ccs, direct coupling and stuff

[beamnet]

Here's my problem

I have an amp using a long tail phase splitter. I am terribly unhaapy with the performance.

I want to use a ccs.

The circuit is half a 6sn7, directly coupled to an ecc99 ltp with a 10k tail


I am not very good with ss design, so i decided to copy the a css

How do i manage to get the gorrect voltage at the kathodes to keep the direct coupling?

Or should i decouple with a cap?

I am not unwilling to do some calcs, i just need a few pointers

[EC8010]

You don't need to worry about the cathode voltage, the CCS will sort that out automatically. Once you add the CCS, you will need to balance the anode load resistors of the phase splitter.

[beamnet]

i know i have to balance. Should i use *about* the same level?

Any suggestions for css? i've been looking at the css in the "coldwar" amp on www.ultranalog.com

gary p makes cool ones too, but they are far more complicated (and i have a sample acount at AD)

[EC8010]

Balance the anode loads. Match them if you like.

You will have to make a CCS from scratch because it must withstand 100V or so. Unless AD have changed their policy, they don't make valve support circuitry...

[beamnet]

err...i know they have to be matched, i meant, what value? about the same as in the circuit? (both at 30k or so)

As i said, i have NO ccs design abilities...any pointers?

Thanks

[Tweeker]

IXYS makes a CCS that is rated at 450V. IXYS 10M45. Sold at Digikey etc.

Texas Instruments makes a linear regulator that can be used as a CCS and will tolerate up to 125V between in and out if you need alot of current. TL783.

Theres some fairly simple CCS designs of good performance using just a couple cascoded mosfets. This picture from Gary Pimm's site illustrates the idea:

[tubelab.com]

The simplest CCS by far is the IXYS 10M45 chip. It works good too. Get the amp working with it, then you can upgrade if you want.

You need two resistors and the chip. The resistor from the "K" pin to ground sets the current, and you need a 1K resistor in series with the "G" lead to supress oscillations. The "stopper" resistor is not shown on the data sheet, but bad things happen if you don't use it.

[beamnet]

i contacted an ixys reseller here in the netherlands, see what comes out...

[ilimzn]

Originally posted by beamnet
Here's my problem
I have an amp using a long tail phase splitter. I am terribly unhaapy with the performance.

Although that is not going to be a perfect splitter, your problem may be elsewhere: your 330nF coupling cap and 47k grid leak frot he GU50 make a 64Hz highpass filter. This REALLY needs to be MUCH lower. Although seeing that you use pentode mode with no NFB or even local FB, are you maybe attempting to equalize the resonant peak of your speaker?

How do i manage to get the gorrect voltage at the kathodes to keep the direct coupling?

Quite easy actually, just set your CCS to the same current that passes through the 10k tail resistor now. That way the DC conditions remain unchanged.

[beamnet]

is it! damn!

That are the things i can't calculate. Increasing the coupling caps to 1uF won't take that away would it? This is a design by a highly respected designer here in holland (in the real design he used kt88's)

I thought he'd done his calcs right...Ii'll drop him a mail...

any suggestions to lower the filter? 100k leak/2uF coupling?

It does have solid bass, but might be increased if the transfer function is optimized below 64hZ.....

I want to experiment with feedback in the future, for now it sound very acceptable to me. The opt's don't require nfb to be flattened out. I want to decide on the final seakers and then go the fb way (or not)

There are recent rumours that UL could be achieved (there's a thrread here that i started or contributed to, don't remember)

Really appreciate the help here, finally my bodged together circuit is analyzed in an uplifing way!

Bas