TGM Amplifier ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am within a few hours of giving up with toner transfer. It's clear that the professional TTP paper from Pulsar is NOT suitable for iron-on, it really needs the controlled pressure of the laminator they sell.

I tried the suggestion of glossy magazine paper. On a small sample, it worked pretty good in terms of transfer, first time. On the full size pcb I didn't get 100% coverage on two tries at it. I suspect the iron temperature needs tweaking up - I had to turn it down for the Pulsar paper. The magazine paper tends to jam in my printer too - it's a bit slippery. The coverage of toner from my printer doesn't look brilliant - they design them not to use up too much toner and there are no controls in the print driver. This begs the question of whether it will work even if I can get it to transfer.

I'm out of time today so I'll have to try again tomorrow. I have 'made' about two dozen boards so far !

Still, the delays allowed a small update to the pcb layout.
 
And here's what I was referring to as an update to the base version:
 

Attachments

  • tgm1.6.gif
    tgm1.6.gif
    34.3 KB · Views: 597
Gareth,
you know, I like your basic view on things and I`m pretty sure that some sunny day you will design and construct good sounding amplifiers.
An active current source for the LTP (without a current mirror) does not mean constant current through the devices but a heavier emitter/source load causes decreased functional accuracy and increased distortion. However, a current mirror and a VAS CCS, without further measures taken, would raise the open-loop gain, lowering THD but also sound quality.
It is supposed to have a front-end that is not designed for measured excellence.
Amplifiers should not be designed for measured excellence.
 
Thanks Lumba, I appreciate the confidence.

I'm still playing catch-up, just think how old the AKSA design is (and it was based on even older designs as starting point), and I haven't yet built my pseudo-clone of it yet.

Things change too fast these days, I already feel that this discrete power amp hobby is today's vacuum tube design. ClassD chip amps will be the only thing we can find soon.

Chasing very low distortion is a lot of work. It's hard for the hobbyist to measure the results because listening is so subjective and most of us don't have the equipment needed.

Perhaps there is some merit in global feedback based on a microphone in the listening room sampling the output our ears receive rather than the electrical signal that the speaker binding post receives?
 
Lumba Ogir said:

Lumba ?



PCB
___

Well, I am able to report that having given up on Pulsar paper I have had more luck with some magazine paper (a big thanks to MJL for that one). The BBQ adverts seem to work well, so too the car adverts.

I'm not completely satisfied with the results, but I no longer have the patience to rub down some copper clad board so into the etch bath it went, just half an hour ago...
 

Attachments

  • dscn1885f.jpg
    dscn1885f.jpg
    80.3 KB · Views: 645
Unfortunately I have a few parts still on backorder. They send me a piece at a time, like a ransom demand.

The transformer is something a bit risky, it's not very beefy, isn't a 'full torroid' so the EMI might try to frighten the amplifier. You can check it out at DigiKey: 595-1346-ND
But I like that these cousins of EI design reject mains HF a bit better, the split bobin gives it low capacitance between windings and the core losses help further. Sounds good in principle, but of course I just wanted cheap and small !

I did update the pcb layout (see attached) with the additional resistor to the LTP. It's an experiment, a last minute crazy thought. It will give me another tool to affect the LTP distortion in addition to the current mismatch. Between them I should be able to wreak some havoc.

My home computer doesn't allow me to do much with Spice, but I have access to spice at work now, so some lunch-hour investigations may help me understand more.
 

Attachments

  • tgmlayout1f.jpg
    tgmlayout1f.jpg
    90.2 KB · Views: 626
Wow, it's nice to have LTspice to play with, no wonder some people never build anything:clown:

So I tested out my amplifier and it works fine. So back to the idea of manipulating the H2 content by unbalancing the LTP (after all, that was my excuse for not using current mirrors).

I am simulating TGM version1, with the simply resistor biassed LTP pair. I'm running it 'cool' with only 0.58mA through the combined LTP. I tweaked the biass resistor to give me this value since it's close to the current that flows through the AKSA LTP (my role model). Picking an E12 value gave me a lucky coincidence that the current balance was pretty good. The input bjt sees 288.054 uA, the feedback bjt sees 288.254 uA. So the currents at dc operating point are balanced to around 0.07%. I obtained an FFT of the output as shown in the attachment top trace.

Then I unbalanced the LTP by reducing the biass resistor and pushing the total LTP current up to 1.1mA (still pretty low). The two current flows are now 288uA and 809uA respectively, which is a huge imbalance. I checked the FFT of the output and saw the lower trace in the attachment.

What gives ? - the harmonic content is not really changed that much, did I miss something here. H2 is at -70dB, H3 at -60dB. If anything H2 is a bit lower when the LTP is unbalanced. I had expected an increase in H2 with the imbalance based on some reading on the web.

Sorry about the image quality, files size limits are the cause...
 

Attachments

  • ltp1_2 copy.jpg
    ltp1_2 copy.jpg
    65.3 KB · Views: 526
So next up I tried the LTP boost resistor, providing some direct feedback from the output to the emitters of the LTP.

The first attachment shows the result when the current through this resistor is around 8% of the LTP dc current. And the H2 content is up, not much though.
 

Attachments

  • ltp5.gif
    ltp5.gif
    32.1 KB · Views: 533
Lovely work Gareth

Enjoy this nice topology sonics.

I made i visit, have not readed all thread..i hope you are having good results with this big current into the drivers.

Floating driver emitter resistance, in this kind of emitter follower, have worked better, at my home, when using 150, 180 or 220 ohms.

This avoids the driver saturates before the output.... reduces distortion and works better this way.

regards,

Carlos
 
Carlos,

Thanks for the compliment, which has value coming from one of the 'Masters' on this forum.

Unfortunately you will be confused reading my thread by all the changes that I keep making to the design as I learn and then unlearn and then learn again... some time ago I have increased the resistor between driver emitters to 180R in the design. I wish this was because I understand fully the need but it is mostly because of good advice received so far.
 
Gareth,

Looking good.

According to an academic friend in the psychoacoustic field, if H2/H3 are slightly above 60dB down, and all higher harmonics are more than 60dB down, there will be masking, and the impression will be one of 'warmth'. Further, the subliminal effects of unhead, objectionable harmonics will be completely masked, and the listening experience will be judged subjectively pleasant.

This is not to say we should try to get the harmonics lower than this, preferably 90dB down or more, but as you have found this is not always easy, though a feedback amp, at the expense of a profusion of harmonics, is pretty good.

Hugh
 
Hugh, thanks for the guidance on optimizing this, I'll have to do some work once I have a prototype by trying some different component values. I think your approach to using low order harmonics to 'mask' other artifacts is an under-appreciated approach.

Power Supply
========
I've started looking at the question of cross-talk. I have to build several channels. For stereo it makes good sense to invest in a dual supply so I suspect not very much effort has been devoted to multiple channels. For 5 channels it gets more cumbersome. People have complimented me on my approach of using mono-blocks, but it is a brute-force solution and I thought it worthwhile to dig further.

I set up a simple simulation. It may be too simple and simulations are not the same thing as real life where there are many more parasitics to include than I have bothered with. But it gave me a tool to explore the relative effects of topology on the cross talk from one channel to another. I looked at a two channel arrangement where I set up one channel with zero amplitude input and the other with a 7kHz test signal (I removed the ground-lift resistor as this messes up the reference point for the output) and looked for 7kHz appearing on the power supply rails and subsequently contaminating the other channel.

I've looked at three basic approaches.
1/ Single Supply to all channels
2/ Dual supply, one for all of the input stages, one for all the output stages
3/ Mono block, each one a separate supply

Attached is an example of the Dual - Supply. The main supply is that for the output stages, the intenal resistance is set to 0.5 Ohm and filter capacitors with a series resistance of 0.05 Ohms. The TGM1 amplifier is represented by the triangle.
 

Attachments

  • dualsupply copy.gif
    dualsupply copy.gif
    50.8 KB · Views: 473
Here's what I found:

3/ Mono-blocks.
zero ripple induced on neighbouring channel. No surprises.
About 55mV of ripple on the power rails of the channel being driven.

2/ Single Supply
With no filtering on the rails to the front end I get cross talk from one channel to the other. In my particular simulation it works out as -75dB of cross talk on the output.

By introducing RC (100 Ohm, 100uF) filter on the +ve supply to the LTP I see very little improvement. This is my basic TGM1 design so this means I have to improve my design if I want to use a single supply.

By introducing RC filters on both +ve and -ve rails to the amplifier front end (LTP and VAS) the cross talk is greatly reduced. I see around -100dB. This maybe good enough.

3/ Dual Supply.
I feed the front end of both amplifiers from a common supply. When I do this on the +ve rail only (so the front end gets it's -ve feed from the same psu as the output stage) there is no great improvement. When I completely separate the front end with independent +ve and -ve supply I see the cross talk eliminated. Nada. It simulates as good as the mono-blocks.


What I conclude is that a Dual Supply is good enough. For stereo it doesn't make sense, since you may as well have dual mono-blocks. But for multiple channels, dual supply is the best solution (in theory).

What I see from the single supply results is that the cross talk can be brought down to very low values. In a design where I may be using increased H2/3 to 'hide' higher order contamination a cross talk result of -100dB is good enough. For HT application this level of cross talk should be fine. Of course the simulation is most likely underestimating the cross talk and I'm running this at low power levels. Nevertheless, a single supply doesn't look to be the dog that it first appears.

Bottom line - Dual supply (front end and output stages) is the front runner for TGM multi-channel.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.