No, the two cases are identical. In both, the closed loop gain is approximately 1 - 1/A, where A is the open loop gain. For an emitter follower A might be 10^2 and largely frequency-independent; for an opamp A might be 10^6 at low frequencies but falling linearly with frequency.tedr said:
No. In both types of buffer the subtraction leaves a small signal. In servo systems this is rightly called the 'error signal' but calling it this in audio circles might give some 'audiophiles' a fit of the vapours.
When it comes to hi-fi it's like it's allways has been with every product; The better it looks,
the better t sounds foer a lor you personally and does it really matteif it's measurable or not. I've just been been testing some midrange vs cheap .I've just been testing some midrange vs normal tonearm cables. the difference is 10% at most...but a lot of gain can be made by soldering... That's why you can't say that a part is much better..; There are more variables which changes if you only win 10% replacing a part. And offcourse it can go the other way around
....
I see thing like a car you can put a turbocharger on a normal car which gains 100hp....and it's costs a lot, but with leaving breaks suspension, etc. standard...it's useless
the better t sounds foer a lor you personally and does it really matteif it's measurable or not. I've just been been testing some midrange vs cheap .I've just been testing some midrange vs normal tonearm cables. the difference is 10% at most...but a lot of gain can be made by soldering... That's why you can't say that a part is much better..; There are more variables which changes if you only win 10% replacing a part. And offcourse it can go the other way around
....I see thing like a car you can put a turbocharger on a normal car which gains 100hp....and it's costs a lot, but with leaving breaks suspension, etc. standard...it's useless
This is my last stand. This is where I call it a day.
1audio and Thorsten have provided good lists of suggestions. The main benefit from them for me is that I now see just how much further this basic design can be evolved to.
Some of their suggestions I adopted "as is", or basically but in a modified form. For example, the input stage CCS is still a 3 transistor job, but they have been completely turned about. Basically, it's a cascode now. But the most interesting part of it is the one lone FET. If I use a plain resistor to terminate the circuit, the amp has a DC offset of -52 mV. With the FET as shown, the offset is reduced to 0.7 mV.
Thorsten pointed out that the BF trannies I use could be replaced by something else. That's true enough, instead of BF 471/472, Wayne could use 2SC3503/2SA1382, for example. The last performance stand is in fact set by one of them, and they do seem better in Data Sheets. It also turned out that they are almost exactly what the BF 720/721 are, especially regardin the C factor, so I checked using them as well. Nothing drastic happened, except that the overall distortion figures dropped a bit - otherwise, the circuit stayed stable in all respects.
Excuse my tendency to throw in bigger caps than really necessary, but there seems to be something wrong with my DNA, so I throw in larger caps. I would NOT advise making them larger still because larger still will also be slower, and that is not really wanted. You hear that, Wayne?
ML means multilayer, and silver mica will do just fine too, TA means tantalum, MKT and MKS are Wima series denominations, do as I wrote and you will be doing what some good names do (e.g. Kensonic Accuphase) and what the Germans have been doing forever.
So, this is is the last variant. Over to Alex now, but I would advise Wayne to consult regardig mechanical outlay, because Wayne is the one who has to decide what it will externally look like. I can help, but I can't decide, I don't want to irritate the lady of the house, and I want even less to answer the doorbell and find a SWAT team looking for me.
Chassis already in play , I will contact you with some questions
None of my business, but transconductance doubling will be a btch (115mA quiescent + 1 ohm load )
Plus you're very likely to hit the saturation button with your monsters (low sensitivity, 1 ohm impedance over a large range, separate & stiff FE rails of higher voltage, sagging OS rails, no baker clamps or active OS+Vas monitoring)
Plus the huge spread in dissipation will make the OS temperature a bucking bull, add to that the 6 x Vbe of a triple EF. (tempco + BD139/140 have moderate Hfe)
Plus you're very likely to hit the saturation button with your monsters (low sensitivity, 1 ohm impedance over a large range, separate & stiff FE rails of higher voltage, sagging OS rails, no baker clamps or active OS+Vas monitoring)
Plus the huge spread in dissipation will make the OS temperature a bucking bull, add to that the 6 x Vbe of a triple EF. (tempco + BD139/140 have moderate Hfe)
DF96 thank you for the corrections, I see I must put my thinking cap back on!
I try not to be concerned about audiophile fantasies, but more concerned with the physics
I am trying to understand the structure of the unity gain device, the combination of the three terminal "subtractor" and the amplifier gain element. Is it correct that the output of the subtractor, that is the signal presented to the gain element, is the input signal divided by A?
I try not to be concerned about audiophile fantasies, but more concerned with the physics
I am trying to understand the structure of the unity gain device, the combination of the three terminal "subtractor" and the amplifier gain element. Is it correct that the output of the subtractor, that is the signal presented to the gain element, is the input signal divided by A?
All true, but:
1. I did my usual bias current, which is 130 mA per device. Onthe other hand, this is adjustable, so if any untoward behavior is observed, it can be reduced or even increased. Some things simply have to be tried live;
2. "Limited Hfe of BD 139/140" is a question unto itself. I like to use them as they have an excellent bandwidth (170 MHz) and tend to follow the heating up rather well. With some selection, which I can do for Wayne, samples can be foud with a Hfe of 100 or better, it just takes time to pair them,
3. The Great Unknown to me here is the temperature rating of the heat sinks. However, when you just look at them, it stands to reason that should be around 0.3 deg./W, given their mammoth area, which I see as at least 1 m2. Puts to shame the Krells and Levinsons of this world.
As for sensitivity, I don't know about that. It seems to me that if under normal listening conditions Wayne peaks ar 17 Vp-p, they are not so insensitive. Low impednce yes, but not so low sensitivity.
Either way, that's what development is, finding the right values. In any normal situation, I'd start with a bias current of 50 mA per device and slowly go up over a period of time.
Thank you for the comments, Jacco, feedback is always welcome.
sagging OS rails ...
Probably not applicable to you Wayne, but generally speaking Jacco is right.
D,
Yes , I'm aware of OS voltage sag and yes , I want , no, need Jacco and others to play devils advocate as much as necessary. As to Big centurion there is enuff to too much of everything, sagging OS rails and SOA is not an issue, use as much class-a bias as necessary , we have all that's needed..
Yes , I'm aware of OS voltage sag and yes , I want , no, need Jacco and others to play devils advocate as much as necessary. As to Big centurion
there is enuff to too much of everything, sagging OS rails and SOA is not an issue, use as much class-a bias as necessary , we have all that's needed..
Last edited:
thank you peufeu, I know the Wikipedia page, it was some help, but I needed to think some more also
Looking at the figure 1 on Wikipedia the circle is shown as an adder and a subtractor, my present understanding is different to this, I believe that the circle has two functions but they are multiplication by A (not addition of A) and subtraction because the amplifier is inverting. That arrangement satisfies the gain equation for a unity gain op-amp buffer = 1-1/A
I read the webpages you wrote about memory distortion, very interesting, but it seemed to end without a conclusion, did you take it any further?
Looking at the figure 1 on Wikipedia the circle is shown as an adder and a subtractor, my present understanding is different to this, I believe that the circle has two functions but they are multiplication by A (not addition of A) and subtraction because the amplifier is inverting. That arrangement satisfies the gain equation for a unity gain op-amp buffer = 1-1/A
I read the webpages you wrote about memory distortion, very interesting, but it seemed to end without a conclusion, did you take it any further?
This thread needs more pictures
An externally hosted image should be here but it was not working when we last tested it.
The circle is a standard symbol for a substractor.
If it was an adder, there would be no need to label the inputs since they are all the same. But a substractor has "+" and "-" inputs so you know which input is substracted from which.
The "+" indicates the positive (non-inverting) input and the "-" indicates the negative (inverting) input. The output of the circle is "Pos Input" - "Neg Input". Inputs and outputs units can be voltage or really anything else (temperature, steering angle, current, etc) depending on the kind of system you're looking at.
In this model the substractor has no gain... in real life the input stage of an amplifier would have a certain gain or transconductance, and output a current instead a voltage. In the model, this is pushed into the "A" gain block along with the rest of the open loop transfer function...
Nope, the triangle is a standard symbol for amplifier (gain), and does the multiplication by A.
not enough time... maybe later... I'm on a ES9018 DAC project right now...
This thread needs more pictures
If memory serves, Wayne did publish some shots a while ago. Anyway, let's all ask for a repeat posting of the materials, you will cry when you see those heat sinks.
Mr Wayne mentioned an impedance of 1.8 to 1 ohm, in the 400Hz-20kHz range.
Plus a sensitivity of 84dB/2.83V/1m (17Vp minus 20dB would put his average listening level under 80dB/1m, so he's not stone-deaf yet)
He also mentioned 100Wrms for 93dB, which would suggest 82dB efficiency, his last quote was 3-5V rms/continuous for average listening (Mr Wayne even posted 0.85 ohm as minimum impedance once)
20A in 1 ohm, with +/-42V rails, equates to a dissipation of ~440W
A quick class AB heatsink calculation for the 16 output devices, suggest a minimum of two 0.30C/W heatsinks per channel.
(a 0.15C/W value is also relatively in-line with the 0.50C/W heatsink of the Elektor high-current amp for the Decca ribbon : 15V rails, 85W continuous in 1 ohm)
You are a party breaker Jacco, those heatsinks will cure the thermall runaway of under-compensated OPS anyway...
Hopefully, Wayne will put up a photo of his heat sinks. Off hand, from the photo, I'd estimate at least 80 cm length, al least 40 cm height. Really, very, VERY big, however, I'd still like to know exactly which ones they are, who makes them and exactly what size are they, so I can work out their temp. coeffcient. I hate not knowing it, although by their sheer size and metal area, they should be more than fine.
Since we are dealing with a lot of current, I have still not given up on forced air cooling just yet.
Last edited:
- Status
- Not open for further replies.