Mike, really good results on your measurements! Better than any of my JLH's. Which version is it, and what transistors are you using?
The 15003's are from a supplier in Sweden(electrokit), I seriously doubt they sell fakes, but would not bet my life on it either. At least the prices are for originals
I have a 69 version with 15003's from the same source, and that one measures much better.
The single supply JLH's that measure pretty good are on 19V laptop bricks (using floating ground with caps).
The dual supply/dual output version I measured using a linear CRC supply around +/- 16V. Supply smoothing (from memory) would be maybe 12mF/0,4ohm/18mF. Only one channel connected when measuring.
I can see the supply ripple 'grass' growing on the FFT when I crank up the Iq. Also approaching clipping there seems to 'grow some grass' around the input signal peaks and their harmonics in the FFT, so I stayed away from these conditions.
I'm getting more and more convinced switched supplies are the way to go, based on FFT spectrum of amps I measured. Especially amps with high constant current, class B and D seem to do ok with linear supplies. This is considering both cost and performance, -I like good and cheap . Downside could be reliability.
I'm slowly putting together a high biased class AB with a switched supply too. Maybe I will change my mind when I run into problems with switched..
The 15003's are from a supplier in Sweden(electrokit), I seriously doubt they sell fakes, but would not bet my life on it either. At least the prices are for originals
I have a 69 version with 15003's from the same source, and that one measures much better.The single supply JLH's that measure pretty good are on 19V laptop bricks (using floating ground with caps).
The dual supply/dual output version I measured using a linear CRC supply around +/- 16V. Supply smoothing (from memory) would be maybe 12mF/0,4ohm/18mF. Only one channel connected when measuring.
I can see the supply ripple 'grass' growing on the FFT when I crank up the Iq. Also approaching clipping there seems to 'grow some grass' around the input signal peaks and their harmonics in the FFT, so I stayed away from these conditions.
I'm getting more and more convinced switched supplies are the way to go, based on FFT spectrum of amps I measured. Especially amps with high constant current, class B and D seem to do ok with linear supplies. This is considering both cost and performance, -I like good and cheap
. Downside could be reliability. I'm slowly putting together a high biased class AB with a switched supply too. Maybe I will change my mind when I run into problems with switched..
Last edited:
I recognize the brightness observation. Here is a suggestion that many will think silly. Get some 0.5 mm bell wire solid core. Be gentle with it as rough handling causes it to go brittle. The few feet I use is about 0.3 ohm. The bass is fine and clarity better than typical. The sound is bringhter than usual except better as a result. I always suspected this was an overlooked quality of valve output transformers. Voice coils also. I discovered this by accident with Bose 901 in a church. To me bell wire sounds poor at first. Give it a day. If you hear what I do it might suit your taste. One friend describes it as hyper fast. It helps MP3 also.
I just did some maths. A class A plus AB of 1 watt A and 90 watts AB might dissipate 60 watts quiescent stereo. This assumes +/- 40 volts. The advantage being the power supply.could be quite modest. Complimentary pair outputs.
I have a 1.7 watt class D amp running low efficiency Teac lsx-8 speakers as almost invisible general purpose little monitors. 1 watt is enough 90 % of the time. 10% it isn't. It is a bridged design using 5 volts. 1.7 watts is at 10% thd. Call it 1 watt 0.06% or better up to 10 kHz. It sounds much like class A or a Quad 303.
A Class A + AB needs a high quality current limiter. A single ended PSU fuse and output capacitor will be ok as protection. As the feedback pair have considerable current gain the VAS can be modest.
JLH made an amplifier like this using bipolar mosfet feedback pair. I remember hearing one. I doubt I ever heard better. Maybe an amplifier using 2A3 valves was better. Better is hearing what the sound engineer thought lost in the mix.
I have a 1.7 watt class D amp running low efficiency Teac lsx-8 speakers as almost invisible general purpose little monitors. 1 watt is enough 90 % of the time. 10% it isn't. It is a bridged design using 5 volts. 1.7 watts is at 10% thd. Call it 1 watt 0.06% or better up to 10 kHz. It sounds much like class A or a Quad 303.
A Class A + AB needs a high quality current limiter. A single ended PSU fuse and output capacitor will be ok as protection. As the feedback pair have considerable current gain the VAS can be modest.
JLH made an amplifier like this using bipolar mosfet feedback pair. I remember hearing one. I doubt I ever heard better. Maybe an amplifier using 2A3 valves was better. Better is hearing what the sound engineer thought lost in the mix.
Just been rereading the JLH original articles on the Class A site. My assumptions aren't too wrong. Although tr1 should have the higher gain transistor the gain of both should be high. JLH says a MJ 3001 can be used as tr1. If so thd is 0.01%. He seems unwilling to say if he would recommend it. It seems to be a drop in substitution even though a Darlington.
He did say the amplifier much to his surprise could work up to 1.5 MHz. I measured similar. This means the output and input must be kept apart. One can always feed signal into the feedback capacitor if the preamp is capable.This should give slightly better sound. Also loop gain can be reduced by tweaking the resistor associated with the feedback capacitor. I could imag aine a JLH could work at unity gain.
He did say the amplifier much to his surprise could work up to 1.5 MHz. I measured similar. This means the output and input must be kept apart. One can always feed signal into the feedback capacitor if the preamp is capable.This should give slightly better sound. Also loop gain can be reduced by tweaking the resistor associated with the feedback capacitor. I could imag aine a JLH could work at unity gain.
Nigel, was there any particular 'format' of the bell wire, e.g. twisting it into a pair or just 'floating in fee space.
Kind regards
Mike
Mike, really good results on your measurements! Better than any of my JLH's. Which version is it, and what transistors are you using?
The 15003's are from a supplier in Sweden(electrokit), I seriously doubt they sell fakes, but would not bet my life on it either. At least the prices are for originals
The single supply JLH's that measure pretty good are on 19V laptop bricks (using floating ground with caps).
The dual supply/dual output version I measured using a linear CRC supply around +/- 16V. Supply smoothing (from memory) would be maybe 12mF/0,4ohm/18mF. Only one channel connected when measuring.
I can see the supply ripple 'grass' growing on the FFT when I crank up the Iq. Also approaching clipping there seems to 'grow some grass' around the input signal peaks and their harmonics in the FFT, so I stayed away from these conditions.
I'm getting more and more convinced switched supplies are the way to go, based on FFT spectrum of amps I measured. Especially amps with high constant current, class B and D seem to do ok with linear supplies. This is considering both cost and performance, -I like good and cheap
I'm slowly putting together a high biased class AB with a switched supply too. Maybe I will change my mind when I run into problems with switched..
Hi Rallyfinnen,
the design I used is the updated Geoff Moss ESL version, but with 4 output pairs. If you look at the data sheet for the MJ15003, (hfe vs Ic) using multiple pairs keeps each transisitor at a collector current below 1Amp, which is the most linear portion of the graph and so keeps their current gain at their optimal value.
I also use a capacitor multiplier in the same case as the amp itself and the mains transformer and initial smoothing are in a seperate case along with a soft start and mains D.C remover.
However, there are a few subtleties to observe: I split the design into 2 parts; the 'front end' and the power transistor connection, linked with a small cable. Also, on the front end PCB I made, I seperated its 0V (input reference and feedback) from the decoupling 0V. The power transistor board has 470uF and 0.1 caps per transistor very close to their connections. All the 0Vs, plus the speaker return, then went back to the 0V point on the Cap Multiplier.
I will try and take some photos to explain what I mean.
Cheers
Mike
Nigel, was there any particular 'format' of the bell wire, e.g. twisting it into a pair or just 'floating in fee space.
Kind regards
Mike
I don't twist it as I feel the wire is fragile. A friend insisted it sounded worse if worked too much. I suspect the 0R3 is helpful. Geoffrey Horn thought it an.optimum value if the power amplifier had a good damping factor. Critically under damped. It's incredibly good looking also. Pro audio friends always seem a little astonished at the sound I get.Theirs often sounds like a PA system at low volume. They ask me what cables I use. They think it will be Notdost or similar. I say Yacsahumi which is bell wire. By the way Notdost is a great substitute for Yaksahumi, it might be better if money not the question. I find multistrand cables can be harsh. From tests I have done it may not be the fault of the multistrand cable. All the same a single core helps. Naim NACA5 cable was a hybrid.
Not silly to me. I use doubled up 0.6mm solid core wire with my JLH, about 2.5m each side, in my case twisted but only loosely so, The dcr is factored into the loudspeaker alignment when necessary. Mine is ptfe/solid silver but I hear little difference with ptfe/copper. Very nice sounding cables at low cost.
I bought the cable to wire some super tweeters. I then used it to knock together a temporary hi-fi. It's years since I used exactly what I recommend. Out of curiosity I read the wire makers pdf. Sure enough copper plated aluminium.
I have moved to Dorset from Oxfordshire. That's a bit like moving back to when the JLH was designed. I like that except it reminds me that the future has things to offer. The first thing I did was build two workshops and a large office. Pure luck my bungalow has outbuildings. Next I measured the mains electricity quality. Absolutely superb. Only -52 dB harmonic at 150 Hz. Woodstock was about -38 dB which I thought unlikely to be bettered.A great start for good sound quality. The sine wave ballooned but without a flat top. The substation 50 metres away. Really ideal. Don't do this if unsure.
All my better things are in process. So it's Thorens TD 160 MK1 Rega R200 Japanese arm. Ortofon VMS 20 EII fitted with a Max Townsend parabolic stylus. Slightly modified NAD 3020 Arcam Alpha plus CD player with Mission 760i. The NAD is my JLH preamp. The system will play any music and has the hook. I was like going to the food cupboard and finding enough to make the perfect omelette.These parts have never been a system. My friend Malcolm found the NAD for me. He has a very expensive Naim system and a NAD one. He never would mix them up. I remember showing him a small pair of
SE valve amplifiers I built using the NAD as pre amp. It absolutely destroyed his main system due to showing every layer. The amplifier 6 watts 1% thd in exponential decay. 0.2% 1 watt.Two valves. No negative feedback not even local where it could be. Damping factor 3. It was a very low distortion sound. Like really clean windows. It has positive feedback of about 50%.
I had a bash at a JLH which was too scruffy. I would like to do a smps version just for the challenge. As the power draw of class A is more constant than AB as long as the spectrum analyser sees no problems there shouldn't be any. That allows a simple RC filter. Often the smps will go into protection briefly with that load. Likely it will cope. If 1 ohm that is 27 amps instantaneous. One could use a ten ohm resistor which is shorted by a relay when the right moment. 1 ohm 2200 uF perhaps? A 24 V relay will cope with 27 volts. It will come on at 16 volts. CR 10 x 2200/1000000 = 22 mS. 1CR being about when the relay will click. The relay should add 10 mS. No fancy timer required.
There is one advantage a NAD has over a JLH. It can briefly get a litre out of a pint pot in terms of wattage output. This helps small modern speakers. The sound distorts in a way that suggests a larger speaker. Ironically the class A often sounds a bit more distorted as the distortion is either there or not there. The advantage of class A is an smps should be possible. Modulation of smps and class AB is a horror story.
I have moved to Dorset from Oxfordshire. That's a bit like moving back to when the JLH was designed. I like that except it reminds me that the future has things to offer. The first thing I did was build two workshops and a large office. Pure luck my bungalow has outbuildings. Next I measured the mains electricity quality. Absolutely superb. Only -52 dB harmonic at 150 Hz. Woodstock was about -38 dB which I thought unlikely to be bettered.A great start for good sound quality. The sine wave ballooned but without a flat top. The substation 50 metres away. Really ideal. Don't do this if unsure.
All my better things are in process. So it's Thorens TD 160 MK1 Rega R200 Japanese arm. Ortofon VMS 20 EII fitted with a Max Townsend parabolic stylus. Slightly modified NAD 3020 Arcam Alpha plus CD player with Mission 760i. The NAD is my JLH preamp. The system will play any music and has the hook. I was like going to the food cupboard and finding enough to make the perfect omelette.These parts have never been a system. My friend Malcolm found the NAD for me. He has a very expensive Naim system and a NAD one. He never would mix them up. I remember showing him a small pair of
SE valve amplifiers I built using the NAD as pre amp. It absolutely destroyed his main system due to showing every layer. The amplifier 6 watts 1% thd in exponential decay. 0.2% 1 watt.Two valves. No negative feedback not even local where it could be. Damping factor 3. It was a very low distortion sound. Like really clean windows. It has positive feedback of about 50%.
I had a bash at a JLH which was too scruffy. I would like to do a smps version just for the challenge. As the power draw of class A is more constant than AB as long as the spectrum analyser sees no problems there shouldn't be any. That allows a simple RC filter. Often the smps will go into protection briefly with that load. Likely it will cope. If 1 ohm that is 27 amps instantaneous. One could use a ten ohm resistor which is shorted by a relay when the right moment. 1 ohm 2200 uF perhaps? A 24 V relay will cope with 27 volts. It will come on at 16 volts. CR 10 x 2200/1000000 = 22 mS. 1CR being about when the relay will click. The relay should add 10 mS. No fancy timer required.
There is one advantage a NAD has over a JLH. It can briefly get a litre out of a pint pot in terms of wattage output. This helps small modern speakers. The sound distorts in a way that suggests a larger speaker. Ironically the class A often sounds a bit more distorted as the distortion is either there or not there. The advantage of class A is an smps should be possible. Modulation of smps and class AB is a horror story.
Very interesting....a fellow Dorset dweller too. It's got me curious about putting a scope on the mains supply as I fitted a dedicated Hi-Fi spur a while back which brought about an audible improvement far greater than I had expected.
My Dorset electricity is better than many low distortion inverters. At 250 Hz it is -40 dB which is still very good. My Garrard 401 runs quietly due to this. The motor is also a filter as it reaches peak efficiency at 60 Hz. 250 Hz will not cause much vibration. As in music even harmonics are less likely to cause problems as to an extent it's more in phase. Live for today because electric cars and low cost generation may ruin it.
The more I try to model the JLH current sink the more confused I get. It's weird it works. I get that the collector output modulates the current sink. I get that this working make for equal and opposite distortion curves. I further understand it can cut down heat. What I don't get is the stability I got when I built it. JLH said working in the nuclear industry taught him to remove as many parts as he could. Each part is a new risk. Oh how I wish that type of genius was common. People think complex is better. It isn't , it sometimes is easier.
I think I will do an an an analysis of a cheap smps and try RC filtering. I think it will be ok as the smps charging should be ≥ 20 !S which is time enough to soften the starting current. It may not be if the smps has other priorities. If positive I will post the results.
The more I try to model the JLH current sink the more confused I get. It's weird it works. I get that the collector output modulates the current sink. I get that this working make for equal and opposite distortion curves. I further understand it can cut down heat. What I don't get is the stability I got when I built it. JLH said working in the nuclear industry taught him to remove as many parts as he could. Each part is a new risk. Oh how I wish that type of genius was common. People think complex is better. It isn't , it sometimes is easier.
I think I will do an an an analysis of a cheap smps and try RC filtering. I think it will be ok as the smps charging should be ≥ 20 !S which is time enough to soften the starting current. It may not be if the smps has other priorities. If positive I will post the results.
SMPSs come in several flavours and it depends what you mean by "cheap SMPS". There are common power bricks and wall-warts at 4.5 -19 or even 24V that are often pressed into audio duty. They're cheap and cheerful but not good at filtering switching noise which I assume you'd prefer not to have to deal with by post-filtering .
Then, there are commercial metal clad modules intended for LED lighting and similar low voltage applications. These tend to be cleaner for a given power rating. There are also complete, metered and adjustable power supplies for safe and convenient use on the bench. These may be available in various quality from trash dressed up in nice cases to very quiet, fully protected and excellent quality that you could use to power almost any audio circuit.
With JLH'69 and variations as a test bed, I admit that I can't always hear the differences but they are certainly there in the noise seen on the 'scope.
Then, there are commercial metal clad modules intended for LED lighting and similar low voltage applications. These tend to be cleaner for a given power rating. There are also complete, metered and adjustable power supplies for safe and convenient use on the bench. These may be available in various quality from trash dressed up in nice cases to very quiet, fully protected and excellent quality that you could use to power almost any audio circuit.
With JLH'69 and variations as a test bed, I admit that I can't always hear the differences but they are certainly there in the noise seen on the 'scope.
Hi Ian. When I get a moment I will test a Meanwell low cost metal box type. They always seem to me to be good value and look quite well made. Sold by many suppliers here in the UK. When I last measured one it easily met it's modest spec. Usually they have slightly more maximum voltage than stated. 28 volts possible. The unit I have is 50 VA if I remember. Also if memory is right it is fine at 33 VA but not at 48 VA which it won't say no to supplying. 55 VA likely maximum. There is a well designed current limiter with Led to indicate that. Quiet handy if the JLH has an error.
One thing I tried that seemed certain to work was a DC PSU before the Meanwell smps. 220 uF 500 V. This became 5 times the original 47 uF when added together. Rectifiers 1N4007. It had no obvious effect. Like many smps Meanwell works fine on DC. Mine was circa 320 VDC. I will try that again as I might see an upgrade. Very dangerous if uncertain of the the dangers. It seems if that is right the problems are in common mode and are 50 Hz modulated along with the switching noise.
Whilst the Meanwell smps almost certainly meets it's obligations concerning noise into the mains electricity I doubt it will be harmless to other equipment. A partial solution for that is battery power to preamps and DACs. Partial as the 0V is also a signal reference point.
One thing I tried that seemed certain to work was a DC PSU before the Meanwell smps. 220 uF 500 V. This became 5 times the original 47 uF when added together. Rectifiers 1N4007. It had no obvious effect. Like many smps Meanwell works fine on DC. Mine was circa 320 VDC. I will try that again as I might see an upgrade. Very dangerous if uncertain of the the dangers. It seems if that is right the problems are in common mode and are 50 Hz modulated along with the switching noise.
Whilst the Meanwell smps almost certainly meets it's obligations concerning noise into the mains electricity I doubt it will be harmless to other equipment. A partial solution for that is battery power to preamps and DACs. Partial as the 0V is also a signal reference point.
Imgur: The magic of the Internet
Hope the link works?
It shows a Meanwell RS50 - 24 is not too bad.
1.7 Amps. 28 V. After a very simple 1R 1000 uF filter about -108 dB below 28Vdc which is reasonable. ( 28 V by coincidence is about 28 dB above 1 dBV ).
Hum was a strange 78 Hz ! It was almost 80 dB below 28Vdc which for the type of PSU is not bad.
A German Voltcraft linear PSU was used as reference. It was surprisingly similar and to be honest not wonderful. It uses 2N3055. 20dB better at 50Hz.
The best PSU I made was - 130 dB. That was using LD1084 with many tweaks. A very standard 7824 would be about -112 dB. Hum is very tricky when any PSU. Exact ground point mostly. TNT Audio on the LM317 a good source.
To use this PSU I would like to get the strange hum down. Hope this helpful. It's my new house so finding things was fun.
Hope the link works?
It shows a Meanwell RS50 - 24 is not too bad.
1.7 Amps. 28 V. After a very simple 1R 1000 uF filter about -108 dB below 28Vdc which is reasonable. ( 28 V by coincidence is about 28 dB above 1 dBV ).
Hum was a strange 78 Hz ! It was almost 80 dB below 28Vdc which for the type of PSU is not bad.
A German Voltcraft linear PSU was used as reference. It was surprisingly similar and to be honest not wonderful. It uses 2N3055. 20dB better at 50Hz.
The best PSU I made was - 130 dB. That was using LD1084 with many tweaks. A very standard 7824 would be about -112 dB. Hum is very tricky when any PSU. Exact ground point mostly. TNT Audio on the LM317 a good source.
To use this PSU I would like to get the strange hum down. Hope this helpful. It's my new house so finding things was fun.
It seems steady. I know the analyser well enough to trust it. The linear supply gave 50 and 100 Hz as usual. The linear PSU was the one I used for my JLH. It's 2.5 amps up to 30 volts. It has very reasonable 50/100 Hz hum and disappointing hiss.
The most interesting thing about the 78 Hz is it is more sine than sawtooth. 78 may be 75 Hz as the analyser reading often isn't easy to define. The rough waveform suggested 5mV RMS. Sawtooth being harder on the ears. Seeing as distortion is -60 dB or better we need-100 dB from the PSU. All the same it's much better than I dared hope. It was nearly at it's limit. Using two smps should be very good. With simple RC filtering very little switch noise was seen.
The most interesting thing about the 78 Hz is it is more sine than sawtooth. 78 may be 75 Hz as the analyser reading often isn't easy to define. The rough waveform suggested 5mV RMS. Sawtooth being harder on the ears. Seeing as distortion is -60 dB or better we need-100 dB from the PSU. All the same it's much better than I dared hope. It was nearly at it's limit. Using two smps should be very good. With simple RC filtering very little switch noise was seen.
Just a word of warning on the Meanwell smps. As best I remember it will not allow earth/ground to connect to 0 Volts. I will check that and see if it can be tweaked. It's just about possible to completely float the system. One item that is grounded will knock out the PSU. I am told this isn't typical.
Imgur: The magic of the Internet
Hope this opens. It would never get NEMCO certification. If built into an enclosure that might be OK. I have tried to make it as safe as possible. In the UK we can have 1, 2 amp plug fuses if we choose. It needs something. This is a junk box special. Good enough to prove the value. I always called ideas like this Magnox after our nuclear power stations when down the pub. Extreme caution if you try it.
The hum is 75 Hz ( 150 Hz /2 ? ).
The improvement is small in statistical terms. I notice the noise floor is nicer. One might imagine it would be much better. The internal is 47uF I think .
I would use a single 100 VA smps as I am not convinced joining 2 x 50VA via 0V is ideal.
Reading up, many smps can not be connected to mains earth/ground. This means your hi fi will have to float. If you own a Linn LP12 disconnect arm to chassis earth. Not the LP12 chassis to ground. Most things will be fine.
Thought it my duty to do this test as too much is written without testing. I do suspect I would do this as I am convinced this DC modification prevents some noise back into the mains. If at 115 V the resisitors need to be 7 watt types, the capacitors could be lower voltage types. One small thing said to me is can the smps rectifier cope with some diodes unused. At this low current it seems likely it's OK. 470 nF class X2 is a certificated part ( Typically 275VAC and 1500 VDC ). The 2 x 150 K meet IEC requirements of 5CR< 2 seconds. The 68K for the big caps is 330uF x 136K at 225 seconds. If made faster it will get hot.
As they say " do not try this yourself ". If I caught my son doing this I would be worried. Fire and electricution. Things going bang also.
Hope this opens. It would never get NEMCO certification. If built into an enclosure that might be OK. I have tried to make it as safe as possible. In the UK we can have 1, 2 amp plug fuses if we choose. It needs something. This is a junk box special. Good enough to prove the value. I always called ideas like this Magnox after our nuclear power stations when down the pub. Extreme caution if you try it.
The hum is 75 Hz ( 150 Hz /2 ? ).
The improvement is small in statistical terms. I notice the noise floor is nicer. One might imagine it would be much better. The internal is 47uF I think .
I would use a single 100 VA smps as I am not convinced joining 2 x 50VA via 0V is ideal.
Reading up, many smps can not be connected to mains earth/ground. This means your hi fi will have to float. If you own a Linn LP12 disconnect arm to chassis earth. Not the LP12 chassis to ground. Most things will be fine.
Thought it my duty to do this test as too much is written without testing. I do suspect I would do this as I am convinced this DC modification prevents some noise back into the mains. If at 115 V the resisitors need to be 7 watt types, the capacitors could be lower voltage types. One small thing said to me is can the smps rectifier cope with some diodes unused. At this low current it seems likely it's OK. 470 nF class X2 is a certificated part ( Typically 275VAC and 1500 VDC ). The 2 x 150 K meet IEC requirements of 5CR< 2 seconds. The 68K for the big caps is 330uF x 136K at 225 seconds. If made faster it will get hot.
As they say " do not try this yourself ". If I caught my son doing this I would be worried. Fire and electricution. Things going bang also.