Xamp-simulations problem_s

[grataku]

HI,
I have simulated a bunch of Nelson's circuits starting from the simple BLS to the X circuits on the patent, they all work (imagine that ) except the most complex, and beautiful looking, Figure 4 on the supersymmetric patent (#5376899).
Has anybody had success with that? I tried all possible variations on the theme without success. As I recall Petter x100 and Uli circuits are implementations of the circuit in figure 3 of the patent.
Does anybody have any suggestions? ideas?

[tschrama]

Hi Grataku,

You'll need to bias the gate's of transistors 120/121/122/123, and stabalize the Iq with a source resistor. I don't understand why there's no bias-voltage shown at the input stage as there is shown for the output stage. You could use JFet's as input trnasistors. Put some resistance parrallel to the I source to keep your simulator from going nuts .... I hope this helps..

Regards,
Thijs



PS
Which Spice programe are you using?

[tschrama]

Hi Grataku,

My simulator starts to puke when I connect a load from the output to groud or differentially across the outputs. The error message says: 'line too lone ... no such vector.....' etc ..

With no load other than the feedback resistor (tried 1k to 47K and a gain of 20dB: no problems there), everything looks fine.. very fine actually ...

Regards,

Thijs

[grataku]

HI,
I am using Simetrix. I'll try to bias the input to see what happens.
Also, I'll try to get rid of 122 and 123 so as to make an hybrid between fig 3 and 4.
The ccs are very touchy.

[tschrama]

Use pure resistors as current sources... this helps getting started getting a feel for the circuit.

Goodluck,
Thijs

[grataku]

Thijs,
i tried the voltage sources and I have not been able to get it to work. Clearly the problem is as you said, the bias of the input pair. I removed the bottom two input devices and split the ccs like in figure 3 and it all works ok.
Also, it looks like the value of r 156 is critical for the frequency response it seems that the lower the better I wonder how low can I go...

tried R in place of ccs and that gives weird results.

[tschrama]

i tried the voltage sources and I have not been able to get it to work. Clearly the problem is as you said, the bias of the input pair. I removed the bottom two input devices and split the ccs like in figure 3 and it all works ok.
Also, it looks like the value of r 156 is critical for the frequency response it seems that the lower the better I wonder how low can I go...

tried R in place of ccs and that gives weird results.

Hmmmm .. I don't use a X-resistor (r156 in fig 4) in my SuperSymm simulations. I think I remeber that the circuits I investigated performed better without it. I would expect that R124/R124/R152/R153 would determine frequency response and phase margin.

I also used a symmetrical resistive voltage devider network from the symmetrical supply voltages to supply the Bias voltage of the input transistors. This works very well and in my experience keeps my computer from getting stuck..

Regards,
Thijs

[tschrama]

I've used 0R1, 47R and 500R as r156. It definnitly influences the openloop gain and therefore has an effect on phase margin and frequency response, but it more or less stays at Fc -3dB = 1Mhz, using 10K/1K feedback/input resistors. Openloop gain was about 63dB, closedloop gain was set to 19.9dB. Hmmmmm I'm getting Spice -addicted again.....

Thijs

[Nelson Pass]

Could it be that simulators are flawed in some way?

[grataku]

Hi Nelson,
that's very possibly caused by what I am doing I am getting a weird frequency response up to 10khz only and no gain.
This is what I am using, I have played with all kinds of values of input resistors, X resistor, voltage sources, feedback resistors etc the other circuits work fine but there is something about this that just doesn't work.

This is the basic circuit I am trying, the values keep changing so if something looks weird is just what I have is right now.