Tube with Power IC Output Stage - JLTi

Status
Not open for further replies.
Jump to Latest
#722
One more graph, I thought it may be useful. The same T network in the feedback but the values are 1k/4.545R/1k, which keeps the overall gain as in the previous graphs (post #706). This pole seems to go down further, thus contributing to yet better phase. May this be too low? Note that there is about 2.5dB roll-off at 20kHz here. In the previous case (5k/118.8R/5k) it was about 0.25dB.


An externally hosted image should be here but it was not working when we last tested it.



Well, what does set this pole?

Pedja
 
#724
Pedja said:
One more graph, I thought it may be useful. The same T network in the feedback but the values are 1k/4.545R/1k, which keeps the overall gain as in the previous graphs (post #706). This pole seems to go down further, thus contributing to yet better phase. May this be too low? Note that there is about 2.5dB roll-off at 20kHz here. In the previous case (5k/118.8R/5k) it was about 0.25dB.

Pedja
Click to expand...


Could this then serve as the LP?

Sheldon
 
#725
Separate bridge for everything?

Hello,

I think there might be similar posts on a related subject before, but I really want to get a clearer answer so that I won't fry the components :xeye:

I understand the need to have two separate rect. bridges for the LM3875 chips in the case of 1 power transformer. However, is it also necessary to have two separate rect. bridges for the tube supply in case I also have only 1 transformer for the tube supply?

Thanks! 😉

Horace
 
#726
I understand the need to have two separate rect. bridges for the LM3875 chips in the case of 1 power transformer. However, is it also necessary to have two separate rect. bridges for the tube supply in case I also have only 1 transformer for the tube supply?
Click to expand...

It is not 'necessary' to have two rectifier bridges but (some say) preferable for the GC PSU.

For the valve supply I would say that one bridge is acceptable and that is what I used with my VBIGC . 😉
 
#727
Terry Demol said:


The ideal topology will present exactly the same R and C impedances to each (+ and -) input.

This will minimise CM distortion and also offset. It will also sound better, we have tried this on opamps and it works. When looking at each RC network treat the opamp OP as ground.

Terry.
Click to expand...

Exactly, in fact that is why I had no problem using 1M feedback value. Look at the original DIY circuit that was proposed many pages back on this thread (not the full JLTi):

An externally hosted image should be here but it was not working when we last tested it.


1M value here results in 100mV DC ofset. That's 0.1uA. Using the two 1M resistors to cancel out (both are going to their respective "grounds") I got less than 10mV offset 8 out of 10 samples and none higher than 18mV. Very stable too. I would tweak the (+) 1M value to get under usually, usually adding 10M in parallel did the trick.

Allowing for 0.1uA and gain of 45, even if poorly balanced, with 1K should contribute (add) less than 10mV DC offset, possibly half that? Since many are using NIGC with no f/b cap to get unity DC gain, say they are quite happy with 50mV - no great harm. But I prefer it lower than that.

Whichever, it shouldn't be difficult to null out by tweaking (+) resistor value. That's not so easy with NIGC!!!

Oh, another good thing, no need to use by-pass cap as shown above.

Good to see you chiming in, you know I value your input.

Re balancing AC as well for best CM result, doesn't just getting Lo Z help here? And nulling using purely resistive values (no by-pass cap as shown above) should get good AC CM as well.

BTW, Peter H. has bought a Canon Pro 1 Camera too. My favourite non-audio toy. :nod:

Joe R.
 
#728
Joe Rasmussen said:
Whichever, it shouldn't be difficult to null out by tweaking (+) resistor value.
Click to expand...

Exactly.
When testing, I use a multi-turn pot on the non-inverting input to ajust DC-offset to... 0 mv. 😀
In most cases I later remove the M-T pot, measure it, and change it for a resistor of around that value.
By doing this, I always have DC-offset under 5mv.😉
:att'n: The resistor value may not be the same on both channels, the chips may have very different DC on the same circuit, they have a huge production drift.:att'n:

Joe Rasmussen said:
That's not so easy with NIGC!!!
Click to expand...

Definitely.😀
 
#730
falcott said:



Forgive my ignorance, just trying to follow along. Are we refering here to the resistor in the position of the 1M-to-ground (from +ve) in Joe's schematic above?
Click to expand...

Yes we are. Whatever feedback value you use and whether you use normal single 1M (to use the above schematic example) or T-Network, this has to be tweaked to get the lowest amount of DC on the output. Usually if you use 1M f/b resistor, you will match that and usually will (should ideally) give you low DC ofset. If not, tweak it. The T-Network gets a bit more complex, but the same apply, tweak it.

Joe R.
 
#731
Thanks Joe!

I'm on my way towards completing an amp based on your tube IGC, and now, before I even get to hear a note from it, I am encouraged to get excited about trying this new-fangled T-network idea thingy. :bawling: Okay, I'll get this amp fired up FIRST and try the T-network AFTER. :angel:

When it all gets a bit more photogenic I will post a few pics. 😀
 
#734
Yes, we may move there for T network issues, just to finish the bandwidth issue here...

I have checked a few things, varied some components and found this loss of bandwidth to be associated to the loss of loop gain caused by low value of the first series resistor of the T network. So, it is loading effect. This loss of gain is mainly defined by the ratio between the input resistor and the first resistor in the T network. Every 6dB should result in one octave shorter bandwidth. In essence this means by the choice of the T network resistors value, since these may be chosen relatively independently of the input resistor values, the bandwidth can be customized. (etc, etc…)

Sheldon said:
Could this then serve as the LP?
Click to expand...
This certainly acts like a LP, but I doubt it can make input passive filter needless.

Pedja
 
#735
I have tried the network on my VBIGC using 10K,100R,10K. I am very impressed the bass seems to benefit the most, been tighter and more integrated with the rest of the music. There are also improvements in the mid (cleaner and more pronounced), and the highs - more natural. A significant improvement over an already excellent amp.

I have a question though. In my original implementation I just had the none inverted input to ground (to no significant ill effect). I have not added a 10K to this pin and have effectively used the none inverting pin as my effective small signal star ground. I am a little concerned that the feedback network, via the 100R resistor might be bleeding into the none inverting input and generating positive feedback.
Do I need the 10K to earth, or am I ok as is ??

Thanks
shoog
 
Status
Not open for further replies.
Top Bottom