concerning power amplifiers for TX and low noise amplifier for RX. Once good power transistors and low noise one were available with good performances, battle was over very fast)
Regards
Laurent
What a funny job we have. (I am about RF, but I accept that close-to-DC work as hobby)
In fact I found this thread as I was hunting some information about MOSFET, your name comes close to it. I read with great interests the various papers you did concerning MOSFET use in audio amp.
As I am on the phase of learning everything about power amp, I am reading many books about this. The last I read was the one from Douglas Self, and as I could find almost no information about MOSFET in his book (guess why (Doug, if you are on this web site, your book is great!)), I supposed you wrote one.
Pfff... couldn't find one. What is your plan for next week end?
Can you or anyone recommand me a good reading about MOSFET for audio power amp?
I think some different opinions are required to promote MOSFET. Up to now, I have more data on bjt and decision is hard to take.
The battle looks funny, sometimes like Tube/Bipolar...
(I work on RF, we had the same battle many decades ago concerning power amplifiers for TX and low noise amplifier for RX. Once good power transistors and low noise one were available with good performances, battle was over very fast)
Regards
Laurent
What a funny job we have. (I am about RF, but I accept that close-to-DC work as hobby)
I have checked them out, to some extent, on paper. I'm not sure they are all that new. Looks like they date back to at least 2006.
They cost less, but an informal search by me did not reveal any place stocks them (non-stock at Digi-Key).
The devices are rated at lower voltage: 180 vs 200.
The devices are rated at lower current: 10A vs 12A
The devices are rated at less power: 130W vs 150W.
The devices are in smaller packages: TO-247 vs. TO-264.
The devices have 1.5X the gate-drain capacitance (not good).
The devices appear to have higher transconductance for a given drain current - this is good if you want to operate at lower bias current.
To first order, higher gate-drain capacitance sometimes goes along with higher transconductance.
Note also that the gate breakdown voltage is smaller, suggesting a thinner gate that may be responsible for the higher transconductance and the higher capacitance for a given die size.
Overall, these look like nice devices that are a lower-cost althernative. I'm sure you can make a fine amplifier out of them. I would hesitate to say that they are overall better devices, however. You may get a bit more silicon and transconductance for your money, but I strongly prefer devices in the larger TO-264 packages for thermal reasons. For one thing, the larger area of the device reduces thermal resistance of the required insulator.
I haven't looked at it yet, but a look at SOA and Rds_on and gate forward voltage required for a given very high current might also reveal some differences.
If you are looking to save money on an apples-apples package comparison, they may be a superior alternative to the IRFP240/9240, which are also in the smaller package.
Anyway, these appear to be appealing for the price. Let me know where they are available.
Cheers,
Bob
Cheers,
Bob
You warned me recently in another thread of increased Miller when VDS<VGS.
This was in regard to IRF(9)240 enhancement MOSFETs. But has occoured to
me to ask, "What of Depletion Mode MOSFETs?" VDS cannot possibly ever be
less than a negative gate threshold???
So, do we still feel the same effect at some low VDS, not necessarily directly
related to -VGS? Or does a negative gate threshold allow these devices to be
free of the sudden increase in Cdg?
This was in regard to IRF(9)240 enhancement MOSFETs. But has occoured to
me to ask, "What of Depletion Mode MOSFETs?" VDS cannot possibly ever be
less than a negative gate threshold???
So, do we still feel the same effect at some low VDS, not necessarily directly
related to -VGS? Or does a negative gate threshold allow these devices to be
free of the sudden increase in Cdg?
The speed at which time passes in reference to one's age seems to be on a par with electronic components, each consecutive decade flows by twice as fast.
First Accuphase model with J618/K3497 : E530 integrated, official release Februari 2002
Ok, the 2SK3497/2SJ608 are maybe not that new, but I haven't seen them used anywhere else than by Accuphase, especially not DIY....
And yes, they're not that easy to get, but Digikey CAN deliver, although the 100 sets I have in the mail, just waiting for european airspace to open, took something like 7 months....
And yes, they're a little lower voltage/power, more like the 2SK1529/2SJ200, but for 1/3 the price you can just use a more of them
It's the Ciss that is larger, but as that have to be seen in relation to Yfs, they're actually a little better there....
The important is Crss, and there the 2SK3497/2SJ618 is MUCH better compared to the 2SJ201/2SK1530, you might be able to connect them directly to a high current VAS, like often done with the Hitachi power mosfets.
How good they actually are in real life I will report on when I get them.
Soren
And yes, they're not that easy to get, but Digikey CAN deliver, although the 100 sets I have in the mail, just waiting for european airspace to open, took something like 7 months....
And yes, they're a little lower voltage/power, more like the 2SK1529/2SJ200, but for 1/3 the price you can just use a more of them
It's the Ciss that is larger, but as that have to be seen in relation to Yfs, they're actually a little better there....
The important is Crss, and there the 2SK3497/2SJ618 is MUCH better compared to the 2SJ201/2SK1530, you might be able to connect them directly to a high current VAS, like often done with the Hitachi power mosfets.
How good they actually are in real life I will report on when I get them.
Soren
Accuphase still seems to favor sticking a pair of A2190/C6072 drivers between the two stages (20W, 200Mhz, TO220), plus A1507/C3902 predrivers.
And high quiescent currents, 30W Class A for both the E530/E550/E560 on 3-pair output stages.
Lower device cost doesn't seem such a compelling issue either for an integrated that retails for $1000 per output transistor overhere.
Last edited:
That's a lot trannies. Was this a group buy? If so, (or if you don't plan to use them all) I'm in for 10 pairs.
Agreed.
But at low drain-source voltages (3V for example), where they tend to make trouble, around 500pF for all types.
Indeed, 'might', but I wouldn't try it. Always use drivers.
Yes, please.
Cheers,
E.
Hello Bob,
In fact I found this thread as I was hunting some information about MOSFET, your name comes close to it. I read with great interests the various papers you did concerning MOSFET use in audio amp.
As I am on the phase of learning everything about power amp, I am reading many books about this. The last I read was the one from Douglas Self, and as I could find almost no information about MOSFET in his book (guess why (Doug, if you are on this web site, your book is great!)), I supposed you wrote one.
Pfff... couldn't find one. What is your plan for next week end?
Can you or anyone recommand me a good reading about MOSFET for audio power amp?
I think some different opinions are required to promote MOSFET. Up to now, I have more data on bjt and decision is hard to take.
The battle looks funny, sometimes like Tube/Bipolar...
(I work on RF, we had the same battle many decades ago concerning power amplifiers for TX and low noise amplifier for RX. Once good power transistors and low noise one were available with good performances, battle was over very fast)
Regards
Laurent
What a funny job we have. (I am about RF, but I accept that close-to-DC work as hobby)
Hi Laurent,
I agree: there is a serious lack of textbook discussion concerning the application of power MOSFETs to audio. This is unfortunate because there are numerous mis-understandings out there about MOSFETs for audio, and there are also pitfalls in using them that must be dealt with. For example, they are fundamentally fast, and therefore it is easier to suffer from parasitic oscillations. In this respect, I think it is fair to say that one may need to be a bit more knowlegable in using MOSFETs.
Crossover distortion issues with MOSFETs are also not as well understood as they should be. Issues like transconductance droop and dynamic crossover distortion (switching distortion) need more discussion and need to be put in better perspective. This general issue of nonlinearity is an area in which Doug Self and I do not completely agree; but it is a matter of how one looks at the problem.
The great superiority of even vertical MOSFETs over BJTs in regard to dynamic thermal stability is not adequately appreciated; indeed, how bad a problem it is for BJTs is also not adequately appreciated.
As you know, I really like vertical MOSFETs. However, I have explored the ThermalTrak BJT's and really like them as well. They are a fundamental advance in BJT power transistors, and I have no qualms about building an amplifier with them. They greatly mitigate the dynamic thermal tracking problems suffered by BJT output stages. More needs to be published about them as well. I will never again build a BJT amplifier that does not use ThermalTrak devices or the functional equivalent.
Cheers,
Bob
I actually do not know of any depletion mode power MOSFETs that exist. In most cases in audio, I associate depletion mode with JFETs. The physics of JFETs are completely different, so this issue of capacitances is not the same.
I should also note that the behavior of gate-drain capacitance vs gate-drain voltage is different for lateral MOSFETs than for Vertical MOSFETs, and the physical structures are quite different.
Cheers,
Bob
The speed at which time passes in reference to one's age seems to be on a par with electronic components, each consecutive decade flows by twice as fast.
First Accuphase model with J618/K3497 : E530 integrated, official release Februari 2002
I have always believed that the perception of the passage of time by humans is on a log scale. When I was young a year was forever. Now a decade passes like a month did when I was young. Where the heck did the 1990's go?
Cheers,
Bob
Ok, the 2SK3497/2SJ608 are maybe not that new, but I haven't seen them used anywhere else than by Accuphase, especially not DIY....
And yes, they're not that easy to get, but Digikey CAN deliver, although the 100 sets I have in the mail, just waiting for european airspace to open, took something like 7 months....
And yes, they're a little lower voltage/power, more like the 2SK1529/2SJ200, but for 1/3 the price you can just use a more of them
It's the Ciss that is larger, but as that have to be seen in relation to Yfs, they're actually a little better there....
The important is Crss, and there the 2SK3497/2SJ618 is MUCH better compared to the 2SJ201/2SK1530, you might be able to connect them directly to a high current VAS, like often done with the Hitachi power mosfets.
How good they actually are in real life I will report on when I get them.
Soren
You are right about the relevent comparison for MOSFETs being the Ciss and Crss compared to transconductance, or even better to transconductance per amp of drain current - the bang-for-buck FOM for MOSFETs.
However, when I looked at the data sheets, Crss at the same Vgd was not smaller than for the 2SK1530 at a given low voltage like 1V.
Cheers,
Bob
I'm not that focused on Crss at that low voltage as the first design I'm doing will not use a separate driver rail so Vgd will never get lower than maybe 3V, where Crss looks like:
2SK1530/2SJ201 - 400pf/650pf
2SK3497/2SJ618 - 500pf/550pf
Not much difference but also remember that when one mosfet is at low Vgd, the complementary one will be at high Vgd.
I'm also more interested in performance at 1/4 to 1/2 at full power as music rarely hits full power, and then it looks different, t.ex. at 25V Vgd:
2SK1530/2SJ201 - 120pf/250pf
2SK3497/2SJ618 - 30pf/80pf
If you then take transconductance into account the factor is even larger, as typically the 2SK1530/2SJ201 is 5 S while 2SK3497/2SJ618 is 12 S....
And of course, its only really matter much if you don't use a current driver stage. My first design using them will have a current driver stage, but I'm exploring the possibility of doing one without that stage for lower power use with just one sets of mosfets.
I also don't need to worry to much about what happens at overload as most of my design are without global feedback. And there the high transconductance is a big advantage, I do expect great results with just a little high bias current.... But then, I'm a optimist
http://www.diyaudio.com/forums/soli...-amplifier-non-feedback-production-ready.html will be the first design, although it's presented as bipolar the Toshiba mosfets fit in as they have the same pinout, just a change of a few other parts....
But as Bob says the mosfets are complex, only actual hardware will tell how good they are....
Soren
Not a group buy, that was the minimum order
But I also have a company to put it though, and are planning to maybe sell some boards. And maybe some special parts.In all circumstances I can spare some for you....
Soren
I was thinking of Supertex DN2535 in 15W TO-220. But didn't remember it
also had a serious 25ohm RDSon limitation working against it. Forget the
question, it relates to a component that probably isn't practical for solid
state output stage anyhow...
1-5pF reverse transfer? I suppose you could parallel a few if cost wasn't
a big concern... But I don't see this one working close enough to the rail
to be relevant to my original question.