1 & 2 I saw that.
3-You do well, Andrew, as 3/4 power is around the point where an AB amp dissipate the most heat. And large enough for any test of high signals.
On my side, i try to do not use any Zobel in my amps if i can.
I just started to bring my prototype (attach. #1) back...
And the IPS is already up and running...
(The IPS output is now shorted and the FB goes to this point as well.)
For starting I removed ALL HF part from the cct.
Even now there is an oscillation by default (attach. #2: 200mV/DIV and 20ns/DIV)
AND: already here when I touch the FB cable it vanishes immediately with an output noise level: 5mVpp band at all timebase setting.
I just posted this because it may help before I start the OPS decoupling experiment (which - don't be afraid..!
- I'll do at all events...)
The input is shorted as well and when I start to open the pot even a slightly the oscillation goes away again. (In this case the noise level is already about 40mVpp.)
Trying the usual HF comp caps dont eliminate it either.
And the IPS is already up and running...
(The IPS output is now shorted and the FB goes to this point as well.)
For starting I removed ALL HF part from the cct.
Even now there is an oscillation by default (attach. #2: 200mV/DIV and 20ns/DIV)
AND: already here when I touch the FB cable it vanishes immediately with an output noise level: 5mVpp band at all timebase setting.
I just posted this because it may help before I start the OPS decoupling experiment (which - don't be afraid..!
- I'll do at all events...)The input is shorted as well and when I start to open the pot even a slightly the oscillation goes away again. (In this case the noise level is already about 40mVpp.)
Trying the usual HF comp caps dont eliminate it either.
Attachments
A square wave (100kHz 5Vpp) with 68pF @ input (nothing other C else) + touching the vas-output-FB wire...
Any idea how to eliminate the oscillation (without touch) and the sensitivity to touch before going further to OPS decoupling?
And how could I check whether my original tiny hum issue is maybe already here at this stage?
Any idea how to eliminate the oscillation (without touch) and the sensitivity to touch before going further to OPS decoupling?
And how could I check whether my original tiny hum issue is maybe already here at this stage?
Attachments
Last edited:
Oh, your amp is a voltage feedback one, right ? So the feedback is high impedance. it is normal to have this behavior, if your feedback path is not shielded well enough.
BTW: Did-you have a serial inductance before your speakers wires, in order to stop any HF to return trough the feedback ?
Wrong.
We were talking about a sqw test signal.
For a Push pull amplifier passing a sinewave the maximum dissipation is when Output Vac ~ 63% of Vsupply.
The Output power is much less than 50% of the maximum power from the amplifier, (probably around 40% of maximum power).
Back to the sqw test signal.
If the sqw wave is perfect with instantaneous rise and fall slopes, then you can do a "simple" analysis of output power and dissipated power.
At maximum output the sqw power would be exactly double the sinewave power and the dissipated power would be zero: [Imax * 0Vce] + [Izero * Vcemax] = zero dissipated power.
But there are losses, the actual is not equal to the "simple".
BTW, switching PSU and switching Amplifiers use that 0Vce & Izero to dissipate very low power during the sqw cycles. That's where they get their very high efficiencies.
Last edited:
Esperado!
I just ran into your website by chance yesterday and found this great article about the CFA
VFA comparison (a BIG THX for it!!) and that you recommend to decrease the FB values
even in VFA amps so I just did it on my (2k + 47R) and the sensitivity got clearly better!!
Why didnt you recommended this approach earlier..?!
It did still oscillate in itself but now I could stop it easily with some caps without any shielding with long antenna wires!
See attach at 100kHz 1V/DIV. (The error at the edges just before the start are already there on my fun.gen...)
Still just with the IPS 68pF+10R @ input and 47R+330R @ LTP.
Another question about these HF filter/comp caps: when and how much series R should be added to these..?
For example I tried to add this 10R to the input cap and it is still works with it...
Attachments
Last edited:
Well, the problem with LTP (voltage feedback) is that it is better to equal the impedances of the two branch. For DC offset. So, if you decrease the impedance of the feedback, or you will mismatch the LTP, or you will have a low input impedance.that you recommend to decrease the FB values
even in VFA amps so I just did it on my (2k + 47R) and the sensitivity got clearly better!!
Why didnt you recommended this approach earlier..?!
After testing it on my prototype I already tried it on my big amp as well and it works great..!
The hum has gone as well and now the amp inst so sensitive any more to touches, heatsink touching, microphony, etc...
However there is still some oscillation so for next step I'll try to "decap" my OPS.
The hum has gone as well and now the amp inst so sensitive any more to touches, heatsink touching, microphony, etc...
However there is still some oscillation so for next step I'll try to "decap"
my OPS.
)
Not so long ago I was just near to be crucified here for the crime to daring to promote CFAs.Esperado: God bless you for this article...
It refer at a schematic where feedback is applied, in low impedance, to the emitter of the input transistor, while the signal is applied to the base of the same device, and out by the collectors.
More on this subject there:
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-new-post.html
Thanks Esperado! (Yet another long thread to read... )
BTW: now I am playing with the stability issues from scratch @ my IPS and I noticed that the VAS/FB cables are touch sensitive.
(Now the IPS output is shorted and thats my FB point, OPS is disconnected...)
Based on that I tried to place a cap @ PCB just between the FB point and GND to shunt any HF from the cables.
I tried different values and with 100pF now it looks stable and way more insensitive to external electric fields.
Now I'll try to check whether this sensitivity is based on the VAS or on the FB cable...
)BTW: now I am playing with the stability issues from scratch @ my IPS and I noticed that the VAS/FB cables are touch sensitive.
(Now the IPS output is shorted and thats my FB point, OPS is disconnected...)
Based on that I tried to place a cap @ PCB just between the FB point and GND to shunt any HF from the cables.
I tried different values and with 100pF now it looks stable and way more insensitive to external electric fields.
Now I'll try to check whether this sensitivity is based on the VAS or on the FB cable...
Attachments
I made a lot of experiments around HF compensation/filtering on my prototype
and it turned out this symmetric kind of soultion works the best..!
The point is: using 2 x 4p7 from both VAS end to FB instead of the original (and conventional) 1 cap (around ~10pF) from VAS+ to FB.
Also the 2 other caps (from VAS out to GND) were neccesary to make it rock stable.
I minimized all the caps untill I could. With this setup it is able to a square wave like I attached:
250kHz @ 2V/DIV @ 6R dummy resistive load.
The ripple on the top is already there on my func.gen.
It is similar @ greater output level but didnt wanted to risk a Zobel R to burn while makeing the picture...
And with this caps finally my basic (white)noise level (without input) is inside of the 5mV band.
Should it be even less..?
OPS: it is not on the picture but I implemented the exact PS layout what is on the SlewMaster OPS.
Decaps: main: 470uF+100nF (separate 22uF for all power trannys) then 4R7 + 47uF+22nF to the pre-driver/driver rails.
33pF @ predrivers CB and base stoppers: 22R@driver, 4R7 @ power trs.
I made a lot of changes so I could not tell the exact effect but I'll do a de-decap at the and just to test only this parts effect.
To me it looks like it helped with the overdriving stability @ clipping now its nice no ringing or any sign of collapse.
Funny: a bit of moderate hum is there again...
I hope it is jus due to the spagetti like wireing at the prototype layout.
BTW: it is amazing how much capacitive sensitivity just a little bit of instability adds to the amp...
Even from 20-30cm the human presence makes a difference without any direct contact.
Touching the wires (even the AC main cables at the outlet...) has a great impact on the state of the amp, etc...
It is very instructive...
and it turned out this symmetric kind of soultion works the best..!
The point is: using 2 x 4p7 from both VAS end to FB instead of the original (and conventional) 1 cap (around ~10pF) from VAS+ to FB.
Also the 2 other caps (from VAS out to GND) were neccesary to make it rock stable.
I minimized all the caps untill I could. With this setup it is able to a square wave like I attached:
250kHz @ 2V/DIV @ 6R dummy resistive load.
The ripple on the top is already there on my func.gen.
It is similar @ greater output level but didnt wanted to risk a Zobel R to burn while makeing the picture...
And with this caps finally my basic (white)noise level (without input) is inside of the 5mV band.
Should it be even less..?
OPS: it is not on the picture but I implemented the exact PS layout what is on the SlewMaster OPS.
Decaps: main: 470uF+100nF (separate 22uF for all power trannys) then 4R7 + 47uF+22nF to the pre-driver/driver rails.
33pF @ predrivers CB and base stoppers: 22R@driver, 4R7 @ power trs.
I made a lot of changes so I could not tell the exact effect but I'll do a de-decap at the and just to test only this parts effect.
To me it looks like it helped with the overdriving stability @ clipping now its nice no ringing or any sign of collapse.
Funny: a bit of moderate hum is there again...
I hope it is jus due to the spagetti like wireing at the prototype layout.
BTW: it is amazing how much capacitive sensitivity just a little bit of instability adds to the amp...
Even from 20-30cm the human presence makes a difference without any direct contact.
Touching the wires (even the AC main cables at the outlet...) has a great impact on the state of the amp, etc...
It is very instructive...
Attachments
The 4p7F doubled is quite a lot of Miller input inclusive capacitance (MIC) is maybe a bit high. It may need a zero (resistance in series to control it).
The 10pF doubled is very low compared to the 220pF and 330pF shown in earlier versions of SymaSym. And even more so since the resistors are also not used.
The AC route from VAS to current mirror is missing. I wonder if that is affecting operation. The unusual sensitivity to your hands seems to be a warning. I can generally touch components and only very rarely do I see any effect of a touch.
250kHz sqw is very high, that's signals out to near 2.5MHz. Your RF filter is set to ~2MHz, so it lets most of that through.
The 10pF doubled is very low compared to the 220pF and 330pF shown in earlier versions of SymaSym. And even more so since the resistors are also not used.
The AC route from VAS to current mirror is missing. I wonder if that is affecting operation. The unusual sensitivity to your hands seems to be a warning. I can generally touch components and only very rarely do I see any effect of a touch.
250kHz sqw is very high, that's signals out to near 2.5MHz. Your RF filter is set to ~2MHz, so it lets most of that through.
Last edited:
But that was only before I stabilized my amp! Now it's completely insensitive to these "inputs".
And my default noise level is acceptable? At 5mV/DIV it takes a whole band (5mVpp).
And what about the caps (~100nF) that are usual at the Vbe and on the emitter resitors at the (pre)drivers?
Are they involved in this or they are only for helping at the class B switching?
I tried them a few times in the past but couldnt see/hear any difference on the scope/in my ears...
What should they improve? Is it visible? Or its just a convention without prrof?
See ? Told you so.
Now you can sit your OPS on the trafo and touch it all day.
I have a test one with wires all over the place , real sloppy ....
plays perfect - no issues at all.
If you copy the slew Vbe , you can put the OPS in the freezer (or outside)
then abuse it (get it hot). 2mv bias change max. Usually it will settle
within .5mv of where you originally set it when thermally stabilized.
PS - that 5mv could be noise pickup from your probes , mine scopes in the uV
range for noise (input stage) , before I fixed it , I got the hum down to where
it just modulated the 20-30uV noise waveform.
OS
Now you can sit your OPS on the trafo and touch it all day.
I have a test one with wires all over the place , real sloppy ....
plays perfect - no issues at all.
If you copy the slew Vbe , you can put the OPS in the freezer (or outside)
then abuse it (get it hot). 2mv bias change max. Usually it will settle
within .5mv of where you originally set it when thermally stabilized.
PS - that 5mv could be noise pickup from your probes , mine scopes in the uV
range for noise (input stage) , before I fixed it , I got the hum down to where
it just modulated the 20-30uV noise waveform.
OS
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.