[qusp]

so you want distortion then? what does 'tubey' even mean? I see that term used quite a lot and still have no idea what people mean when they say it and discrete solid state sure doesnt sound 'tubey' is it perhaps used to describe a sound like music is being played down a tube?

 

[diditmyself]

When you wish for a portable MOSFET amp, do you mean MOSFET output, or an all MOSFET amp?

When MOSFETs are used in the output stage there are several issues that make them unsuitable in a portable amp:

1. The size. You'll have a hard time to find small complementary parts. Most of them are TO220 (or larger) like IRF610/IRF9610 that would be the parts of choice, i think. There are some integrated N-MOS/P-MOS SOIC parts, but I don't know if they're good for audio. I've been thinking of trying them just for fun. My local dealer ELFA stocks PHC21025 from NXP.

2. The voltage loss, several volts compared to the 0.65 V in bipolars. To get a decent voltage swing, you'll need at least 18 V supply. That means 2 x 9 V batteries.

3. As stated before, there seems to be an oppinion that MOSFETs need higher biasing to give their best and 6F22 batteries wont do.

So, if you want to make a good portable MOSFET amp it will be quite large. IMO a bipolar diamond buffer is the best option for output stage in a portable amp. Discrete or monolithic like LMH6321, LME49600 or BUF634.

 

[qusp]

I dont think the OP had any idea what he really wanted when he started this thread and certainly had no idea of the choices involved or $$$. he actually just meant discrete when he said mosfet. I already suggested some SMD discrete diamond buffers for output in his P3+

 

[diditmyself]

I've built some discrete portable amps. Sometimes I think they're very good. Sometimes I find opamp based amps superior if opamps like OPA627, OPA827, AD797 or LME49720 are used.

The simplest yet OK sounding, and stable as the voltage drops, would be a class A biased JISBOS with gain. Only 7 transistors per channel. 8 AAA batteries will give an acceptable voltage swing. It's questionable whether this is better than an opamp, but it could be a fun project. I have something like this in a Hammond J1201 with 10 AAA's. Unfortunately I removed the ground channel in favour of two extra batteries and now use a discrete rail splitter instead.

The problem is that to improve on the sound of the best opamps, the topology has to be complex, has to be run at high currents and needs at least 12 V supply to get 4 - 5 volts output swing att full charge. 18 - 20 transistors, about as many resistors, some diodes and capacitors per channel, 10 AAA batteries, pot, jacks, trimpots will make it impossible to be made very small. It could be made smaller than a LISAIII though. It'll be interesting to see how Triad's discrete amp will look like. I think it's called the Portacode.

 

[TzeYang]

1 Ohm degeneration for the VAS stage seems to be a risk for thermal runaway

Sure they are stable IRL?

 

[qusp]

no, the portacode is another thing altogether, much of the space in lisa III is batteries, but there is a small amount of space left. they have gained space back by using small signal transistors instead. its simply called lisa III mini

 

[diditmyself]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif 1 Ohm degeneration for the VAS stage seems to be a risk for thermal runaway Sure they are stable IRL?

I don't know. In real life I use something like 10 - 20 R, I can't remember the exact value. Lower values yields better voltage swing wich is important if going portable.

Have a look at le Monstre The Class-A Amplifier Site - Hiraga 'The Monster' No emitter degenaration on the output transistors either.

 

[diditmyself]

Just to show what's possible with throughhole parts. This is a class-A 3 channel amp powered by 8 AAA's. There's a built in charger and voltage regulator. The circuit is a "standard" folded cascode, like an OPA-Earth or Burson with diamond buffer output stage for more current, or a discrete AD823. It's made for Hammond J1201, but since the price of them has gone up, I've DIY'ed boxes as well. The circuit is suboptimal. To get it the way I really want, I'd need 10 AAA's and some more parts. It might be possible with some SMD parts and two boards on top of each other.

 

[qusp]

nice work, pretty much what i'm doing but with a dac/amp. just ordering the parts now, what regulation circuit did you use for the charging if you dont mind me asking? i'm putting together a build with COD from twisted pear its a PCM1794A dac with i2s input and ASRC, discrete diamond buffer output and if I can find enough current i'd love to put a folded cascode in there too, mine will be balanced output, so i'm struggling a little with volume control of any quality. i'm hoping 10 cells will be enough, but I think it might have to be 12 aaa NIMH

 

[diditmyself]

Quote:

Originally Posted by qusp /img/forum/go_quote.gif nice work, pretty much what i'm doing but with a dac/amp. just ordering the parts now, what regulation circuit did you use for the charging if you dont mind me asking? i'm putting together a build with COD from twisted pear its a PCM1794A dac with i2s input and ASRC, discrete diamond buffer output and if I can find enough current i'd love to put a folded cascode in there too, mine will be balanced output, so i'm struggling a little with volume control of any quality. i'm hoping 10 cells will be enough, but I think it might have to be 12 aaa NIMH

I use an LM317T with 100 mA constant current charge. I don't have the skills to do an "intelligent" charger. The regulator is a simple L7815 or L7818. To make an amp smalller or to cram in more amplifier parts, one could have the charger and regulator on the outside.

10 cells will be plenty enough, especially if your amp is balanced. With a diamond buffer output stage and 1.2V drop from positive rail to the folded cascode bases, my amp swings to within 1 V from rails.

To improve the "standard" folded cascode topology, I'm using floating current mirrors à la AD797 in one of my amps. I think it's an improvement soundwise, and the distortion is lowered. In one amp I have a CFP input and in another a cascoded input. Both of these tweaks improve the sound. What I would like to do is a "no memory distortion" = constant power input stage with both CFP and cascode, and use a floating current mirror for the common base VAS. To do this I need higher supply, at least 12V. I've found that AD797 (both the real deal and my clone) behaves weird at lower voltages. It's possible to lower the measured distortion by using active current sources/current mirror on top of the LTP, but IMO it seriously degrades the sound.

 

[diditmyself]

This is what I have in mind. I think this complexity is needed to match the best monolithic opamps, and it's not only complexity that's needed. The components have to be choosen carefully, especially the transistors. I never liked the amp with BC549/559, blurry in a way. With SA970/SC2240 it was too bright. SA1016/SC2362 is perfect for my taste. I know that BD137/138 like I use now aren't really up to the task either, but I haven't had the energy to order some nice Sanyo output transistors.