TGM4 amplifier - diyAudio
Go Back   Home > Forums > Amplifiers > Solid State

Solid State Talk all about solid state amplification.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 27th August 2011, 01:45 PM   #1
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
Default TGM4 amplifier

I seem to have too many projects on the go. My JLH10W amp is my current project, but I've decided it's best suited to the transformer in my HT amplifier.

So my HT amplifier, based on TGM1, needs upgrading (perhaps to something inspired by the LazyCat SSA) with some higher voltage trafo's I have. But it currently houses my TGM amp which I need to keep my HT in service.

So first, I need to re-build my TGM amp into a smaller box for temporary service in the HT system before it retires as a stereo. And so I arrive here….

If I'm going to rebuild the TGM, I may as well tinker with it first !

The TGM1 [TGM Amplifier ?] is a simple Lin topology and sounds very nice, but it has some weaknesses - it doesn't have the greatest bass punch and it trades crisp treble for a softer sound with no harshness. it has some H2 and H3 engineered into it.

Next I modified it to TGM2 [TGM2 amplifier] which used CFPs in the LTP. This brought noticeable improvement in bass punch and treble clarify, but at the expense of losing the easy listening pleasure - it was to harsh in the treble for my ears with my source and my speakers. It had no H2 or H3 engineered into it. I tried introducing some 2nd harmonic [TGM amp goes 'tubey'] but this didn't go far enough to 'fix' it. A very nice amplifier otherwise but it's not the way forward now.

Next I went to the TGM3 [designing TGM3 - output Triples] which uses a singleton input and low impedance voltage feedback. I put a CFP in the driver stage to shield the VAS from the non-linearity of the Class AB output stage and to provide a low impedance drive to the output. This sounds very good, I thought that was the end of the road. But thermal stability and dc offset are a nuisance with singleton inputs and I had to resort to a dc servo. It has a bit more start up 'cone movement' and the CFP driver requires a compensation cap on the negative driver which adds capacitive loading to the VAS. But this is the one to beat.

So, TGM4 will start again, from the TGM1 design.
Attached Images
File Type: jpg tgm1v8_forum.jpg (76.8 KB, 571 views)
File Type: png sch2v9.png (65.8 KB, 553 views)
File Type: jpg TGM3asbuilt.jpg (265.4 KB, 557 views)
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.

Last edited by Bigun; 27th August 2011 at 01:49 PM.
  Reply With Quote
Old 27th August 2011, 02:21 PM   #2
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
Here's the first sketch up incorporating some of the ideas from the old TGM threads.

The key features are:

- no CFP
- current mirror added to LTP to improve balance, accuracy and slew rate
- no CCS for LTP or VAS to keep signature 'sound' of TGM
- folded cascode + EF buffer for VAS
- separate front end power rails to reduce x-talk when using common trafo for both channels

The VAS is the main change from TGM1, it was discussed in the original TGM threads but my interest was renewed in common base with the lazy cat SSA thread. I remember it was part of the Reoender amp which I admire. But I feel it needs to be buffered because unlike TGM3 I'm not proposing to use a triple output stage. With an EF buffer I can lower the output impedance of the VAS and reduce the impact of Class AB output stage non-linearity.

It's a bit sketchy as to how the VAS should come together, in detail. Where should I tie the collector of the common base device - to the +ve rail ? My worry is what happens if this device turns on hard during clipping ? In a suggestion on a DX thread I suggested tying the collector of the beta-enhancer device to the output so that the voltage across the device was lowest when it was most stressed. So I'm thinking I should tie the collector of the common base device to the output.
Attached Images
File Type: jpg TGM4v0.jpg (77.2 KB, 526 views)
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.
  Reply With Quote
Old 27th August 2011, 11:54 PM   #3
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
A couple more thoughts came to mind today.

Firstly, the input impedance of the CB stage is rather low compared with a conventional CE VAS. I remember seeing a schematic on this forum that uses simple bootstrap of the LTP leg to address this (first/left attachment)

Secondly, I'm still a bit unsure i like the use of current mirrors. But I don't think I'll get adequate current balance without one with this kind of VAS to drive. The other way to build an LTP is the Rush Cascode. It does require careful set up to avoid d.c. offset issues and thermal drift. I'm not sure if I have the answer, but I've put a couple of diodes in the input side to address this. The TGM-Rush version is attached 2nd/right.

So this would be another way to realize TGM4. I'm not sure which version I like best, the Rush input or LTP input.
Attached Images
File Type: gif guru003001_amplifier.gif (7.6 KB, 468 views)
File Type: jpg TGM4-Rush.jpg (56.9 KB, 234 views)
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.
  Reply With Quote
Old 28th August 2011, 02:24 AM   #4
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
Giving credit where it's due:

I would like to give credit to Greg for the bootstrap idea, the thread is here: Simple Killer Amp!

and

I would like to give credit to Keantoken for drawing attention to the Rush Cascode, the thread is here:
Rush Cascode Headphone Amp + JLH Output Stage
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.
  Reply With Quote
Old 29th August 2011, 01:23 AM   #5
AKSA is offline AKSA  Australia
diyAudio Member
 
Join Date: Sep 2001
Location: Melbourne, Australia
Hi Bigun,

In your RC circuit, how about a bootstrap cap from the junction of the two resistors on the collector of the bottom device to output, then the present collector connection from the same device moved to the base of the EF driving the output stage?

That way you remove the common base device, and maintain all the correct phase relationships for nfb.

Cheers,

Hugh
__________________
Aspen Amplifiers P/L (Australia)
www.aksaonline.com
  Reply With Quote
Old 29th August 2011, 01:56 AM   #6
diyAudio Member
 
Join Date: Apr 2010
Location: Coffs Harbour
Hi Bigun
Take a look at the LTP biasing on this old Rush cascode amp.
4QD-TEC: Low distortion Audio amplifier
__________________
Ian
  Reply With Quote
Old 29th August 2011, 11:02 PM   #7
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
Thanks for the input on the RC version - I will give this some thought.

With the LTP input version, using common base VAS with current mirrors isn't obvious, so I invented a way to connect the VAS in a Lender topology using a current mirror.

Of course, we never invent here, we just reinvent and an hour later I found it here: Borbely-Lender True Symmetry

However, I've been shy to use current sources and their associated ancestor, the current mirror, on the grounds that in real life they add non-linear devices with phase shifts into the amplifier.

I worked up a quick simulation of the common base VAS (attached) without a current mirror and it seems to work.

I've noticed a h.f. roll-off in the simulated OLG and my question is for this topology - from where does the roll-off arise ?


I also simulated a version with the grounded base device connected with the base taken off the 'left leg' of the LTP, a kind of simplified Lender (simplified because there's no EF buffer before the base connection). It also worked well but as Lumba Ogir pointed out in the original TGM thread, this also introduces miller capacitance back to that leg of the LTP unless the EF buffer is included which adds more devices into the chain. Nevertheless, what is attractive about the simplified single-device Lender topology is that it cuts out more parts. This might be something that can only be decided by listening since the difference between the simplified Lender and the version posted below are difficult to appreciate through simulations alone. I did find comments on this forum that Hugh has tried the Lender approach and didn't hear any sonic benefit.

I did a mental check on the bootstrap idea of Greg's. Of course, it doesn't work here. Firstly, the quick sketch I posted of it above has the bootstrap signal from the wrong side of the VAS slave device, and secondly this VAS is not a simple buffer like Greg's so the amplitude of the bootstrap signal doesn't match that of the LTP leg. Cross that idea off.

Did you notice the 'SandyK' resistor ?
Attached Images
File Type: jpg TGM4folded cascode (2).jpg (195.0 KB, 209 views)
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.
  Reply With Quote
Old 30th August 2011, 11:36 AM   #8
Account disabled at member's request
 
Join Date: May 2006
Hi Bigun

Looking at your sketches I think the schematics of NAD 306 would be of some interest.
  Reply With Quote
Old 30th August 2011, 03:19 PM   #9
diyAudio Member
 
Join Date: Apr 2011
Location: Pretoria
Shouldn't the driver transistors (Q1 and Q4 in the LTspice simulation in post 1) be on the heatsink too?
  Reply With Quote
Old 30th August 2011, 05:23 PM   #10
Bigun is offline Bigun  Canada
diyAudio Member
 
Bigun's Avatar
 
Join Date: Jan 2009
Blog Entries: 2
I thought I'd post up an schematic of my simplified Lender VAS idea. I like it for the fact it removes the need to bias the common device base even though it potentially introduces some miller capacitance to the left leg of the LTP. Simulations indicate an increase in OLG but stability looks very good - no Cdom required.

I also like that it makes use of both legs of the LTP, something that the Rush cascode can't do ( I haven't give up on the Rush ) and has the elegance of a design that takes a single ended input, converts it to a differential and then converts it back to single ended all with the purpose of providing a good error amplifier for gnf that maintains a stable dc offset over temperature. The amplifier can be dc coupled if desired (something else the Rush version can't do without additional devices).
The collector load of the common base device has been taken up to the top of the Vbe multiplier. This allows me to use a larger resistor which improves performance slightly. In fact, based on simulations this is a good take-off point.
I may need some clamping diodes between the different voltage supply rails so that the front end can't reverse bias the driver transistor base-collector junctions.

Quote:
Originally Posted by homemodder View Post
Hi BigunLooking at your sketches I think the schematics of NAD 306 would be of some interest.
If you mean the NAD phono amp with RC - yes I noticed this in the thread Keantoken started but I don't know that they still use it anymore.
Quote:
Originally Posted by ingenieus View Post
Shouldn't the driver transistors (Q1 and Q4 in the LTspice simulation in post 1) be on the heatsink too?
If you mean the Vbe multiplier device - yes, it needs to be in intimate contact with the output devices. In the past I have mounted them on the same heatsink. But my last build I used silicone to glue the Vbe device directly onto the package of one of the output devices and it gave me a more stable bias during warm up.
Attached Images
File Type: jpg TGM4_LenderVAS_sch.jpg (207.4 KB, 185 views)
__________________
"The test of the machine is the satisfaction it gives you. There isn't any other test. If the machine produces tranquility it's right. If it disturbs you it's wrong until either the machine or your mind is changed." Robert M Pirsig.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Wanted: DIY amplifier builder with experience. I need a 0.7 ohm stable amplifier Audiophilenoob Swap Meet 34 16th May 2005 11:58 PM
Questions about how to decrease the PASS A-40 amplifier into a class AB amplifier. novtango Pass Labs 2 21st October 2002 01:50 AM


New To Site? Need Help?

All times are GMT. The time now is 08:11 PM.


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