This can be done with BJTs as well. It requires relatively low resistance base resistor (on the lower OP) and a matching load on the upper OP. This means that the VAS current is not defined by the OP gain as much as a typical JLH. In either case the VAS current should be temperature compensated, which a classic JHL is not. The addition of a diode in the upper side is a mixed blessing and was a common topology in early chip amps and op-amps. A common mistake was to use too much feedback, and they sounded pretty good if you got it right.
The JLH Memorabilia Part-4 concludes the survey of the 1980s:
https://jlh-memorabilia.blogspot.com/2024/12/jlh-memorabilia-past-perfect-4.html
Do read, share and comment !
https://jlh-memorabilia.blogspot.com/2024/12/jlh-memorabilia-past-perfect-4.html
Do read, share and comment !
Well, dipping back into this thread after a break I am not surprised that the JLH fan club is keeping it going.
Having spent some time building and simulating JLH's (mainly, I have to say, for interest as I did not much like the sound it produced) it may be worth summarising some observations givin that points are now buried in nearly 10k posts.
- the original JLH used 4MHz MJ480/MJ481 transistors. I suspect these were devices from Motorola's 2N4904/5/6 2N4913/4/5 series which did not quite make the gain grade. Even if that is not the case, these devices would have provided a moderately respectable sound due to the relatively low junction capacitances and good frequency response.
When folk started using the 2N2055, at the time this would have been RCA's 800kHz device and this, to both simulation and listening, gave inferior results. Modern 2N3055's build with epi technology may somewhat improve the response but they are only rated at aroud 2.5MHz. Some may reach or exceed 4MHz.
Simulations have confirmed that the choice of bias current in the input stage provides the lowest distortion, with the specified design, although increasing the feedback by reducing both the feedback resistor and its grounding resistor can give lower at possible cost of stability, although I have not seen this.
Using an LTP does not improve the situation at all, because some of the distortion characteristics rely on counter-phased cancellation of the single input stage and the following stages. As mentioned a couple of posts above, JLH noted this and quite rightly mentioned the additional phase shifts that the LTP would introduce.
What can be done is to use a current mirror to allow the input device to operate at higher current with the second input fed from a constant current, and to minimise any phase shifts here that is best achieved with a high value resistor powered from a filtered and stabilised rail. That can be used to reduce distortion too, but the input stage current has to be optimised if distortion cancellation is still to work, depending on the driver and output transistors used.
Perhaps the best performance I have heard is when using modern high ft high gain devices although I have to say that I still prefer the performance of a driven (driver-output) pair operating in Class AB with local phase lead (VAS inclusive) stabilisation. Simulations of comparisons between the JLH and the Class AB design I use shows that the JLH PNP input can still clip off or saturate on high level, high frequency signals whereas the inpust stages of the AB design I use do not.
Having spent some time building and simulating JLH's (mainly, I have to say, for interest as I did not much like the sound it produced) it may be worth summarising some observations givin that points are now buried in nearly 10k posts.
- the original JLH used 4MHz MJ480/MJ481 transistors. I suspect these were devices from Motorola's 2N4904/5/6 2N4913/4/5 series which did not quite make the gain grade. Even if that is not the case, these devices would have provided a moderately respectable sound due to the relatively low junction capacitances and good frequency response.
When folk started using the 2N2055, at the time this would have been RCA's 800kHz device and this, to both simulation and listening, gave inferior results. Modern 2N3055's build with epi technology may somewhat improve the response but they are only rated at aroud 2.5MHz. Some may reach or exceed 4MHz.
Simulations have confirmed that the choice of bias current in the input stage provides the lowest distortion, with the specified design, although increasing the feedback by reducing both the feedback resistor and its grounding resistor can give lower at possible cost of stability, although I have not seen this.
Using an LTP does not improve the situation at all, because some of the distortion characteristics rely on counter-phased cancellation of the single input stage and the following stages. As mentioned a couple of posts above, JLH noted this and quite rightly mentioned the additional phase shifts that the LTP would introduce.
What can be done is to use a current mirror to allow the input device to operate at higher current with the second input fed from a constant current, and to minimise any phase shifts here that is best achieved with a high value resistor powered from a filtered and stabilised rail. That can be used to reduce distortion too, but the input stage current has to be optimised if distortion cancellation is still to work, depending on the driver and output transistors used.
Perhaps the best performance I have heard is when using modern high ft high gain devices although I have to say that I still prefer the performance of a driven (driver-output) pair operating in Class AB with local phase lead (VAS inclusive) stabilisation. Simulations of comparisons between the JLH and the Class AB design I use shows that the JLH PNP input can still clip off or saturate on high level, high frequency signals whereas the inpust stages of the AB design I use do not.
Last edited:
Am I not happy to see a senior member like John Ellis chiming in once again.
Over the years I have been taking notes from the posts by stalwarts like Geoff Moss, Nigel Pearson et al, and of course, Mr Ellis too appears frequently in my notes. Of late, I have been once again trawling the nearly 10,000-strong Forum posts to try and gather the nuggets hiding therein for inclusion in my 'JLH Tributes Gateway' blog. I hope to post the distilled wisdom of the many capable designers and circuit honchos who contributed to the continued interest in the JLH Classic amplifier, and their many tweaks and mods and adaptations that made it easy for today's JLH fan to build one with ease and satisfaction.
On this note, permit me to take a minor issue with Mr Ellis when he says "...I am not surprised that the JLH fan club is keeping it going". Sir, on second thoughts you too will agree that the so-called 'fan club' is NOT taking any special efforts to somehow keep the thread alive and moving ahead. Rather, it is the inherent timeless goodness that has been there in the JLH Classic design that keeps it going. You are right when you mean the 'fan club' mostly consisting of jaded old cronies like me, for example, as many of us go for a repeat build of the JLH Classic, perhaps prompted by nostalgia.
NO sir, my experience here is exactly the opposite. It is the younger set (with keener hearing, and also owning modern hi-fi systems that cost, say, half a year's salary of a well-employed young man) who build and listen to, AND like, simple amplifiers like the JLH Classic, Death of Zen by Rod Elliot etc. It is the niggling queries of such a large group in my personal circle here that triggered my lazy soul into completing and posting the detailed blog posts on the JLH Classic amplifier. My stand is that, let them have access to all the original material, and the 'filtered wisdom' of the many stalwarts from the Forum before deciding to build it. The build is their's, as are the ears that do the final assessment. We will have absolutely nothing to do with the choice as diehard members of the old 'fan club'. Period.
On another angle, I understand that there are quite a few who are planning a build of the famous JLH MOSFET amplifier, which the Editors of ETI magazine termed 'simply the best amplifier' announcing it in the May, 1989, issue.
It is the perspicacity and sharp acumen of JLH that make these designs hold their own despite the passage of time and the progress of technology, to say nothing of the infusion of new design concepts by gifted modern designers.
Permit me to also share some of the findings by the 'group' here. Many have had excellent subjective results with the BEL (Bharat Electronics Limited)-manufactured 2N3055 metal TO3 devices ( which I believe was what Nigel Pearson too had found out, IIRC.) -- in the JLH Classic as well as the D-o-Z, which closely resembles the JLH amplifier in circuit and sound signature.
Another Indian-made output device is the C5200 made by CDIL, which gave mixed results, inviting more detailed investigation. ( I don't know if I should be happy, or otherwise when I say that a majority of us here are not Simulator jockeys. An able designer turning to the SIM to fine-tune his design is absolutely fine. But every Tom, Dick and his pal playing the fool with SIM software to up his ante in Forums is a deplorable sight IMHO. Here we choose to build and listen by ourselves, rather than let the software be the dubious arbiter.)
Our approach here is to make one small 'tweak' at a time in one channel, and then do a series of comparative auditions. Let me share the results of some of these forays.
The CDIL C5200 devices did yield good results when the bootstrap capacitor was retained, perhaps as that served to keep them 'reined in'. However, the use of a current source or a high value resistor fed from a stabilized DC supply with the input transistor gave the best noise rejection and overall stability. Some have even tried using an SMD LM317 for this purpose, but a simple transistor+zener does the job better. (Reference is to the 1969 single supply design.)
Another observation was that making up both C1 and C3 (Classic 1969) from a few smaller value caps and paralleling it with a high quality non-polar made a positive subjective difference. So did making the output capacitor a similar combo.
Some have reported better subjective results when they played around with the output node connection on the PCB, and others swear by the P-2-P implementations.
One thing is certain -- the JLH Classic design is not going to be forgotten anytime soon, fan club or not!
Over the years I have been taking notes from the posts by stalwarts like Geoff Moss, Nigel Pearson et al, and of course, Mr Ellis too appears frequently in my notes. Of late, I have been once again trawling the nearly 10,000-strong Forum posts to try and gather the nuggets hiding therein for inclusion in my 'JLH Tributes Gateway' blog. I hope to post the distilled wisdom of the many capable designers and circuit honchos who contributed to the continued interest in the JLH Classic amplifier, and their many tweaks and mods and adaptations that made it easy for today's JLH fan to build one with ease and satisfaction.
On this note, permit me to take a minor issue with Mr Ellis when he says "...I am not surprised that the JLH fan club is keeping it going". Sir, on second thoughts you too will agree that the so-called 'fan club' is NOT taking any special efforts to somehow keep the thread alive and moving ahead. Rather, it is the inherent timeless goodness that has been there in the JLH Classic design that keeps it going. You are right when you mean the 'fan club' mostly consisting of jaded old cronies like me, for example, as many of us go for a repeat build of the JLH Classic, perhaps prompted by nostalgia.
NO sir, my experience here is exactly the opposite. It is the younger set (with keener hearing, and also owning modern hi-fi systems that cost, say, half a year's salary of a well-employed young man) who build and listen to, AND like, simple amplifiers like the JLH Classic, Death of Zen by Rod Elliot etc. It is the niggling queries of such a large group in my personal circle here that triggered my lazy soul into completing and posting the detailed blog posts on the JLH Classic amplifier. My stand is that, let them have access to all the original material, and the 'filtered wisdom' of the many stalwarts from the Forum before deciding to build it. The build is their's, as are the ears that do the final assessment. We will have absolutely nothing to do with the choice as diehard members of the old 'fan club'. Period.
On another angle, I understand that there are quite a few who are planning a build of the famous JLH MOSFET amplifier, which the Editors of ETI magazine termed 'simply the best amplifier' announcing it in the May, 1989, issue.
It is the perspicacity and sharp acumen of JLH that make these designs hold their own despite the passage of time and the progress of technology, to say nothing of the infusion of new design concepts by gifted modern designers.
Permit me to also share some of the findings by the 'group' here. Many have had excellent subjective results with the BEL (Bharat Electronics Limited)-manufactured 2N3055 metal TO3 devices ( which I believe was what Nigel Pearson too had found out, IIRC.) -- in the JLH Classic as well as the D-o-Z, which closely resembles the JLH amplifier in circuit and sound signature.
Another Indian-made output device is the C5200 made by CDIL, which gave mixed results, inviting more detailed investigation. ( I don't know if I should be happy, or otherwise when I say that a majority of us here are not Simulator jockeys. An able designer turning to the SIM to fine-tune his design is absolutely fine. But every Tom, Dick and his pal playing the fool with SIM software to up his ante in Forums is a deplorable sight IMHO. Here we choose to build and listen by ourselves, rather than let the software be the dubious arbiter.)
Our approach here is to make one small 'tweak' at a time in one channel, and then do a series of comparative auditions. Let me share the results of some of these forays.
The CDIL C5200 devices did yield good results when the bootstrap capacitor was retained, perhaps as that served to keep them 'reined in'. However, the use of a current source or a high value resistor fed from a stabilized DC supply with the input transistor gave the best noise rejection and overall stability. Some have even tried using an SMD LM317 for this purpose, but a simple transistor+zener does the job better. (Reference is to the 1969 single supply design.)
Another observation was that making up both C1 and C3 (Classic 1969) from a few smaller value caps and paralleling it with a high quality non-polar made a positive subjective difference. So did making the output capacitor a similar combo.
Some have reported better subjective results when they played around with the output node connection on the PCB, and others swear by the P-2-P implementations.
One thing is certain -- the JLH Classic design is not going to be forgotten anytime soon, fan club or not!
Hi,
I have some questions about gain adjustment for the JLH 1969 circuit.
I’m working with a typical JLH 1969 kit. From what I understand, the 2.7k and 220 ohm resistors form a negative feedback network, which determines the gain.
Here are my questions:
I have some questions about gain adjustment for the JLH 1969 circuit.
I’m working with a typical JLH 1969 kit. From what I understand, the 2.7k and 220 ohm resistors form a negative feedback network, which determines the gain.
Here are my questions:
- By adjusting the ratio of these two resistors, I can change the gain and then simply readjust the bias, correct?
- I want to lower the gain, and ideally, I’d like to turn it into a buffer stage (or nearly). Are there any recommended resistor values for this? Also, are there any other components I might need to modify?
- Are there any downsides to lowering the gain?
Last edited:
Removing the 220 ohm (or the cap) gives you unity gain. No need to alter the bias as you still need the same current to deliver any given voltage to a given load. A simple R/C filter at the input of say 2k2 and 330pF might not go amiss although its more a theoretical than practical concern I suspect.
Do not alter the 2k7 as that will upset DC conditions.
Do not alter the 2k7 as that will upset DC conditions.
Well, almost. As I suspected, the input does not like high amplitude signals and there is phase inversion when you overdrive the unity gain version. There can be stability issues when you increase feedback but I don't see that in simulation.
Rather than removing the 220 ohms, I wouldsuggest disconnecting the 220 uF and reconnecting it to the output rail so that the 220uF-220 ohm pair are in parallel with the 2.7k feedback resistor. This maintains (nearly) the original open loop gain and will give a correspondingly lower distortion.
Omitting the 220 ohm resistor makes the 2700 ohm resistor the effective input loading, and will reduce the OLG because the input transconductance will now be only 1/2700 A/V instead of 1/220, approximately, as the internal emitter impedance is not included in this estimate.
The polarity of the capacitor will have to be swapped, though, so that it is correctly polarised (positive to the output).
The massive reduction in input stage gm with the 2700 ohm probably accounts for the stability steveu reports.
In my simulations there is some awful switching distortion when overdriving the unity gain option with either approach and the shunted RC version gives a high frequency component in the first few cycles while the unbypassed option gives a large second harmonic switch, but these were without an RC input filter which may help, along with a preamplifier clipping limit.
Omitting the 220 ohm resistor makes the 2700 ohm resistor the effective input loading, and will reduce the OLG because the input transconductance will now be only 1/2700 A/V instead of 1/220, approximately, as the internal emitter impedance is not included in this estimate.
The polarity of the capacitor will have to be swapped, though, so that it is correctly polarised (positive to the output).
The massive reduction in input stage gm with the 2700 ohm probably accounts for the stability steveu reports.
In my simulations there is some awful switching distortion when overdriving the unity gain option with either approach and the shunted RC version gives a high frequency component in the first few cycles while the unbypassed option gives a large second harmonic switch, but these were without an RC input filter which may help, along with a preamplifier clipping limit.
As a postscript to the previous post, the JLH shows some tendency to latchup which is the second harmonic switching I referred to.
This is not unlike the old Class AB with quasi-complementary outputs with a PNP/NPN output pair where the emitter of the pnp was connected to the output rail instead of the "emitter resistor" of the combination.
These effects can be mitigated by suitable current limiting, but I 've not explored this in any great detail for the JLH10.
This is not unlike the old Class AB with quasi-complementary outputs with a PNP/NPN output pair where the emitter of the pnp was connected to the output rail instead of the "emitter resistor" of the combination.
These effects can be mitigated by suitable current limiting, but I 've not explored this in any great detail for the JLH10.
Steve, could you please post your LTspice ASC files to help us lazy wannabe LTspice gurus. I think you have to rename the to *.txt for this forum.
OK. This is based on a post from S. G. Ingram some time ago. I have added a third version which is a net ~unity gain with no phase inversion (initially) by attenuating the input by 1/2 and then 2x gain.
THe problem I was seeing with overdrive is due to the PNP current being switched on hard, following the half-period when it has been off.
The emitter voltage takes some time to recover, and in the meantime, in the switching period, a high curent pulse is actually sucked out of the base (this is the negative-going phase) which prevents the output stage being fully driven.A base resistor of 1k reduces this problem, but it seems that the basic circuit is not ideal as a unity gain stage.
With the resistor, as shown in steveu's circuit the overload is mitigated and with normal listening levels should still perform adequately. Whether an input capacitor is also needed is dependent on whether you need it for RF filtering or for slew rate limiting.
Using the series RC feedback ararngement (220 ohm-220u in series, in parallel with 2700) also minimises the effect by allowing the maximum current to increase.
This is perhaps a case for a differential input stage with current mirror and possibly Miller stabilisation - but that would not be a JLH -10W (AKA JLH69).
The emitter voltage takes some time to recover, and in the meantime, in the switching period, a high curent pulse is actually sucked out of the base (this is the negative-going phase) which prevents the output stage being fully driven.A base resistor of 1k reduces this problem, but it seems that the basic circuit is not ideal as a unity gain stage.
With the resistor, as shown in steveu's circuit the overload is mitigated and with normal listening levels should still perform adequately. Whether an input capacitor is also needed is dependent on whether you need it for RF filtering or for slew rate limiting.
Using the series RC feedback ararngement (220 ohm-220u in series, in parallel with 2700) also minimises the effect by allowing the maximum current to increase.
This is perhaps a case for a differential input stage with current mirror and possibly Miller stabilisation - but that would not be a JLH -10W (AKA JLH69).
JLH1969 is a single ended current feedback amp. The NFB ratio solely depends on the 2.7K feedback resistor. 220 Ohm is only to set the voltage gain. It won’t affect the feedback ratio. Thus, the voltage gain won’t affect THD very much. Why do you want a unit gain JLH?
If the C in the feedback is your concern, just use a high quality capacitor.
If the C in the feedback is your concern, just use a high quality capacitor.
From simulations I have run, the JLH 10W design relies on distortion cancellation between the stages. This is obvious in the OPS where the push-pull configuration balances, to a large extent, second harmonic between the output pair.
However, the PNP input stage operates with a relatively large current swing, almost all of its DC value when lower gain output and driver transistors are used. The relatively high value resistors in the driver base and emitter reduce the nonlinearity of the exponential base-emitter responses and thus mostly, but not entirely, current drives the driver.
Any nonlonearity in the PNP stage current also flows in the feedback resistor, which is going to cause distortion which the feedback cannot compensate. When the 220 ohm is employed, this non-linear voltage is much reduced compared with across 2700 ohms.
Therefore, if the 220 ohm is removed, the THD will increase. If it is used to bypass the 2700 ohm resistor, with DC blocked by the 220uF capacitor, the THD will be lower even than the original circuit with the gain of around 11.
However, the 220 ohm, as mentioned in my earlier post, keeps the OLG high, and with a lower closed cloop gain (unity, or thereabouts) the bandwidth will increase and could give rise to some instability, although this might be cured with a local phase lead compensation capacitor and/or input "filter" capacitor to keep the input impedance at high frequences low - another aspect to check.
It's almost worth considering a 10:1 input attenuator and leave the original circuit as was!
However, the PNP input stage operates with a relatively large current swing, almost all of its DC value when lower gain output and driver transistors are used. The relatively high value resistors in the driver base and emitter reduce the nonlinearity of the exponential base-emitter responses and thus mostly, but not entirely, current drives the driver.
Any nonlonearity in the PNP stage current also flows in the feedback resistor, which is going to cause distortion which the feedback cannot compensate. When the 220 ohm is employed, this non-linear voltage is much reduced compared with across 2700 ohms.
Therefore, if the 220 ohm is removed, the THD will increase. If it is used to bypass the 2700 ohm resistor, with DC blocked by the 220uF capacitor, the THD will be lower even than the original circuit with the gain of around 11.
However, the 220 ohm, as mentioned in my earlier post, keeps the OLG high, and with a lower closed cloop gain (unity, or thereabouts) the bandwidth will increase and could give rise to some instability, although this might be cured with a local phase lead compensation capacitor and/or input "filter" capacitor to keep the input impedance at high frequences low - another aspect to check.
It's almost worth considering a 10:1 input attenuator and leave the original circuit as was!
I'm considering the JLH 1969 low-gain option because I want to use a 12AU7 tube as a preamp paired with a solid-state power amp. However, no matter how I calculate, the total system gain is way more than I need.
Given my current budget constraints, the JLH 1969 is one of the few kits I can afford (just $15(USD) for the main PCB and components!).
Thanks for the suggestions—I’ll give those setup a try and see how it turns out. (When will the retailer finally ship my order...😪)
By the way, has anyone tried using this crazy capacitor for Cout?
YAGEO/KEMET
Aluminum Organic Polymer Capacitors 35V 3200uF
PHH227JLP4320ME4
It doesn’t specify the DF/tanδ in datasheet, but aside from that, all the other specs seem perfect. I’m not sure if I should take the risk.
YAGEO/KEMET
Aluminum Organic Polymer Capacitors 35V 3200uF
PHH227JLP4320ME4
It doesn’t specify the DF/tanδ in datasheet, but aside from that, all the other specs seem perfect. I’m not sure if I should take the risk.
JLH 69 has a couple problems. Firstly, it uses the limited gain of the output transistors to set the idle current and limit the output current. So, using "better" transistors gets you into trouble. But the exact gain of any BJT is unpredictable, so you probably have to fiddle with the pull-up, bootstrap resistors (~100x2) to get an appropriate operating current. And as the OPs heat up, that current increases. 1.4 Amps on a 24V supply is appropriate for a 4 Ohm load (1.4x2x4 ~= 24/2) but it is a bit excessive for 8 Ohms, and the OPs are dissipating 1.4x12 =16.8 Watts each all the time, ie hot. Secondly, the THD will be about 0.1%, although it is largely 2nd harmonic which is a sound many people love.
JLH made some better amps later in life, and the 69 reflects the limited parts options available at that time, 55 years ago. Today, the 10W range is mostly cheap chip-amps, but you can roll your own with a handful of penny transistors. Even if you insist on a class-A amp, there are better options today if you are willing to DIY. Others have tried to tame the JLH69 bias problem. Attached is my cut on that idea:
PS: Note the bias transistors need to be larger due to excessive heat, but you get the idea.
JLH made some better amps later in life, and the 69 reflects the limited parts options available at that time, 55 years ago. Today, the 10W range is mostly cheap chip-amps, but you can roll your own with a handful of penny transistors. Even if you insist on a class-A amp, there are better options today if you are willing to DIY. Others have tried to tame the JLH69 bias problem. Attached is my cut on that idea:
PS: Note the bias transistors need to be larger due to excessive heat, but you get the idea.
Last edited:
The heating problem of the JLH has been considered by several folk. My solution did not use a current source as you show, because that limits the peak positive output swing. I used a transistor mounted on the heatsink to monitor the temperature and provide a variable curent shunt to bypass a fraction of the driver current in a current mirror. This could be tuned to give an almost stable quiescent curren but needed trimming pots to get the compensation correct, which would be different for different heatsinks etc.
However, these schemes detract from the original simplicity of the '69 circuit, and the distortion of the 4ohm version can be quite high which I measured as 0.2% with one build. It is to be noted that in the original article, JLH seems to use 15 ohms as his standard. Readers could infer that that the distortion would be similar (around 0.05%) for 8 or 4 ohms but JLH never explicitly said so as far as I recall.
Modern transistors with high gain are recommended, as these give the best frequency response and lowest distortion, except to say that near clipping the distortion increases again due to quasi-saturation. Which drives the operating voltage up if it is to be avoided, with the implication that you should not drive it into near clipping (depending on the current output the preferable minimuim Vce would be around 2-4V, for example) ... and an increase in power dissipation.
I would agree that there are better Class A designs to use today.
However, these schemes detract from the original simplicity of the '69 circuit, and the distortion of the 4ohm version can be quite high which I measured as 0.2% with one build. It is to be noted that in the original article, JLH seems to use 15 ohms as his standard. Readers could infer that that the distortion would be similar (around 0.05%) for 8 or 4 ohms but JLH never explicitly said so as far as I recall.
Modern transistors with high gain are recommended, as these give the best frequency response and lowest distortion, except to say that near clipping the distortion increases again due to quasi-saturation. Which drives the operating voltage up if it is to be avoided, with the implication that you should not drive it into near clipping (depending on the current output the preferable minimuim Vce would be around 2-4V, for example) ... and an increase in power dissipation.
I would agree that there are better Class A designs to use today.
Yep, with a simple bootstrap, the bias current is proportional to the power supply voltage. As long as the power supply is sloppy enough, the bias is kinda stable, even without the emitter resistors at output.