Has anyone seen this front-end before?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm guessing this means a heavily buffered and/or error corrected output stage? Better stability and speed?

Hi Keane,

Nope. I was referring to the front end, which has three distinct features:
1. Low subtractor error of the IPS.
2. Well defined standing current of the VASes (or TISes)
3. No 'fighting VAS' issues (they can't fight, as there's only one Cdom)
Please, see also post 116 and 120.

Cheers,
E.
 
a masterpiece of poetry


Thanks for the links Steven.
Of course, I know of this circuit, though under a different name. I thought, erroneously!, it was named after Borbely. This circuit differs in many ways from the super TIS. The NPN/PNP combination you mentioned for example, serves a different purpose. Digging further into the links, I found this schematic of the Lender VAS. :rolleyes:
Next, I decided to have a closer look at the origin of this incorrect implementation, and stumbled on a folded cascode configuration (see below). This circuit comes more closely to the super TIS. However, the author dismissed the idea:

'Firstly, there would appear to be a lack of overall open-loop gain because the common-base stages Q4, Q5 do not give any current gain.'
It's precisely this 'lack' of gain that makes the super TIS stable! (at DC and HF)

'Secondly, there is no obvious way to apply the Miller dominant-pole compensation that is so very useful in linearizing a VAS.'
Oh really? Apparently, the author has never heard of an 'input-inclusive-compensation '. :tongue:
Perhaps he is too much indoctrinated by his own 'blameless' ideas.

Then I dug a bit further and found a highly amusing discussion about IPS subtraction errors between Fred Dieckmann and another guru whose "semiconductor physics background is beyond reproach" (his words, not mine), which culminated in a masterpiece of poetry.

Once again, thank you so much for the links!

Cheers,
E.
 

Attachments

  • Folded cascode.png
    Folded cascode.png
    24.8 KB · Views: 366
Disabled Account
Joined 2006
Hi Edmond,

I know Yamaha uses them in a completely different (traditional) way. It was just because Baxandall pairs are almost never used in commercial designs and some discussion arose about stability issues, that I wanted to show these applications of the Baxandall pair.
BTW Already in 2004 these stability issues were discussed on this forum, both as TIS (VAS) and as cascode stage: http://www.diyaudio.com/forums/solid-state/25172-baxandall-super-pair.html#post293504. But going through that thread I just saw that also Yamaha's use of it was already mentioned.

Cheers, Steven

The biggest problem here is the simulators, these always show baxandalls as oscillating, but in real life they can be stabilized. When you look at for example the thread you mention youll see every comment there is based on simulators, why didnt anyone actually go and build it. Does anyone really believe yamaha sell amps that are oscillating ?? I too have used traditional baxandalls with success in some circuits, theres another member here providing diy power amp schematics with baxandalls for builds, and none of his builders have mentioned any problems as yet, from what I gather 87 builds so far. Yamaha are still using them and many more recent Analog Devices opamps are full of them, look at AD quadcore technology.
 
oscillations

Hi Alex,

My sims don't show any oscillations. At only one occasion -under very unrealistic conditions- I observed peaking at 20MHz. That was with an output completely left open, i.e. without any load. However, as soon as I put a small cap (22pF or so) at the output, the peaking was gone.
Nevertheless, I've blocked the pos. FB of the base currents (at UHF) by means of 470pF caps, just to be on the safe side; better safe than sorry. Most likely one may omit these caps (C10 & C11).

Cheers,
E.

PS: Did you read the poem?
 
Last edited:
Disabled Account
Joined 2006
Hi Alex,

My sims don't show any oscillations. At only one occasion -under very unrealistic conditions- I observed peaking at 20MHz. That was with an output completely left open, i.e. without any load. However, as soon as I put a small cap (22pF or so) at the output, the peaking was gone.
Nevertheless, I've blocked the pos. FB of the base currents (at UHF) by means of 470pF caps, just to be on the safe side; better safe than sorry. Most likely one may omit these caps (C10 & C11).

Cheers,
E.

PS: Did you read the poem?

Edmond the easiest way I found to keep from trouble was having the second transistor having much lower beta, I see you have it as well, I dont think youll have problems either although your design is a little different, I was refering to using the baxandall in traditional way.

Btw, I tried a sim of your design with just a triple outputstage, no problems with stability and 4 ppm distortion at THD20 40v p-p, some members find this dissapointing ?? :confused:
 
The biggest problem here is the simulators, these always show baxandalls as oscillating, but in real life they can be stabilized. When you look at for example the thread you mention youll see every comment there is based on simulators, why didnt anyone actually go and build it. Does anyone really believe yamaha sell amps that are oscillating ?? I too have used traditional baxandalls with success in some circuits, theres another member here providing diy power amp schematics with baxandalls for builds, and none of his builders have mentioned any problems as yet, from what I gather 87 builds so far. Yamaha are still using them and many more recent Analog Devices opamps are full of them, look at AD quadcore technology.

Hi homemodder,

But if they can be stabilized in real life, then that same stabilization that is used in real life should be applied in the simulation. Then it had better be stable. I would never ever build a circuit that simulated in any reasonable simulation. I do know of cases where designers depend on real-life parasitics that may not be modeled in the simulation to have a stable circuit (either knowingly or not), but this is an unreliable practice.

We should not generally depend on real-world "imperfections" for a circuit to behave. Another example is building a circuit that is not happy if transistor beta is too high; just not a good idea.

At the same time, we must of course be cautious with simulators. For example, unless you go out of your way, all of the same type of transistor in your simulator will be perfectly matched. Some circuits might like this and work fine in simulation, but work poorly in simulation.

Finally, there are some circuits that legitimately will not work in a simulator unless coaxed to do so, and which depend on real-life "imperfactions". Multivibrators and some other classes of oscillators come to mind. In a simulator, you can stand a pencil on its end indefinitely, but you usually cannot do that in the real world.

Cheers,
Bob
 
To all,

Regarding the circuit as shown below, I like to know whether anyone has seen this front-end before. If so, I'm curious to learn more about it and how it behaves in real life. More details can be found here.
(BTW, I'm still editing and updating that page, so it's not yet finished)

Cheers,
E.
Not due to you Edmond old chap...
This front end has been done before by Shinichi Kamijo here:

Evolve Power Amplifiers
 
Last edited:

Thanks Michael!
By now it's a bit late. So tomorrow I'll have a closer look to it.

Yes. The compensation described as input inclusive compensation by Self in linear audio is simply phase lead compensation with the output stage excluded.

I already expected you would call it 'lead compensation' (I still remember the discussion about the compensation of Bob's HEC amp), but what's wrong with calling it 'input-inclusive-compensation'? In fact, the compensation loop does encompass the inverting input of the IPS. So.....

PS: Given the context of post 123, I think it was more than appropriate to use D.Self's own terminology. :D
 
Last edited:
Thanks Michael!
By now it's a bit late. So tomorrow I'll have a closer look to it.



I already expected you would call it 'lead compensation' (I still remember the discussion about the compensation of Bob's HEC amp), but what's wrong with calling it 'input-inclusive-compensation'? In fact, the compensation loop does encompass the inverting input of the IPS. So.....

PS: Given the context of post 123, I think it was more than appropriate to use D.Self's own terminology. :D

Actually, I termed it Miller Input Compensation. So-called input compensation has been around forever, and has always had the property that it did not impair slew rate. I employed MIC in the early 1980s because it also does not impair slew rate and it does not reference either side of the compensation capacitor to the supply rail.

It is wrong to describe MIC as lead compensation, as Edmond has explained. MIC is fairly well-described in my book.

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