P3A Bootstrap and NFB Capacitors

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello all,
I built a stereo ESP P3A amplifier with my own boards and am very satisfied with its sound.

I am issuing a 2.0 board with more room for (physically) bigger resistors. At this point I am asking if it could make sense to place a footprint near C4 and C5 to place a small (or not so small) MKT capacitor to bypass them.

I even thought to use, instead of the usual 0.1/0.22uF bypass value, a Wima MKS2 6.8uF/50V (only 5mm pitch) in parallel with a 220uF/35V Elna Cerafine.

In the FB leg for example, 6.8uF with 1k (R4) would make an LF -3dB frequency of about 25Hz, bringing the contribute of the MKT down to medium/low frequencies and leaving the electrolytic alone to pass only very low frequency signals. What do you think about this?

Thank you and cheers from italy!
 
An externally hosted image should be here but it was not working when we last tested it.


You are talking about C3 (100uF) and C5 (100uF) in this schematic.

I am not sure much to improve by using smaller film caps to parallel those
... I am not the one to parallell capacitors ;)

C3, FB blocking, has got a good value: 100uF
which matches C1 x R2 = 4.7x22

To me, the Bootstrap C5 100uF electrolyt seems a bit high in value
I would like to experiment with 10, 22, 47uF here
For example 10uF (or 6.8uF) MKT, MKS, MKP film cap might work well as Bootstrap.
But high quality electrolytics might do.
 
My first question would be why you want to do this. Do you feel you are missing something in the sound?

I have by-passed big caps in the NFB path with lower value polypropylene before but I cannot say that it made any audible difference.

I am sure if Rod felt the need for a low value by-pass cap he would have included one.

If you want to do this, there is no harm.

Nico
 
i have a question also

somewhere you mention about fysicly biger resistors ....would you like to tell us more about that ?????

( i keep in mind something i read somewhere that many exotic designs that use exotic huge caps have some trouble since big exotic caps are working more like antenas compaired to normal capacitors due to size ......eventhough might sound better but also might be more sensitive to anything ....)

as about bypassing in this areas i would go with Nico Ras ......

thanks
 
Thank you for your replies guys.

I really like the sound of this amp but when I'm in the phase of PCB design, I always feel nervous about missing something that I'd regret later. So I asked about those bypass caps.

I understand bypassing electrolytic capacitors would not bring a great improvement.

About bigger resistors, nothing exotic. I wanted to try some Holco H4 0.5W resistors and I'm laying down a PCB with 12.7mm pitch against 10mm of standard 0.5W metal layer resistors.

I understand bypassing electrolytic capacitors would not improve the sound.

Being also in the market for new components, if you don't mind I'd also like to know what are your choices about some other things about this amp and the like.

1 - I am using silvered mica caps for C2-C4-C6.
Do you think some other types would be worth trying? What about something like Wima FKP2 (polypropylene film and foil)? I read polystyrene is a very common choice for C4 and C6.

2 - What about reducing C2 and reducing/removing R1/R3?

3 - What kind of resistors do you use for R13/R14? I'm using 4W MOX now.

4 - I'm planning to use a 2SA1406/2SA1306/2SA1209 (depending on what I find) as Q4 since their low Cob. Is it also worth trying other transistors than BC546 for the input differential pair?

5 - Has anyone ever tried modding this particular amp to use lead compensation? I am very curious about this. Would 10 pF for both C4 and the new capacitor between collector of Q4 and base of Q2 be two good starting values? Simulations tell they're ok but I wonder if anyone already tried this in real life.

6 - Now something hardcore. Using a tube preamp and having another power amplifier (a SE DHT with lower gain than the P3A) I'd really find useful to have less gain from this amp. Would bringing R4 to 1.5K be too harmful for stability? Spice tells everything is ok with 1.5K for R4 but only using lead compensation. With standard 100pF Miller compensation I get an awful dip in magnitude response at about 1MHz.

Thank you for your help!

Christian
 
Peak at 1MHz?

That suggests you have too low phase margin

Maybe the circuit depends on the internal miller capacitance of the VAS to be stable? What transistor did you simulate with?

If the current through LTP is increased, the circuit gets more unstable too because of increasing transconductance of the input stage.

Slow output stage (wrong bias current?) or capacitive loading of the whole amp will also make that happen.
 
Please note that I get this peak only in simulations using 1.5K for R4, that means increasing the global feedback value by about 3.5 dB. I'm simulating using MJE15035 or MJE350 as VAS transistor (I can't find 2SA1406-2SA1306-2SA1209 spice models!), BC546 input pair, MJE15034/5 drivers and MJL4281A/MJL4302A output.
 
christian.lenci said:
Please note that I get this peak only in simulations using 1.5K for R4, that means increasing the global feedback value by about 3.5 dB. I'm simulating using MJE15035 or MJE350 as VAS transistor (I can't find 2SA1406-2SA1306-2SA1209 spice models!), BC546 input pair, MJE15034/5 drivers and MJL4281A/MJL4302A output.

Increasing R4 increases loop gain as you say. It also moves the unity gain crossover frequency up by ~50% which makes the circuit more unstable.

I don't see much of a problem with those models mentioned (assuming they are good models!), they are pretty similar in speeds to what was originally used and should not cause the instability.

A way to fix point 6 if you want less gain is adding emitter resistors to the LTP - they decrease the GBWP without decreasing slew rate like when you increase cdom. It also makes the input stage more linear.
 
megajocke said:


A way to fix point 6 if you want less gain is adding emitter resistors to the LTP - they decrease the GBWP without decreasing slew rate like when you increase cdom. It also makes the input stage more linear.

I think this would only reduce open-loop gain, closed loop gain still relies on (R5+R4)/R4 ratio, isn't it?


I managed to get matched couples of 2SA1406 (VAS) and 2SB649/2SD669 (drivers) so I'll test them directly with a scope since I couldn't find spice models.

-Just wanted to know what kind of resistors (MOX, wirewound, carbon) do you use at the emitters of the complementary pairs in the output stage and which you prefer. I'm using MOX now.

-Also, what kind of capacitors do you prefer for C2, and C4 and C6.


Thank you!
 
christian.lenci said:


I think this would only reduce open-loop gain, closed loop gain still relies on (R5+R4)/R4 ratio, isn't it?


True!

But reducing closed loop gain by changing feedback without doing anything else will increase loop gain and loop unity gain frequency, decreasing phase margin.

If you also lower the open loop gain-bandwidth-product by adding/increasing emitter resistors you can keep loop gain the same and not compromise stability. If the topology is kept, the other option is increasing Cdom or decreasing LTP standing current (decreasing transconductance) but this will lead to reduced slew rate and worse distortion.
 
I've simulated this a few times, and it always turns out unstable without further compensation. Also, the minimum bias current is always crazy high, meaning I have to reduce the value of R16, or make VR1 larger.

I suspect the PCB layout is a very critical part of this amp's working, although I did build a version on Veroboard that worked well as a test - I only used MJL21193/4 outputs though. This design also benefits from larger decoupling, say 330uF near to the output transistors, which Rod's PCB does not have.

Personally, I don't think CFP output stages are worth the hassle - i'm a Triple man myself.
 
Class AB cfp isn't my favorite - too much complication in turn on/off dynamics

of course with what I'd consider adequate bias - say 50-100 mA in Q5,6 (in addition to the 100 mA output Q bias) local stabilization measures like base R might be needed - (any guess why no collector C on the upper half to match C6?)
I'd try smaller/faster types for Q5,6 as well, the BD139/140 could be good with a little heat sinking

C5 should hang off Q6 emitter for better bootstrapping - no reason to accept the .33/Rload division ratio

I'd also add load isolating R||L

current mirror load for diff pair collectors is the best way to insure diff pair collector current balance - at a minimum add some way to trim R6 and measure the current balance in Q1,2

of course I would also change much of the compensation (TMC or 2-pole), darlington VAS, Vbe multiplier bypass...
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.