Tube with Power IC Output Stage - JLTi

[fedde]

Millwood: have a look at my gainclone page:
www.fedde.tk

It will maybe not answer all your questions, but...

Fedde

[Joe Rasmussen]

PART TWO:

Now for the NON-Tube version, which also incorporates similar approach, but actually has more complications.

Two problems: One, keeping feedback constant (not totally possible but acceptable) when we still have a pot directly in circuit. Actually this did not turn out as troublesome as was thought.

The second problem is major: How do we employ the low pass filter to eliminate HF brightness when component values are shifting about and hence also the turn-over frequency of the filter. This will be covered in this installment.

So this is our REAL PROBLEM, the filtering/bandwidth limiting. How do we get our target of –1.25dB at 20KHz? Oh yes we can get it but only when the pot is set at one discrete position (all you need is adjust 1n3* to get it, simple) but as soon as we move the pot position the filter shifts turn-over frequency and our target is lost.

Clearly it has to come down to a compromise. Please note the suggested compromise works better with lower value pot, so 10K is better than 25K and likewise 25K should be better than 50K (max).


Here is the basic recommended circuit using a 25K Pot:




Below is the basic circuit using a 50K Pot:




Below is the basic circuit using a 10K Pot:



They are much the same EXCEPT using a different value pots means adjusting the cap value after the 2K2 resistor. Look again and you will see what I mean.

This cap value is the most tricky part of this NON-Tube IGC and what I’ve shown is based on getting the desired –1.25dB value when the pot is set at its –20dB setting.

I ran some simulations (aren’t computers great?) for the three different pot values and measured each at –14dB, –20dB and –26dB. This gives us an indication what happens to the HF response over a window of 12dB adjustment in an area (pot position) where you will most like be when listening to a 2V RMS output CD Player.

THESE ARE GUESTIMATES FOR EACH VALUE POT:









Caption for all graphs: The middle line in each graph is the desired minus 1.25dB with the cap value shown and pot in –20dB position. The lines above and below is varying -6dB and +6dB, a total of 12dB variation of the volume control/pot.

They all look similar but the 10K is more stable whereas surprisingly not much difference between 25K and 50K, provided cap value is made. Please note this is only a guide but also a good starting point when building it first time. Just choose the pot and its recommended value.


Here is the FINAL suggestion:

FIND the volume at which you generally listen and make a note of the Volume pot setting. Find an average setting from a reasonable range of CDs (assuming that is your source). Now vary 1n3 (or whatever matches the pot) cap value by listening carefully – or if you have the instrumentation, adjust to minus1.25dB @ 20KHz. In which case you will still be able to further experiment by ear.

----------------------------------

Getting Stable Gain:

Let’s examine GAIN variation with the three different pot values. I made the calculations/measurements and came up with these results:

25K: Circuit Gain variation between +31.2dB and +33.3dB. Quite a good result. Over-all recommendation?

10K: Circuit Gain variation between +32.2dB and +33.3dB. This is close to the ideal BUT with one BIG drawback: The 10K Log pot is not really suitable for a Tube source, but if from a direct coupled (also sometimes called DC coupling) like most CD Players, should be OK. Again vary the 1n3* by ear.

50K: Circuit Gain variation between +29.5dB and +33.3dB. This is just acceptable but loads the source the least. Otherwise I don’t see much problem here.

I would not use 100K.

This then sorts out having too much feedback, much lower than the original IGC. Because of this we now revert to normal Log operation. All three pots are acceptable, but 10K load needs to be kept in mind.

In Part Three we will deal with recommended Power Supply, Wiring and Earthing, so that our little wonder-amps (?) work properly.

END OF PART TWO.

[millwood]

Originally posted by planet10


From all reports -- i've been gathering parts for Joe's version -- these things sound way better than anyone would think they have a right to on an absolute scale and in terms of bang-for-the-buck they are a big winner. Do a search on GainClone... in particular the megathread called "Not just another GainClone" will give you plenty to read.

I guess I am trying to figure out 1) if it sounds so good? has anyone done any (semi?)scientific testing to confirm its quality? and 2) why it sounds so good.

I have tried some power amp ICs (from my stk days in the 1980s to tda7294) and I wouldn't rave about any of them the way folks here raved about the gainclone. so naturally I am quite curious as to it.

[Joe Rasmussen]

PART THREE:

We have covered a lot of ground, let us leave the more theoretical behind and deal with more practical matters.

I think it can be appreciated that a circuit diagram only tells so much about wiring and especially the technique that a designer or DIY’ourselfer might employ. Earthing is now taken for granted as very important not just for best sonics but also to keep an amplifier from going unstable. The latter is certainly the case here. A badly wired Gainclone can have all sorts of buzzes and noises, also the heatsink might feel too warm even though there is no signal passing through – most certainly a case of oscillations and watch out for your tweeters!

I certainly have my own ideas, both practical and intuitively. I think in terms of AC and DC currents ALWAYS running in loops, and different loops must not interfere with others. This is difficult to explain in a few words and this won’t be the place.

In our IGC circuits we have TWO Earths, we will call them “S” for signal and “G” for ground, indeed “G” can also be called Power Supply Ground or “PSG”.

This is how it is applied:

<img src="http://members.ozemail.com.au/~lisaras/DIY/Layout.gif">

The TWO earths can clearly be seen here, they are the two black round pads. It can also plainly be seen that the Power Supply uses “G” but the input Signal uses “S”. They are tied together by a single wire.

Thick lines indicate heavier duty wiring. One that I found particularly critical is the wire from between the two 1000uF electro caps (0V) going back to “G”. Use good quality multi-strand wire with very low inductance.

It is important that IF using the Tube Version, that it too gets its Signal ground from “S”. So in technique in the two versions are much the same:

<img src="http://members.ozemail.com.au/~lisaras/DIY/Tube_Wiring.gif">

Re the power transformer, follow the usual recommendations. Some like using separate 160VA toroidals for each channel. You can do that or use a common toroidal for both but may then want to use 300VA.

Whatever you do, the Bridge and 1000uF caps must be separate for each channel. So the diagram above must be repeated for each channel. My strong recommendation is the 1000uF/50V Panasonic FC cap. The Bridge I used is four 10A 100V Schottky Diodes (in TO-220 packages) and make a discrete Bridge.

Here is my suggested Power Supply – keep in mind I am NOT supplying a PAINT-BY-NUMBERS project, this is for the serious DIY’ourselfer.

<img src="http://members.ozemail.com.au/~lisaras/DIY/Tube-IGC_PS.gif">

Note that I use a feature also used in the commercial JLTi Hybrid Tube Integrated Amplifier. When power is connected to the IEC socket the Tube Buffer powers up and stays in Stand-By mode. The Power Switch then only turns On of Off the Gainclone output stage. This so-called ‘Permanent Stand-By Mode’ means that when amp is plugged into AC, make sure that the Power Switch is set to OFF. In this way we avoid rude noises of the Tube Buffer as it stabilizes.

Tube Buffer Supply:

The 2 x 1000uF/50V may I suggest that you also use the same Panasonic FC caps here. The 1uF/100V must be a suitable film type, NOT an electro. I recommend a 30VA Toroidal Tx, 25V-0V-25V. I use four 1A 1000V FRDs as a discrete Bridge. Together with the Panasonic FC caps, this makes a Split Rail Power Supply, plus 35V and minus 35V.

The Filament Circuit: This can be powered by a small transformer rated no less than 5VA. The 5R6 1W slows down the initial filament current, better for long term life.

The rest should be self-explanatory. Be sure to apply all Power Supply ZERO volts to the “G” ground, and NOT “S” signal ground.

For NON-Tube version IGC, just pare down above.

Shall we end Part Three here? Part Four will be used to mop up.

END OF PART THREE.

As a teaser for Part Four, here is one of THD measurements under discussion: Are there good REASONS WHY these little devices can sound as good as they do when implemented well. There may well be...



IF Part Four is not posted within 24 hours it will not be until Monday or Tuesday, as I'm going away for a well-deserved break. The wife and I are depositing the kids with the in-laws and going to visit an audiophile friend in Canberra (capital of Oz-stralia) and he is going to escort us around the hills around there where there are twenty-somthing wineries, all providing free tasting and... buy some we like!!!!

Till then.

Joe

COPYRIGHT JOE RASMUSSEN 2003 - NOT PUBLIC DOMAN PROPERTY - FOR INDIVIDUAL USE ONLY

[Joe Rasmussen]

Originally posted by millwood


I guess I am trying to figure out 1) if it sounds so good? has anyone done any (semi?)scientific testing to confirm its quality? and 2) why it sounds so good.

I have tried some power amp ICs (from my stk days in the 1980s to tda7294) and I wouldn't rave about any of them the way folks here raved about the gainclone. so naturally I am quite curious as to it.

Hi Millwood,

Re testing, come back here Monday or Tuesday and I'll have something for you. Being one of those fortunates, I've got the equipment to do a variety of tests, THD (as total or spectral graphs), IMD (likewise), MLS & FFT. Serious toys.

The tda7294 I've not tried BUT I have it from good sources that it is no way up to par sonically with the NS LM3875 or the other 'Overture' (which is what NS calls them) chip amps. Not long ago I would have been sceptical too, but when you get results, then - hey - that's what it's all about!

Other than the tests I will be posting, have a look at my "Tubes & The Gainclones" essay - this also has some pointers why?

http://members.ozemail.com.au/~joera...gainclones.htm

Joe Rasmussen

Webmaster : www.vaccumstate.com & http://members.ozemail.com.au/~joeras

[Nielsio]

Terrific work, Joe. Good job on the grounds wiring too.

Hopefully I will be able to turn your ideas into practicion soon.

[leadbelly]

I've got the parts for the tube version on order, and I have a question about the PS.

For the +/-35V for the chips, I ordered a 50VCT 2A transformer. I assume this is is enough to power 2 x 3875's?

Now for the tube, I already have a small 24VCT transformer. Will +/-16V be enough to supply the 6922? I prefer to use this if possible since it would make for a nice low-draw standby mode.

Or would it be better to just power the tube from the main transformer? And if so, do I have to worry about implimenting minimal path distances and mounting caps right on the chip, or is on-chip mounting only an issue for the signal path around the chip? Or is it better to put 2 sets of filter caps, 1 on the chip and 1 on the tube?

Any other ideas welcome too.

[planet10]

Originally posted by leadbelly
For the +/-35V for the chips, I ordered a 50VCT 2A transformer. I assume this is is enough to power 2 x 3875's?

100 VA for both chips? 300VA is recommended. I'm going to use a 600 VA i have been looking for a use for (should be good for 4 channels if it works out)

Now for the tube, I already have a small 24VCT transformer. Will +/-16V be enough to supply the 6922?

Tubes like lots of B+. The 70 V speced is already kinda low.

Or would it be better to just power the tube from the main transformer?

Yes, And then get a beefier trafo for the 3875s...

dave

[leadbelly]

Well, I already ordered 2 of the 2A xformers, so I guess I'll plan on using 1 each per channel.

So, if I power the tube using the same xformer, I guess I should put a set of 1000uF/1uF filter/bypass caps right on the 3875, and another set right on the tube socket?

[planet10]

Joe,

The configuration for the tube B is one i've not seen before. How does this work?

dave

Note: AudioFreak went thru and reset all the graphics settings in all the forums and we seem to be back to normal... i took the liberty of moving all the posts relating to that out of the thread.