Allen Wright said:
Good for you!
Allen Wright said:
That's your opinion as you say, it is respectable but questionable as well. Too many variables in the audio chain where the amp is usually the last to worry about!
So?
Comments on how things sound are always questionable and it was not me starting to say how nasty a CF is!
45
Allen Wright said:
In the first place, it's important to know what it does and what it does not do. You might come up with a Zo= 500R for a particular design. It might look as though this can drive a 500R load, but that's not necessarily the case. The CF is good for presenting a low impedance to a Hi-Z load. It doesn't necessarily work with Lo-Z loads since the heavier the loading, the less NFB it receives. The result is rising distortion.
Lots of the complaints about CF's seem to arise from this particular misapplication.
As for the actual design, I use these as grid drivers for finals to insure adequate grid drive to avoid slewing, and to provide a bit of grid current on overdrive peaks to improve overdrive behaviour. The first thing is to determine what peak current you'll likely see at ~30KHz at the finals. Then apply the "Rule of Five" borrowed from SS design and make the Q-Point current at least five times greater than the peak current.
Then select a loadline that gives an acceptable THD estimate from the plate characteristics. The better the estimated THD performance, the better the CF operation. Best not to rely completely on NFB to clean up your open loop messes.
Miles, I think that Allen is aware of the difference between source impedance and what a cf can reasonably drive. The configuration he uses doesn't have a lower source impedance or any more current drive capability, and if you read his book, he's driving appropriate loads. He just plain and simple doesn't like cfs. I happen to disagree with him on that point (we agree on many more points), but it does not come from him trying to make a cf do what it's not supposed to do.
I can quite understand where this approach would be the correct one. If inadequare drive leaves the grids flapping in the wind then its going to introduce more distortion than the CF. I however have always designed my projects to run in class A. If I need hard drive I choose something like a 5687 ,ECC88 and 6H10 in anode follower. So if you don't need to drag the amp into class B, and you choose a beefy driver, you can generally save yourself an extra stage.
It has to be said as well, the issues that the CF brings are very similar to the issues brought by gNFB. Therefore if you are designing your amp with gNFB it is very unlikely that you will be able to tease out of the mix the particular contribution the CF is making to the overall sound.
If thats your design philosphy then alls well and good. My design philosphy doesn't easily allow for either of those things and I am certain brings its own issues.
Shoog
Shoog said:
No!
The ECL82 is a nice tube and features characteristics that other tubes simply can dream of.... like the gold plated grid for the best of them. However it remains a cheap tube even if Telefunken made and they are subject to the same laws just like any other device, however.
First: the output is self-biased using resistors which are not small and the grid current must pass through them.
You always have some degree of modulation which is never as small when the driver is low output impedance (where can really be practically negligible in comparison to other things), direct-coupled and at right potential to bias the output tube without the need of cathode resistors.
Second: no matter how much current the drivers have they remain high impedance sources and the distortion still remains proportional to the source impedance. Also the grid currents can be different from tube to tube, so you are just lucky if the LTP sees to identical loads. There is no electronic splitter which remains symmetrical if the loads are not the same.
45
In this particular design (Partial feedback) the higher the output impedance of the driver the better it works. Place a really high impedance pentode there and its nearly perfect.
The ECL82 is a crap tube really and presents real difficulties to getting it sounding good in modern applications. Most people seem to overdrive the triode and complain that it sounds harsh. I would guess that there are very few designs which can drag the best out of this tube without surrounding it with additional tubes. That was never mine or Jaaps original intention so you have to work with what you have.
LTP's given a stiff enough tail (which admittedly mine isn't stiff enough) will give perfect balance because there is nowhere else for the signal to go than through the anodes, irrespective of load. They work if used correctly. I could have improved the design of my headphone amp by implementing a stiffer tail - but the results simply didn't warrant it - it works and sound very clean.
Lets face it our design philosphys are completely different. I doubt I would ever build an amp to your methods and I doubt you would use mine. Shame, but no loss to you or me in the end. However don't try to give the impression that either is the perfect way to go as both have their limitations and someone out there might just take us seriously.
Shoog
The ECL82 is a crap tube really and presents real difficulties to getting it sounding good in modern applications. Most people seem to overdrive the triode and complain that it sounds harsh. I would guess that there are very few designs which can drag the best out of this tube without surrounding it with additional tubes. That was never mine or Jaaps original intention so you have to work with what you have.
LTP's given a stiff enough tail (which admittedly mine isn't stiff enough) will give perfect balance because there is nowhere else for the signal to go than through the anodes, irrespective of load. They work if used correctly. I could have improved the design of my headphone amp by implementing a stiffer tail - but the results simply didn't warrant it - it works and sound very clean.
Lets face it our design philosphys are completely different. I doubt I would ever build an amp to your methods and I doubt you would use mine. Shame, but no loss to you or me in the end. However don't try to give the impression that either is the perfect way to go as both have their limitations and someone out there might just take us seriously.
Shoog
Shoog said:
Sorry, it is not difficult to make ECL82's work. This is simply ridicoulus!
The problem is your design.
Shoog said:
And the output grids? Are you driving them or just the anode resistors????
Shoog said:
That is sure.....
Shoog I really think you should study some basic electronics before speaking about philosophy. This is not an impression, it is written in your posts!
45
The ECL82 triode biases up to very low voltages and passes very small currents. In a two stage design these are both problems. It is not a good triode for modern applications. The triode of the ECl86 is much easier to use and that is why most people pass over the ECl82 in favour of the easier to use ECL86.You can try the obvious approach of wrapping loads of gNFB around the two stages to try to lower the input sensitivity - but I tried that and I got a dull and lifeless sounding amp - it didn't work.
You are claiming then that LTP's cannot do the job of driving outputs. I am certain there are a raft of people who would strongly disagree.
Do you understand the operation of a partial feedback amp, have you built one ?
Finally - and for the last time - there are no problems with my design - it sounds fine and perform well. Others have listened and agree that it was superior to other well designed headphone amps. Build it, see for yourself and report back. Then I will respect your opinion.
Shoog
You are claiming then that LTP's cannot do the job of driving outputs. I am certain there are a raft of people who would strongly disagree.
Do you understand the operation of a partial feedback amp, have you built one ?
Finally - and for the last time - there are no problems with my design - it sounds fine and perform well. Others have listened and agree that it was superior to other well designed headphone amps. Build it, see for yourself and report back. Then I will respect your opinion.
Shoog
Shoog said:
It doesn't seem to me. The triode section of the ECL82, in rather large area of its curves, is basically half of the 6SL7!
You can get more than enough bias without 1) overloading problems, 2) any other "acrobatic" request, if used according to its characteristics in plenty of "modern" applications.
Shoog said:
This is what I am saying from the beginning: the ECL82 is just not suited for that application.
However this is not equivalent to defining it as a difficult tube to work with!
Rather, you are making things complicated where it is not necessary.
Shoog said:
I know exactly such circuit and it is very common in SE using an ECC83 to drive EL84.
Not my favourite, anyway....
Shoog said:
Your and other listening impressions have no value for me. Do you still believe in the reviews??
I think that now I have enough of this discussion.
45
Jaap said:Wel, perhaps I better built a chipamp because there is only one way to connect the different parts (?)
Unfortunately not, Jaap... there is the inverted and the non-inverted (which is, with the 'non' and 'in', the straight version, me think). Then there is also the quest for which resistors to use, which de/un/cross-coupling caps, binding posts, power supply (here again which de/un/over/cross/coupling - caps.
it ain't easy...
Shoog said:
When bending a Pentode or Sand into a Triode using Schade's
partial feedback method, the effective "grid" has a much lower
input impedance than usual. It can completely mess up linearity
and the computation of Mu (for that virtual Triode) when driven
from an impedance that is permitted to vary.
The virtual network that sets Mu in a real Triode is intrinsic to
the space charge (which is a distributed virtual grid to itself).
And I'm sure its actual impedance must be quite a bit higher
than any formed by external resistors, except perhaps for the
Miller effect.
That said: It matters not if the driving dynamic resistance is
extremely high and flat as Shoog suggets, or extremely low
and flat. Anywhere inbetween and flat. Just plain flat is all
that really matters. And flat with frequency as well, at least
over the full range of audio ...
If you compare Schaded FETs as present by myself and MJK
for comparison. You see I use high impedance Drain to buffer
the drive. Michael uses low impedance Source. Both achieve
the same result by slightly different external networks.
Would you care to share a schematic of your FET buffer implementation.
Could you suggest a suitable implementation for Jaap since this seems to be the way he wants to go to get the maximum clean power out of his PP ECL82 build.
I think a lot of the issue of linearity are absent from my implementation because it is expected to work at really low output power levels in headphone duty. If you want to drag the theoretical max power out of the PP ECL82 distortion may rise significantly. Having said that driving my speakers and it still sounded sweet.
Shoog
Could you suggest a suitable implementation for Jaap since this seems to be the way he wants to go to get the maximum clean power out of his PP ECL82 build.
I think a lot of the issue of linearity are absent from my implementation because it is expected to work at really low output power levels in headphone duty. If you want to drag the theoretical max power out of the PP ECL82 distortion may rise significantly. Having said that driving my speakers and it still sounded sweet.
Shoog
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.