The Ultimate Transistor Amplifier

[rogerk8]

Just want to post this final design which I have decided to build thus giving people the same chance for success as for me.

Take care!

Best regards, Roger
PS
The major difference is power dissipation. I somehow missed that the dissipation is calculated before any signal is applied and not at the dynamic working point.

I have also modified the bias adjustment possibillities somewhat.

The transformer kit and the LU1014Ds are ordered and I am awaiting their arrival very soon.

 

[rogerk8]

Now I have designed the component layout of my FET amplifier.

A depth of less than 400mm seems possible.

Best regards, Roger

 

[rogerk8]

I now have 12pcs of the amazing LU1014Ds in my hand.

I have thus decided to build my FET-Amp with these only.

I have changed the bias somewhat so that the J-FET gate load is now less than 15%.

I have also designed a preliminary layout for the pre-amplifier. I have decided to use a RF-approach which means the use of a copper laminate (hope this is the right word).

I recently built a RF-experiment using this approach and I liked it so much that I now want to use it for LF purposes also.

I hope I can send you a picture of the pre-amp very soon.

Take care!

Best regrds, Roger

 

[rogerk8]

I have decided to skip my preliminary bias arrangement due to (probably) different gate input impedances and therefore different VLF roll-offs.

I have decided to use Williamson-bias instead giving constant input impedances to the power transistors and thus equal VLF roll-offs as well as tube compliance.

The only problem with this method is the availability of low resistance high current potentiometers. And their cost...

I have however found a 1,5R/6A (50W) potentiometer at my swedish retailer (ELFA) at the fascile price of 1000 SEK (approx. $140) a piece.

These I will use!

All because I totally love the idea of being able to design Williamson-style with modern components

Best regards, Roger

 

[CDRIVE]

I have however found a 1,5R/6A (50W) potentiometer at my swedish retailer (ELFA) at the fascile price of 1000 SEK (approx. $140) a piece.

These I will use!

All because I totally love the idea of being able to design Williamson-style with modern components

Best regards, Roger

I'm ... uh ...........speechless!

Chris

 

[rogerk8]

I'm ... uh ...........speechless!

Chris

In a positive or negative way?

 

[CDRIVE]

In a positive or negative way?

Well, it's obscenely expensive but I love your work! So if that's what you needed then so be it.

Chris

 

[rogerk8]

Thanks Chris!

I must say that I am kind of having second thoughts. The price is obscene as you say and 1.5R is far from perfect (very unlinear adjustment in my application). Wish I could get me a pair of 0,1R/6A instead.

Perhaps you could help me with this one too

Best regards, Roger
PS
I recently checked Vishay, they did not have lower reostat values than 1Ohm (RT100).

 

[(*steve*)]

One option would be to get some current sense resistors (these are available in resistances in the milliohm range).

You could make a long chain of them and then tap off the appropriate number (sort of a manual rheostat).

However, you'll quickly run into the problem that solder joints have resistances as do copper traces and wire.

As a possibly useless note, I have some 2 milliohm resistors like these. In theory they're good for over 20A.

 

[rogerk8]

Hi Steve!

Actually, my collegue told me just this today and I thought it was a good idea.

I have however not descided yet exactly what to do.

These are my options:

Fixed Bias:
Benifits:
1) Simple
2) Cheap.
Drawbacks:
1) Different VLF roll-offs most certainly
2) Loading the PP driver stage considerably

Automatic Bias using reostats:
Benifits:
1) Equals standard PP tube biasing (nostalgic )
2) Presents a high resistance as load to PP driver stage
Drawbacks:
1) Expensive
2) Very unlinear regulation

Now to yours and my collegue's nice tip:

Automatic Bias using set of fixed resistors:
Benifits:
1) Cheap
2) Perfect load on driver stage
Drawbacks:
1) Long chain of resistors (unless I match the JFETs first )

I do however not wish to match the transistors first. I wish to do it "In Circuit".

I think I like your descrete resistor approach best. I think I will do so. Thanks!!

Best regards, Roger
PS
I love your resistor collection 2mOhms/1W at 1% tolerance...not bad and very useful indeed

I'm attaching a couple of pictures on my FET pre-amp which I built today. I had to filter the input due to crappy signal generator but otherwise it gave a SE gain of 2X35 times at 100mVp in. And as Chris already has simulated, the (unfeedbacked) distorsion is hardly visible (<5% according to him). This at 12V.

To set the feedback resistor and the overall loopgain (BF+1) to 5 times I do however need to test the gain after the power JFETs and thus at transformer output. This transformer kit will be delivered to me shortly.

 

[(*steve*)]

I'm attaching a couple of pictures on my FET pre-amp which I built today.

Looks good.

You should look up "manhattan PCB construction. It looks like you have connections floating in mid-air. This method allows you to do what you have done here, but in a more rugged way.

 

[rogerk8]

Thanks!

Today I decided to continue using my original thought. This means that I will use a combination of fixed bias and self bias. All because it got rediculously expensive, complicated and actually not that perfect using (available) reostats.

My current methode will enable all biasing options (both magnitude and difference) in a more cheap and simple manner. The only drawback is actually the (different) load on the driver stage but any difference in the push-pull amplitudes may be adjusted with the means of P1. What happens at VLF will however be nonadjustable.

Attaching my latest upgrade of the FET Pre-Amp. Which includes a glowing lamp

Best regards, Roger

 

[rogerk8]

I now have a transformer kit in my hand.

Attached picture only shows the copper wire and the bobin.

The bobin has two chambers. I might reduce that to one by a cutter. Perhaps after winding 2X100T. Or else I will split the secondary also to 2X71T which will give 8 Ohms drain-to-drain impedance due to my 4 Ohms speaker.

The shown copper wire is of 0,71mm. 3A per winding would yield some 8A/mm^2 which is quite high. My former transformer manufacturer told me a long time ago that 3A/mm^2 is common practice. Fortunatelly I have 1,0mm copper wire also even if I do not like it's stiffness.

What do you think? Will 8A/mm^2 get my transformer too hot?

Because I prefere 0,71mm.

Best regards, Roger

 

[rogerk8]

I have now soldered my LU1014Ds to some small copper pieces to be able to mount them on a cooler.

The IPAK capsule is almost as funny and unique as the Power JFETs themselves because who doesn't need to mount them on a cooler while there is no screwholes in them

The project is progressing very slowly but it is progressing.

Instead of building only one prototype unit I now feel so confident that I will build two. Because let's face it, a mono demonstration is not that attractive

Next step is to (linearly) regulate and trim my switched motor supply (24V/5A). Or perhaps, buy a couple of serious switched supplies (12V/6A).

Best regards, Roger

 

[CDRIVE]

Yeah, I've seen that package before and they don't lend themselves well to easily heat sinking. Your other option would have been a sandwich configuration.

Chris

 

[rogerk8]

A language note:

A capsule should be called a package?

A cooler should be called a heat sinker?

Is this right?

Best regards, Roger

 

[rogerk8]

These are the heat sinkers I will use.

I had some other thoughts at first but these suits me fine.

Each will cool away some 30W.

Hopefully, anyway

Best regards, Roger

 

[CDRIVE]

A language note:

A capsule should be called a package?

A cooler should be called a heat sinker?

Is this right?

Best regards, Roger

Hi Roger,

I think "Heat Sink" will suffice A sinker is something you use when bottom fishing or a food that feels like one. .

Chris

 

[rogerk8]

Hi Chris!

I am trying to understand this new fantastic user interface. Uploaded an avatar but that didn't work. Neither did it work to simply list my threads...

Anyway, I have tried to design a switched power supply for my JFET amplifier. I am using a motor drive from work where there is much ripple but I am combining it with a linear regulator. Preliminary tests show bad results but I am refusing to give in. With my 6A (into 2 Ohms) and used heat sink(!) no more than 30W may be cooled away. This means some maximum 5V drop across the Darlingtons.

This is what I will try do build on Easter Evening. I think a combination of 12Vz+3X1N4148 will suffice instead of Rt while R will be changed to 1k and some 16V to supply it all (Iz=2mA). I will also add a 10mF capacitor on the motor drive output.

What do you think?

Attaching pictures.

Best regards, Roger

 

[rogerk8]

Hi!

I have now cleaned up the SMPS that now gives 12V@3A with only 10mVpp/100Hz pipple (the drawing is wrong).

However, I think the ripple is too large so I have decided to add some filtering in the driver stage.

The choice of the time constants and the individula attenuation is not made by random. I have written an article about how this may be done optimally but that one is in swedish.

I have measured the dynamic parameters for a bogey type of BF245A and they are my=50 and rp=12k.

The initial results actually showed that loading 15 pcs of BF245A with 2k9@12V (Rs=200 Ohm) gave the mean value Vd=4,8V with a standard deviation of 30%.

Anyway, I have not been able to finish my driver stages due to lack of two 1mF capacitors.

Attaching pictures of my projects and a drawing of the supply.

Enjoy!

Best regards, Roger