JFET Power Amplifier With Output Transformer

[rogerk8]

Hi!

I have designed, measured and tested an amplifier which pre-amplifier consists of JFETs (BF245A) only, the transistor type for the power stage (2SK175) has however been chosen to be made of MOS-transistors, this is mainly due to the fact that I want Class B only and to "strangle" Power JFET (which I actually own a couple of due to the kindness of this nice forum) is more difficult than to strangle MOS because all you need to do is to go below pinch-off.

I am a bit proud of this design because it shows several good behaviours.

The whole thought with it all is to use it in a guitar amplifier (which I am building).

This is however even more interesting (if I may say so), it seams like the warm and nice sound of a clipping tube-amp actually has NOTHING to to with tubes, it seams like it is the output transfomer that gives this nice sound (I have FFT analysis's showing this)

To me it is kind of strange because when you calculate with it, ONLY odd harmonics is being generated while I have heard that the nice harmonics are even.

Once upon a time I built a 2X6W Williamson amplifier and it was reallly cozy to hear it clip at those tiny 6W, this is also an experience that is, power is over estimated (because I had a blast with my 6W).

Anyway, I have managed to make my KJA work, I started by using a current generator to crank the gain up but it was so unstable that I switched to a common (drain) resistor instead, this did however make my gain drop about a magnitude but a friend of mine at my swedish forum told me that a guitar amplifier actually does not need to have so much gain, input sensitivity could be as high as 500mV, he told me.

Of course this amplifier (KJA) uses an output transformer (which is kind of unique för semi conductors), the transformer is thus eliminating even harmonics and passing on odd harmonics (or so I have been told).

The output transformer (OPT) is not that perfect, I have wound 2X100 turns of 0,7mm Cu as primaries on a 3F3 ferrite toroidal core, a stupid choice actually because it saturates early with B, so I have bought a pair of iron-powder toroidal cores also, they can withstand some 3 times as much B but the 3F3 is readily wounded and I really don't want to wind again (even if I have some tools, maybe you can call it needle-shuffles).

Right now I don't know what to do, the only thing I have learned is that when it comes to electric guitars, there's not much information below some 300Hz.

300Hz is also about the saturation frequency for my 3F3.

Best regards, Roger

 

[Harald Kapp]

the warm and nice sound of a clipping tube-amp actually has NOTHING to to with tubes

Rumor has it that this sound comes from teh difference in clipping characteristics between tubes and bipolar transistors. FETs have a clipping characteristic more similar to tubes, therefore the similarity in sound.
The output transformer also seems to play a role, as you observed.
http://www.hawestv.com/amp_projects/amp_solid_tube/tube_sound1.htm

 

[WHONOES]

You appear to be measuring the output without the transformer connected. If that is so then what you are showing on your scope trace apart from the clipping on the + rail is crossover distortion form the output transistors not being properly biased.
How many watts are you expecting to get from your amp?
If I remember correctly, bottom E (the top string) on a guitar using standard tuning is E2 which corresponds to 82Hz.
Sound consists of a fundamental frequency plus many harmonics which can be higher or lower than the fundamental. So your 300Hz is a bit out. I used to build my guitar amps and speakers so that they would go down to about 40Hz or a bit lower for Bass.

 

[rogerk8]

You appear to be measuring the output without the transformer connected. If that is so then what you are showing on your scope trace apart from the clipping on the + rail is crossover distortion form the output transistors not being properly biased.
How many watts are you expecting to get from your amp?
If I remember correctly, bottom E (the top string) on a guitar using standard tuning is E2 which corresponds to 82Hz.
Sound consists of a fundamental frequency plus many harmonics which can be higher or lower than the fundamental. So your 300Hz is a bit out. I used to build my guitar amps and speakers so that they would go down to about 40Hz or a bit lower for Bass.

Hi WHONOES!

You are right, I am temporarily measureing how the output Power MOS Transistors are working, I have carefully adjusted the bias so that the transistors just about conducts and while it is pure Class B crossover distorsion is expected, then I have assembled two power resistors of 18 Ohms in place of each primary half of the output transformer.

Then I have tested using single-ended 18VDC as supply (actually, my plan is to use two rechargeable 9V HR22 batteries to supply this emerging guitar amplifier of mine called the KGA) and putting 120mVpp (without the current generator) as input signal with the result of hardly output clipping at som 30Vpp, sensitivity is thus some 120mVpp.

How many watts are you expecting to get from your amp?

Each transistor will be loaded by 2 Ohms, maximum current will be some 6Ap, maximum voltage swing will be some 15Vp and in an ideal world this would mean 15X6/2=45W (because total swing is 12App and 30Vpp).

But really, I don't believe these figures myself

Especially when it comes to those tiny batteries supplying 6Ap

If I remember correctly, bottom E (the top string) on a guitar using standard tuning is E2 which corresponds to 82Hz.

This is extremely interesting information, I thank you for that!

Sound consists of a fundamental frequency plus many harmonics which can be higher or lower than the fundamental.

I'm sorry but you are wrong here, harmonics are always higher frequencies, the only time they can be lower frequencies is by using a mixer or another unlinear component.

Once again I thank you for your interest in my project (looks like 300Hz is a bit optimistic).

Best regards, Roger

 

[rogerk8]

Rumor has it that this sound comes from teh difference in clipping characteristics between tubes and bipolar transistors. FETs have a clipping characteristic more similar to tubes, therefore the similarity in sound.
The output transformer also seems to play a role, as you observed.
http://www.hawestv.com/amp_projects/amp_solid_tube/tube_sound1.htm

Hi Harald Kapp!

I was about to show you with formulas that you are wrong if you think that clipping characteristics of the active element is relevant, because how can it be if you have a perfectly balanced transformer which I wanted to show you ONLY gives odd harmonics.

Of course these harmonics can be more or less due to the active element but there will still be only odd harmonics.

It is thus questionable if the similarties between the characteristics of penthodes and semicondutors would be even slightly audioble, for triodes I do however reserve myself.

Finally, why isn't it possible to code in LaTex?

Many discussions, like this one, can be made clearer by using formulas.

Best regards, Roger
PS
Please change the title on this thread to "JFET Power Amplifier With Output Transformer", thank you!

 

[WHONOES]

No. The harmonics can be lower as well as higher.
Some transformer distortion is caused by the non linear region when the magnetising current passes through zero and produces harmonics.

 

[rogerk8]

I am not certain that I believe you but remeber that I am not an expert, just a glad amateur.

That being said, the only way I can think of to produce frequencies below a certain frequency is to have TWO frequencies and mix them, ordinary distorsion like clipping produces HIGHER frequencies, NEVER lower frequencies.

You have no mixing properties in a power amp built around an output transformer because you have ONE frequency at the time like every periodical signal can be described as containing a waighted sum of harmonics.

Remember, this is just what I think.

Best regards, Roger
PS
I think I contradict myself here

But, as almost always is a forgotten fact, can you hear it?

 

[WHONOES]

Google sub harmonics.

 

[Harald Kapp]

Finally, why isn't it possible to code in LaTex?

I don't know. Possibly because this forum software assumes most people wouldn't know how to do it anyway?
You can always upload the equations as image.

 

[Harald Kapp]

You have no mixing properties in a power amp built around an output transformer because you have ONE frequency at the time like every periodical signal can be described as containing a waighted sum of harmonics.

While I'm with you that clipping alone will not prodice subharmonics, you should consider that the non-linearities in the amplifier (and the transformer, if you insist) act as multipliers and thus are able to create subharmonics from the various frequency components in the clipped signal.

 

[WHONOES]

Also, having a quick gander at your transformer, there is way to little magnetic material in it to produce any substantial output at low frequencies. Ferrite toroid's are not really suitable either. You really need a laminated transformer. Breaking down and re-winding an old mains transformer would be a better prospect or, go for a more conventional output stage that doesn't need one.

 

[rogerk8]

Also, having a quick gander at your transformer, there is way to little magnetic material in it to produce any substantial output at low frequencies. Ferrite toroid's are not really suitable either. You really need a laminated transformer. Breaking down and re-winding an old mains transformer would be a better prospect or, go for a more conventional output stage that doesn't need one.

Thank you for your reply!

Using my small ferrite was kind of a mistake but the reason was the heresay that guitars do not produce frequencies below some 300Hz and if that is so, even my tiny ferrite core will do the trick.

BUT if it is as you say, not even a iron-powder core of roughly the same size will make it work, I would need too many turns and the loss will thus be too high.

Breaking down and re-winding an old mains transformer would be a better prospect or, go for a more conventional output stage that doesn't need one.

Here you are missing the whole point.

I will even use an adjustable regulator for the centre-tap (and thus the drains) of the transistors, actully this feature will be my only volume feature and volume is thus set by the level of voltage over the drains giving the possibility to adjust volume while keeping the same amount of smoot clipping push-pull distorsion (the volume knob of the guitar is the real volume, this knob thus only regulates distorsion AT that certain volume level).

You may view this feature as a "volume independent distorsion" which I actually know from a friend of mine who plays in a blues band that, as he pointed out to me, he had to buy a guitar amplifier with less power to get the wanted distorsion at some lower level, he actually did that and was happy about it.

So volume independent distorsion seems to be a wanted feature among musicians.

Best regards, Roger
PS
He actually experimented with fibreglass walls around the amp to lower the level of the sound while keeping a certain distorsion that he liked (that made his trousers flutter, he said), he showed me youtube videos about this and I almost laughed and built him an L-Pad attenuator which he however refused to try.

 

[rogerk8]

I don't know. Possibly because this forum software assumes most people wouldn't know how to do it anyway?
You can always upload the equations as image.

I think this is a nice and also kind of fun answer, I thank you for this!

On the other hand I kind of wonder what company I am in that do not know how to code in some basic LaTex, the whole Wikipedia is built around it.

The lack of LaTex in this nice forum draws down the grade from 3 to 2, which still however is good

Best regards, Roger
PS
Where are you placed? I mean, am I talking to USA och England now?

 

[Harald Kapp]

Where are you placed?

Last time I looked at my Avatar it said Germany.

 

[WHONOES]

When I played many years ago, I used a clean amp and then used distortion generators to give required effect.
If I remember correctly, the Kinks got the distortion they wanted by cutting slits in the speaker cones.

 

[rogerk8]

Last time I looked at my Avatar it said Germany.

Diplomatically and funny aswered, but I was kind of thinking of the headquaters of this nice site.

When people around the world can be members (which is amazing and fun) you do lost track of where the site is located/created/stored.

I mean, we talk english here

So a better question is problably, where is the server placed?

Best regards, Roger
PS
I wish to have my swedish flag in my Avatar also, how do I do that?
PS2
Thank you for changing the title!

 

[Harald Kapp]

The site is based in Great Britain, Ian is the boss here (or maybe his wife ), whereas members are from all over the world.

To add your country's flag to your avatar:

On the top right of any paint hover the mouse over your name until a pop-up menu appears.
Select 'personal details' from the list (top left menu entry)
Select your country from the list of known countries in the lower half of the page, then save your changes. That should do it.

 

[rogerk8]

My greatest inspiration when it comes to (tube) amplifiers is someone called D.T.N Williamson.

Sadly I do not know so much about the guy but I really think that his design is the best in the world, ever.

I have built several "copies" of his design since 1991, finally I got tired of copying his (and others) design so I have now designed a tube power amp of my own.

I does however use some of Willamson's tricks.

I wish to say this because my KJA (this tread) uses his topology but with JFETs which are very similar to pentodes, I wanted to use JFETs as power devices also but firstly they are hard to find (I do have some LU1014Ds though) secondly they are hard to "strangle" with auto-bias because my KJA is supposed to be a pure Class B amplifier for street musicians.

A MOS is much simpler to bias for the region of zero drain current because it is a voltage above zero.

I have thought a lot about this and I am really not so keen on using MOS, but if I wish to use my power JFETs I think I really need fixed bias (negative voltage that is).

However my KJA works with single ended voltage only so I would have to incorporate an inverting regulator (which however is not so difficult to build, especially if current draw is small).

A reson for me to tell you this is that I think Williamson was english, I know that he recommended the 807 as output tube instead of his preferred Osram KT66 because USA did not have KT66 at the time (1947).

My KJA is built very similar to his tube amp but I have to differences and the first is that I use a BJT current source for the first stage to crank up the gain, this however made it difficult to trim the pre amp stage, a 10-turn trimmer made it almost work but it was not satisfactory so I switched to a simple resistor and lost more than 10dB of gain BUT it was stable.

The other difference is that his patented (?) Cathodyne splitter could not be used with semi conductors (JFETs) so I used a circuit which I don't know the name of but it involves two resitors i series as source resistor, tapping to gate in between.

In the end it did however got clear to me (due to my swedish forum) that the gain in a guitar amplifier really doesn't have to be so high, a sensitivity of some 500mV seems to suffice (my sensitivity is 120mVpp, with resistor only, but I am planning on some 12dB of feedback, reducing effektive output impedance some 14dB).

Now I need to decide what type of OPT I will use.

Today I calculated a kind of interesting thing (while googling on sub harmonics), a string on a guitar can not give a frequency below a wavelengt of 2L, let's say that the guitar strings are 1m (=L) then the lowest POSSIBLE frequency is 340/2=170Hz and this has to do with a swinging string needing a MINIMUM of two nodes.

So to design a guitar amplifier that goes below 170Hz is not really useful.

The character of the strings are very different BUT this has to do with harmonics ABOVE the fundamental frequency I just calculated.

In other words, sub harmonics does not excist in the practical world, if I may say so.

Look at it this way, the transformer core is of course unlinear and it can mix signals and while a periodic signal can be described in a Fourier series the harmonics are ALWAYS higher than the fundamental.

To give rise to say f/2 you will have to mix some f with 3/2f but the strings on the guitar gives trise to signals that ALWAYS are higher than the fundamental and I doubt that one string can have a frequency of 3/2f in comparison to the fundamental, frequecies are more like f, 2f, 3f and so fort.

If there even are "sub harmonics" they will never be audible because the mixing feature is not that effective.

Best regards, Roger
PS
Harald Kapp, I don't understand what you are saying. Look at it this way, I am "datadyslectic"

 

[rogerk8]

I have now studied my more than a year old notes on EF (my swedish electronics forum) when it comes to the toroidal transformer I have manually winded for this purpose.

A not so bright thing is that I am using a ferrite core while ferrites can not widthstand so high a B (Magnetic Flux Density) as iron (powder) cores can.

The difference is somewhat trefold (0,5T for ferrites, 1,5T for iron) which mean that the core of my 3F3 ferrite will saturate at a frequency (keeping voltage the same) some tree times higher than iron.

Now, I winded my ferrite core almost for fun, using a needle-shuttle a friend of mine from EF manufactured, by thinking that (out of experience) I needed to cover the whole windings (to avoid resonances) with wire so due to this I could not design the exact turn-ratio I wanted so I was stuck with a turn-ratio that wasn't "perfect".

But somehow this unplanned turn-ratio actually got perfect anyway

For one thing, the load-line intersects a lower current which is more possible for two 9V (HR22) batteries to supply but of course maximum output power is less yet it is some 20W.

Some other data of this winded ferrite core is:
1) 2Np~180 turns
2) D(wire)=0,7mm
3) Ns~180/2,4=75
4) fl=20Hz@1,5Vpp
5) fh=360kHz@1,5Vpp
6) fl(B)=250Hz@36Vpp

For lower voltages 6) gets lower, normal listening level I think is somewhere around half the voltage, this means that core-saturation starts around 125Hz.

So down to some 100Hz KJA will work very well and core-saturation as distorsion feature might even be wanted in a guitar amplifier, i don't know but it would be an interesting experience.

In other words, I don't feel a change to iron powder core as that eminent.

I would very much like to change but I will actually try with this ferrite first.

Best regards, Roger

 

[rogerk8]

Today I managed to do a big thing.

Recently I attached my ferrite-transformer to my KJA in hope of evaluating frequency performance.

However, all I got was crap.

It oscillated with both HF and VLF.

I am not so sure about these things so I got confused but my old tube-amp friend Williamson helped me a lot.

To get rid of the HF-problems I just needed to put a capacitor around the drain resistance of the first gain stage, which I did and the HF-part of the problem disapperared.

This capacitor is however probably a bit too large but at this stage I don't care.

To get rid of the VLF-problems I had to think for a while but once more my dear friend Williamson helped me because he shows an eletrolythic capacitor over the push-pull driving stage supply.

So I thought that maybe if I put some 1mF over the push-pull driving stage supply my problem would disappear.

Guess if I was sceptical of this BUT to my great surprise it worked!

It worked and I am very pleased about this, now I only have to create a new transformer because the current transformer works but "flux density distorsion" begins at such a high frequency as 500Hz (full power).

I am however not totally disappointed in my 3F3 ferrite toroidal transformer and here is the data:
1) f(B, 4Vpp out, 4 Ohm)=200Hz (core distorsion)
2) f(B, 15Vpp that is max out, 4 Ohm)=500Hz (core distorsion at maximum output level)
3) fh(4Vpp out, 4 Ohm)=27kHz (decoupled by preliminary capacitor, this capacitor can probably be smaller)
4) fl (4 Ohm)=83Hz (I have decreased the level very much to not make the core go into saturation)
5) Pmax(1kHz, 4 Ohm)=14W (not so bad )

Best regards, Roger
PS
I insist that the lowest frequency a guitar string of some 1m can generate is 170Hz so i need not make my system work for much lower frequency than such and while iron powder have some three-fold higher B capability some 500Hz/3=170Hz will be possible, this is almost perfect (except for the winding of the toroidal transformer which is kind of awkward).