Member
Joined 2009
Paid Member
I believe JLH used SOT fixed resistors, i.e. 'select on test' which maybe what I'm going to do too. There are big pots out there, but they don't tend to be 'hi-fi' and I'm reluctant to put anything nasty into the current path.
I was considering a switch, to allow me to select 'hi current' and 'low current' settings by switching in additional resistance in parallel with the fixed resistor. This does mean putting switch contacts in the signal path...
I was considering a switch, to allow me to select 'hi current' and 'low current' settings by switching in additional resistance in parallel with the fixed resistor. This does mean putting switch contacts in the signal path...
Ah yes, fair call. Sounds good although it might take some fiddling to get two channels exactly the same. What a nice "pure" amplifier topology though
Member
Joined 2009
Paid Member
any thoughts ?
I have decided on a capacitance multiplier for the power supply in line with JLH himself, but it will not be regulated.
The existing heatsinks which have the power devices riveted on don't look big enough so today I used my circular saw to cut 4 pieces of heatsink off a 3m extruded piece that was lying around. It ain't got big fins but I plan to stick them all together.
Hi all, BIGUN-- if it's worth anything I have used both cap multiplier and CRC supplies for my prototype JLH to see which one I prefered. I am now building the final ( I hope ) version and am going to use CRC. 45000uF 1 Ohm 45000uf per rail per channel. eg.8 caps and 4 resistors.
My opinion only- best of luck with your build
Alan
My opinion only- best of luck with your build
Alan
8.2k means only ~0.2 mA in input transistor.
Now the lowest I know is 0.65V/4k7=0.138mA by D.Self in his lectures.
I think 8.2k gives not too low current.
I think there is nothing to gain by increase current.
But in the place of 2N1711 we should put a transistor with good gain.
This will take load off input transistor.
I have recommended replacing 2N1711 with 2SC2911.
It improved the amplifier much
Now the lowest I know is 0.65V/4k7=0.138mA by D.Self in his lectures.
I think 8.2k gives not too low current.
I think there is nothing to gain by increase current.
But in the place of 2N1711 we should put a transistor with good gain.
This will take load off input transistor.
I have recommended replacing 2N1711 with 2SC2911.
It improved the amplifier much
Member
Joined 2009
Paid Member
Alibear - I know Hugh Dean isn't a fan of capacitance multipliers either (and he is careful to listen to this design) and I am always a bit concerned about introducing additional active devices. But, in a Class A amp with constant current demand (meaning adequate rail decoupling capacitors after the cap multiplier) I wonder why a cap multiplier would have much impact at all except to reduce psu noise. When you tried the cap multiplier did you have plenty of capacitance after it ? I'd be interested to hear more about why you prefer the CRC - I assume based on some listening impressions, so spill the beans, what did you 'hear' ?
Lineup - I've always worked with the 'rule of thumb' that I want about 1mA minimum through the input device, or any BJT for that matter. I have noticed that most of the blameless style amps around here follow that rule of thumb. But I am thinking the issue for a singleton input is that this current must flow through the feedback resistor which at 2k7 produces a voltage drop that you might not want to be too large. Also, the gain of the input device - does it not fall if the collector load resistor is reduced in value and therefore OLG suffers ?
Lineup - I've always worked with the 'rule of thumb' that I want about 1mA minimum through the input device, or any BJT for that matter. I have noticed that most of the blameless style amps around here follow that rule of thumb. But I am thinking the issue for a singleton input is that this current must flow through the feedback resistor which at 2k7 produces a voltage drop that you might not want to be too large. Also, the gain of the input device - does it not fall if the collector load resistor is reduced in value and therefore OLG suffers ?
wrong conclusion.
A ClassA amplifier, whether a Power amplifier or Pre-amplifier or a single stage single ended amplifier is only constant current when there is Zero output current, i.e. in the quiescent state.
Surely one of the worst aspects of the JLH amp is the singleton input transistor. I've used this amp happily for years, and as the rest of my system has improved I'm left thinking the sound is a bit "transistory". I wonder the effect of replacing the singleton with a JFET (which is linear in operation as opposed to the exponential BJT). It would alter the biassing, especially in relation to thermal changes, but that may still be within acceptability. It would remove the "transistory" distortion products that the singleton BJT introduces. Anyone tried this?
Yes, chaniging 8.2k for more cureent will cause a larger drop across 2.7k.
You then may have to change the resistor and CAPACITOR that goes with gain setting.
In the end you must have a large value capacitor.
And maybe this is not so attractive.
But if I had designed it, I would have used more current than ~0.200 mA
Sure is.
Yes, though the class A stage has a steady average current. This reduces its sensitivity to source resistance and makes the capacitance multiplier an appropriate choice, I think.
Bigun said:
I didn't have to try too hard to get low enough noise, and I have sensitive speakers.
Member
Joined 2009
Paid Member
I'm not convinced that the Singleton is a disadvantage here. In the main alternative, an LTP, you still have a non-linear gain stage, with the same exponential dependence on the control voltage. The main difference is that the LTP is symmetrical and therefore, theoretically, has lower even order harmonics. It does nothing better with regards the symmetrical (odd order distortion). It also stops short of achieving it's theoretical advantages of symmetry if there are errors arising from any mismatch between the characteristics of the two devices and in a simple topology it's very difficult to match the current profile through the two devices without the brute force of a current mirror - adding perhaps unwanted additional OLG. The LTP requires more stability enhancing Cdom capacitance. I think a Singleton input is as good as if not better than an LTP providing you don't need the automatic dc-offset correction that an LTP can provide. A JFET has a different characteristic and I haven't tried this approach, it would be interesting to hear the difference. It's still non-linear of course and a JFET has much lower trans-conductance than a BJT, so what it might gain through it's less exponential characteristic it potentially lost through a lower nfb factor.
Last edited:
I do not believe that it is the case. JFET it is not linear by itself. Our dream linear device should have parallel, equal distance , straight output characteristics, and even a JFET is not so. The only way to go is linearize them with Local Feedback. So gain is an advantage, hence, like Bigun correctly says is BJT territory. I do not see JFET use unsess you have to deal with high impedance sources (that should not be the case in a power amp). Take also care that pursuing distortion you shuld not forget noise, that beahave differently in BJTs and JFETs
@Audio
Isn't that what the emitter/source resistor does?
I once removed a BJT singleton gain stage from a Creek Audio amplifer and the effect was revelatory. Of course there may have been other things involved, but I have bad vibes about singletons. The feedback provided by the emitter resistor helps to linearise the inherently non-linear (Ic/Vbe) BJT - remember, it is really Ic that determines the output voltage in this configuration - but it can't make it perfect. I will try a JFET at some point if no one tells me they've already done it.
No. No transistor is a perfect amplifier of voltage.
Not even with heavy use of emitter resistors.
Some additional negative feedback or help is required.
JFET. I compare one 2 resistors voltage amp in spice
with two JFET SRPP.
THe SRPP showed THD 0.010% while simple JFET went like 0.500% THD.
What a SRPP Shunt regulated push-pull stage does
is that it suppresses 2nd harmonic in a very effective way.
This is why distortion is low. Highest harmonic is 3rd. 2nd below.
Hi all, Bigun, the cap multiplier I used was one from the class A website. There was no extra capacitance fitted after it, large value caps were used before it as per the diagram. Both the cap multiplier and then CRC were used to power the amplifier at different times. Driving into BC1's ( not very efficient I know )no PSU noise could be heard even with the ear almost touching the bass driver cone.
I believe simple is best, again only my opinion, but I know the Krell boys get along OK with just CRC.
Alan
I believe simple is best, again only my opinion, but I know the Krell boys get along OK with just CRC.
Alan
CastleJohn
I suspect you are looking at the wrong end for "transistory" sound of the JLH. As I mentioned (probably in this thread a few hundred posts ago) the bootstrap output stage is probably the main issue if you are using slow output devices. Or old devices which have non-linear gain, even the fairly fast MJ480 which JLH prescribed. The lower device can only work by stealing current from the upper, so with gain-droop transistors you have to run at a higher quiescent current than you would like. The slow response may cause transient intermodulation distortion (dare we mention this these days?) but both problems could be cured, it seems to me, by using modern transistors like MJL3821 2SC5200 etc. Reports from people trying these indicate oscillation. I suspect this can be cured with a small feedback capacitor on the feedback resistor and possibly sprog stoppers on the base-collector junctions of the output pair.
These types of transistor also have the typical gain discussed in the original article (80-100) and probably will work well.
John
I suspect you are looking at the wrong end for "transistory" sound of the JLH. As I mentioned (probably in this thread a few hundred posts ago) the bootstrap output stage is probably the main issue if you are using slow output devices. Or old devices which have non-linear gain, even the fairly fast MJ480 which JLH prescribed. The lower device can only work by stealing current from the upper, so with gain-droop transistors you have to run at a higher quiescent current than you would like. The slow response may cause transient intermodulation distortion (dare we mention this these days?) but both problems could be cured, it seems to me, by using modern transistors like MJL3821 2SC5200 etc. Reports from people trying these indicate oscillation. I suspect this can be cured with a small feedback capacitor on the feedback resistor and possibly sprog stoppers on the base-collector junctions of the output pair.
These types of transistor also have the typical gain discussed in the original article (80-100) and probably will work well.
John
Hi all
oops... just resimulated a JLH. Input transistor Vbe swing is several millivolts, will definitely be contribuiting to distortion (for the original 4-transistor, with BD139 driver/MJL3821's in the output), at 10W/ 8ohms. Higher input stage current did not seem to bring the distortion down as much as expected from the gm increase... so now looking at lower feedback resistor, etc .
John
oops... just resimulated a JLH. Input transistor Vbe swing is several millivolts, will definitely be contribuiting to distortion (for the original 4-transistor, with BD139 driver/MJL3821's in the output), at 10W/ 8ohms. Higher input stage current did not seem to bring the distortion down as much as expected from the gm increase... so now looking at lower feedback resistor, etc .
John
Member
Joined 2009
Paid Member
I'm not looking for significantly lower distortion from the JLH 1969 except that which might be gained from the dual rail supply and possibly a bit more current through the input device. There are many ways of reducing the distortion, the easiest being to increase OLG and hence increase nfb factor. I don't consider more nfb a step in the right direction although I know some people do and that's OK with me too. There are some obvious ways to increase OLG through the use of some constant current sources which also reduce the Vbe variation of the devices so you win twice over - but in my version of the JLH I don't want any (active device) CCS's near the signal which means they are out of the question for the input device or the phase splitter device. I know it has been tried (see Geoff's website) but to me, this is no longer the JLH.
At the same time I don't see TIM as an issue in the JLH either. The OLG extends well past the audio region even using the MJ480 device and there's nothing to suggest it would be slew rate limited within the audio spectrum. I'm not in favour of using 'modern devices' that require additional compensation to remain stable. I view compensation as an unfortunate consequence of nfb. Faster output devices would allow more OLG at higher frequencies and hence extend higher nfb factors further up the spectrum. But I'm not looking to push the JLH into such territory, I'll leave that for you fellas.
At the same time I don't see TIM as an issue in the JLH either. The OLG extends well past the audio region even using the MJ480 device and there's nothing to suggest it would be slew rate limited within the audio spectrum. I'm not in favour of using 'modern devices' that require additional compensation to remain stable. I view compensation as an unfortunate consequence of nfb. Faster output devices would allow more OLG at higher frequencies and hence extend higher nfb factors further up the spectrum. But I'm not looking to push the JLH into such territory, I'll leave that for you fellas.