diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Solid State (http://www.diyaudio.com/forums/solid-state/)
-   -   TGM2 amplifier (http://www.diyaudio.com/forums/solid-state/145652-tgm2-amplifier.html)

Bigun 10th June 2009 05:07 PM

TGM2 amplifier
 
I want to continue my learning about amplifiers, started with the TGM1 [http://www.diyaudio.com/forums/showt...hreadid=140461], by designing and building a new design, the TGM2.

I'm going to step back and start with just one change to the LTP at first, which is to introduce the complimentary feedback pair. It's not new (e.g. Frugal amp, DX precision) but it is a good next step for my 'education' and of course it will still be different from these other designs.

Your help and suggestions/corrections etc. are much needed and very welcome ! :)

Since first posting a TGM2 design I have started to look at how the CFP works. The first thing I realized is that you need to choose the current distribution between the input device and it's CFP 'slave' and set the collector resistor appropriately. I started with things the wrong way around, so I have increased this resistor to better distribute the current. I will post a schematic.

nelsonvandal 10th June 2009 05:28 PM

CFP input is a good idea. My favourite is JFET input and bipolar "slave". I use a 680R resistor and the same current (1 mA) goes through both transistors.

destroyer X 10th June 2009 05:57 PM

The amplifier looks alike the Aksa published in this forum
 

What is the point to make almost the same?

hehehehe... strange that...maybe because you want to have fun building something may have your personal touch.

Your driver emitter resistance seems very small... drivers will be hot because of that...maybe you want this way.

Roender have used 33 ohms or 47 ohms (do not remember exactly, and i imagine that was the floor value..under that may create you problems.

Drivers may need heatsinks, or to be installed into the main heatsink

We already have Aksa published, mine Dx amplifier, maybe the GEM (do not remember exactly), MJL21193 has something alike (a little different, more to HRII), Ostripper made one, and now producing the RCA 1972 with other name having modifications that made it the same as mine, yours, Hugh and everybody.... you have TGM1 and now TGM2.... 8 amplifiers that will sound almost the same

Our forum may have to change the name of our forum.

www.diybootstrappaudio.com

hehehehe.

Very crazy thing fellows.... very strange.

Well...go ahead!

Carlos

Bigun 10th June 2009 07:39 PM

Carlos,

I like you because you push us to do better. Of course some people should try to be original, to push the knowledge forward. We should not copy other amplifiers and just put different clothes on them. When my slowly increasing knbowledge allows, I may come back here with something new, or at least less familiar than this one.

But for now I have still to learn and I am choosing a well trodden path as the best way for that.

If I may be allowed to add my humble opinion... AKSA was an inspiration for people to revisit old designs, Hugh showed that old designs offer some fertile ground for great results. Some of these other amps we are making might share topology with the forefathers of AKSA but they are not AKSA. The DX has some key differences which doesn't let it sound excactly the same as AKSA (as you have said many times with regards bass and treble) and since you have seen the real AKSA design you understand why they sound differently. You and I can both say we know something of why these differences in design result in their different sound and why the BTF amp also will not sound like AKSA but will only share the characteristics of the topology. So let's not get to hung up on the curent interest in this topology as being in anyway a bad thing, it's a great starting point for many DIY'ers and perhaps also sufficient ending point for many of them too. :)


NelsonVandal,

The TGM3 uses the JFET as you mention. I will look at that option once I have a proper understanding with the BJT. Can you offer any opinions on the sound of the JFET/BJT input verses a BJT/BJT input ?

destroyer X 10th June 2009 08:15 PM

I see..you have agreement about all you said
 

regards,

Carlos

Bigun 10th June 2009 09:24 PM

2 Attachment(s)
I have made a first schematic with the CFP but no other improvements. I have not added a CCS because I am wondering if it really makes a lot of difference ? As far as I see, it gives some improvements in PSRR and increased OLG but I'm not sure how much these improvements give me in terms of sound quality ? - I didn't see anything significant in the sims.

I also don't have emitter degeneration, the resistors are there for convenience of measuring current balance in the sim. At one point I thought I had pushed the LTP into instability without these resistors so maybe they are important for that alone. But again, nothing significant in the sims.

The distortion performance improved over not having CFP so this is already good. Maybe it's enough.

p.s. Carlos - there is a new transistor a friend of mine has recommended to me for the output transistors so there may be hope of something new that I will post here if it turns out OK.

Bigun 11th June 2009 03:09 PM

Emitter Degeneration ?
 
I need some help with a real newbie question !

Looking at the CFP LTP I can't see a need for the emitter degeneration - which is a popular improvement. First off, the reason to include it is to improve linearity through some local feedback. But the CFP has a built-in level of local feedback already, linearity is already vastly improved. Emitter degeneration is used when you want to drive more current throught the LTP to increase the slew rate (i.e. curent drive into Cdom) but need to pull back the gain this generates. But the CFP nicely uses the 'slave' devices to provide a good increase in current drive to the VAS without pushing up the gm of the input devices. I also find that Self reports most of the gains in linearity from emitter degeneration to occur at 20kHz or above, not in the 'meant and potatoes' range of listening frequencies.

One could say that the degeneration resistors are inexpensive and have no real downsides - but why bother if you don't need them ???

p.s. I don't know how much credence you guys put into the 'memory effect' theory ? [http://peufeu.free.fr/audio/memory/m...-circuits.html] which also favours a CFP input stage, but it appears key not to use a current mirror or you lose most of the benefits.

nelsonvandal 11th June 2009 03:47 PM

I think the memory distortion theory is very appealing. If I build a new amp, I'll definately make room for both CFP and cascode.

Regarding the difference between JFET/bipolar and bipolar/bipolar, I find J/B more "forward" sounding, having more "texture" and "body". Most of all it adds "presence" and it's more involving. B/B is softer, smoother and a bit vague.

One problem is that the only JFET input transistors I've used are SK170 and SJ74, so I don't really know if it's JFETs in general that sound this way or just the tested transistors. I've tried a lot of different opamps though and I find the ones with JFET input sound different from the ones with bipolar input in the same manner, with one or two exceptions. The downside with JFETs is a grainy and harsh sound, but I think the use of CFP cures a lot of this.

One of the pros with J/B CFP is the low input bias currents so you don't need input capacitors or balancing the pos and neg inputs.

I've also found that CFP input amps are unstable without emitter degeneration in the LTP.

I don't hav any good experience from using current mirrors on the LTP. The distortion is lowered but I don't like the sound.

Bigun 11th June 2009 05:29 PM

2 Attachment(s)
Thanks Nelson - a most interesting comparison. I will want to try hearing the differences so I should try and design things to allow me to try both BJT and JFET input.



I had a play with Spice at lunchtime again:

I decided to look more closely at a CCS fed LTP as an option. I'm not sure I see enough benefit, would be grateful for some real-life comparison in term so the sonic benefits of a CCS vs resistive feed to a CFP LTP - anyone ??

One touted benefit of the CCS is a reduced turn-on thump. I tried to simulate this in Spice with zero input and a quick ramp in supply rails. The simulations show how the voltage at the LTP power node rises as the rail filter cap charges up. But the turn-on thump was no different at the output compared with a resistive LTP feed. ??


However, I was able to initiate some nasty oscillations which started some time after the initial turn-on thump.

I found that the LTP oscillations occur only when I have a CCS (a perfect CCS in Spice). With a resistive feed to the LTP there are no oscillations. With emitter degenration of the LTP (33R - 47R wasn't enough, 82R was marginal, 100R OK) there are no oscillations. The triffer was the interaction with the VAS. I see that as the LTP power rail comes up the LTP output voltage ramps away from the -ve rail, as it should. This output feeds the base of the VAS and when the voltage reaches a point where the BE junction of the VAS is forward biassed the LTP oscillations start.

So why do I want a CCS at the top of the LTP ???



edit: attachment shows VAS input (green is CCS loaded LTP, blue is resistor loaded)

AKSA 11th June 2009 10:28 PM

Gareth,

I'd guess the reason is that you have a phase reversal before the OLG has come back to unity - that is, you have not met the Bode-Nyquist criteria. The fact more degeneration seems to correct it is clear indication your lag compensation may be the problem. Increase Cdom!

In a standard LTP, as one device turns on, the other turns off, with a powerful link between the two. In a CFP LTP, each control device is run at constant current (that is the strength, yielding constant Vbe), and current swings are absorbed in the slave devices. If the collector of each device is connected to the control emitters and the CCS, then there is no feedback from the slave back to the master emitter; you need to signal to the control devices any imbalance by using some degeneration. You don't need much, 10R will do the job, but CFP LTPs in my experience do need it.

Cheers,

Hugh


All times are GMT. The time now is 03:28 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2