It seems to me that what Mr. Broskie writes has often never been built. When I breadboarded his earlier solutions (like the one above) they were quite dissapointing.
But don't let me stop you from trying them. 😀
Member
Joined 2009
Paid Member
Hi Soulmerchant - I haven't tried anything from Broskie myself and was always quite curious about his version of hum cancellation on the output. When you say it was 'disappointing' can you say a bit more - was the hum reduction incomplete?, how did you adjust the capacitance?, was it frequency dependent ???
The problem with Broskie's scheme is that the cancellation depends on a capacitive divider. But caps have high tolerances, and there are no variable caps to adjust. (At least not for the capacitances we're talking here)
The cancellation in practice will never be as good as L-W.
The cancellation in practice will never be as good as L-W.
How can I explain this...... it just didn't sound like there was someone playing live music in my living room.
Believe me I made many adjustments and tried many things before simply scrapping that topology altogether. There is no thread here on it because I was never satisfied to share anything of the experience.
Not quite. The mu is 60 or so, not 15. Not going to get there with this pair - not without some creativity.
Sheldon
Last edited:
Whoops, you're right! I read if off the wrong side of the chart.... Its even worse! Guess this is why I always re-do my calcs 3x before thinking of building... 😉
That makes the "triode characteristic factor" only 3.2 (which is not nearly good enough).
Honestly, I don't know what people have against the bootstrapped follower. It works beautifully. I seriously doubt anyone here could "hear" it...
Hi Sheldon,
MANY thanks for contributing to our questions, very nice !!
You are overlooking something, which I just want to KINDLY and respectfully point out to you, and others.
You state " But you add unnecessary components and gain nothing in the bargain."
I suggest we DO gain much with the Input stage decoupled from the Finals stage, IF you want to optimize the amp's performance, and not just " have it pass a pleasing signal".
Decoupling the Finals and the Input stages frees us from designing a B+ filter that has to run TWO stages, each having competing electrical requirements.
(1) The 2A3, etc., Finals need quick FAST Peak Instantaneous Current delivery, hence the need for very careful PSUD simulation and filter optimization, with low DCR chokes ( under 20 Ohms) and small amounts of capacitance in the filter ( 50 uF maximum ). Low ripple to the 2A3 Finals is NOT a problem, I know guys who run 2A3 amps happily with 850 mVAC of ripple, on ALE compression driver systems.
(2) The Input stage, however, needs superbly low amounts of ripple, to get proper definition and understandable diction on voice, eg : a Gilbert and Sullivan Opera, on any London OSA-series LP record. I always shoot for 1 or 2 mVAC as maximum ripple to the Input stages, to keep the music intact.
If and when you are forced into having ONE supply B+ filter to fit the needs of two stages with OPPOSING filtering needs, you are put into the position of having to " rob Peter to pay Paul", and nether of the two will be totally satisfied.
THIS is why I am suggesting and looking for a way to apply separate filtering between the two stages, so EACH can be optimized and perform at its best !!
If ANYONE has a bright idea of how to do this, AND employ LW hum cancellation on a two stage ( not three ) amp, please weigh in !!!
BUT, all of the above is just PART ( part one ) of the design problem as I see it.
Part two is as follows :
WE all do not know how an optimized two stage amp, with great Rk bypass caps, will compare with a similarly run ( same tubes and same op points ) LW implementation, with hum cancellation and NO Rk bypassing. I suspect, maybe only Ciro Marzio in Italy may have experimented with these two topologies, in which case, I hope he will comment !!
What people FAIL to ever discuss is Rk bypassing, done well, has certain distinct and undeniable ADVANTAGES to the final result we obtain at the speaker voice coil, and hence, our ears.
Let me try to explain. Caps are the WORST part in audio !!! Only the very best are even remotely acceptable. A film cap may measure wide band on a bench, but none of them ( well, maybe one, a Multicap ) will PLAY wide band into the speaker voice coil. All caps "tune" to a certain range of frequencies, and it is a function of their uF value !!!
**** If you try to bypass a Rk with a single cap, the amp will tune to that film cap's value !!! It will exclude other music ranges that WE WANT to hear !!! ****
There have been only two people who I have seen address this, but I am certain others have.
In 1989, Dennis Fraker artfully used MULTIPLE Rk bypassing, film caps, in his Serious Stereo 2A3 amp prototypes, and he continues to do so today, I am sure !!
Also, in Sound Practices 1994, Ciro Marzio introduced to "us unwashed masses" the need for PROPER multiple bypassing, with a rudimentary use of HIGHEST QUALITY main film cap, PLUS a carefully selected-for-sonics 0.22 and a 0.01 uF .
I don't think ANYONE really knows HOW to bypass a Rk cap in 2015, except possibly the few people I have mentioned, and some of my audio buddies.
If I wanna get "all the nuances" and "every little vibe, musical layering, recorded ambiance" and a LINEAR presentation at the voice coil / our EARS, out of the SE DC amp, we need to bypass caps properly !!!
This takes time and often lots of money to do, because only the best caps are remotely usable. AND, we need to figure out three things (1) the numbers of caps required to get a linear presentation, (2) and their values, (3) and their makes !!! Its a chore, and a costly one in time and money. There is no free lunch.
So the question is, has anyone ever compared a "true LW with hum cancelling " to an Rk bypass cap-optimized two stage DC triode amp. Ciro ?? Anyone ??
I am, in 2015, very OPEN to exploring hum reduction ala Darius, but the overall result has to be better than what some presently do with Rk bypassing !! Now you all know from where I come !!
Have fun, enjoy the music, and THANK you for contributing to this cool thread !!!
I still will likely have more Q's about implementing two stage amp ( LW / Darius ) hum reduction... I image. Lets see how things progress up here !! Thanks to everyone for contributing.
Jeff Medwin
Last edited:
Jeff,
You are missing the entire point to my response, and of the unique design of this amp. You can't "decouple" the input stage from output stage. They are coupled plate to grid. There are only different ways of coupling. You have to think in terms of AC coupling and DC coupling.
Yes, in a typical AC coupled amp, you might want to provide additional filtering of B+ (not decoupling) to the input stage. In fact, you could use a completely separate power supply for the input stage.
In a typical design with the cathode bypassed to common, both the DC and AC (signal and noise) current components circulate in a loop from power supply, through the OPT primary, and back through the power supply via common. So the power supply is directly in the signal path.
In this design. First, the output cathode is returned to B+, not to common. So the signal path does not circulate through the power supply, so the power supply is not influence by the signal (or, it acts more like a constant current source). However, noise from the power supply still creates ripple at the plate of the input and output tubes.
Any ripple through the transformer and output tube has to show up also at the output cathode. So, secondly, an exact fraction of this noise is instantly injected at the (unbiased) input cathode, which exactly cancels this noise at the output.
Another way of saying it, is that for current to change in either tube, the voltage relationship between plate, grid, and cathode must change. It doesn't matter if the supply changes with respect to common, only what happens across the tube. This design keeps those in virtually perfect balance.
But, because it's using opposing changes to counter balance, they must be kept in perfect phase with each other. Any filtering you put between the input tube supply and the output supply will take them out of phase. Therefore you will add extra components and gain nothing. If you doubt that it works, I'm using mine as a headphone amp - essentially zero hum.
The design, as every good design, works as a whole. You can't blindly substitute pieces of something from an entirely different design concept and expect benefit.
Sheldon
Edit: As this design takes much of the power supply out of the equation, you are pretty free to do what you want there - but quality never hurts. You do need to have the final cap in the supply about the same size as the cathode cap. And yes, a film cap. I would recommend an indirectly heated rectifier for slow start.
Last edited:
Here's the .asc of post #217, zipped with triode models used. Use it for educational purposes.
Hi Jeff,
look at the attached pic. It illustrates a simple and most straightforward way to decouple the input and output stages, kind of, PSU hum wise.
Note that you need to adjust R4 in the hum cancelling network to achieve optimum performance, since the hum level on the input plate changed.
However, no matter what you do, the best achievable hum cancellation is now 5 times worse (for this schemo, in simu at least), which supports Sheldon's point. It just works better when you keep PSU hum in synch on all stages..
That being said, I'm aware there are other ways to decouple the input stage. One would be a stiff CCS load for the input triode, another one would be to feed only the input triode from a decent HV regulator (can be run from the same PSU). Both would effectively kill all ripple at the input triodes plate, rendering the ripple phase matter a moot point.
Both would however be more complex, and require the use of SS parts (in the audio path😉) for effective designs.
Cheers,
GB
Attachments
Member
Joined 2009
Paid Member
I think it's been agreed that the input stage can and should be receiving a filtered supply from B+ and not simply connected to the same direct line as feeds the finals - and that this filtering is not in conflict with the L-W topology.
I would like to offer the opinion that the L-W topology does have a level of cathode bypass on the input tube. As I mentioned in my comments on the bootstrap - ground is a moveable concept. If you look at a traditional common cathode stage the cathode bypass cap goes from the cathode to ground. In the L-W there is a cap + resistor which goes from the cathode to the cathode of the output tube which in turn has a cap to B+ and which in turn has a cap to ground. It's all connected. The cap + resistor on the cathode of the input tube in the L-W topology also has the potential to impact the sound as does the bypass cap in a traditional common cathode stage. Therefore, we should be asking if the L-W topology eliminates the cathode bypass altogether. Perhaps poor Dennis would need to spend some Serious (pun intended!) time finding the right combination.
Last edited:
Agreed by whom? I would restate it thus: For the LW, the input stage SHOULD be simply connected to the same direct line as feeds the finals - additional filtering after the finals IS in conflict with the L-W topology.
All caps (none being perfect) have the potential to impact the sound. But not all caps in all topologies have equal impact. I assert that the caps in this configuration impact the sound as little as possible. Simulate caps with distortion in the two different topologies, and or switch the connection from common to B+ and see what you get.
Sheldon
edit: See this post and read down a few more regarding a filter between stages: http://www.diyaudio.com/forums/tubes-valves/258790-3-direct-coupled-2a3-amps-15.html#post4201171
Last edited:
Use the ltspice model I posted, apply an audio signal of say 500 Hz (V1) and psu hum at 100 Hz (V3). You can easily observe which cap or resistor shunts the hum, and which shunts the audio signal (or both).
What NO ONE Considers, Discusses with Loftin White, Darius, etc etc
Hello ALL,
When you reverse-out the 120 HZ component ( 100 HZ Europe ) of the hum, you also reverse-out the 120 HZ component of the music !!!! The LW circuit can not discriminate between 120 HZ hum and 120 HZ music, none can.
Dennis Fraker, this morning, pointed this out to me, and I am indebted to him, and ALL of you for contributing openly to this thread.
If you are into tubes, you accept a bit of hum, otherwise, build with mosfets and a battery !!
KISS rules!! And when "push comes to shove", a simple non LW, done well, will outperform a LW.
It sorta funny, Ciro stayed out of this discussion. Smart fellow !!
Thanks to all, lets stay in touch...
Jeff Medwin
Hello ALL,
When you reverse-out the 120 HZ component ( 100 HZ Europe ) of the hum, you also reverse-out the 120 HZ component of the music !!!! The LW circuit can not discriminate between 120 HZ hum and 120 HZ music, none can.
Dennis Fraker, this morning, pointed this out to me, and I am indebted to him, and ALL of you for contributing openly to this thread.
If you are into tubes, you accept a bit of hum, otherwise, build with mosfets and a battery !!
KISS rules!! And when "push comes to shove", a simple non LW, done well, will outperform a LW.
It sorta funny, Ciro stayed out of this discussion. Smart fellow !!
Thanks to all, lets stay in touch...
Jeff Medwin
You comment :
" Both would however be more complex, and require the use of SS parts (in the audio path😉) for effective designs."
No sir. I regulate with a single part , and a non solid state part !!!!!!!!!!!
From a single power supply, I shunt regulate the decoupled Input stage with a SINGLE Mills MRA-12 12 Watt resistor. ONE PART, a ONE PART SHUNT REGULATOR.
It provides a very low Z for the Input stage . It shunts to ground MANY more times the current of the Input stage.
Jeff
Dennis is wrong. They are two different signals, one from the input, one from the power supply. The cancellation changes the music signal not one bit, just subtracts the noise. This can be easily shown by connecting/disconnecting the noise canceling circuit. Only the noise component is eliminated.
Or you design circuits that don't hum.
And how is this LW not KISS. Well, apparently not easy to understand, but component complexity?
Ah, so you've heard both? If you have and you like the conventional version more, then be happy. Why think too hard?
Sheldon
I assume you're talking about a voltage divider, and not a regulator then? How does the resistor provide any voltage stabilization against mains fluctuation, or any psrr?
Sure it can. The music signal enters thru the grids, the psu hum thru the plates. Both add up linearly.
At the output tube, the music signal is then directly shorted from the cathode thru the ultrapath cap back to B+. The hum component is obviously shunted thru the cathode resistor to common. This means that the voltage on top of the cathode resistor varies with psu hum only (and not with the music signal).
L-W feeds back part of the psu hum from the output cathode to the input cathode, to cancel (better: balance) it out at the opt primary.
Member
Joined 2009
Paid Member
If you are correct then I am not appreciating how the L-W works - can you help me, please explain why the L-W can not have an RC filter in the B+ supply feeding the input tube - assume a perfect RC filter so that the B+ supply the input tube is perfectly clean - why is this not good ?
- Status
- Not open for further replies.