Simulation Analysis of several unique Allison-based output stages.

[keantoken]

Thank you Paul, I was hoping for a post from you.

Originally posted by mfc
Looks like the sims are missing 2SD669_CHRISTER and 2SB649_CHRISTER

These are Christer's models for said devices. I renamed them to give him credit and so I know what models I'm using. Sorry, I forgot to include them.

Attached is the simulation file with the models included in the netlist. Currently, it is set up to simulate with 10k source impedance, so to remove the source impedance just set parameter Rser (the directive line I added) to zero.

TOINO:
So, I'm guessing that the voltage sensing is meant to detect and cancel the effects of the base current of the output transistors in order to linearize the output stage?

- keantoken

[PaulBysouth]

Keantoken,

I like your `Allison-Keane' variation to the standard Allison circuit, which overcomes the need to use large emitter resistors. However quite a bit of LTspice simulation leads be to believe that your design, and a very similar variation that I came up with, will both suffer from two problems:

- Temperature changes in the Allison bias transistors effect the output bias currents considerably and this is made worse by the small emitter resistors of the output transistors. Your design would probably benefit in using one of the THAT packages that contain 2 NPN and 2 PNP transistors. My variation was even worse.

- The transition from Class A to Class B is very abrupt, and hence will almost certainly produce high order harmonics.

Inspired by your `Allison-Keane' design I tried to come up with a design that would allow small emitter resistors but give a gradual transition into Class B.

The attached design uses large resistors (R3, R4) to set the bias current, and then bypasses these with diodes to allow high currents without the large voltage drop across R3, R4. The diodes cause a very gradual transition into Class B, with gradual turn off, but not `Non Switching Class B style (but almost). Note that Q3,Q4 will need to be thermally bonded to D1,D2.

So far this design appears to be promising, and I hope to build and test it some time next month (my day job is getting in the way at the moment). I fear that I may have swapped one set of problems for another, namely diode switching harmonics destroying the THD figures.

I have some questions that maybe someone can answer ...

- Has anyone tried putting diodes across emitter resistors like this before, and if so, do they generate unacceptable switching harmonics?
- Is a MUR820 the best choice for the diode in this application?

Paul Bysouth, March 2009.

ps - Sorry about the delay in posting this, however my Internet router power supply decided to break just before I pressed the submit key, and as it was midnight I went to bed. I've spent this morning fixing it. I hope my cut and paste above works.
pb.

[h_a]

Hi Paul,

I like your ClassA-variant (1st post)! I hope somebody finds the time to build it and report back

Have fun, Hannes

[keantoken]

Originally posted by PaulBysouth
Keantoken,

I like your `Allison-Keane' variation to the standard Allison circuit, which overcomes the need to use large emitter resistors. However quite a bit of LTspice simulation leads be to believe that your design, and a very similar variation that I came up with, will both suffer from two problems:

- Temperature changes in the Allison bias transistors effect the output bias currents considerably and this is made worse by the small emitter resistors of the output transistors. Your design would probably benefit in using one of the THAT packages that contain 2 NPN and 2 PNP transistors. My variation was even worse.

- The transition from Class A to Class B is very abrupt, and hence will almost certainly produce high order harmonics.

Inspired by your `Allison-Keane' design I tried to come up with a design that would allow small emitter resistors but give a gradual transition into Class B.

The attached design uses large resistors (R3, R4) to set the bias current, and then bypasses these with diodes to allow high currents without the large voltage drop across R3, R4. The diodes cause a very gradual transition into Class B, with gradual turn off, but not `Non Switching Class B style (but almost). Note that Q3,Q4 will need to be thermally bonded to D1,D2.

So far this design appears to be promising, and I hope to build and test it some time next month (my day job is getting in the way at the moment). I fear that I may have swapped one set of problems for another, namely diode switching harmonics destroying the THD figures.

I have some questions that maybe someone can answer ...

- Has anyone tried putting diodes across emitter resistors like this before, and if so, do they generate unacceptable switching harmonics?
- Is a MUR820 the best choice for the diode in this application?

Paul Bysouth, March 2009.

ps - Sorry about the delay in posting this, however my Internet router power supply decided to break just before I pressed the submit key, and as it was midnight I went to bed. I've spent this morning fixing it. I hope my cut and paste above works.
pb.

Can you attach the simulation file and any models you used that may be hard to find?

All that I know about the diodes is that if they have high Cj, they may cause stability problems in the Allison. Also, Distortion may benefit if you decrease R9 and R10 and increase C9 and C11, as these resistors cause a miniscule (probably insignificant in most cases) amount of distortion. Also, with some carefully placed caps you may be able to quiet any switching harmonics.

What about these THAT packages? I Googled and found nothing.

About temperature stability, the bias wanders about 47mA per 10 degrees Celsius, which is 4.7mA per degree Celsius (according to simulation). About the only thing this affects is clipping point, which won't matter much if it is being operated within its limits.

There wouldn't be any bias wander if the Allison was thermally insulated. I'll bet this could be done without a fan but I have no experience here.

- keantoken

[jwb]

Originally posted by h_a
THD at 20kHz has also only academic interest since humans can't even hear the first higher order frequency (40 kHz).

The ultrasonic distortion products can intermodulate with audible fundamentals to produce IMD in the audible band. So I believe you need to measure distortion products at high fundamentals, or measure IMD. Either one provides valuable information and is certainly not of academic interest only.

[atiq19]

keantoken,

The 'THAT Series' transistors are monolithic array of pnp/npn transistors. These are new in the market, catered for tightly matched applications.

http://www.thatcorp.com/300-series_M...Array_ICs.html

-Atiq

[Lumba Ogir]

Hannes,

THD at 20kHz has also only academic interest since humans can't even hear the first higher order frequency (40 kHz).

That old-fashioned interpretation needs a complete revision. All harmonics contribute to the tonal balance effecting the audible area. The scanty industry standard 20kHz bandwidth does not even cover the range of the human voice.

jwb,

The ultrasonic distortion products can intermodulate with audible fundamentals to produce IMD in the audible band. So I believe you need to measure distortion products at high fundamentals, or measure IMD. Either one provides valuable information and is certainly not of academic interest only.

Of course.

[Lumba Ogir]

keantoken,
I`m a little bit disappointed. Time rolls by but nothing convincing so far. I`ve been hoping that you would come up with something really great. Take all the necessary precautions in order to limit the damage and losses this experiment will inevitably bring about and get going. The guru says, achievement is born of sacrifice (manifesting infinite wisdom).

[keantoken]

Originally posted by Lumba Ogir
keantoken,
I`m a little bit disappointed. Time rolls by but nothing convincing so far. I`ve been hoping that you would come up with something really great. Take all the necessary precautions in order to limit the damage and losses this experiment will inevitably bring about and get going. The guru says, achievement is born of sacrifice (manifesting infinite wisdom).

Of course! Infinite wisdom, my friend. I think we must all be patient. I'm a fast learner but I'm 15 years old, I've only been at electronics seriously for 4 years, and even then mostly on the simulator (which I think has actually been quite an advantage save for a few quite outstanding issues).

If no one so far has noticed, I'm pretty much on a quest to get as low distortion as possible but also to have a nice stereo amp as a souvenir. Class A should probably make this easier, and I believe this output stage is a good start. I do worry about what effects interference might have as a result of the high input impedance of my enhanced version. I will have to start point-to point soldering things together as currents get too high for a breadboard.

My immediate struggle has to do with the Cdom cap. It works fairly well but tends to destroy 20KHz THD if you use one large enough to get the recommended -100 degree at 0db OLG phase plot.

My aim at the end is to get this together with a Rush cascode at the input, to give some nice 2nd harmonics in places where distortion is unavoidable. In simulation I've noted that the Rush Cascode also tends to be more stable than the common LTP, which in the end means that we can have higher OLG and lower THD. What THD there will be will mainly be 2nd harmonic, supposedly pleasant sounding. Unfortunately, this is theoretical (the part about me getting it to work). So I cannot make guarantees.

If you are interested, this is what I am currently playing with in the simulator. It simulates unimaginably well at full output power if you can just get it stable (EDIT: the 100n cap across the speaker is overkill, maybe I can be safe without it)! This is also an excellent example of how a large Cdom can magnify LTP nonlinearity at high frequencies. I have tried a zillion different ways to get better stability without loading the LTP down at high frequencies. The only time I have had success with this, is when I was simulating with the Rush cascode. But I've done enough greasework to experience the difference between simulation and reality. In general, my circuits were more stable in real life than in the simulator. If I want to use the simulator exclusively with real life, I will have to ensure that they are comparable.

I will next try simulating with the Rush cascode instead of the normal LTP and see if I have more success. But until I get a decent power supply I can't start experimenting with the MJL devices.

- keantoken

[Lumba Ogir]

keantoken,
I`m just teasing you. Maybe I`ll make a comment later on, although seemingly, you don`t take advices and don`t follow leaders. A strategy of independence.