designing TGM3 - output Triples

[Bigun]

I maybe after you for a pcb then !

 

[keantoken]

Aye! May have to be a joint effort if you want it to be soon however. It's the first PCB I've ever done, and there are some things I don't like... PCB'ing the Allison should be easy though.

- keantoken

 

[AKSA]

Gareth,

Here is a temp compensated bias schema for a singleton input stage,


Hugh

 

[Bigun]

Thanks Hugh !

ah ha ! I see it needs four Vbe drops to compensate adequately, whereas I've used only one. I'll have to see if this can be accommodated on the current pcb or if I have to use it when I respin for the last two channels.

 

[keantoken]

I like Hugh's scheme, I would try that before my follower, since it has a chance of generating 3rd harmonics.

- keantoken

 

[AKSA]

Gareth,

When you do a .temp step on this bias schema, you find that from 15C to 30C the difference in output offset varies about 12mV, which is acceptable. All four diodes are needed to narrow down this variation. I can buy 1N4148 for around 1.5c apiece; this is inexpensive, though lots are used. Another option would be a Vbe multiplier, similar to a bias generator on a Class AB amp.

I failed to mention that the 200R trimpot is routed via a 10K resistor to the negative rail. In my prototype this does about 4mA.

Cheers,

Hugh

 

[keantoken]

Bigun, I realized I made a really bad mistake when I suggested the bootstrap mod; the two large capacitors would form a large capacitor across R15 (in Bigun's schematic).

Here is the mod fully realized.

- keantoken

 

[Bigun]

Well I've taken some time to fix the dc-offset instability that I was unhappy about before. In the end I thought I'd do something different than beefing up the passive thermal compensation since there are several sources of dc-offset drift and not just the thermal drift of the VAS device to compensate for.

My decision was a dc-servo. Never made one before so it was an opportunity to learn something new. Trouble is, I don't want to be using any ic's as this is against the code of the discrete amplifier builders guild

So it had to be a discrete dc-servo which didn't need a specialized supply but could function in an amp with whatever supply rails were available (most OpAmps seem to need some extra effort to provide them with the right power rails). And I didn't want to add any active devices to the design because I like things to be simple and I didn't want to have to make another pcb but rather reuse what I have. Sounds impossible.

However, I was able to meet the challenge by re-purposing one of the transistors already on the board that was originally set up as a diode in a futile attempt to compensate for temperature drift in the dc operating point. With the addition of a couple of caps and resistors I was able to devise a single-ended dc-servo. And the really good news is that I didn't get a problem with turn-on thump as a result.

The attached schematic shows how it was done, and if you compare it with earlier schematics you'll agree it was a trivial modification (T8 on the schematic)

The dc-offset is now rock solid and more or less non-existant. The remaining thermal drift is produced only by the dc-servo device it'self, but it sees little current to cause internal heating and so it suffers only from the effects of ambient temperature changes which have a tiny effect in this implementation.

I've detected no adverse impact on the sound performance.

 

[Bigun]

I conducted some more listening tests of TGM3 against TGM1 and Cellini using my FM tuner (on a high quality french channel as the all the English ones where I live are horribly compressed). It confirmed some thoughts. TGM3 is cleaner than TGM1 (which has that familiar sound of amplifiers of old) and it's more open at the top than Cellini, my tube amp.

I've simulated a *lot* of potential improvements but nothing really seems worth the change, except that I can remove the shunt capacitor from the feedback network and lower the impedance of the feedback network as a result - using the dc servo to keep everything under control. I've dubbed this version TGM4. It's the last all-solid-state gnf amplifier I will ever design. The TGM3 is very similar to the JLH Class B amplifier and it sounds great to my ears.

For higher powers I might want a cleaner version: Cascoding the VAS drops the simulated distortion into the noise - so for those wanting a very clean amplifier this would be a possible approach - with a very simple front end compared to some other designs I've seen around.

My journey with these amplifiers is now a couple of years in the making and is now at an end and if I build more SS class AB amps it will be TGM3(4).