DIY Audio Projects Forum

Strange, I never have problem with 10X probe at work or at home. It must be the capacitance of the 1X probe cause your circuit to oscillate. I can assure you the 10X probe will see the oscillation.

If any audio amplifier oscilaltes with few pF load on output then I consider it very bad and instantly fail to my own quality control standards.
When I make new op amps sometimes I encounter these scehamtics, I always do check to see if amps can take at least 220pF pure capctive loads.

It's easy to test. Hook the 1X probe and observation the oscillation. Then with the 1X probe still attached, put the 10X probe and look at both channels. If the 1X show oscillation and the 10X doesn't, then you are right.

I bet you anything you will see oscillation with the 10X probe this way. OR if the oscillation stop, both probes will show it.

I test all my amps with 3300pF, 6600pF, 9900pF, 22,000pF, 32000pF. I sweep frequency at different amplitude to saturation to make sure no oscillation. I even turn power on and off with frequency to make sure no sustain oscillation.

Speaker cable is as important as the power amp in high end system. I did a lot of experiment with it and I design my own cables. The best I found is the multiple pairs of small cables twisted together. These cables are very high capacitance. Measured about 2500pF for 8ft long.

Again, it's not my original idea, people sell "knitted" cable like this, they just don't explain why. A lot of amps are NOT stable using those cables. My PA-7 amp oscillates with those cables.

Stability is very tricky. You just don't put a small cap, then say it's good if no oscillation. I design a lot of closed loop feedback system and I was the go to person to fix oscillations. It's not that easy. That's why when you claim you have 4 stages, 163dB open loop gain, I seriously question this. Then you say you only have about 150KHz BW. Nothing adds up. I can't comment before because I don't have the schematic.

Even a normal 8ft Monster cable might have more than 220pF, and you said you see oscillation at the output with 1X probe. These are very alarming symptom.

Some cheap amps took a short cut by put an inductor between the speaker output and the amp to isolate the capacitance. That hurts sound quality. No high end amps put an inductor. That's what make taming amp so hard.

Now you getting into the heart of the power amp. How to put more gain, but remain stable under trying condition. Like most of the electronic, designing the circuit is easy, keep it stable is what differentiate the good and the bad.

JMHO

Hi

I want to make sure that you don't take me as attacking you. This is what I meant of not making sense:



I take your comment of 164dB open loop gain at low frequency. I take you said BW of about 140KHz and back extrapolate the Bode Plot as shown. It is logical to assume you use 2 poles roll off to get the roll off fast enough so you can get down to loop gain equal zero at about 140KHz. You need to introduce a zero to ensure one pole crossover at the -3dB point.

You did say your closed loop gain is about 3.3, so I use 10dB as your closed loop gain. I assume your -3dB frequency is 200KHz and draw the Bode Plot. You did say you use 2 compensations.

In order to make this fit, You have to start rolling off about 15Hz. It goes -40dB/dec all the way until the very end (60KHz in the plot) and a zero kick in to bend it back to -20dB/dec to get phase margin of 90deg. If you look at the LOOP GAIN ( open loop gain minus closed loop gain). at 10KHz and 20KHz, they are only 40dB and 30dB resp. That is not a lot of head room. You are not gaining any advantage over lower open loop gain. Based on the fact you see distortion at high frequency tells me you don't have loop gain head room.

To get better loop gain at higher frequency, you need wider bandwidth. That's what I was concentrated on, getting large signal of 350KHz and small signal even higher. then you can have simpler compensation and more loop gain at high audio frequency.

For one, the TIP36 is only 3MHz, it's too slow in my book, that will put a lower pole frequency. I use 30MHz transistors for output. The drivers are 60MHz and 100MHz.

This is all the derivation from what you said as there's no schematic.

Regarding to stability under capacitance load, some high end cables are very high capacitance. Like this one:

https://www.audioadvisor.com/prodinfo.asp?number=KK8PR&opt=649|655|657|660

They use a lot of small wires "knit" together to get very close coupling to reduce series inductance. I have my own design by using 5 pairs of 16 gauge speaker cables in parallel and twist them together to get tight coupling. Result is very amazing.

But the down fall is very high capacitance. Mine is about 2500pF, some of those on the market can go as high as 5000pF. A lot of amps will oscillate using these kind of cables. They miss out the best cables.

Not only you have to test at least to 5000pF, you cannot use any load resistor in parallel. Speakers are like open circuit at 300KHz and above as they are inductance. So at oscillation frequencies, it's like an open circuit with only 5000pF capacitor across the output.

Also, you cannot connect the capacitor by a cable, it has to be right at the output of the amp.

It would be a whole lot easier to stabilize the amp if the cap is hook up through a cable, OR if there is a load resistor in parallel with the cap. The cable serves as series inductor and resistor that isolate the output of the amp from the capacitive load.

This is where it gets difficult in designing power amp. designing the circuit is very easy, so many different ideas. Getting it to behave is where the rubber hits the road. It took me like 2 weeks to design the compensation done.

You cannot rely on closed loop feedback to get low output impedance ( less affected by the capacitive load) because as your loop gain goes down, output impedance goes up. At oscillation frequency, you have no loop gain. The OPS determine the output impedance.

I have other things that I want to talk about, but I'll leave that to later quite bussy latley.

If the op amp response was like the one you've drawn then it would sound terrible

The openloop freq response is about 6-8X higher than 140kHZ = 840+kHZ

The frequency resposne is reduced at feedback path externally to preserve open loop gain while reducing distoriton.
All this ment was that at high frequency the gain is unity and at audio gain is flat.
This helped reduce distortion.

My unique 3 stage op amp can withstand a pure capacitive load of any value 1pF to 100uF without showing any extra disotrion on sine waves, less than three oscillations on square and little overshoot
Have not tested 4 stage with pure capacitive load stablity but not stable at 0.1uF load, cause I was in a mega rush building it.

As mentioned before the loop gain at 24khz is much higher on my 4 stage design compared to my 3 stage design.
3stage was higher than standard 2 stage designs on loop gain at 24khz.

The 4 stage design had so much more gain that didn't use any tricks on output stage and gave me lower distortion than of 3 stage.

I normally test with short wires comming out of the amp directly on bench table.

I'm using tip41c for head amps and they are only about 0-15% faster than tip36c. Since power disipation is much lower

You actually look at on the scope to get 840KHz. I doubt you can get it with TIP36 or TIP41. For fT=3MHz, your beta drops to about only 4 at 840KHz. I don't even trust my amp that use much faster transistor can get there. I never measure mine because I have a low pass filter of 350KHz at the input, so I see large signal BW of 350KHz. 350KHz is plenty. Next time when I hook it up, I'll take off the low pass filter and see what is the true BW of my amp.

That's the reason you need the Spectrum analyzer at least like the one I have to really tell. You cannot see distortion, when you see distortion, you are getting into 1% high distortion region.

Did you try the scope probe test hooking both 1X and 10X to see whether you see the oscillation on both? I bet you anything that if there's oscillation, you'll see on both.

I'll get back to you with some testing and photos of scope on both capacitive load and frequency resposne of the 4 stage op amp.
In the mean time I need to get my stuff done.

Heres what the 4 stage looks like, one shot of non frequency reducted sine response with 8r load
other shots are pure resistive 8r loads

Without frequency reduction the slew rate is about 8v/us with a output voltage of 3v pk-pk
With larger output voltages slew rate is slightly higher.

I haven't showed capacitve testing but it dosen't oscillate if you have 8r load in parallel with any capacitance.
Have not tested pure capcitive loads but takes at least 1nF to make it oscillate.

I actually tried out a new type of emitter and collector resistorless output stage biasing method, but didn't work out well.
Distortion figures were pretty much same, zero gains were observed.
I ran out of time so I skipped the capacitve testing.

Yeah I think emitter and collector resistiorless amplifiers aren't good in anyway other than being unique and that I like the sound of it.
Distortion I think is higher than standard output stages however I have not built any for a long time to do comparison with emitter resistorless.

I am trying to understand

Is this 23KHz sine wave as said on the lower right of the scope?
You see a kink circled, you need to check whether it's the scope problem or it's oscillation.


Is this 123KHz sine wave?
Again, I see some kinks.


Is this still 123KHz?
I see the sine wave starting to look like triangular wave that indicates slew rate limit.


Is this square wave 123KHz?
The crossover reagion has problem.

You definitely have crossover problem, quite bad.
8V/uS is quite slow.
You need to try large signal, this is small signal in my book. You need to run close to rail even if you have low supply voltage. Like taking pictures again, if you have +/-12V, take picture of about 18Vpp. Then you can see a whole lot more the deficiency of the design.

I am bringing up my new amp, I'll be in position to take some pictures in a few days.