Help: Mosfet SE Amp

[mac]

I have a simple 18v SE amp in the breadboard.
The fet preamp is like the Epi Valve Junior, two fets BF245A, Q1 and Q2.
The power amp is a MTP3055 mosfet Q3, and a 3:1 OT I wired for this project.

I read that lateral mosfets are better for audio, and once the gate "is driven correctly", whatever that means, mosfets sounds very good, at least for hifi.
I couldn't get laterals, I have a 3055 and a IRF740 to experiment.

I have a question for the gurus,

1) If I connect Q3 gate to the drain of Q2 (15k from drain to Vcc) through a 22nf, and bias Q3 with a 220k to gnd and a 1M pot to Vcc, it sounds bad, fizzy, not in focus, and with some ringing. With a IRF740 is like a Green Ringer
A series resistor with the 22nf cap is even worse.

2) But if I use a 3rd fet Qb as buffer, Qb gate to Q2 drain, Qb drain to Vcc, a 15k from Qb source to gnd, a 22nf from Qb source to Q3 gate, and bias Q3 in the same way as above, it sounds good.

The parallel resistance of the bias network is 150k. Is the lower impedance of the buffer making the difference, or is it something else?

mac

[PRR]

It is probably oscillating.

Layout may be a factor. Keep OT leads *far* away from the rest of the amp, including the power stage grid/gate.

[Morocotopo]

Don´t know if it´s any help, but, I made an amp with a B+ voltage regulator using a big power MOSFET, and it oscillated like hell until I put a 10K or so directly in front of the gate, then it behaved perfectly. Seems they are quite finicky regarding gate capacitance.

[mac]

It is probably oscillating.

Layout may be a factor. Keep OT leads *far* away from the rest of the amp, including the power stage grid/gate.

Don´t know if it´s any help, but, I made an amp with a B+ voltage regulator using a big power MOSFET, and it oscillated like hell until I put a 10K or so directly in front of the gate, then it behaved perfectly. Seems they are quite finicky regarding gate capacitance.

I did what Paul suggested and the sound improved, at least for clean tones. Distortion is still bad.
Maybe I need a lateral mosfet designed for audio, not for switching?

I tried the SS tube idea, a fet buffer direct coupled to a TIP41, instead of a power mosfet. Easier to work with. And it sounds way better this way.

BTW, I tried a 36v and a 45v supply. As voltage goes up it sounds better. It's like a camera with more pixels.

mac

[tca]

Can you post the schematic... it would help the discussion.

The way I usually do is to bias Q3 not with a voltage divider (from Vcc to ground) but with a typical self bias arrangement (resistor from drain to gate no more than 10k).

These types of class A does ring a lot!

I bet that sounds really good clean and really good VERY distorted The frontier is some how muddy and ringing?!

You need to limit the bandwidth of the amp. Cut the bass at 150Hz before the IRF and cut the treble after it. To cut the treble the simplest way is to put a 10u to 100u (to ground) from the secondary that connects the OT to the speaker.

Questions: Why a 3:1 OT transformer? Made some experiments with a 1:1 and works nicely.

What is the quiescent current? What's the primary DC impedance?

Cheers.

[mac]

This is the Mosfet version,



R10 begins to bias above 220k. It's working at about 200ma. Almost 7watts across the mosfet.

The OT was designed for 100-200ma @ 36v-45v, and it begins to sound good at those currents. IIRC it has 210:70 turns of thick wire, a tiny gap, and the size is like the one in my Valve Jr.

You need to limit the bandwidth of the amp. Cut the bass at 150Hz before the IRF and cut the treble after it. To cut the treble the simplest way is to put a 10u to 100u (to ground) from the secondary that connects the OT to the speaker.

I tried caps across the speaker, across the primary, at the mosfet gate, etc.
The mosfet becomes a 555 with a small cap from drain to gate: toc toc toc

These types of class A does ring a lot!

I bet that sounds really good clean and really good VERY distorted Wink The frontier is some how muddy and ringing?!

Exactly.

You can hear how it begins to break when the amp volume is at 12 o'clock.
I had to remove Q2 cap because it has too much gain.
With a TS808 (drive low, vol high, tone high) it's annoying the whole building.

I have to draw the *SS tube*.

mac

[tca]

^ Thanks for the schem.

My amps run at .5 to 1.5A class A but with a resistor load, much more heat.


You get 2W out of it?

Try these mods:



As long as your ears are used to ringing check the following sounds.

There is enough ringing there in any Plexi, that's the kind of distortion you are going to get from a class A mosfet, but you need to drastically reduce the bandwidth of the power amp.

Note also that sometimes we expect to much from one single device

Cheers.

[mac]

TCA,
I tried your mods.
A lower input Z helps to reduce lows, but C2 must be higher.
Taking 10k from the drain makes it hum a lot above 50ma. No hum at all from Vcc to gate.
As R10 goes up hum goes down, and Zin increases...
I ended with a 1M from D to G and a smaller C2.

A 4.7-10 ohm before the OT is fine.

Compared to a BJT, this thing is too gainy and its gate too sensitive that I can hear the wings of a butterfly in the Pacific Ocean

I'll try to get another mosfets.

Note also that sometimes we expect to much from one single device Wink

one 12ax7 and one EL84 work

mac

[tca]

^ Thanks for testing that.

Taking 10k from the drain makes it hum a lot above 50ma. No hum at all from Vcc to gate.

I think you need to RC decouple the first gain stages from the power amp and also reduce the gain of that BF245A stages.

What kind of iron (E's and I's) did you use for the transformer? A salvaged power transformer?

P.S.

Note also that sometimes we expect to much from one single device Wink

one 12ax7 and one EL84 work

ah, but those are completely different animals.

[mac]

I think you need to RC decouple the first gain stages from the power amp and also reduce the gain of that BF245A stages.

I was thinking of another RC decoupling. Not needed with a BJT.

What kind of iron (M's and I's) did you use for the transformer? A salvaged power transformer?

Yes.
5x5x2cm or so. It sounds creamy with a TIP. And even filling the tiny gap it does not saturates at 300-500ma.

mac

[Rob Strand]

What about just upping the bias current does it improve or even change the sound?

Sometimes adding a resistor in series with the gate of the MOSFET can help tame oscillations.
Depending on you layout you might need to add small caps to the earlier stages to roll-off high frequencies.

[tca]

BTW, what speaker are you using? Bass roll off? And pickups? From my experience single coils works best with large IM distortion.

Cheers.

[mac]

What about just upping the bias current does it improve or even change the sound?

Above 300ma I get that "mmmmmmmmmm" sound.

Sometimes adding a resistor in series with the gate of the MOSFET can help tame oscillations.

I did it but no, it sounds muddy.

Depending on you layout you might need to add small caps to the earlier stages to roll-off high frequencies.

I'm using 47pF at Q1,2 gates. I tried small pF at Q3 too.

BTW, what speaker are you using? Bass roll off? And pickups? From my experience single coils works best with large IM distortion.

Greenback and/or a Laney HH, Ibanez+3xDiMarzio.
I roll basses off with C3 or a clean TS808. C1 filters some lows too.

Clean is fine, and superdistorted... well, some metal head might like it. The problem is in the middle.

*******************

Today I tried something different to tame highs.
Are you familiar with those small inductors used in compact fluorescent lamps? With the core they are about 3-5mH, and 1-3 ohms.
Instead of R11, I'm using a coil without the core and things are way better.
This mod and TCA idea of a resistor before the OT made a difference.

mac

[tca]

I've been thinking about your transformer build. Can you measure the inductance of the primary? One should use the bass roll-off of the transformer to kill the low freq intermodulation. With a cut-off of 80Hz I would say that you need a 50mH primary inductance (or maybe less would do).

[mac]

Can you measure the inductance of the primary?

I just have my cheap made in china DMM
But i think it can be estimated using a simple RCL resonant circuit tuned to a known freq, say 1khz.
RCL network --> LM368 -->speaker

One should use the bass roll-off of the transformer to kill the low freq intermodulation. With a cut-off of 80Hz I would say that you need a 50mH primary inductance (or maybe less would do).

It's working better at 0.3A. Let's say at 0.36A. Z = 36V/0.36A=100ohm
Np = 3.5*Ns. The real thing is 3:1 IIRC, a higher current will match, but higher than 0.3A I get hum that maks the sound.
Anyway, above 0.15A there is more and more volume but no noticeable sound improvement.

I guess I calculated the primary inductance as
L = Z/w = 100ohm/(2*pi*80hz) = 0.2H
That's 4x the value you suggest.

Since I didn't know Km of the core, I used a value of 1000, and with the geometrical data I estimated N around 200-300 for the primary. I wired 210 and 70 turns IIRC.
If Km is higher, L is bigger too.

SO, without any measurement, by ear I can say the OT passes lots of lows.

mac

[tca]

It's working better at 0.3A. Let's say at 0.36A. Z = 36V/0.36A=100ohm
Np = 3.5*Ns. The real thing is 3:1 IIRC, a higher current will match, but higher than 0.3A I get hum that masks the sound.
Anyway, above 0.15A there is more and more volume but no noticeable sound improvement.

I guess I calculated the primary inductance as
L = Z/w = 100ohm/(2*pi*80hz) = 0.2H
That's 4x the value you suggest.

Since I didn't know Km of the core, I used a value of 1000, and with the geometrical data I estimated N around 200-300 for the primary. I wired 210 and 70 turns IIRC.
If Km is higher, L is bigger too.

SO, without any measurement, by ear I can say the OT passes lots of lows.

The primary of the transformer has two impedances: Rp +Xp

Rp is what you are measuring, the voltage drop (DC voltage, f=0) at the primary, this quantity is not Xp=2*pi*f*L. The turn ratio relates the Xp of the primary to the Xs of the secondary Xp/Xs=(Vp/Vs)^2=9 not Rp/Rs; Xp is frequency dependent, Rp is not. This voltage drop is a loss and it is released as heat at the transformer.

In an ideal transformer Rp=0 so at zero frequency the current flowing is only determined by the bias of the mosfet (have to check the datasheet, for the exact value). This means that reproducing low frequencies in a class A configuration requires a lot of current.

When you say that the turn ratio is 3:1 this means that for AC signals with a 8Ohm load the value of Xp=8*3^2=72 (I'm correcting my last calc., sorry) . For a cut-off of 80Hz this gives L~140mH.

The calculation of the number of turns is somehow more complicated... but check R.G.'s text for more details:  http://www.geofex.com/article_folders/xformer_des/xformer.htm


(edit) Corrected some bads.

[mac]

When you say that the turn ratio is 3:1 this means that for AC signals with a 8Ohm load the value of Xp=8*3^2=72 (I'm correcting my last calc., sorry) . For a cut-off of 80Hz this gives L~140mH.

You started from the secondary. I did it from the other side.
If you use a 3.5:1 instead you get Xp = 8ohm*3.5*3.5 = 98,
almost 100 ohm and 200mH, the numbers I calculated above.

I guess the mosfet impedance should be equal to the OT primary Xp. Rp can be neglected in this case as is less than 3 ohm.
Since most of the drop is in the mosfet,
Xp = (8 ohm)*(turn ratio)^2 = (mosfet voltage drop)/(bias current) = 36V/Ibias

Mine it's a rough approach, but it takes AC into account, ie, Xp.
As I can't go above 0.3A due to hum, my OT is forced to work as a 3.5:1 @ 0.3A.

Using your numbers, the optimal bias current should be higher to meet the 3:1 condition,
Ibias = 36V/72 ohm = 0.5A

The calculation of the number of turns is somehow more complicated... but check R.G.'s text for more details:  http://www.geofex.com/article_folders/xformer_des/xformer.htm

I downloaded that page to my HD years ago.

This is how I wired an OT some ago, units are in MKS,

http://www.diystompboxes.com/smfforum/index.php?topic=80047.0

mac

[tca]

So every thing adds up correctly.

Thanks for the refs on your builds, good reading.

I've tested several class A mosfet amps with OTs, but using a 1:1 turn ratio (two 12VA power transformer secondaries) @ 1A. It works nicely for "hi-fi", but it stills rings at moderate levels of distortion if used as a guitar power amp. Same happens with a resistor load at the drain.

I *guess* the capacitances of the mosfet is playing tricks when the mosfet clips. Ringing did not occur, at these levels, on your BJT power amp. Did it?

I still think that the input impedance of you power amp is to high and reducing that 33k of your common drain (Qb) to 10k or 1k may improve the driving of the mosfet gate.

[mac]

After several rewirings, I could bias with half an amp without much hum.
The mosfet is sounding much better now, less ringing.

I *guess* the capacitances of the mosfet is playing tricks when the mosfet clips. Ringing did not occur, at these levels, on your BJT power amp. Did it?

Amptramp posted something about mosfet capacitance some days ago, that it has to be driven hard to just to charge the internal caps.
And yes, BJT distorton is more natural and tube-like.
The mosfet has more gain, humbuckers distort too fast, even with no bypass caps.

I still think that the input impedance of you power amp is to high and reducing that 33k of your common drain (Qb) to 10k or 1k may improve the driving of the mosfet gate.

I'll try that at Qb.

mac