DIY Audio Projects Forum

Well if the op amp(gain stage) is designed well then it should be tollerant to capacitive loading from the output stage internal capcitances. Poorly designed op amp then I'd agree that it may be a plorbem with highly capcitive output stage.

I'm using half fake Chinese tip41c as my head amps output stage(takes only 32W max), also tried tip3055 2955 fake transitors which both case I get same frequency response and slew rate.
My amps gain stage not excatly tuned for high frequency but with 90 degree phase margin slew rate is arround 15v/us.

If I was to convert to power amp All I would be doing is resetting resistor values to suit higher voltages or use parallel transitors for my op amp stage to stand higher voltages without reducing bias current.

What do you mean by opamp gain stage? You use IC opamp or you mean type opamp design in discrete form.

No, power electronics is nothing at all this simple. I thought is simple too before. It is not simple like you described by just scaling up. It described in Cordell's book about input capacitance of transistor proportion to current. When power goes up, capacitor loading the front end change drastically. You get into conditional stability problem. Amp don't oscillate that easy in small signal, only in large signal. You see oscillation riding on part of the sine wave!!!

Power electronics is really a separate topic. I used to think how hard can it be in audio frequency after years of designing RF, ultra low noise high speed analog, analog IC design and all other analog circuits. I was surprised by how power electronics different. That's where simulation models fail big. They cannot account the change of characteristics accurately.

Look up the thermal derating of TO220, when temperature is up, you don't get 20W or anything close. That's the reason people use many pairs of big transistors that rated 250W each and they still burn transistors if not careful.

I actually thought of using many pairs of TO220 as output stage, it did not pan out at all. The reason I thought about it is I want fast transistors of over 100MHz to get better stability. Long story short, I need like over 20 pairs or more. Can you imagine the work to drill holes and mount the transistors with isolation pads. Let.s just say, I spent the last night mounting my 9 pairs on the pcb and heatsink. Just one amp, hours. I spent one night drilling and tapping hole on the heatsink. Imagine triple the amount of transistors.

TIP42 is too old, too slow. It's not easy to find good transistor. I use 2SC4883 and 2SA1930 TO220 for pre-drivers, over 100MHz, high beta 150V. They are all discontinued, I bulb buy 100 each as I use 2 pairs only for each amp. The faster the transistors, the easier to tame the stability.

It is not that simple, that's why I take on power amp as a challenge.

You have the book by Bob Cordell? It's a very good book. I talked to him a lot on the forum before. We both were analog IC design engineer before ( power amp is like a big discrete opamp inside). We both have a lot of insight in this subject. He talked a lot about these subject in the book.

I go on a separate post as it gets too long.

We only talked about the theoretical amp design so far. There is the real world. When you run like 10A signal, a wire is not a short circuit, the grounding and power distribution becomes a big problem. The ground becomes a design by itself, so is the power distribution. You need to think like the current, where does in go. Designing the power supply and distribution is a separate subject. I have to say, there are still things I cannot explain in the power amp. I thought I've seen enough electronics. But when you are pushing to 0.001%THD, it's like RF, nV level of noise, High voltage etc. There comes a point of time it's hard to explain.

Imagine the simple bridge rectifier gave me problem, the turn off is too slow and it affect THD reading. I have to resort to pay $25 each of the bridge rectifier to lower the THD and make the wires less sensitive to where it rest. I can explain it now, but that alone took me like 2 or 3 days. It's like voodoo that I can change the THD reading by waving my hands on top.

When you get down to 0.001%THD, everything matters. To get 0.1dB flatness to 20KHz, you need an amp that response way into hundreds of KHz. To prevent large signal distortion, you need like slew rate of 30V/uS.

I meant op amp type design in discrete form
I have not looked at the book by bob cordell.

Further more all I had talked about is what happens in my real life testing, as said before I do not belive in circuit simulators nor use them.

I had emphasied head amp able to drive 8 r headphones to be same as power amp, becase mainly what made most of my head amps have cross over like distortion was from a large load like 8 or 4 r.
I think this is because of the current drawn out of the op amp when voltage swings it reduces gain and causes it to mailfunction.

Fruther more when there are multiple output devices normally the predriver handles the capcitance at ease.
This is becasue its low output impedance and has high current sink source capaciblity, this allows like 10 pairs internal capacitance to charge up quickly, therefore 10 and 1 pair should hair fairly simliar response.

I've not tried more than 3 pairs yet on scope, but I would be making a class a amp with 8 pairs of output when I have time I already have the case.

I like everything Japanese but not the transistors
I normally not chose high frequency and low capacitance capacitors because japanese transitors 2sa 2sc 2sb 2sd has poor sound quality compared to just a geniune tip3055, pre drivers are also worse than a tip41c.
I like mj14003 14002 and tip36c and 35c +3055. I like mpsa06 or mpsa42 for small dudes.
MJ14003 has arround a BE BC voltage at 0.45v measured on meter BE voltage is also like 0.02 0.03 volts higher than BC voltage, this kind of high power transitor really had improved my lisetning expreience.

Small signals to-92 I find fairly simliar to non japanese ones however I still prerfer to stay away.
Just every time I test japansese transistors they give me poor sound quality. Its hard and mid range sounds distorted like mp3.
I've got like 10pairs or so of diffrent japanse transitors from geniune broken down yamaha and luxman machine, amplifier board was fortunate enough to not be dead. Just sound quality not up to standards.

Normally cross over like distortion and higher order harmonics are at higher frequencies, its not auditable theoriotically. I gave one of my audiphilie mates a amp with oscillations, he couldn't tell the diffrence.
From contiuous and precise listening though, high frequencie distorions are actally slightly heard, the oscillations or cross over like distortion are at high freq, the high frequency noise some how is able to get distortion into the 20-20khz range.

I do have some of douglous self amp books, only found it slighlty usefull when I was still a noob. I don't find it usefull anymore though because I am getting into pro range.
I think the main plorbem with self is that he trust and uses the circuit simulators too much. Where my real build has different from sim a few times or even not work I stayed away from sims.
Virtually everything is talked with reference to circuit simulators, it needs more parcticall and more verbal therotical explnantions on whats going on.

I personally don't think transistors make any difference to the sound. I choose transistor for stability. I don't like book by Doug Self at all.

I was on DIYaudio, I was talking with Cordell, Self joined in and got into big argument between the two. Cordell and I both design analog bi-polar IC, we see eye to eye in a lot of things. Power amp is just a big discrete IC. I don't agree with Cordell 100%, they are all human, nobody is perfect. I just look at his explanation and think about it. I did verify a lot of his writing with the IC design book by Grey and Mayer, and from my own experience of designing opamp, I agree a lot with Cordell. I think Self is a little full of himself.

I think for small signal like preamps, simulation should be quite accurate as devices are working in linear mode. LTspice is quite good and is free. Just download from Linear Technology site. It just not accurate with large signal that device becomes non linear.

Ha ha, when you add more drivers and predrivers, that's when trouble starts.

When you said you use 20 transistors with the tube for preamp, are you trying to shape the sound, frequency response. That's the reason that got me curious. Because normally, all the circuit I saw in preamp are just straight pass, no tone circuit, no shaping the response. I just don't see how you need all the transistors to do a straight pass.

I found all the tone control, presence, loudness control contaminate the sound. I always put them on bypass. total passive preamp with just a volume pot gives the most fidelity. I am at the process of finding room in the power amp to put in the motorize volume and remote control to put the volume pot close to the power amp so the short coax won't load down the volume pot to get the best frequency response.

Yes, crossover harmonics are higher order, but remember it down mix and get right back into the audio band.

For my pre amp zero eq control, just direct input and pot on pre amp output.
20or so transitors are to get high gain charteristics and get lots of feedback so distortion=0, this allows me to test weither or not valve sound is distortion to the highest accuracy
Conclusion valve sound is not distortion but a phenomina when electornics flow through a air gap

I personally not like eq either, thats why all my stuff is zero eq.
I feel like eq leads to distortion

This is the first amp I designed and built. It is Blameless design ( single LTP input stage). Here are the 3 pages schematics.

First page is the OPS section with SS output relay for speaker protection. I use 9 pairs of output transistors, total bias current is 0.9A ( 100mA per pair) to give the first 13W of pure class A into 8ohm. I have 0.1uF and 10uF right at the collector of all the emitter follower transistors. Together with ground and power plane, this minimize the parasitic for better stability. I use triple EF output stage to increase open loop gain ( avoid loading the VAS stage).


Page 2 is power up turn on delay and output DC detection for speaker protection


Page 3 is the IPAS section. IPS is Blameless design. I have the page number for the Power amp book by Bob Cordell labeled in different section. They are described into the book on reasoning of using the circuit. The design is very standard, the only difference here is I put ALL the bells and whistles in. I have never seen any amp on the market that put everything in. Even the high end amps like Krell and Mark Levinson leave out a lot of the features. Only thing I can think of is the more complete it is, the higher the open loop gain and the harder it is to tame and make it stable. I worked on a lot of closed loop feedback system, so I put in the time to tame the amp. I found LTSpice simulation is not accurate at all. Simulation might show a lot of phase margin, but the real amp oscillate and I have to tame it by feel. The difficulty is the transistor model is not accurate for large signal. Took me over a week to tame the amp.


This is picture of the filter caps. I have about 2.5F of capacitors using a lot of smaller 10,000uF capacitors to lower the ESL and ESR. I have 100uF, 10uF and 0.1uF in parallel to cover full range of frequencies.




BW of about 330KHz, slew rate of about 30V/uS. THD of 0.001%@1KHz, progressively increase to 0.0035%@20KHz at 46Vpp or 66W@4ohm or 33W@8ohm. Because I use 25V-25V power transformer, the rails are +/-34V only. This is a low power amp of only about 55W@8ohm. But it can drive low impedance. 110W@4ohm and 220W@2ohm.

I have to put this in the second post as I can only upload 6 files max.

This is the finish build



This is the TOP layer of the main pcb, I have complete ground plane for OPS section and power planes for the IPS/VAS section. Care was taken to make sure the planes are not accidentally cut up by through holes and vias.


This is the bottom layer, I have power planes for the rail voltages. The output has it's own plane instead of a trace.

The above amp was designed and built over a year ago, I am currently working on another amp and I improve the slew rate to 45V/uS large signal. But I run into a strange problem. One channel have very good result, 0.004%THD @ 20KHz, but the other one is 0.011%. For the life of me, I cannot figure this out. Both module were tested and characterized on the test jig, but putting in the amp change one channel. Like the grounding is different, the wire placement of the input signal and the power feed of the front end has to be different from the other amps. The good channel behave exactly the same as all the other amps.,

It feels like a grounding problem, but I have no idea why as one channel works perfect. The grounding design is proven in the other two amps to be solid. It's a cookbook at this point. The pcb is very much the same as the first amp above that you can see the ground plane( blue) is one very solid piece with no cut up. It's 2oz copper, very beefy. AND it was tested and passed on the bench. Only other thing I can do is take it out of the amp and run it on the bench again. I have a suspicion it's going to test out good. I measure all the voltages inside the amp and are all good already.

Strange. There are things that I still discover and learn all the time. On the spectrum analyzer ( on the laptop), things are very sensitive, you literally breath on it and things change, Still have no explanation for all this. Put it in the system, sounds fine ( not a real test, just listen to it at lower level, big boss is sleeping). I am stumped.

Sorry about the ranting, I just run out of idea.