I think 5 watts 6 ohms is a better option than people might think. An input sensetiveity of 300 mV might be helpful. 50 watts is noticeably better, 10 watts not. I have a 1.5 watt amplifier with JPW Minims. It sounds much better than it should. It is in my workshop. It can go delightfully loud. 1.5 watts at 10% THD! 1 watt 0.05%. The JPW is a clone of another make with slightly less good details. This combination seems to elevate them.
In my current JLH 1969 I use transformer regulated power supply, individual for each channel. The transformer is EI type from older Yamaha amp rated 180 Watts. With 33 volts DC and 1,2 amps current it gets very warm after long operation. After Shottky bridge rectification, there are 6800uF capacitor cans and linear regulator with LT1084. Then about 2500uf more caps and a capacitor multiplier with 2sc5200 plus some more hundreds uf capacitance. After this double filtration the supply voltage is 26 volts, very clean, I cannot see any ripple with oscilloscope.
Yesterday I tried to power the circuits from two cheap Chinese SMPS power supplies. I powered the boards including the 5200 multipliers and skipped the 1084 and capacitor cans. What I heard from the speakers had noticeably more dynamic and rich sound, with more alive presentation. There was not so much clearance on highs compared to regulated supply, but overall I was very impressed with the sound.
The SMPS configuration has some drawbacks. There are “needle” spikes in supply at approx. 60kHz of about 200mV. Also, the earthing becomes very difficult and the amp oscillates an HF easily.
So I want to ask in this thread, if anyone had similar experience with SMPS? Is there any scientific explanation to why it sounds more dynamic with SMPS supply? Can it be that in my current transformer supply I use too little capacitance, or the transformer is too small?
Any ideas appreciated.
Yesterday I tried to power the circuits from two cheap Chinese SMPS power supplies. I powered the boards including the 5200 multipliers and skipped the 1084 and capacitor cans. What I heard from the speakers had noticeably more dynamic and rich sound, with more alive presentation. There was not so much clearance on highs compared to regulated supply, but overall I was very impressed with the sound.
The SMPS configuration has some drawbacks. There are “needle” spikes in supply at approx. 60kHz of about 200mV. Also, the earthing becomes very difficult and the amp oscillates an HF easily.
So I want to ask in this thread, if anyone had similar experience with SMPS? Is there any scientific explanation to why it sounds more dynamic with SMPS supply? Can it be that in my current transformer supply I use too little capacitance, or the transformer is too small?
Any ideas appreciated.
Not directly with a JLH, nor mains driven SMPS.
But I have used a power inverter to drive a 50W amp from 12V battery. Switching spikes got into the amp. All it needed was a couple of RF type power chokes in the supply lines to suppress the switching noise. (Plus another smoothing cap on the other side - what used to be called "pi" filter). Not seen many of those for a long time, but still effective.
RF choke(s) could be wound with a ferrite core + bobbin with suitable wire (often parallel wires of a thinner gauge are easier to wind than one thick wire) and air gap to avoid saturation. You can get cores pre-ground with gaps - just need to determine the inductance needed.
Or you could look into commercially available chokes. I would recommend avoiding any open magnetic cores to prevent or minimise inductive coupling from the external field.
At least something to try?
But I have used a power inverter to drive a 50W amp from 12V battery. Switching spikes got into the amp. All it needed was a couple of RF type power chokes in the supply lines to suppress the switching noise. (Plus another smoothing cap on the other side - what used to be called "pi" filter). Not seen many of those for a long time, but still effective.
RF choke(s) could be wound with a ferrite core + bobbin with suitable wire (often parallel wires of a thinner gauge are easier to wind than one thick wire) and air gap to avoid saturation. You can get cores pre-ground with gaps - just need to determine the inductance needed.
Or you could look into commercially available chokes. I would recommend avoiding any open magnetic cores to prevent or minimise inductive coupling from the external field.
At least something to try?
Last edited:
I have heard this also. A simple RC filter like 0R1 and 10 000 uF or even 1 uF can clean up an smps. The range is to say experiment. Some smps won't like big capacitors. Voice quality might improve. Like old valve amplifiers this capacitor is not critical on ripple current. They had red and yellow terminals with black common. A resistor between red and yellow to be the π filter. I think yellow was the output side.
Hypex Smps sounds very dynamic. It only measures moderately well. It is slightly bright. My own PSU after some experiments sounds to have more tonal range on the Hypex amplifier.
Hypex Smps sounds very dynamic. It only measures moderately well. It is slightly bright. My own PSU after some experiments sounds to have more tonal range on the Hypex amplifier.
PS- you need a choke which is able to operate at D.C. currents appropriate to your amp. Probably 2A rating is adequate for the standard JLH if used one per channel. The chokes I constructed were about 350uH as I recall.
At this point I am trying to figure out the nature of this phenomenon. Actions to improve the quality of SMPS voltage is the next step if the decision is made to switch to SMPS.
I can only imagine that large capacitors that work in regulated PSU take time to charge up. So SMPS operating at higher frequencies and with relatively small capacitance can deliver "faster" current on impulses. This is just guessing.
I can only imagine that large capacitors that work in regulated PSU take time to charge up. So SMPS operating at higher frequencies and with relatively small capacitance can deliver "faster" current on impulses. This is just guessing.
I have measured SMPS so many times. Each seem to have their own solutions as to filtering. I suspect SMPS are in someways rather good and that's the reason they can surprise. There is very little transformer loss. Also 320 VDC side might be doing a good job. An E&I 1970's type transformer twice the size you think you need most likely is better. It will cost > £200 I guess.
You can run most SMPS on DC ( 140 to 320 VDC ). I am not sure if it helps. It should. Some say it could overheat a diode. I doubt that. It looks different on the spectrum analyser.
I have been testing IRF640 N type enhancement MOS FET today. Some very interesting observations. It's on resistance is 0R15 so it isn't like audio FET like BUZ 900 ( circa 1 R ). One myth is no thermal runaway. Not true and rather rapid especially if the 0R33 source resistor removed. One sample got so hot it unsoldered itself. It still works! I sort of doubt a T0220 bipolar would after that. I nearly gave up, glad I didn't. Doubtless I could do it better.
I clamped the seemingly OK heatsink in a small vice ( 2.5 inch ). That worked. Rock solid. Including 0R33 0.5A 3.7V 1 amp 4.0V gate voltage ( 4 - 0.33 V for 0R33 ). I wouldn't have said any of this except both sample FET were almost identical. Douglas Self says 2 to 5V variation. I shall use these as they are workable. What I like is they need very little current to work ( 1.3 mA here ). I will have to see how they work at > 10 kHz. I am tempted not to bother and use my ears. 1 watt 10 kHz is more than music needs. This is a simple constant current source. To have the current source drop at > 20 kHz might be a good thing.
I spent an hour cleaning this keyboard. I must be bonkers as I took it apart. My son got loads of food in it. In this time clean is good. Labtec cheapy. Very well made!
I read that the difference between science and religion is that science will in the end change it's mind if wrong. I find the JLH is a science thing with plenty of mystery.
You can run most SMPS on DC ( 140 to 320 VDC ). I am not sure if it helps. It should. Some say it could overheat a diode. I doubt that. It looks different on the spectrum analyser.
I have been testing IRF640 N type enhancement MOS FET today. Some very interesting observations. It's on resistance is 0R15 so it isn't like audio FET like BUZ 900 ( circa 1 R ). One myth is no thermal runaway. Not true and rather rapid especially if the 0R33 source resistor removed. One sample got so hot it unsoldered itself. It still works! I sort of doubt a T0220 bipolar would after that. I nearly gave up, glad I didn't. Doubtless I could do it better.
I clamped the seemingly OK heatsink in a small vice ( 2.5 inch ). That worked. Rock solid. Including 0R33 0.5A 3.7V 1 amp 4.0V gate voltage ( 4 - 0.33 V for 0R33 ). I wouldn't have said any of this except both sample FET were almost identical. Douglas Self says 2 to 5V variation. I shall use these as they are workable. What I like is they need very little current to work ( 1.3 mA here ). I will have to see how they work at > 10 kHz. I am tempted not to bother and use my ears. 1 watt 10 kHz is more than music needs. This is a simple constant current source. To have the current source drop at > 20 kHz might be a good thing.
I spent an hour cleaning this keyboard. I must be bonkers as I took it apart. My son got loads of food in it. In this time clean is good. Labtec cheapy. Very well made!
I read that the difference between science and religion is that science will in the end change it's mind if wrong. I find the JLH is a science thing with plenty of mystery.
Last edited:
Were you able to set a desired bias current and the output to 0 Volts?
I built the version Mamak referred to and did not encounter any issues. In fact I'm listening to it right now.
Here is the calibration procedure, from Geoff Moss, who designed this modification to the JLH and the PCB pattern I used. He is also the creator of the 'Class A Amplifier Site".
If you are building the high power version with parallel output transistors, with VR2 set to maximum (50R), I would expect each output transistor to be passing about 300mA (ie there will be about 30mV across each of the 0R1
resistors) when the supply rails are at +/-10V. This does not vary very much when the rails are +/-22V. The actual current depends on the gain of the individual output transistors. The Q7/Q8 ccs will supply a current to the bases of the output transistors which is equal to the Vbe of Q7 (~0.6V) divided by the value of VR2, so with VR2 at 50R the ccs current will be ~12mA. This gives 3mA into the base of each output transistor and if it's gain is 100 (a typical value) the collector current will be ~300mA.
As for initial setting-up, before power is first applied I would adjust both VR1 and VR2 for maximum resistance. After power is applied, VR1 is adjusted to set the output dc offset to near zero and the quiescent current (and
current sharing) checked by measuring the voltages across the 0R1 resistors. If all seems well, VR2 is adjusted to raise the quiescent current to, say, 1A. After a short period to allow the transistors to warm up, the output dc
offset is reset to near zero by adjusting VR1. Back to VR2 again and raise the quiescent current to, say, 2A. Wait a while and reset the output offset.
Unfortunately, the adjustments of VR1 and VR2 interact and so the setting of the quiescent current and output dc offset is an iterative process but after about four steps you should be able to get the current and offset somewhere
near to the required values. At this point, it is necessary to let the amp idle for an hour or so in order to ensure that thermal equilibrium has been reached and that the amp is at its normal operating temperature. The
quiescent current and output dc offset are now trimmed to the required values. These are best rechecked after a further settling period (say about 30 minutes).
It's probably worth a thread of it's own: what are the tricks/traps of SMPS use for audio amplifiers.
I've had mixed success but the guitar crowd report some work pretty well (if without the growl-inducing 100Hz/120Hz supply rail ripple when amps move from class A to class AB)
Which I have to say, is one of the unsung (in the hifi forums) benefits of Class A - no IMD from power rail ripple / sag when the watts go up.
I'm also using SMPS and floating GND on my JLH, and it sounds great. I have abt 15mF on the supply, and the SMPS does not complain. I think I measured abt 2mVp-p ripple at 66kHz on the output. No oscillation problems. FFT (noise) is very clean in the audio band. I think that is a factor for sound quality.
Current variations are relatively small in class A compared to AB, so 'sag' should be less significant?
Last edited:
I tend to know about Smps being fussy about connecting to signal ground. Meanwell Smps in a metal cage like RS50-24 isn't happy having 0V connected to mains earth. Also additional output filtering can upset it's protection circuit. It's worth persisting as it does have a good basic spec. I do find a good conventional PSU can have noise at -130 dB and hum at -110 dB If 1.5 amps. Smps seem to be -85 dB and -65 dB at the switching frequency. The later can very easily be reduced to about -80 dB. It does seem if the noise is white it is acceptable. It might even compliment the sound of a 1960s amplifier that usually is a bit magnolia.
I was talking typical. The cage tends to be an indication of better quality. They are low cost. The problem with Smps is finding a place for the noise to go to. Just connecting a capacitor sometimes does very little. I can't remember if my -85 dB is dBV. I will see is I can find the measurements. You can add 30 dB improvement if dB absolute. I have seen -111 dBV from a LD1084 regulator at 1.5 amps. It requires a ground plane PCB to get that. There is a difference if using 100 VAC. You might be able to do other stuff if so. The yellow building site transformer rented to try it might be an idea. Very cheap to buy. 3000 watts cheap enough.
This is a 10 000 uF capacitor on a delayed start with a small resistor ( 0R33 ? ) on a Meanwell SMPS at 45 watts. This simple idea gives about 1/10000V rms noise at 67 kHz. 2.2uV Vpk to pk without seems very likely ( - 65 dBV and spread ). For such a simple modification it's not bad. Notice ripple is OK but not as good as the LD1085. 10 VDC adds 20 dB. 20 VDC 26 dB an so on. Notice the hum is 75 Hz. It's still about 90 dB below 26 V.
It took days of work to get the LD1084 this good. It isn't possible to do this with switchmodes. Notice the noise floor is - 111 dBV or better. It is - 84 dBV for the SMPS. About - 110 dB absolute.
I was wondering if or when you were going to say more about your SMPS adventure, Nigel, considering your moving house and all that goes with it. I'm impressed that you managed to get the residual noise down so low with the LT regulator too. Do you have any tips from what you did to get the noise floor down that low?
The most important thing I did was read a text by TNT audio on the LM317. The LD1084 is very similar only low drop out. Extremely stable. I also have a crowbar comparator using LM324 in the noise figures. Useful extra current also.
The ripple is very tricky. It's basically a ground plane and the capacitors on a T shaped section. I nearly pulled my hair out when the first PCB tried with a ground plane. It was rubbish. Out of desperation I removed the capacitors and built them on solid wire in the T shape. Instant success. To my delight the MK2 PCB was better still. 20 dB from memory bad to best.
7812 LM317 LD1084 all seem to be a LM741 with bandgap and pass transistor. LM317 has less hiss because the gain resistance can be noise shunted. 10 uF is usual 47 uF perhaps 2 dB better. A clone regulator using NE5532 would be better.2.5 nV per ✓Hz if paralleled. LM741 18 dB worse. The inside op amp in LM317 is generic.
LM324 is ideal in this use. The reason LM324 is still around is it has useful ways of working. I was paid to do this so no circuits.
The LD1084 is a superb JLH device. Class A suits it's compromrises.. It will do exactly what a capacitance multiplier will do if thought of as much the same thing.
The ripple is very tricky. It's basically a ground plane and the capacitors on a T shaped section. I nearly pulled my hair out when the first PCB tried with a ground plane. It was rubbish. Out of desperation I removed the capacitors and built them on solid wire in the T shape. Instant success. To my delight the MK2 PCB was better still. 20 dB from memory bad to best.
7812 LM317 LD1084 all seem to be a LM741 with bandgap and pass transistor. LM317 has less hiss because the gain resistance can be noise shunted. 10 uF is usual 47 uF perhaps 2 dB better. A clone regulator using NE5532 would be better.2.5 nV per ✓Hz if paralleled. LM741 18 dB worse. The inside op amp in LM317 is generic.
LM324 is ideal in this use. The reason LM324 is still around is it has useful ways of working. I was paid to do this so no circuits.
The LD1084 is a superb JLH device. Class A suits it's compromrises.. It will do exactly what a capacitance multiplier will do if thought of as much the same thing.
Last edited:
Hi everyone, first time posting in this thread. Looking to build the original JLH 1969 amp and looking for quality PCB. Would you recommend just getting random off ebay or any other options available (here via GB etc)? Not interested much in the kit, probably going to source parts separately or use some leftovers from other projects.
Ian. As to moving house. This house is smaller yet has more gardens and more workshops. I finally fixed the plumbing two plumbers wouldn't touch. It's a most remarkable little oasis and suntrap. It's just as good for tranquilly to do work as Woodstock was. It makes lockdown easier.
Going back to my smps drawing. It's a 24 volt relay and perhaps 10R resistor. 10 X 10 000 gives about 62% voltage in 0.1 seconds which often defeats the smps current limiter nicely. If the relay needs 80% charge it's a little longer. That leaves the other resistor in place. 0R33 is about right. 0R1 perhaps for class AB. I find the 10 000uF does one thing. It makes the noise less spiky. I speculate that when a PSU is this reasonable it might even sound nice. The BBC found hiss improved FM sound quality if subtle.
When building PSU all one can do is experiment and use a spectrum analyser. To an extent any book you read will be wrong. This is because the exact best place for reference earth is not always obvious.
Going back to my smps drawing. It's a 24 volt relay and perhaps 10R resistor. 10 X 10 000 gives about 62% voltage in 0.1 seconds which often defeats the smps current limiter nicely. If the relay needs 80% charge it's a little longer. That leaves the other resistor in place. 0R33 is about right. 0R1 perhaps for class AB. I find the 10 000uF does one thing. It makes the noise less spiky. I speculate that when a PSU is this reasonable it might even sound nice. The BBC found hiss improved FM sound quality if subtle.
When building PSU all one can do is experiment and use a spectrum analyser. To an extent any book you read will be wrong. This is because the exact best place for reference earth is not always obvious.
Noise reduction.
Finesse Voltage Regulator Noise! |
https://www.edn.com/simple-circuits-reduce-regulator-noise-floor/
Simple Voltage Regulators Part 1: Noise - [English]
From these I discovered what to do and retain the simplicity with linear regulators. Alas these solutions dont work for SMPS. However -85 dBV for the example I show is not worse than LM7812. If for example we say 12V is 21.6 dB higher it depends where we reference from. If 20V and 2.2mV pk to pk we can say for simplicity 1mV rms noise. That's -60 dBV. 20V is + 26 dB ( 20 = 20 log 1.30103 or 10x2 = 20 dB + 6 dB, Log2 = 0.30103, Log10 = 1 ). We can call that 86 dB below 20V. Or do it this way 20 000 mV/1mV 20 x 4.4 = 86 dB. dBV is a good way of saying how good a thing is.
When designing always check your input ripple and output ripple are to book spec. If not find out why. It should meet minimum spec. Star earths etc help. PSU's work like musical devices in a bad way. If you like a flute with the wrong holes in use. Ripple harmonics. It is usual for 100/120 Hz to be the big one.
Why I like LD1084/85 is the results should be possible without using an oscilloscope. I hate doing that. I know most people don't have one.
-144 dBV is a target.
Unregulated Power Supply Design
Finesse Voltage Regulator Noise! |
https://www.edn.com/simple-circuits-reduce-regulator-noise-floor/
Simple Voltage Regulators Part 1: Noise - [English]
From these I discovered what to do and retain the simplicity with linear regulators. Alas these solutions dont work for SMPS. However -85 dBV for the example I show is not worse than LM7812. If for example we say 12V is 21.6 dB higher it depends where we reference from. If 20V and 2.2mV pk to pk we can say for simplicity 1mV rms noise. That's -60 dBV. 20V is + 26 dB ( 20 = 20 log 1.30103 or 10x2 = 20 dB + 6 dB, Log2 = 0.30103, Log10 = 1 ). We can call that 86 dB below 20V. Or do it this way 20 000 mV/1mV 20 x 4.4 = 86 dB. dBV is a good way of saying how good a thing is.
When designing always check your input ripple and output ripple are to book spec. If not find out why. It should meet minimum spec. Star earths etc help. PSU's work like musical devices in a bad way. If you like a flute with the wrong holes in use. Ripple harmonics. It is usual for 100/120 Hz to be the big one.
Why I like LD1084/85 is the results should be possible without using an oscilloscope. I hate doing that. I know most people don't have one.
-144 dBV is a target.
Unregulated Power Supply Design