JLH 10 Watt class A amplifier

[Bigun]

I'm not going to use split supplies, so I will have an output condenser. This means the load current goes through the output cap and the 'last' psu cap. Even with split rails I'd have my load current going through the psu caps, both rails. But in contrast to the split rail version the single rail version has the disadvantage that the nfb network does not take off just before the load and reference to the other (gnd) side of the load - rather it takes off before the output cap and can not correct for any non-linearities this cap introduces. The split rail version is better, but I want to stick with the 1969 version and also have the benefit of speaker protection against dc.

A possible tweak to the '69 circuit would be to take a small portion of the nfb signal from after the outptut cap. How would I do this - has anybody found it worthwhile to do this ?

A related circuit is the Leak Delta 70, it has an interesting arrangement that bootstraps the input but there is a feedback resistor from the load side of the output cap - I can't quite understand what it is doing ?
You can find schematic here: Leak Delta 70 - Attempt DIY, Overhaul or buy another amp?

Can you help me understand this ?

[mikelm]

Originally Posted by Bigun

I'm not going to use split supplies, so I will have an output condenser.

If in 5, 10 or 15 years you get around to making a more recent, DC linked version of this amp you may be kicking yourself for not having done it originally - the difference in sound quality is not small.

An output "fuse" made of suitable size constantan resistance wire soldered between two PCB posts 0.1 - 0.2 ohms will also protect your speaker and in my opinion does not compromise sound.

Anyway, whatever route you choose I wish you well

mike

[AllenB]

One of the better iterations of this amp, IMO is the plain '69 version with a dual supply tacked on.

[alibear]

Wow the thread has burst into life again, keep it going, the more I read the more I learn.
With you on that mikelm.
My project has slowed down due to too many other boring jobs. That's life.
Thanks all.
alan.

[Bigun]

I still need to understand something here.

I've attached a simple dual rail supply schematic. It's not the conventional way to draw an amplifier with nfb but I want to make some points.

The signal input is a voltage, sharing a common point of reference to the output. The signal input is a single ended signal over an RCA-connected cable and we assume that the shield of the cable is, practically speaking, connected to the chasis of your source and the chasis of your amplifier (potentially via a ground isolating resistor bypassed by a diode to reduce ground loop hum).

The d.c. current from the power supply flows through the amplifier as shown. With no input signal the amplifier output should be at the same potential as the Common Point so that no dc flows through the speaker and the power supply.

The a.c. signal current in the JLH flows through both upper and lower output devices - call it push-pull or not. Since current flows in closed loops there must be two of them, each including the load and one of the output devices. I've shown these a.c. signal loops. We see that a.c. current must flow through both of the power supply rail capacitors. So we see that a dual rail amplifier does not remove large electrolytic caps out of the load current path.

But we don't find this is a problem because distortion introduced by these caps is reduced to negligible levels because the amplifier has plenty of nfb to correct for it. The nfb network samples the voltage signal at the output from the amplifier - shown by the arrow - and feeds a portion back to the inverting input of the amplifier.

OK so far ?

[Bigun]

Now I've attached a similar schematic, but this time the original single rail JLH amplifier.

This time we can see that again, the a.c. signal current flows in two closed loops. One of them flows through the output capacitor. The other one flows through both the output capacitor and the power supply rail capacitor. As with the dual rail version, we have two capacitors in the speaker load a.c. current path.

But people say the single version sounds worse. Why ?

I think it may because the distortion in one of these capacitors is no longer adequately corrected for by the nfb. The nfb is sampled at the output of the amplifier as before, shown by an arrow, but this means it doesn't sample the distortion introduced by the output capacitor and therefore can not correct for it.

So I go back to my earlier question - if my understanding of how the amplifier with dual rails and single rail works, the issue is the nfb take off point.

A possible solution is to move the nfb take off point to the load side of the output capacitor, shown by the red arrow. In principle, this should allow the nfb network to correct for the distortion of the output capacitor. There are problems with phase-shifts when capacitors are part of the nfb loop (and potential motor-boating), so a practical design uses a combination of both nfb take off points.

Has anyone tried this - can it make the single rail amplifier sound as good as a dual rail amplifier ????????????

[mikelm]

I think you have analysed the amp topology quite well but I think that even if you figure out how to have a dual feedback system the resultant o/p will still be the product of the two f/b paths and not purely the f/b from the far side of the o/p cap - so the result will also me a mix and not as pure as a single f/b loop in an amp without an o/p cap.

Also in both of your diagrams there is an electrolytic cap in the feedback path which will introduce it's own limitations to how accurately the feedback loop can function. This is clearly audible when you remove it.

Have you checked out Geoff's site to see the different versions he shows there ? I think there is a link above somewhere a few posts up - well worth a read

cheers

mike

[mikelm]

To make a single rail amp as good as a I would take a balanced signal all the way from the DAC or other source.

Have two amps per channel, one for each phase with an earth defined with two equal resistors across the rails and smoothed with an electrolytic.

this means:

speaker is then driven by two amplifier o/p's with no caps
DC feedback with no caps
i/p signal DC linked also with no caps
1/2 - 2/3 rail voltages only needed for same o/p swing.
perhaps also common mode noise will be cancelled

how's that sound to you ?

[john_ellis]

Hi Bigun
ONe reason that the SE version might sound worse is due to poor LF response. It is not clear to me when people say that capacitor coupled designs are worse what the problems are - whether they are HF or LF part of the audio spectrum.
Modern low-ESR, ESL capacitors should be far better than the older types. I've even had some old types which sing along with the music -that has to be introducing distortion.
Even with good caps the size is too small. With the usual 2.2 mF (or 2200 mfd as it used to be known) leading edge bass transients are hopelessly attenuated. YOu need larger ones, as was point out to me some time ago. 22 mF starts to get close.
THe second point about the original JLH is that the output transistors need to be faster than 2N3055 was back then. A modern epi 2n3055 should be better, but the bootstrap can only work as fast as the output - so if the output is slow, so is the amp. WIth the second version of JLH with a current source the output can be pulled faster.
John
John

[Bigun]

I don't fancy going the whole hog with a balanced signal - I don't really have a suitable source or cables and it would be incompatible with the rest of my gear. Perhaps viable in the long term but not for my first JLH.

If we agree that the single rail can't completely replicate the dual rail set up in terms of nfb correction then it comes down to a choice between a) virtual earth supply - since my trafo has no centre tap, b) better quality output capacitor.

The original JLH uses relatively small value output caps compared to those used for rail decoupling so this is already an issue - D Self has shown that distortion from the capacitor remains low until the a.c. voltage across the cap starts to increase (i.e. the l.f.. roll off). Very large value caps would therefore diminish the distortion problem - Nelson Pass recommends such an approach for his Zen amplifiers.

Ever considered good quality double bi-polar ? - if not, read these:

http://www.diycore.com/patrick/files.../capsound5.pdf
http://www.diycore.com/patrick/files.../capsound6.pdf


Who has tried really good output caps and still found dual rail better ?


Edit: just saw John's post. Looks like we're both thinking the same about the need for Big capacitors. John - I'm not considering the 1996 design with current sources at this point, just the original '69 although I will entertain dual rails. I guess there are many changes and improvements possible so I have to draw a line somewhere and using a CCS is a casualty of this choice. I have some objections to the CCS too - it places another active device into contact with the signal (if you think carefully) and increases OLG which I don't wish.