Mismatched drivers in output stage - is this a problem?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am currently building an amp using a quasi complementary output stage (with Baxendall diode). I have a nicely matched set of output devices (MJ15003), however the drivers are MJE15030 and MJE15031 and have very different current gains - 70 and 200 respectively.

Is this likely to degrade the sound of the amp?

The schematic can be found here:

http://www.users.zetnet.co.uk/avondale/aaModnPcbsPowAmps5.htm
 
David,

I would not have thought so, because the two output circuits are quite different anyway as one is common collector and the other common emitter. These quasi circuits always introduce some asymmetry in the output signal which manifests as H2 but disparity between drivers should not worsen this.

The amps look very interesting. Let me know how they compare to the AKSA!

Cheers,

Hugh
 
Thanks Hugh. After measuring the MJEs, I had a read through Doug Self's papers - he stated that the quasi + Baxendall scheme can give almost as low distortion as fully complementary. However Self does not go into any detail on gain selection to optimise distortion for each o/p stage, he also uses spice simulation to generate his linearity plots thereby completely ignoring the real world issue of component tolerances.

The Avondale circuit is basically the same as that used by Naim with a few tweaks. Naim reputedly select parts very carefully and I was wondering which parts and to what criteria. Maybe they encourage 2nd H distortion by deliberately mismatching the halves??


OT - My family spent 3 weeks in Oz last winter (this summer for us!) visting my brother-in-law and family in Sydney. We went to Heron Island and spend a few days at Noosa as well. We also visited the Blue Mountains where we had an inch or 2 of snow!!
There was a bizarre incident where my kids were feeding a wild kangaroo a banana (indigenous Aussie food!!??) in the snow. A local couple drove up and were amazed by the snow, they had never seen real snow before. They were raving about how it crumpled underfoot and how cold it was - we exclaimed "Yeah, but look at that kangaroo, it's so tame etc etc". They glanced at the roo, shrugged with complete disdain, and got back in the car because it was to cold for them.
I guess it's all down to your POV.

Oz is a truly great country though, think we made a mistake shipping you chaps over there!

Cheers,
Dave
 
David,

Consider the output stage configuration. The upper stage is comprised of two cascaded emitter followers, driven at close to unity gain directly from the VAS. The lower stage, however, is a CFP with diode/resistor degenerated driver, which ensures the collector of the lower NPN always tracks the driver emitter. This pair too is very close to unity gain.

However, when you look inside the CFP, you see 100% local feedback applied from output through the driver emitter, and the combined betas of both devices - since there are two 180 degree phase shifts - creating a huge OLG with corresponding very low Zout and distortion.

Effectively the negative (lower) output block is a closed feedback loop; nested within the global feedback of the amp. Therefore, since the CLG of a CFP is unity, just like the upper block, the relative betas between the upper and lower drivers are not relevant. In the upper block, the driver is an emitter follower; in the lower, it's a common emitter.

I interpret this to mean that their betas need not be matched. In fact, the output impedance of the lower section would be somewhere around ten times lower intrinsically than the upper section because of the huge CFP feedback factor; but since the entire amp is wrapped in global feedback, it's not too important anyway.

It's worth noting that an asymmetrical Zout from each half of the output stage creates H2 in the interaction with real loads, and this, along with production of intermodulation products, has always been the main criticism of this topology. But to my ears they sound pretty good, so I suspect the criticism is, well, a little academic........:clown:

What is important is the stability of the CFP. At switch on and switch off, this configuration is generally not very stable and liable to throw up a spray of artefacts. To dampen this problem, JLH added a small 10nF cap across the driver's emitter diode to simulate the capacitance of the base/emitter diode of the output device, and I suspect this is a very good move as it will soften the transitions.

OT: Kangaroos in Oz are a damn nuisance, as on the roads they can easily kill the occupants of a car as they fly through the windshield if struck on the hop. Now snow is a big deal, we so rarely see it, particularly in NSW, hence the reaction of your locals.

I often muse on Britain sending all its thieves and vagabonds to Australia. This left an indelible mark on Australia, giving us a love of equality and a close association with the underdog. It also made us somewhat anti-intellectual, sport loving, and outgoing. However, since we were so badly used by the Brits in the First War and in the atomic tests of the fifties we have become very suspicious of the old country. When we were declared aliens in Britain in 1968 that finished it for most of us. We are strategically important to the US, not to Britain, so our associations are rather faint now even though we still have the Queen as titular head. There is quite a bit of strong feeling about that, too!

You must also factor in that a great many Brits came to Australia as free settlers, including my family (from Lincolnshire, 1843), and since about 1900 this has been an immigrant destination from more peoples than you could name. 60% of the children at my daughter's school are from Asian and sub-continent background!

Cheers,

Hugh
 
Thanks for the explanation of the CFP + darlington structure of the quasi complementary o/p stage - it all makes sense when you look at it like that.
I will try the JLH 10nF trick - very crafty indeed!

OT: Interesting comments about Oz and I wholy agree.

Regards,

Dave
 
Hi Dave,

The Circuit is a Naim Clone....I am a huge Fan of Quasi Complementary Configuration...Because in this you only have to search for NPN types for output..indeed which are of low cost and are also available with various specs. to suite your needs...

regards,
K a nw a r
 
AKSA said:
...I interpret this to mean that their betas need not be matched. In fact, the output impedance of the lower section would be somewhere around ten times lower intrinsically than the upper section because of the huge CFP feedback factor; but since the entire amp is wrapped in global feedback, it's not too important anyway...
It may not have a significant effect on the distortion contributed by the output stage itself, but what about the asymmetrical loading on the Vas? This is a large factor in total distortion in the best of cases, but does a mismatched beta increase it significantly or is it swamped by the generally nonlinear input impedance of each half individually?
 
Mr E,

You are right; in fact there is always asymmetrical loading on the VAS, particularly at the crossover disjunction. Loading from the top of the positive half cycle down is successively high, low, almost zero, low and highish. Swamping the VAS load with a resistor to ground, typically 47K, is a good idea; I notice Michael Bittner did this in his Symasym and I would expect it to be more load tolerant as well since the reactive reflection in the load is less influential in phase shift. However, the downside is that it does cost OLG, reducing feedback factor.

Generally, the Sziklai (CFP) has a greater input impedance than the Darlington, but the asymmetry would likely be greatest as the upper and lower halves hand over. You would expect that the NFB would ameliorate these non-linearities, of course.

But we are wading in murky waters; this stuff is subtle and the psychoacoustic effects of different circuit topologies are not well understood, particularly with all the emphasis on unidimensional parameters like THD.

Cheers,

Hugh
 
Hi, AKSA,

How are you?:D
To dampen this problem, JLH added a small 10nF cap across the driver's emitter diode to simulate the capacitance of the base/emitter diode of the output device, and I suspect this is a very good move as it will soften the transitions.
Is it 2 x 10nf's, one in B-E of MJE15030 and one in B-E of MJE 15031?

I remember that the minimum RC/RE resistance is about 3x to make stable CFP, but here it uses 100/100 or gain=1x. Is putting IN4148 // to 100R (in lower CFP driver) enables us to use 1x gain CFP with good stability? What is the main purpose of this IN4148 there?

If I'm not mistaken, I think I've seen a quasi complementary design that is not using 100//IN4148 in the lower section. The emitors of the driver connected directly to collectors of output stages directly, and the collectors of the driver has 100R to rail (like this one). This means very big gain on the cfp section. Is this OK too?

Sorry, so many questions, I wanted to learn :D
 
Hi David,

Baik2 saja, baigamana saudara?

I just wrote a long response, and somehow deleted it. Very frustrating.

Caps across the BE junction prevent rapid change of the bias voltage which prevents oscillation without much affecting speed. Yes, it can be done here, but I'd prefer a base stopper.

The cap across the baxandall diode simulates the large BE junction capacitance of an output device, so gives the VAS electrically similar upper and lower blocks. From a drive POV, this is important, as Mr Evil pointed out.

Don't use a light duty signal diode in this role; UF4002 would be far more suitable.

If instability persists, you can reduce the internal gain of the CFP by adding to re with a small resistor on the emitter, typically 10-22R.

CFPs with gain are usually more stable, but their Zout is not as favourable as the emitter follower on the positive rail, though with global feedback this may not be too important. They are easier to bias, however.

Congratulations on your website, and your marvellous auto amplifier! It seems you don't need to ask too many questions!

Ciao,

Hugh
 
Hi, AKSA,

Thanks for the advice.

I just wrote a long response, and somehow deleted it. Very frustrating.
Yes, you are right. Happens to me several times also :D

It seems you don't need to ask too many questions!
No, you are wrong. :D A few days ago, I got a chance to meet Manuel Huber of FM Acoustics. One thing I learn from him, that nothing beats the "flying time" on anything, including making good audio amps. Textbook and theories really help, but that's only small portion of what is really needed (in making good amps). The real information comes from people like you, NP, John Curl, etc.
 
It lives!

Quiscent current stability with temperature is non existent - as is predictable from the layout/schematic.

Square wave response into various loads looks fine - there is a bit of ringing into 2u2 as per normal.

I will report back when I have had chance to listen to it (kids watching DVDs at the moment - aaargh!).
 
Dave,

Quiescent adjustment, temperature stability and switching artefacts are the three reasons I don't like the CFP in Class AB. It's a very different matter in Class A, however.

Adjusting bias and designing a suitable Vbe to accurately track OP device tempco is very difficult.

The switching instability colours the sound, and the lower CFP tends to be worse.

This quasi-comp design is only half a CFP so likely is a bit better.

Siemens did a very good scalable 50-400W quasi-comp back in the eighties which drove the VAS from the input LTP collector and a phase inverter from the feedback LTP collector. Since the drive for each output block was derived from the LTP with much closer signal delays on both sides, this amp had very good HF response and complete freedom from cross conduction, a problem on many quasi-comp designs. For the life of me I can't find a schemat but I'll have a look around.

Cheers,

Hugh
 
David,

You really should let these things settle for a few days. I'd suggest living with it for a week, walking into another room to see if the image holds, listening in the background on vocals, violins and piano, and generally introducing the amp to its new environment. You have to get used to each other.

The caps definitely take time to settle down. I no longer have any doubts about that.

Does the amp have realism on human voice, and sweetness on violin? What's the image like?

Cheers,

Hugh
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.