and don't forget, it is free for anyone to build....
You betcha!!
And since we're all collaborating on the design and nobody is selling kits, then there is no fear of conflict of interest.
I know a lot of people out there are building the leach amp, some, like me, wish they had him as a teacher... So many have never even met the guy but are missing him! I haven't had the privilege to exchange emails with him, but I could have, if only I had known...
Marshall! We miss you!
Yes, quite a lot. This year only I made six different amps, last year a few, also the last one using bipolar outputs in BIGBT form.
Spooky, I have a separate chassis for my F5 so I will be comparing the two amps, not rebuilding my Leach.
I still think that you can do away with the INA134 and use the ground end of C12 as your negative input with a bit of fiddling with your input resistors. Saves a 15V PSU and the potential signal degradation using the INA134.
I still think that you can do away with the INA134 and use the ground end of C12 as your negative input with a bit of fiddling with your input resistors. Saves a 15V PSU and the potential signal degradation using the INA134.
I would like that, and I didn't do it because I wasn't sure exactly how to do this right. The input needs to easily be converted to unbalanced if need be, so that broken connection to signal ground on the negative input needs to be made right so it gets reconnected when going unbalanced. I'm not sure how to do this properly.
It would save space and several parts, without that extra psu for sure. That ina134 has rather good performance, so I wasn't too worried about its impact, however I am always in favor of having the least amount of parts in the signal's path, so obviously this qualifies.
There is one thing that I do want to be able to do though. I want to add a muting on that input, and I don't see how this can be done if we are in such a balanced arrangement. It must work either way.
I want to use a solid state input muting, avoiding relays and this circuitry needs to be easily triggered. I would make some kind of arrangement to mute the input at powerup, then the output mute is lifted first before the input, but then when powering off, the input mute should be done first.
It's just a shame to have a project with such high level design standards and so average sound quality.
I don't think of the leach amp as "average" sound quality. I don't think there is something that much better out there. So I'm curious and I'd like to know what you propose. For the moment, what I've seen on the threads you mentioned, although I have yet to see it all, was all mosfet based, so that doesn't appeal to me at all. That's why I was asking about non mosfet based amps.
I'd like to take a closer look at what you're referring to and get a sense of that supposedly superior sound quality. I need to be convinced. As far as I'm concerned, I'm not using a design that's substandard, so let's compare, objectively if possible.
Many people's consensus on their's subjective opinions becomes objective.
BTW I recommended other's guy design, not intended to promote anything implying on me.
It's just a shame to have a project with such high level design standards and so average sound quality.
imho, the test of time is the verifier, will people still build your amps say 20 years from now?
the leach amp designed in the 70's and still being built even today, isn't that something?
Do following steps for better SQ:
- if needed use Lundahl input transformer for un/balanced input operation
- completely omit T1, T6 input cascode
- replace R5, R20 by j-fet CCS
- replace R1, R23 with zero ohm wire
- change FB network values to 2,2 k/100 Ohm/2200 uF
- change R2, R21 to 680 Ohm
- change R9, R10, R14, R15 to 22 Ohm
- change R25, R33 to 22 Ohm
- overall use of better modern transistors
There's more but for the start is OK.
Let's say enhanced VAS, etc.There's more but for the start is OK.
Attachments
Now talking about sound quality, I've always been very wary of using transformers in the signal path. I see those are extremely expensive, even on ebay, and to me this type of thing is more what we'd find in tube amps, which to me is not something that rimes with sound quality, although some people swear by it.
So you don't like the zeners there. What don't you like about that cascode config?
What would using a jfet there bring? How can that help the sound?
This offers a little bit of separation between the input stages and the more ripply rails. Why would you do that? How could more noisy rails on the input stages improve the sound? I don't get that one.
All this would simply undo what makes the leach amp a low tim. You would change the feedbacks, reduce the local feedbacks in several stages, which would increase the gain on each of them, and then change the main feedback loop as well. All of this is going exactly in the opposite direction from what leach was doing to reduce TIM. I'm sure you would likely reduce further the THD, but it would increase the TIM in the process. How can this be better? I don't see it. Can you explain the reasons behind those choices? And have you done any sims to compare all this?
Well, that's in my plans for further versions, but this one for now is done with the idea of using what I have on hand.
I will eventually move on to the flat packs, although they're inferior for heat dissipation, we'll just use more pairs.
The leach amp was designed with one important thing in mind, "low tim", and that's a much harder distortion on the ear than HD. This goes way back to Matti Otala and Leach was applying what Otala has found out, and this obviously was right on target and it's done well.
What you propose seems to me to be aimed at further fighting the THD and not TIM. I don't think the majority of people are able to distinguish a 0.00.... distortion difference, and not if it's HD. However lowering the TIM has much more of an effect on what the ear can distinguish and it does thus improve sound quality. Have you read what Leach was explaining about how the design was done and why the choices were made that way?
I think using a different type of protection that allows removing the VI limiters' action on the drivers could improve things, because the potential remains there for the protections acting when not necessary. Even well calculated, there is a chance the protections can have an effect on sound.
Plus Obviously removing the output coil would improve things a bit, at least on measurement apparatus, most likely nothing a normal human can detect. Or the output coil could be included inside the feedback loop maybe, which would reduce its influence somewhat.
Beyond that, we may be able to improve the current sources as you suggest, but I'm not sure this is enough to really make any real difference.
I don't think a 0.000xx reduction in THD will matter. No one can hear that. But a 0.1 reduction in TIM is a lot more noticeable.
Please prove I'm wrong!!
Let's say enhanced VAS, etc.
Looks interesting, and complicated. How does this work and what does it bring?
How can this be brought in the design properly?
Steps to be taken improves SQ, each step practically tested, compared to your's basic sch configuration, confirmed by listening tests many times in recent years, now it's up to you what you will choose. 
For instance if you don't know what CCS brings over plain resistor in LTP common node, than I see no reason to explain any further. Just telling you what will elevate this amp to higher SQ level.
For instance if you don't know what CCS brings over plain resistor in LTP common node, than I see no reason to explain any further. Just telling you what will elevate this amp to higher SQ level.
Extremely expensive
It is in perfect synergy with your's design price level.
Lundhal line transformers are located in top high-end equipment like Linn Klimax, DCS, Jeff Rowland etc. and you're questioning SQ
OPA134 is a
compared to Lundhal. your enhanced VAS falls into the 3transistor VAS category.Let's say enhanced VAS, etc.
Many experts on this Forum tell us that 3transistor VAS are very difficult to stabilise.
But guys all these features I counted are practically tested, no problems found with them, I would not lead spookydd to some odd ways. Aren't we all strive for better SQ, at least that's my priority. For instance if listening tests clearly shows degradation in SQ by using input cascode than there's no place for it to be there.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.