F4 Beast Builders

Ian: Good points:


  • Stability: The LTSpice simulations show that the phase margin is "on the edge" of instability, properly measured within the feedback loop rather than with respect to the input.
  • Degeneration: That is a good question. Using a wye-delta transformation, a two resistor feedback network can be converted to a lower impedance two-resistor network with a series resistor at the output. This is not exactly equivalent to the original network since it presents a lower impedance to the source driving the network. In any case, the resistive feedback network presents some level of degeneration.
  • Doubling up: I agree that my circuit does not support putting multiple output FETs in parallel.
 

Attachments

  • wye-delta.jpg
    wye-delta.jpg
    72 KB · Views: 567
Thanks for the rapid reply :)

If stability is on the edge as you say, that raises the question of what is the output impedance or damping factor of your circuit? If it is well in excess of 500 then I suspect that Nelson must be using some feedback round the front end in his XA25. That might be a good idea here too although I know that many variations are possible, some mentioned earlier on in this very thread.
 
The IXYS SPICE models are designed to reflect the behavior of those FETs for use in HIGH power applications, rather than in puny 50W per FET amplifiers. It is also the case that those SPICE models do not reflect the low bias (sub-threshold) device behavior which will influence the level of bias needed for a given class-A wattage.

Therefore I am not sure how accurate the SPICE simulations are regarding phase margin and distortion.
 
Agree that the ISYS models are pretty useless for our purposes. As usual, building is the only way to find out for sure. Still interested in the simulated damping figure though...

I finally measured the simulated damping factor. The results are hard to believe.

Test conditions: input is a .8V P-P sin at 1kHz

Measurements:

  1. 2R 8.19326V
  2. 4R 8.19326V
  3. 8R 8.19343V
  4. 16R 8.19351V
  5. 100R 8.19358V
Using the formula df(v1, v2, r1, r2) := 8*(v2/r2-v1/r1)/(v1-v2)


Taking 2R and 100R
df(8.19326, 8.19358, 2, 100) = 47934.55761184867
Taking 4R and 16R
df(8.19326, 8.19351, 4, 16) = 49159.06000011457
 
Official Court Jester
Joined 2003
Paid Member
dunno ...... it really depends of execution ( yours and mine are different ..... pretty much as everyone else's here) ...... decent amount of overall feedback combined with insane outputs (and drive being firmer than A class drivers in Quad 405 .... you know that entire heat on 405 heatsink is from 4 TO220 devices ?)
 
I finally measured the simulated damping factor. The results are hard to believe.

Taking 2R and 100R
df(8.19326, 8.19358, 2, 100) = 47934.55761184867
Taking 4R and 16R
df(8.19326, 8.19351, 4, 16) = 49159.06000011457

They are roughly what I would expect with all the feedback applied globally. Rdson for the IXTN40P50P is 0.23R giving a damping factor of ~35 into 8R and if your open loop gain were around 63dB, which is quite possible in practice, then the damping factor is going to be around 5000. This suggests to me that the XA25 either has a lower open loop gain or more local feedback, or perhaps both. My concern here is not so much the damping factor but the worry that so much global feedback may well result in instability.
 
I will check them over.

Dear 2 picoDumbs,

Did you have the change to check these models? I have used these devices(216/79) as drivers in a previous project and as VAS in a current running prototype.
In both projects, these were compared to bipolar devises and the 2013/313 Fet's. The 216/79 came out each time as the winner soundwise.

However I would like to have a good idea about OLG in my current running prototype using LTspice. With the variety of models you find in the attachment, I get all sort of results. On top of this, the capacity parameters are this time also rather important as I'm using them now in a VAS.

Can you, or anyone else help me out regarding these models. Which from the list enclosed would be the better ones? Could you put your created/updated models on the forum?


Thanks in advance
 

Attachments

  • 2SK216 & 2SJ79 spice models.pdf
    44.2 KB · Views: 221
Has anyone who has a working prototype tested the amp's thermal stability under a heavy load (i.e. push it into its "class AB" range)? Time has shown my (overly) simple solution is adequate to have stable bias at my power uses (2W or less), which is likely to be expected since my power requirements are less than the bias point. But, I am left to wonder if I'd get thermal runaway if I pushed it into higher output. Unfortunately, I cannot test anything right now since I moved a few weeks back, and all my audio/DIY stuff is currently in boxes and/or dismantled.

Just curious what others have done with their amps. I tried to model the thermal characteristics of my solution in SPICE as Lynn has done with several designs, but it appears that modeling temperature variance isn't just as simple as running a simulation with .temp=100. This command seems to set every device's temperature simultaneously, which is not what I want. Would you, Lynn, be willing to share details of how you check a circuit's thermal stability?