The Ultimate Transistor Amplifier

[rogerk8]

Hi!

Below is my planned transistor amplifier for car usage.

It involves push-pull FET transistors while making use of the limited car voltage (12V).

With this use the major consern is amps and not voltage.

The circutry is copied from the Williamson amplifier.

Hope you find it interesting!

Best regards, Roger

 

[Fish4Fun]

rogerk8,

I have left this alone for several days thinking someone else might point out that your schematic is a simplified drawing of a standard 1/2 H Bridge, traditionally used in audio amplifiers in the linear region configured as Class AB amplifiers. The problem with this type of amplifier is the dependance on linearity and tremendous heat dissipation in the power devices. The "bleeding edge" in audio amplifiers is Class D amplifiers which employ a PWM output that eliminates the need for dependance on the linear region of the switch. The main difficulty with Class D amplifiers is finding switches that can handle the high switching rates required (typically 400kHz or more.) All high-power audio amplifiers require tremendous attention to detail and careful component selection.

Fish

 

[CDRIVE]

Other than the fact that it's solid state I see no difference in this basic design than commonly used with tube amps dating back <1940's. It's a textbook Push-Pull amp. As Fish stated they were typically class AB.

Chris

 

[rogerk8]

What can I say? I drink too much!

There is however a serious part to all of this. A few years back I actually tried building my favorite Williamson amp with J-FETs (BF245A). I actually copied it. And it all worked well except for the "Cathodyne stage" which means that I could not connect the drain of the first transistor directly to the gate of the next. It was somehow impossible. But inserting a capacitor made it work excellently. I think I even built the push-pull driving stage with BF245As and no distorsion whatsoever was visible on my scope. Then I don't know what I did but I think I tried using two 2SK175 and hooked it up onto some mains transformer which of course did not work.

The interesting part is however that up to the actual push-pull drive of the power MOS, the signal looked very good!

To me it is kind of sad that power MOS are not available in depletion mode. Which means that I cannot use Williamson bias.

The thought of building such an amp has reappeared in my mind due to the simple fact that I am about to design a couple of small loudspeakers and I want to be able to demonstrate them at work without the need for a car to transport one of my superb Williamson amps.

I have put some more thought in the output transformer (OPT). My thought now is to buy a discrete transformer assembly and wind some 100 turns manually for both transistors and some 100 turns for the secondary (abandoning the information in the picture). This way I will probably get more than

wl*L=2*pi*7>4Ohm=>L>100mH

But most importantly, I will still run the output FETs in Class A which means that the transformer need not handle any DC-curents (and thus air-gaps).

Roughly speaking I will then have a quisent dissipation of

P=12*1,5=18W

Considering push-pull and the I/V-characteristics of a MOS then some

Pout~10W

would not be impossible.

The only thing that is really boring, as always when it comes to semiconductors, is the cooling arrangement. Using tubes you never have to consider that.

Finally, there will of course be some feedback applied also. I mean, we are talking semiconductors here. I will use feedback around the OPT and add my original push-pull driving stage to get some raw amplification which I think is needed.

Thank you both for replying to my stupid thread

Best regards, Roger

 

[CDRIVE]

What can I say? I drink too much!

Then you're in good company here.

The only thing that is really boring, as always when it comes to semiconductors, is the cooling arrangement. Using tubes you never have to consider that.

Who told you that? Are you aware that some high power transmitters used circulated liquid coolant to keep the vacuum tube finals from melting down? Here's an example on a much smaller scale.
http://www.realhamradio.com/liquid-cooling.htm

Thank you both for replying to my stupid thread

Best regards, Roger

I don't think your thread is stupid at all.

Chris

 

[rogerk8]

I don't think your thread is stupid at all.

Chris

Thank you very much Chris!

I was mainly thinking of audio amplifiers. I was even visualising 845s as output tubes and not even them are cooled by liquids.

But you have pointed out a for me new field of interest.

I have built several of these linear amplifiers but I know almost nothing about unlinear technologies like radios.

I have recently designed a J-FET oscillator of some 4MHz and it was rewarding when I first got it to oscillate because I did not know what I was doing. In the beginning I hand-wound some 20 turns of uninsulated wire and tried using the Hartley-concept but got no oscillations. Then someone in my swedish electronics forum (by the way, I like writing in english better) told me that the coil had shorted turns so I turned to insulated wire instead, used the ferrite-rod I had winded it around and moved the feedback turn from 20% to 25% and then suddenly, it started to oscillate!

To me it was like magic!

My next setup is actually to design two different Hartley oscillators, sum them up unlinearly (which equals multiplication) and select one of the mixing products by an IF-stage.

If I get this to work, my plan is to build a radio with absolutelly nothing but ECC83s as tubes.

And yes, I have too many projects going on at the same time

Best regards, Roger
PS
Interesting link by the way!

 

[rogerk8]

Hi!

I have now done some calculations regarding my OPT-FET amplifier.

It looks like the gain is insufficient so I will add yet another push-pull stage like Williamson did.

Attached is my schematic.

Best regards, Roger

 

[rogerk8]

Hi!

Now I have added this additional push-pull stage. The gain is now sufficient.

The only question that remains is T1 and T2.

I have looked at datasheets for different kinds of Power FETs but they are diappointing.

The I/V-plots does not make much sense to me.

Not so much as the I/V-plot of a pentode, anyway.

But it should!!

Best regards, Roger
PS
And yes, I will have to use some feedback.

 

[Fish4Fun]

See, now that looks more like an amplifier. Hehe. Now the only question I have left is: Why? Don't get me wrong, I like doing the already done as much as anyone, but there have been Audio amplifiers "on a chip" readily available since the early 1980's. I am not suggesting any of these are the "ultimate amplifier", but some of them are pretty darn good. If you want to use discrete FETs to create an audio amplifier you might consider looking into class D amplifiers.

I haven't really followed audio amplifiers since the late 90's when it became readily apparent to me I couldn't build anything even close to what was commercially available for a fraction of the price, but audio amplifiers were my first real interest in electronics, so they hold a special place in my heart. Sadly my turn-tables, cassette decks, amplifiers, pre-amplifiers, tuners, equalizers, speakers, mixers etc have all been replaced by a PC. I am not suggesting this is a "good thing", just kinda telling it like it is, lol.

Anyway, right after I finish re-inventing the thermostat, I plan to re-invent the H Bridge, so please don't take my comments as any type of criticism, or be dis-swayed from your task

Enjoy!

Fish

 

[CDRIVE]

IMHO, and from an educational (design and engineering) standpoint, much more is gained from your design than sticking a chip on a board. This certainly doesn't contradict anything Fish said though.

Chris

 

[rogerk8]

See, now that looks more like an amplifier. Hehe. Now the only question I have left is: Why? Don't get me wrong, I like doing the already done as much as anyone, but there have been Audio amplifiers "on a chip" readily available since the early 1980's. I am not suggesting any of these are the "ultimate amplifier", but some of them are pretty darn good. If you want to use discrete FETs to create an audio amplifier you might consider looking into class D amplifiers.

Thank you Fish for your comment!

If you aim at designing stuff that hasn't been designed yet you end up sorry because there aren't any! So I design stuff simply because I like to design stuff and what I like to do most is to apply my own rules, so to speak. I like to come up with a solution of my own. It need not be the best. It just needs to be my design.

You should not be fooled by on-chip solutions. They of course work well but in my humble opinion they often use too many transistors, simply because they are so small, so that it might have an impact on the quality of the sound.

Class D amplifiers are getting better and better. It is amazing that a 1kW amplifier might not be much bigger than a credit card. But PWM-solutions are cheating in my opinion. To get real good audio quality you need Class A amplifiers. And there's really no way around that fact

I haven't really followed audio amplifiers since the late 90's when it became readily apparent to me I couldn't build anything even close to what was commercially available for a fraction of the price, but audio amplifiers were my first real interest in electronics, so they hold a special place in my heart. Sadly my turn-tables, cassette decks, amplifiers, pre-amplifiers, tuners, equalizers, speakers, mixers etc have all been replaced by a PC. I am not suggesting this is a "good thing", just kinda telling it like it is, lol.

You obviously missed the fact that tube amplifiers are the best audio reproducing equipment humans can build

PCs are just convenient. Never better. Because let's face it. Who believes that by quantisizing an already perfect analog signal and then reassemble it would be any better?

Anyway, right after I finish re-inventing the thermostat, I plan to re-invent the H Bridge, so please don't take my comments as any type of criticism, or be dis-swayed from your task

Enjoy!

Fish

Regardless if your project is readily available or not, you learn a lot from trying to design stuff by your own.

Take care!

Best regards, Roger

 

[rogerk8]

Hi!

I have now fine-tuned my OPT-FET amplifier.

I came to the conclusion that BF245C was a bit unlinear and suddenly it struck me that I could use a different bias arrangement (stolen from the Tube Amp field) and use the more linear BF245A instead while this means that the whole "pre-amplifier" consists of one type of J-FET only.

Everything now looks good. I have carefully selected the roll-offs and bias points to get the most linear amplification from BF245A as possible. At least that is what I think

Sensitivity is now some 0,2V for 5V output at pp-drive. With a planned global feedback of some 14dB (5 times) input sensitivity should end up at some 1V for full output power.

My main problem now is to choose output power FETs. I have a couple of complementary audio MOS-FETs (2SK175/2SJ55) which can withstand 8As. And Class A biasing means some 1,5A quisent current (considering the planned OPT turn ratio and the reflected impedance, 4Ohm/FET).

But their behaviour in the low voltage region is cathastrophic(?)

I am attaching both my new preliminary schematic of the FET Amp as well as BF245 datasheets and 2SK175 datasheets. Meaning that someone might be able to give me a tip on how to proceed.

Take care!

Best regards, Roger
PS
This datasheet was too large for EP (BF245):
http://www.nxp.com/documents/data_sheet/BF245A-B-C.pdf

 

[rogerk8]

Finished

The only problem now is the OPT.

I am planning on ordering a 15VA E-core transformer from Svebry.se. They actually sell transformers as kits

Take care!

Best regards, Roger

 

[rogerk8]

I how now maximized the power output (and changed the OPT turn ratio).

6W felt a little to low. Now it's up to 10W. But the cooling arrangement is huge

Best regards, Roger

 

[CDRIVE]

Hi Roger, I confess that I took a peek at your profile because I was curious about your age. Would you care to share that info with us? You don't have to be precise. The age group, teens, 20s, 30s, 40s, etc. will suffice.

Chris

 

[rogerk8]

Hi Roger, I confess that I took a peek at your profile because I was curious about your age. Would you care to share that info with us? You don't have to be precise. The age group, teens, 20s, 30s, 40s, etc. will suffice.

Chris

Hi Chris!

I have now updated my profile.

Thanks for your interest in me!

Best regards, Roger

 

[rogerk8]

Hi!

Does anyone see the small and "insignificant" difference of this final edit?

Best regards, Roger

 

[CDRIVE]

Yup, you changed the output transformer phasing.

Thanks for posting your personal info. I think you're extremely talented and see you as a great asset to our forum! If you had posted that you're 20 I'd have fallen off my chair.

Curious; What's the DC resistance of your primary / secondary and what's the value of Rf?

Chris

 

[CDRIVE]

Yup, you changed the output transformer phasing.

Thanks for posting your personal info. I think you're extremely talented and see you as a great asset to our forum! If you had posted that you're 20 I'd have fallen off my chair.

Curious; What's the DC resistance of your primary / secondary and what's the value of Rf?

Chris

Hey, are you still there? You didn't answer this. Also, what is your THD goal?

Chris

 

[rogerk8]

Yup, you changed the output transformer phasing.

Thanks for posting your personal info. I think you're extremely talented and see you as a great asset to our forum! If you had posted that you're 20 I'd have fallen off my chair.

Curious; What's the DC resistance of your primary / secondary and what's the value of Rf?

Chris

Hi Chris!

Thank you very much for your nice words!

I have not calculated the DC resistance. It will be what it will be using 2X100 turns of copper wire (1mm) on the 15VA EI-core which I will buy from Svebry.se.

The use of Class A does not only give the best sound (according to me) it lessens the requirements on the OPT due to no DC-magnetization but I suspect you already knew that. The only drawback is the huge cooling arrangement

Due to the power FET's high output impedance, the primary inductance will be defined by almost only the reflected impedance which in this case is 2 Ohm/FET.

This means that some total primary inductance of 0,2H will suffice (using wL>8 Ohm, 7Hz). Which I think is achievable with 100+100 turns.

The THD will also be what it becomes. But testreults below show that the THD seem quite small. Surprisingly small actually

But this is before the power FETs. Which will work in a quite unlinear region. The reason for feedback is thus mostly due to them, the OPT (and sensitivity). Even though my wish is not to use more feedback than some 14dB because I think this will have negative effects on sound quality. Which, by the way, is why I believe OPs sound "bad"

So sensitivity will become dependent on raw gain (which I will have to measure first) and my demand on BF+1=5. But I think it will end up on some 1Vp for full output power.

I am attaching some photos. The upper channel is set at 0,1V/DIV and the lower at 1V/DIV which measures the top transistor drain in the push-pull stage.

Enjoy!

Best regards, Roger
PS
Sorry for not answering earlier. Personal issues.