F4 Beast Builders

Jump to Latest
#1,602
SPICE yes, so all the usual caveats apply. That said, and although I accept your point about perspective, the increase in noise is still x10 even for a resistor <10R. If the resistor is increased sufficiently to equalise the AC current in the output FETS, the noise becomes worse still (~ x100 over the no resistor case). This may or may not be important in practice but I would prefer not to add noise at all and hence and I am wondering if there may be better way to achieve the same effect. Unfortunately so far I haven't found any that don't have their own issues.
 
#1,603
Lynn: have you considered this alternative for gm adjustment? it has its own pros and cons of course but does allow a smaller resistor in series with the input which benefits both noise and damping factor. The downside is reduced rejection of noise on the negative supply.
 

Attachments

  • Gm adj.PNG
    Gm adj.PNG
    10.9 KB · Views: 718
#1,604
The other downside is that the drain of the OS FET has a very high K2 component, thus changing the harmonic character of the signal to the gate. I would need to run simulations to determine how significant that effect would be and also the PSRR issue that you suggest.
 
#1,609
Ian Macmillan said:
Lynn: have you considered this alternative for gm adjustment? it has its own pros and cons of course but does allow a smaller resistor in series with the input which benefits both noise and damping factor. The downside is reduced rejection of noise on the negative supply.
Click to expand...
I have not yet run any simulations on that circuit, but after closer examination it looks like the R17 C17 combination could connect to ground with the same effect, and C17 could probably be eliminated. Something looks very "fishy" here. Why would 10K to ground at that point make any significant difference to the gm of the PFET?
 
#1,610
You may be right. I hoped to achieve some measure of negative feedback from the connection to the PFET drain but it seems that it is negligible and connecting to ground has a very similar effect. I can't explain why a 10k resistor to ground would have the desired amount of attenuation though so perhaps something is indeed fishy as you put it. Interestingly, both forms of connection appear to result in both an increase in NFET signal current and a decrease in PFET (dc values are unchanged) which is not what I expected. Further investigation is required.
 
#1,612
maartentje said:
...
Im currently working on the F4 beast pcb design. I will show some stuff in a few weeks probably. Im going to try the opto bias parallel version.

Martin
Click to expand...

I recently did bench tests of the opto bias parallel version and have found some issues bias adjustment procedure. That circuit has a positive temperature coefficient, meaning that the bias current increases with heatsink, optocoupler, and zener diode temperature. In my build the bias current increases about 4mA/C of heatsink temperature. That is not necessarily bad, but it does increase the time before the heqatsink temperature stabilizes, and makes it difficult to guess the initial adjustment of the bias current. I can provide more details as needed.
 
#1,614
maartentje said:
Is there a substantial benefit in using the Cbias capacitors?
Click to expand...

Referring to the schematic in post #1594, there might be three options regarding Cbias (C1 and C2):

  1. Cbias as shown in post #1594, connected between top of R10 and J313 source pin. FET has only DC source degeneration. Gain is insensitive to pot adjustment.
  2. Cbias connected between top of R10 and rail (VF+). This is the same as in the BA3 Front End, shown below. FET has full source degeneration. Gain is insensitive to pot adjustment and about 15dB lower than option 1.
  3. No Cbias. FET has full source degeneration. The open-loop gain is sensitive to the pot adjustment. Poor choice.
My amplifier was originally built for option 1, but recently modified to allow easy switching between options 1 and 2. After 3 or 4 days listening to option 2, I reverted back to the original option 1 configuration, whose sound I preferred. Since 3 or 4 minutes are required to switch between the options, my listening tests were not really A/B tests where one can quickly switch.
 

Attachments

  • BA3-FE.jpg
    BA3-FE.jpg
    50.7 KB · Views: 908
#1,616
Ian Macmillan said:
There is also a fourth option: Cbias connected across R8 (Rs). FET has only DC source degeneration. Gain is sensitive to pot adjustment. Probably not a great choice but worth noting for completeness.
Click to expand...

You are correct. I actually tried that option about 1 year ago. The main problems I saw were with the RC time constant and with the series resistance of the capacitor, and variability between that resistance between the caps on the positive and negative rails. In order to get a time constant equal to option 1 requires the capacitor to be about 5 times larger.

If you were to use that fourth option, you would also probably want to have the capacitor of option 2 so that pot adjustment does not change the gain and thus the even harmonics. If the cap series resistance is not a problem, that should result in essentially the same behavior as option 1.
 
Last edited:
You must log in or register to reply here.
Top Bottom