Hi Jan, thanks for the reply. As I said, I did understand your basic intent to be positive, regardless.
Let's make peace on this. I suggest you edit your post 16333, to point to my post 16368, which has the path to the full article. Anyone interested in what has been discussed on Ampzilla's front end should read it all. I especially like the part about those "unnecessary" changes.
It turns out that I already have a lot of it already scanned, Ed. Just never posted to my web site. This includes reviews of some Heath amps, and the book reviews of the Ben Duncan and Doug Self power amp books. Those two reviews have a lot of historical info on power amp developments.
Walt Jung
Hi John,
You're right, about 100 V/us should be fine, especially if the open loop amplifier is quite linear. Awhile back you seemed to agree that 1uH or less was OK for the output coil, and that such a value would not compromise sound quality. I like 1uH or even 0.5uH in there to assure stability into virtually any load, including a low-ESR capacitive load at the amp terminals. Maybe in this regard I am being too conservative.
You also need to recognize that the output stage stability issue when not using a coil can involve the global feedback, the local output stage stability, or both. Output stages tend to be emitter followers, and emitter followers naturally like to oscillate into capacitive loads. This is even more the case when you use fast output transistors like RETs or MOSFETs. Some of the things one may do to allow stability into a capacitive load can compromise other areas of amplifier performance.
In your case, those very high-value 10-ohm base stopper resistors in the JC-1 may have helped you be reasonably stable into capacitive loads, but they compromised your HF performance by making the output transistors brain dead. They also compromized your thermal stability by introducing beta-dependent base voltage drops. That's part of the reason you had to back off from 0.1-ohm emitter resistors to 0.15-ohm emitter resistors. Pick your poison. At least you didn't throw out the zobel network
.Cheers,
Bob
I might update on what we are talking about to others looking in here.
Back in the early days of complementary amp design, we didn't NEED any 'stinking' base resistors! We ran direct with relatively slow power transistors like about 4MHz. Just look at the schematics of the JC-3 and the Ampzilla, and see for yourself.
Around 1980, the first really good 'ring emitter' transistors because available, and I proceeded to make my 'super speed' FQ-2 power amplifier with these devices, AND I did NOT originally put in base resistors. The amp worked, it was stable with any capacitive load that I could throw at it on the test bench, but when Brian Cheney took it home and played it, it would blow up! What to do? Well, a former designer, now probably deceased, Jon Iverson of Electro-Research, gave me a 'tip' when I discussed my problem with him. He said to add 10 ohm base resistors and see if that fixed the problem, AND IT DID! No more problems. So, in future designs I specified the 10 ohm base resistors, and they are now in 1000's of Parasound amps of dozens of different designs and power points, from 100W-400W.
The resistors keep the amp stable with the high speed output devices that we now use, but they require better matching of both the output device betas and heat sink temperature, in order to keep one of the output pairs from 'running away' and blowing up the amp. We went to: better beta matching, a more uniform heatsink assembly, and ultimately 5 ohm resistors, instead of 10 ohms. This does help, BUT it does allow for the possiblity of output device self oscillation, if we go too low in value. I stopped at 5 ohms or so. Others have been more adventuresome, to 3.3 ohms. I don't dare go that low, without a paid project to worst case the amps once again with these lower value resistors. So be it, I will stick to 5 ohms or so. Works for me.
Back in the early days of complementary amp design, we didn't NEED any 'stinking' base resistors! We ran direct with relatively slow power transistors like about 4MHz. Just look at the schematics of the JC-3 and the Ampzilla, and see for yourself.
Around 1980, the first really good 'ring emitter' transistors because available, and I proceeded to make my 'super speed' FQ-2 power amplifier with these devices, AND I did NOT originally put in base resistors. The amp worked, it was stable with any capacitive load that I could throw at it on the test bench, but when Brian Cheney took it home and played it, it would blow up! What to do? Well, a former designer, now probably deceased, Jon Iverson of Electro-Research, gave me a 'tip' when I discussed my problem with him. He said to add 10 ohm base resistors and see if that fixed the problem, AND IT DID! No more problems. So, in future designs I specified the 10 ohm base resistors, and they are now in 1000's of Parasound amps of dozens of different designs and power points, from 100W-400W.
The resistors keep the amp stable with the high speed output devices that we now use, but they require better matching of both the output device betas and heat sink temperature, in order to keep one of the output pairs from 'running away' and blowing up the amp. We went to: better beta matching, a more uniform heatsink assembly, and ultimately 5 ohm resistors, instead of 10 ohms. This does help, BUT it does allow for the possiblity of output device self oscillation, if we go too low in value. I stopped at 5 ohms or so. Others have been more adventuresome, to 3.3 ohms. I don't dare go that low, without a paid project to worst case the amps once again with these lower value resistors. So be it, I will stick to 5 ohms or so. Works for me.
Last edited:
Actually MOSFETs don't suffer from this problem. It seems that there is an extra pole (I think that's what it was) in the BJT that is lacking in the MOSFET. This is covered by Feucht, who gives the only English-language analysis of why the follower oscillates into a capacitive load.
I have built thousands of MOSFET amps with no output inductor and no problem. My first BJT amp worked great on the test bench, but oscillated into a real world load. I took me two months to figure out a solution that didn't rely on an output inductor.
I don't think removing the output inductor and still having a stable amp is that difficult to do. What you have to do is make sure the output device pole remains well above the ULG frequency, so that when you do hang a 2uF load on the output, the pole does not drop below the ULG frequency. A judiciously selected lead capacitor across the feedback resistor (not too large) can also get you an additional 10dB of gain margin, although some people frown on this because it can inject RF back into the inverting node of the amp. However, these points are related to loop stability, and not local parasitics, where you need additional steps if you are to avoid HF problems.
PMA, burnout was not due to parasitic oscillation. You can run my amp without base stoppers no problem (but NOT without the base RC network in the drivers). I have not run it without the output coil though or without the Zobel. The output devices are slow 21193/4. No doubt I could not do this with fast modern devices.
(BTW, I am just ordering boards for my new amp - hope to post up in the next few months)
(BTW, I am just ordering boards for my new amp - hope to post up in the next few months)
Hi John,
That's fine. To each his own. I was just pointing out that as designers we each have to make our own choices about which dragons we will fight. Its never a simple black-and white thing. That's one of the things that makes this hobby so great.
Cheers,
Bob
Good point, Bonsai. Gate Zobels on my MOSFET power amplifier helped a great deal in enabling the use of smaller gate stopper resistors.
Cheers,
Bob
These are good points, John. I might add that I have seen more use of base stoppers in output stages using paralleled output devices. I suspect that there is also an oscillation mechanism that exists among the paralleled output devices. A long time ago Ed Oxner wrote an article on the need for gate stoppers when putting power MOSFETs in parallel.
Cheers,
Bob
Hi Charles,
Power MOSFETs in source follower mode suffer from other effects as well that can cause them to oscillate. I describe these mechanisms in my original MOSFET power amplifier article, and I think I touched on it in my book as well. Many of these circuits can be reconfigured to look like a Colpits or Hartley oscillator if there is a little bit of inductance present just about anywhere. There are also multiple sources of poles in power MOSFETs and we need to recognize the contributions of both Cgs and Cgd, along with gate resistance internal to the device structure. It is also important that we not neglect the contribution of Cds in a MOSFET.
Cheers,
Bob
Bonsai. 2uF cap in parallel with 4 or 8 ohm really does not say much. It may put current stress on output devices, but does not reveal oscillations. 2uF tends to create RLC ciruit with lead (wire) inductance, nothing more.
To investigate oscillations due to capacitive load, you better try values between 10nF and 200nF.
Regards,
To investigate oscillations due to capacitive load, you better try values between 10nF and 200nF.
Regards,
Jan, I am glad that you could talk to Ron Quan. As you might know, he worked for me on some professional projects and the Symmetry Crossover, back in the late 1970's when he was still an undergraduate engineering student at UCB, years after I took classes there. He showed a high discernment of audio problems, early on. In some ways, he later paralleled my path, such as working at Ampex, but he went mostly into video, yet he kept his eyes and ears open about audio developments. Ron has been VERY helpful to me, first to restore much of what I lost in the fire, but now, Ron has pursued many of the questions that we discussed so many years ago. I think that you will find him very methodical and correct in his research.
- Status
- Not open for further replies.