That's a tough question, I really don't remember because I have thrown away all LJM products. I have tested four different LJM boards and then abandoned them alltogether when I've realised how pointless was to try to improve them with better components and adjusting parameters. The real reason was poor low frequency performance, besides tendency of these boards to fail/burn suddenly without an obvious reason. There are many cases arround here of burnt amplifiers including destruction of expensive speakers when output transistors failed. I know an example of a diyer who had 10 burnt pairs of L20.5. These are less than prototypes. But people don't care because these LJM boards cost like a better breakfast.
Good luck
Is there any improvement to be had by replacing the 5 main power capacitors, in my board it's 5x 100 uF / 100 V Aishi NB series? It's advertised as high temperature high reliabilty, but not low ESR. I could easily install 470 uF low ESR caps there, although nothing fancy - I can only get Samwha and Hitano capacitors.
Or am I better off leaving the Aishi caps and bypassing them with PP capacitors, or with ELNA Silmic / Cerafine? I have a bunch of of those (4.7 uF and 1 uF).
Or is this all silly waste of time that makes to audible difference?
Or am I better off leaving the Aishi caps and bypassing them with PP capacitors, or with ELNA Silmic / Cerafine? I have a bunch of of those (4.7 uF and 1 uF).
Or is this all silly waste of time that makes to audible difference?
I replaced them all with slightly larger (220uf I think) brand name caps, and also added 0,47 decoupling on the underside of the board.
Don't know if the difference is audible as such, but it should help with stability, and there won't be any problems with "unknown quality" chinese caps.
Don't know if the difference is audible as such, but it should help with stability, and there won't be any problems with "unknown quality" chinese caps.
I see people measuring bias current across R13/R20, but shouldn't it rather be measured across R22/R31 and R33/R34, that is, individually for each transistor?
There are two good reasons you may want to consider.
1. Model the amp in Micro-Cap 12 or similar. Fix the temperature of all the front end, current source and VAS amp transistors to 27C using the T_ABS model parameter to approximate a free air case environment. Let the temperature of the bias sensing transistor, the drivers and the output devices ride with whatever spice wants, ie leave them alone, so they will model a heatsink environment. Set the bias current to 30mA per device under "Dynamic DC" analysis at 27C. Now do a "DC analysis" and plot Bias current while stepping temperature through 0 to 100C or 10 to 80C or whatever you are happy with. Look what happens to the bias current as temperature rises. Pop. Sadly, it takes one extra resistor in the bias chain (as placed by D. Self) to avoid the problem and have well behaved bias current.
2. The designed D669/B649 driver transistors are rarer than rocking horse p**p because they haven't been manufactured for, what?, nearly 2 decades? There are not enough available at any price to make production designs with. Loads available from far eastern outlets, but have a really good look at them and the markings on them and draw your own conclusions. Pop. Again, sadly, there's no need for this as there are modern alternatives, from contemporary manufacturers, which out perform them, have a very high hfe linearity, are cheap, and off the shelf from Mouser, Digi-Key, etc.
When setting for the first time, I would go round all of them, always measuring one at a time, until you are happy they are stable and the same +/-. Now give it a load an a signal to give about 1/3 power or so. They it settle down for at least 30 mins and more on a bigger heatsink. Now remove the signal and check them all individually and if they are vastly different, increased by more than 5% or so, or cold, then you have a fault. As I'm sure you have found out, you need to do the adjustment in small nibbles with settling time between.
Have a look at Self's chapter on Thermal Compensation.. It's the addition of a small resistor in the top right leg of the bias circuit in series with the bias sensing transistor. So the take-off to the top driver comes from a resistor rather than the top of the chain. Value needs to be around 10 to 20 ohms. If the bias curve goes the wrong way but stronger, then try the resistor in the bottom right leg of the circuit, still in series with the bias sensing resistor. That resistor may have to be 20 to 50 ohms. There are also a lot of other methods in Self book, but they would all need a modified PCB really.
"Audio Power Amplifier Design, Sixth Edition" Douglas Self, Focal Press 2013, ISBN: 978-0-240-52613-3.
Chapter 22, p505, "Thermal Compensation and Thermal Dynamics". Fig 22.28b p530.
Thank you very much. I've heard about Self's amplifier(s), but I didn't know he also wrote a book.
Hi Git,
quite a damning review of those boards! I am sure you know what you're talking about, but do we really want to throw in the towel? There are quite a few of us that really like the price/quality ratio of the l12-2.
Is there anything we can do to prevent the amplifier from blowing up? Maybe put a heat sink on those Chinese 2sd669 and 2sb649? Would originals secure the board (not easy but they can be found, UTC made)? Should we try to find a working modern replacement transistor?
To that extent, anyone willing to share a microcap sim file for this amplifier ?
Thank you for sharing your thoughs.
Tibia
quite a damning review of those boards! I am sure you know what you're talking about, but do we really want to throw in the towel? There are quite a few of us that really like the price/quality ratio of the l12-2.
Is there anything we can do to prevent the amplifier from blowing up? Maybe put a heat sink on those Chinese 2sd669 and 2sb649? Would originals secure the board (not easy but they can be found, UTC made)? Should we try to find a working modern replacement transistor?
To that extent, anyone willing to share a microcap sim file for this amplifier ?
Thank you for sharing your thoughs.
Tibia
Hi Git, your message is that bias current drifts way too high after the amp has worked for a while and has warmed up, is that correct? What kind of current, do you think, is dangerous? I'm thinking of constantly monitoring the bias current (via voltage drop on the emitter resistors).
Hi Alexium. Your message made me think... Currently I have installed a protection circuit on the output of the L12-2. It uses one of those ubiquitous upc1237, which , in my DIY kit , is configured for delayed start (working) and DC protection (working, tested with +- 1.2V DC). This chip has also a pin for over-current sensing. Could that be a way forward ?
Guys,
This is much worse. Once an honest manufacturer becomes aware that his product is faulty, he immediately issues an apology and withdraws the product from the market to protect customers. But a petty thief sees an opportunity in continuing to sell faulty product in hope to increase sales. Have you noticed that LJM sells a »mono« version, a single channel product because he knows very well what’s going on. He is selling mono under the headline Save up to 10% when you buy more. But, of course, buy 10 pieces in advance to »save«; you will need them all 10 sooner or later. There is a member of this community who has 10 blown L10.5. I have thrown all LJM »products« into garbage where they belong.
I am aware of another fraud who persists selling faulty KSA 50 boards with wrongly connected Zener and the amplifier blows inevitably. No chance to use it. He persists selling it. Business is business.
Every such product you buy for a cost of a breakfast is overpaid because it is what it is: trash.
And, by the way, there is no chance that bias will be properly self-adjusted because the biasing transistor must be at the same heat sink close to power transistors. There is no feedback effect to control the bias and you cannot do anything else to solve this.
Forget it and move on. 😎
This is much worse. Once an honest manufacturer becomes aware that his product is faulty, he immediately issues an apology and withdraws the product from the market to protect customers. But a petty thief sees an opportunity in continuing to sell faulty product in hope to increase sales. Have you noticed that LJM sells a »mono« version, a single channel product because he knows very well what’s going on. He is selling mono under the headline Save up to 10% when you buy more. But, of course, buy 10 pieces in advance to »save«; you will need them all 10 sooner or later. There is a member of this community who has 10 blown L10.5. I have thrown all LJM »products« into garbage where they belong.
I am aware of another fraud who persists selling faulty KSA 50 boards with wrongly connected Zener and the amplifier blows inevitably. No chance to use it. He persists selling it. Business is business.
Every such product you buy for a cost of a breakfast is overpaid because it is what it is: trash.
And, by the way, there is no chance that bias will be properly self-adjusted because the biasing transistor must be at the same heat sink close to power transistors. There is no feedback effect to control the bias and you cannot do anything else to solve this.
Forget it and move on. 😎
Yes, Mr. Picard, that was the first class nonsense from my side indeed, and I do appology for my ignorance.
All participants in this conversation are amatheurs, not real experts, prone to errors, no exception. All. That's normal.
All participants in this conversation are amatheurs, not real experts, prone to errors, no exception. All. That's normal.
Tibia
Sorry for the late reply. I don't visit here often and I didn't realise that you only get notification for the first reply to your message. I didn't know a sub-thread had formed.
Firstly, please don't get me wrong. In no way was I throwing mud at LJM or the overall design, I have used part of the input stage in my own home amp. I'm trying to point out how the amp could so easily be improved and made more manufacturable by a few changes. It is made to be low cost and in fact has amazing distortion performance for such low cost. The biggest area in need of improvement is the bias circuitry. I think most people would agree that adjustment here is quite important. You will see in the attached MicroCap file that with default parameters throughout, the contribution to quiescent current from the driver transistor is considerably more than from the output devices. To be fair, I could not find spice files for the 2SA1186 or the 2SC2837 so have used replacements 2SA1837 and 2SC4793 as recommend here. It takes only a few very small tweaks to the R10/R19 bias chain values to restore something like normality. But there are still two problems there. 1-The bias current increases with temperature which can lead to thermal runaway. This can easily be fixed with a resistor of 10 to 20ohms in the collector of Q7 (stll take driver from Q7.C). 2-bias current imbalance between the two members of a pair (eg, Q2 and Q10). Unless you make deliberate changes, a simulation like this uses identical models for two instances of the same component. Translated to the real world that means it uses perfectly balanced pairs everywhere. I didn't realise myself just how important it is to match the members of a composite output set. Using the standard Vbe/Ic formula and a while with excel shows that a mismatch of only 5mV to 10mV can lead to one of the pair taking several times more current than the other, though the emitter resistor does partially mitigate this. I had heard the original manufacturer uses matched output pairs, but this could explain why some of the clones pop output devices so easily.
Some words on the attached MicroCap file. After I first used it I made changes and additions all over the place and haven't been very disciplined about keeping a backup chain. So I have recreated the file from the latest schematic I could find, which the poster here said was the same as on Calvins site. So be aware it is a brand new untouched implementation with no changes or configs anywhere. All of the transistor models that I've used should be incorporated into the file, so it should load and be ready to go. You can also reuse those models - search File Portability in the MicroCap help. I strongly suggest you put a pot between R10 and R19 and reduce there values to keep the total resistance of the chain roughly the same, and take the pot wiper to Q7 base.
Sorry for the late reply. I don't visit here often and I didn't realise that you only get notification for the first reply to your message. I didn't know a sub-thread had formed.
Firstly, please don't get me wrong. In no way was I throwing mud at LJM or the overall design, I have used part of the input stage in my own home amp. I'm trying to point out how the amp could so easily be improved and made more manufacturable by a few changes. It is made to be low cost and in fact has amazing distortion performance for such low cost. The biggest area in need of improvement is the bias circuitry. I think most people would agree that adjustment here is quite important. You will see in the attached MicroCap file that with default parameters throughout, the contribution to quiescent current from the driver transistor is considerably more than from the output devices. To be fair, I could not find spice files for the 2SA1186 or the 2SC2837 so have used replacements 2SA1837 and 2SC4793 as recommend here. It takes only a few very small tweaks to the R10/R19 bias chain values to restore something like normality. But there are still two problems there. 1-The bias current increases with temperature which can lead to thermal runaway. This can easily be fixed with a resistor of 10 to 20ohms in the collector of Q7 (stll take driver from Q7.C). 2-bias current imbalance between the two members of a pair (eg, Q2 and Q10). Unless you make deliberate changes, a simulation like this uses identical models for two instances of the same component. Translated to the real world that means it uses perfectly balanced pairs everywhere. I didn't realise myself just how important it is to match the members of a composite output set. Using the standard Vbe/Ic formula and a while with excel shows that a mismatch of only 5mV to 10mV can lead to one of the pair taking several times more current than the other, though the emitter resistor does partially mitigate this. I had heard the original manufacturer uses matched output pairs, but this could explain why some of the clones pop output devices so easily.
Some words on the attached MicroCap file. After I first used it I made changes and additions all over the place and haven't been very disciplined about keeping a backup chain. So I have recreated the file from the latest schematic I could find, which the poster here said was the same as on Calvins site. So be aware it is a brand new untouched implementation with no changes or configs anywhere. All of the transistor models that I've used should be incorporated into the file, so it should load and be ready to go. You can also reuse those models - search File Portability in the MicroCap help. I strongly suggest you put a pot between R10 and R19 and reduce there values to keep the total resistance of the chain roughly the same, and take the pot wiper to Q7 base.
Andrew Eckhard has stated many years ago and I do agree that 0.01-0.03% of adequate THD is just fime if everything else is done right.:
My question is:
Last edited:
Looking at the schematics, it seems to me that it's easy to measure individual bias currents of each individual power transistors across R22, R31, R33 and R34. In other words, between each transistor's emitter and VCC / VEE. Is that right, or am I missing something?