TGM Amplifier ?

[MJL21193]

Originally posted by Bigun
However, I read a posting in a forum that it uses 68k as the default load for the LTP. Looking at the limited data I have for 2N5401 I don't see why the lower current (0.5mA) would be a problem (does it hurt linearity or frequency response ?)

Looks like you are starting to have fun
Here's my take on it:
There isn't a default load for the LTP. You need to determine what the current through the stage should be and make it so. To do this, you look at the device current gain plot. Ideally, you want to be smack dab in the middle of it's most linear range, given the operating temp and the dissipation. See the attached chart.

[Bigun]

That makes sense - you don't want the gain varying with the signal current. This seems to be one of the limitations that it's necessary to work around. As you say, you have to pick a spot. It would be interesting to put a pot in here and listen to the difference it makes to the sound. But the current impacts the dc operating points of other parts of the amplifier.

It would be nice to simulate this in advance, but.....

There are virtually no spice tools for the Mac. The only one which runs and uses a decent graphical interface (rather than some unix command line mode) is called MI-Sugar. It's no longer supported, you will be hard pressed to find it. And so far I have been unable to get it plot any results or run complex analysis.

Ccan't give up.

[Bigun]

update...

seems that 'ac voltage source' and 'sinusoidal voltage source' are different things. Now I can run an ac analysis (small signal). No plots though (darn). I think what I see is that if I inject a 1V amplitude signal to the input I see that it appears on the base of T1/2. About 1mV of this appears at the base of the PNP. None of it appears on the supply rail.

[ostripper]

It's a shame you have just have a mac. With windows I can't run
mac programs. For linux ,"wine" can virtualize a win32 program.
(I CAN run LT4 on my Ubuntu Linux)

Here is something that might help,
http://www.codeweavers.com/products/cxmac/

It let's win32 programs run on a newer mac (intel based)
never tried it but it might work.

MJl was right, the 2n5401's work far better at 2mA.

I never built a amp without a current source , so your
TGM got me curious. I took out one of my CCS's and
cranked the other channel... I could definitly hear
crosstalk(the other channel).. so , you might want to reconsider current
sources unless you are going to use totally separate
PS's for each channel . also , Distortion is 5X without
the CCS because too much "garbage" re-enters at the LTP
emitters , modulating the gain,
which even negative feedback can't fully compensate for.

Some amps use a separate regulated supply for just the VGS, (LTP and VAS) and a 'brute" supply for the output section.
My new one will be adaptable for that. (attached) note the
split on the side where the V- goes...there I could take
your route, as the supply would already be "clean" with
little load (ripple).

The original destroyer X amp is like yours , but carlos added
a zener to the current "side" of the LTP that acts as a primitive regulated source.(I think his original inspiration was the aska55
as well).

OS

[Bigun]

Good catch! Power Supply: I only need 25W channels (so I should be thinking about +/-24V supply?) to drive my single driver DIY speakers; a single toroid would work. I don't think AKSA, without it's CCS was designed for such an environment. For HT I don't worry about some cross-talk, operated bridged for stereo and I may have to worry.

But...the update to AKSA 55 is LifeForce 55, which does use a CCS and people have reported that it doesn't sound as sweet as the AKSA 55 ? I am convinced that one of the 'secrets' is having distortion, just has to be the right kind. Actually, I haven't read anybody claiming to have an amplifier based on this topology that sounds as sweet as AKSA at the hi end so it's a tough target !


I will keep CCS as an option.

p.s. I'm checking out that link...

[Bigun]

So here we go with improved current source. I didn't like the zener idea, not because of any personal experience but because I read that they are noisy - especially those that operate above 5.6V where they rely on tunneling instead of breakdown. There is a way to fix this by adding a transistor, but then you have to add another diode for temperature compensation. Needless to say I picked the non-zener approach. I have thrown in a pot so that I can 'play' with the LTP current and use this to fine-tune the o/p offset.

[MJL21193]

Originally posted by ostripper


I never built a amp without a current source , so your
TGM got me curious. I took out one of my CCS's and
cranked the other channel... I could definitly hear
crosstalk(the other channel).. so , you might want to reconsider current
sources unless you are going to use totally separate
PS's for each channel . also , Distortion is 5X without
the CCS because too much "garbage" re-enters at the LTP
emitters , modulating the gain,
which even negative feedback can't fully compensate for.

?
Funny, I have an amp design with NO active current sources.
I must have been lucky (or did something right) - PSRR plot below.

Plays out in real life with 6 REAL amps on the same PS, no crosstalk.
And distortion? I think we already know the answer...

[Bigun]

I've made some progress in that I've persuaded my Mac to do a dc operating point analysis on what is nearly the full schematic.

1) Trouble is, to get balance (offset at output still too high at 97mA) the LTP it's running at 19mA (too high no doubt) so is the recommended method to address this with a collector resistor on T2 ?

2) I don't have Spice model parameters for several devices - can anyone help me here ?
2SC5200
2SC1943

2SA1837
2SC4793

2SC1819

p.s. I've tried running small signal 'ac' analysis. I need to read up on this a bit more because it looks as if it's telling me I have unity gain - oops I think I have 100% NFB

Thanks for the help so far !

[ostripper]

By mjl- Funny, I have an amp design with NO active current sources.

But you do have a cascode/current mirror. that is not the same
as having nothing, as on the TGM.
OS

[sajti]

Originally posted by Bigun
1) Trouble is, to get balance (offset at output still too high at 97mA) the LTP it's running at 19mA (too high no doubt) so is the recommended method to address this with a collector resistor on T2 ?

Change R9 to 47k. It will help to keep the DC balance.

Sajti