My first amp design - comments

[blueskynis]

There is a better solution for R20 problem, I could split it in half and from there to take it to the base of Q7. In normal conditions it's base voltage should be at about 200mV. (thanks to the Andrew T.)

[jacco vermeulen]

Originally posted by blueskynis
lowpass

Any motivation for selecting that corner freq. in the V/I-limiter, Riedl picked 1KHz for the PA-300 iirc ?

[AndrewT]

Hi,
I would set a transient peak current of 14Apk for an 8ohm speaker. Ipk~=Vout/Z/0.35~=40/8/0.35~=14Apk. If 4ohm speakers are used then this becomes 25Apk (150W into 4r0).
Assume the output devices can supply 7Apk each then the driver output is about 400mApk.
At this current the driver gain may be down to around 70 or 80 requiring around 5mApk from the VAS.
That requires 8mA+-5mA just to run the amp into a severe 8ohm speaker. See what I mean about running the VAS hard. And why your protection Q7 will trigger well before peak current is achieved.

The output will source current through the VAS and the triggered protection. If the output is shorted or connected to a low impedance speaker then the EF preceding the VAS is overloaded.
The other thread decided that the best way to protect the EF was to add a collector resistor to limit the current passed from the ground into the EF.
The output base stoppers are a good addition. They may not be needed in which case a 0r0 link can be inserted. The higher this value the worse the distortion becomes. Most settle for a value between 1r0 and 2r2 as a compromise to minimise oscillation and minimise the added distortion. 10r might be OK, but be prepared to experiment with the value.

[blueskynis]

Originally posted by jacco vermeulen


Any motivation for selecting that corner freq. in the V/I-limiter, Riedl picked 1KHz for the PA-300 iirc ?

No, I haven't paid much attention to that. Current values give corner frequency of about 2kHz? Maybe it's not so bad selection after all ...will see later.

Originally posted by AndrewT
Hi,
I would set a transient peak current of 14Apk for an 8ohm speaker. Ipk~=Vout/Z/0.35~=40/8/0.35~=14Apk. If 4ohm speakers are used then this becomes 25Apk (150W into 4r0).
Assume the output devices can supply 7Apk each then the driver output is about 400mApk.
At this current the driver gain may be down to around 70 or 80 requiring around 5mApk from the VAS.
That requires 8mA+-5mA just to run the amp into a severe 8ohm speaker. See what I mean about running the VAS hard. And why your protection Q7 will trigger well before peak current is achieved.

The output will source current through the VAS and the triggered protection. If the output is shorted or connected to a low impedance speaker then the EF preceding the VAS is overloaded.
The other thread decided that the best way to protect the EF was to add a collector resistor to limit the current passed from the ground into the EF.
The output base stoppers are a good addition. They may not be needed in which case a 0r0 link can be inserted. The higher this value the worse the distortion becomes. Most settle for a value between 1r0 and 2r2 as a compromise to minimise oscillation and minimise the added distortion. 10r might be OK, but be prepared to experiment with the value.

Wow, Andrew, those currents are way larger than I expected. I loked at the datasheets once more and got 4mA max current, so you were right. As I can see, there are multiple ways to solve this problem:

- add more output devices (adding one more OP pair, current should be about 2-3mA)
- rise VAS current (are specified 2SA/2SC types any good for Ic more than 20-30mA? - I don't think so )
- add a unity gain buffer stage (is it really necessary?)

I need more time to think about this...

About protecting the EF, isn't that Q7's job? Sorry, but I can't locate the thread you are referring to. The search engine is just to simple...

In Symasym project they used 4,7ohms for base stoppers. Maybe I should start with that value too.

Anyway, about oscillation problems I had a few posts back... I managed to kill them (I hope so) using a feedback network which looks pretty much like feedback used in the Leach amplifier.

[blueskynis]

Last night I have finally got it how to measure open loop gain and phase and it showed a open loop gain of approx 80db! I never expected such a high open loop gain. Is it true, or multisim is going crazy again? Is the current mirror responsible for high gain?

About the buffer... I added the buffer like shown in schematic to see if anything changes, also I added another CCS set to 30mA. THD was not changed during normal conditions. Then I lowered the speaker load to 2 ohms (P=100W) and THD is improved by only 1db, so there is a little difference between buffered and unbuffered version.

Maybe the models of output devices found here on the forum are too optimistic?

Is there a more elegant way to add a buffer? Maybe is triple EF the answer?


Cheers

Nesa

[AndrewT]

Originally posted by blueskynis
- add more output devices (adding one more OP pair, current should be about 2-3mA)

For the current guess I took 14A output and output gain at 35 for 7A per device. This might improve to gain of 45 for a 3pair output stage.
That reduces the peak driver current from 400mA to 311mA. Their gain could improve slightly to around 85 giving a VAS drive of around +-3.7mA (quite a bit down from the 5mA estimate).
But these are all estimates based on driving a severe 8ohm load.
A triple EF output stage is easier to get current gain from than trying to select the best gain devices for a double EF stage. Worth experimenting. D.Self and others got it to work.

D.Self offers the EF after the VAS as an option instead of the EF before the VAS. He also said the post VAS EF was compulsory if the VAS was cascoded. I have seen others do it this way and even then follow the Cascode+VAS+EF with a triple EF.

[blueskynis]

Hi Andrew

Here is what I am going to do: make a compromise. I will add another OP pair and increase VAS current to 10mA from previous 8mA. That should do the job just fine.



Cheers,

Nesa