L12/2 version 5. FIXED.
Вставка
- Опубліковано 1 січ 2025
- At last, we have a solution to fix the instability of this new version 5 of the famous L12/2 amplifier.
Out of the box it has a number of issues, we think we have fixed them all. We also show a couple of simple mods you might like to make.
PLEASE WATCH PARTS 1 & 2 BEFORE WATCHING THIS. Links below.
Part 1. • L12/2 Version 5 Amplif...
Part 2. • L12/2 version 5. The U...
Not sponsored, paid for with my own money.
I hope you give this video a "LIKE" I cannot tell you how devastated I will be if you don't!
Mike,
Wishing you both a happy and healthy new year
Thomas
I'm sure we all appreciate the full schematic picture you provide as well as the picture of the component IDs on the PCB. This way when you discuss the components by their IDs we can make better sense of what you're talking about as opposed to the Accuphase C-3850 videos where we lack any useful frame of reference for your conversations across that series of videos. You've earned the thumbs up that you're asking for.
Methods of description are always subject to improvement... 😁 Thanks for the feedback!
It's very time consuming to make and edit a simple 15 minuet video. In fact, about 8 hours in total. We are not professionals. I had to learn video editing from scratch. While my videos are a long way from being perfect and professional, they are getting better in my humble opinion. I do make errors, forget important details. I rely on you guys to point out my errors.
Well done Mike sorting this issue out with Ron !. I'm more than happy with the previous L12/2 working at 50-0-50 being driven by the Accuphase 3850 preamplifier with switchable input. I listen to mainly classical CDs or Radio 3 via internet streaming. My speakers are TDL studio 1M transmission lines on 15 inch stands. The sound is powerful, detailed and very very clear. An excellent stereo image. Many thanks for your pupilage !!!!
It's always a pleasure to watch your work and learn.
Happy new-year to you and Ron.
Thanks Michael. I'll read up and start the build. Hope you have a happy new year.
Happy New year Michael and nice to hear from you too Ron have fun guys....cheers.
Thanks Andy, good to see you here also.
@MichaelBeeny Thanks to Ron and you for doing these videoes and the large amount of work to find these modifications. I wish you all a Happy New Year.
Thank you for your kind words, most appreciated.
Howdy Michael , another great video from you and your friend in the US. Great work and very well presented as always. Just love you videos (and that t-shirt). Merry Christmas and a Happy New Year (at least in a couple of days time...) I will continue to follow this adventure, and i might even get one of those some time during this winter. Best regards to you and you family (and your followers) , from the snowy cold North.
Thankyou Oie, we much catch up on Skype. Our time never seem to match up. Like now, I'm in bed. lol
A definite like from me!
I'd like to say it's surprising how you need to repeat that you did not design the circuit. However, when I used to lecture, I found that good listening is a skill often missing in people, especially young people, but sometimes older people too.
Good work, you two! 👍
I was shocked to find that not every one of my viewers watch the complete video! I was in shock for days! Some people ask questions that are covered in the video.
@@MichaelBeeny I'd noticed that. It's the age of impatience. Everyone wants stuff instantly.
Another thing seems to be that people can only grasp one fact at a time, probably because they don't hang around long enough! 😏
Thanks a lot Sir Beeny for your verry informative video.For me you are one of my best You Tube Electronic Teacher....God bless you and your whole family.
I didhit the like, and merry christmas and happy new year, and thnkyou and Ron for your work, great job
Really nice piece of work there, very educational!
Отличный результат! Вы и Рон молодцы! Красивая футболка! С наступающим Новым Годом!
And Happy New Year to you as well.
Well done gentlemens!
Unfortunately, my kit was provided with so many bad parts I'm questioning if it worth the effort to invest money and time in this. As far I'm concerned this version is stillborn and should be removed from the market. Thank you for your work! I wish you both a healthy and happy new year.
Agreed, and I have been told that this is not the only LJM design that was unstable out of the box, though I have not confirmed that myself.
And a happy and healthy New Year to you and yours as well. 😀
What parts did you find bad? I had no components that I would describe as "poor"
@@MichaelBeeny
Sorry Micheal I should have said not matching or out of specification. LJM talk about few bad transistor, but most of them need to be replaced in the kit I purchase. Let's hope some get better luck than me.
Happy New Year Mr. Beeny!🎉
Thank you, Bob, it will be new year here in NZ in about 3 hours' time from now. Looks like I will be celebrating answering comments and e mails lol
The shorting of those transistor pins suggests a good medicine for the wrong problem. Maybe replacing that transistor would do the trick. I don’t have the board, so I should just keep quiet, though.
We have 4 samples of these boards, they all have exactly the same issues.
@ and those 4 FETs do not behave as FETs? I don’t know what’s more likely “a large batch of fake parts” or “a mislabeled board”
I should ask LJM. Thanks, Michael.
Yay front row seat!
Sorry, you don't win anything lol
@@MichaelBeeny 😆😆
Very good work, thanks Michael! Did you mention the small cap on one of the VAS transistors?
Hi, thanks for taking the time to fix this board, helping out all of us laymen. I really enjoy your videos. I saw a post on the DIYaudio board from Vit4min4s where he moved a 33uf capacitor to between the two 100 ohm resistors that are linked and claimed that it fixed things. I can link the thread if you like. What are your thoughts on this potential fix? The reason that I ask besides curiosity is that it would be easy to implement.
Also, what voltage power transformer would you recommend for someone that wanted to swap in various sets of speakers from time to time that can range from 4 to 8 ohms?
Thank you for these videis Michael. What's the thing about these DartZeel clones? Looks like a rehash of a lot of the stuff we tried in the early 80s?
Very little work these days is original. No money in originality. Look at all those overpriced valve amplifiers. The circuits go back to the 40s
The schematic shown at 6:15 has a new capacitor C12 that is not included in any of the other schematics shown in this video. What's up with that?
Mike showed a slightly earlier version of the schematic at 6:15.
We experimented with various velues of capacitance across the VAS stage current source because of an asymmetric behavior observed during square wave testing.
The C12 was intended to suppress a very short leading edge spike, and does in fact help to do that.
However, it it is made any much larger than 12pF the board will become unstable in a different manner.
The spike itself is only evident when using very fast rising square waves and there is none present with sine waves or actual music waveforms.
And as it is so small anyway (12pF) that it is not really worth adding.
Particularly so because when you add an LR network to the output (feeding the speakers) the spike is completely suppressed anyway.
So- In the final circuit C12 was just omitted.
@@ronschauer839 I left that cap in circuit just to see if my viewers were paying attention lol. Sorry guys.
Why never put a pot on the input? Is that to keep the input impedance constant?
Granted, a preamp will introduce a pot on the input.
Hi Mike and Ron
If I understand the video correct you suggest that the gain can be lowered by changing the bottom resistor in the feedback loop. Please be aware that you a getting very close to loop instability.
If my schematic is correct, LTSpice shows that 11dB gain is 1-3dB from loop instability (180 degree phase shift with 0dB loop gain at app. 1.2MHz) Likewise, phase margin at 0dB loopgain is -178 degree. A no go in my world.
You could increase the miller capacitor to app. 200p, but that will destroy slew rate. It is better to leave the gain as is. Sorry, but there it is.
NB! You often in opamp specs have a minimum gain setting. Not all opamps can run in unity gain mode. This is the same thing.
Please crosscheck my findings eg. from Doug Self or Bob Cordell.
BR/Claus from Denmark
Hi again
I was to fast with my reply and did not check that every mod you made was included in the simulation. I missed the 33pF you have added to the VAS current source. With this mode the amplifier is much more stable with normal gain. (Gain margin 14dB and Phase margin 75 degree). I would perhaps not use 11dB gain mode, since the margins will go close to badness.
One thing to note is that adding capacitance to the VAS will degrade slewrate and perhaps also influence distortion at high frequencies.
Sorry for the to fast triggerfinger
BR
Claus
@@ClausDreisig We did all those things. And tested for stability. Reducing the gain causes no harm that we can determine, at least to the levels shown in the video. If anything the amplifier performs better. It is what it is, and in its as-shipped form this board is dreadful.
@@ronschauer839 I do believe what you did help a lot. The second sim run showed that. Great job. My comments are made of interest in your work.
happy new year
Claus
Great vwork Mike, love your videos! Whooohoooo front row seat. How do i buy one of your T shirts?
They came from TEMU, very low cost, even a one off. I just sent a JPEG of the artwork, 2 weeks I had them. If you really would like one, send me your e mail, I will send a link and the artwork.
Hi, great video as always, which is better for you? this version or the older versions (version2, 3..) or are they the same,? i built the C3850 preamp (waiting for 33k resistors,i asked for 33 ohms😓) would it be a good match whith these boards?thanks.
It's too early yet to do any real comparison, either technically or audibly. We have only just got the boards working on the bench. We will report back with an answer ASAP.
Hi Michael! Thank you for this nice presentation!
At 6:49 you show a schematic which has an additional Miller-C of 33p at the active load of the voltage gain stage and a lowered supply voltage of +/-34.5V instead of 37.5. Are these additional improvements for an upcoming version?
Sorry, my error the 33pf was not used in the final circuit. As for the different voltages. We tried many different voltages to see if it effects the stability. I will be running mine at 50 volts. Ron will be using 37 ish volts. The only real difference is power output and heat.
@@MichaelBeeny Thank you for the explanation!
@@MichaelBeeny Hi Michael. I know it might be annoying with all these "good advice" comments. Just want to be sure that you are aware that putting 33p from base to collector in the current source seriously load the VAS, as charging and decharging of have to take place via the base components. My guess is that the risetime of the amplifier is affected by this. Anyway. Guess there was no other way.
BR and happy new year
Claus
@@ClausDreisig Sorry, my error the 33pf was not used in the final circuit. We tried so many things to get this stable this was one of them. Made thing worse in the end so not used. I put up an earlier circuit. Sorry.
I was contemplating using the L12/2 in my DIY amplifier project with the Accuphase C3850 and a Raspberry Pi music streamer. Now I'm not so sure. The Aiyima store on Aliexpress is currently selling the 4.2 version of the L12. But I'm also considering the Dual 7293 amplifier or the LJM MX50se. If I buy a cabinet with heatsinks on either side then it seems the L12 would make more efficient use of the space because of the way it's mounted. Maximum output power is not my first priority. I'm looking for reliability and good sound. I don't have the budget for expensive speakers and will be looking for a decent set of bookshelf style. Not sure if I should avoid the L12 altogether. For me, the version 5 L12 is out of the running.
I'm still using, at present the Dual 7293 I used in my original videos. I really like it. Ron uses the same in one of his systems. This is somewhat better than the MX50 which is also very low cost and decent sound. The best of all has to be the L12 vers 4.2 however.
As to the latest ver 5. You MUST do the update mods or it's e waist. As to its sound...too early to really say yet, only just got it stable. Listening tests will start soon.
Why not tpa3255?
Good vid Mic. And thanx we could hear the mistified Ron on this vid.
Bty, can u post a complete pic of the schematic of the rectified/modded amp ?
The full circuit is at 4.42 in the video.
Mystified? 🤪😆 👍👍 (I do not care for the sound of my own voice all that much...)
Thank you for your work Michael! That inductor shown at 12:00 was mentioned as having a 100uh value in 12:04. Are you sure about that? I worked with air coils/inductors in the past: that inductor in your hand seems to me having at maximum 2 or 3uh...
Actually that was my error from very early discussions. The ones I am using for example are in fact about 1 microhenry and that is the value shown in the "simple mods" schematic. Good catch. 😀
Great video. Are you planning to make your own design amplifier?
At one stage we were talking about making a L12 super amplifier, then this came out. I think we both might need a holiday from L12s lol. A design from scratch needs hundreds of hours of design, prototyping, money and then more money. We are both retired, we just do this largely for fun. Certainly, no money in it, in fact it costs US money. 25 years ago, maybe!
@@MichaelBeeny Never say never... 😆
Thank you very much grandpa l ❤❤❤❤
Thanks for your hard work Mister Beeny. Now, if somebody want buy a L12 board, what is the reason to take version 5 with problems? The oldest version is not a much better choice?
We both agree with you, the older versions are better and if you can obtain a set of them, don't buy version 5. The problem is that listings for L12 modules on eBay (for example) that show the older versions may not be accurate. If you try to order them you are very likely to actually receive version 5 boards. This happened to me when I attempted to order version 4 boards recently. Fortunately I asked the seller before placing the order. So our advice is to ask the seller to verify that you are going to receive version 3 or 4 boards, and NOT version 5. If they cannot verify that, do not order them from that seller. However, if you receive version 5 boards it is not the end of the world. Just do these simple mods and it is not bad.😀
@@ronschauer839 Ok, thank you very much for the clarification.
Thank you for another great video and the work you and Ron have put into improving this power amplifier module. ❤ Just a question about a terminology you used regarding the parallel network of the inductor and resistor wired in series between the output of the amplifier and the loudspeaker. You referred to it as a Zobel network but I always thought it was called a Thiel network??? Unlike the series RC network wired across the amplifier output which I was lead to believe is a Zobel network or Boucherot cell. Are the two terms interchangeable for the different networks?
A Boucherot cell (or Zobel network) is an electronic filter, used in audio amplifiers to damp high-frequency oscillations that might occur in the absence of loads at high frequencies.
I think we were BOTH right. lol
Just out of my curiosity, how can you tell Jfet from bipolar transistor ? Jfet also has enhanced mode and depleted mode, where the latter is more commonly available. For example, tl032 is enhanced mode jfet opa, tl071 is depleted jfet opa.
Enhanced mode Jfet bahave same as bipolar if measured by current. The only difference is Fet has one major carrier while bipolar has two. They will react differently with temperature, frequency or voltage. This is bacause electron and hole has different property,such as mobility.
The device in question ("FET1" on these boards) measures with steady junction drops from what is clearly the base connection to either the emitter or the collector and yet nothing (open circuit) in the reverse polarity.
This is precisely the same behavior as basically any bipolar transistor, right down to the typical Vbe voltage drop of around 645mV.
And it holds true when measuring between only two terminals of the independent half of the package at a time.
Any FETs I have on hand measure quite differently from that.
Also, even a simple component checker identifies it as being a bipolar NPN transistor whereas a FET or MOSFET will be identified correctly as FETs.
I will not say that it couldn't be some unusual FET.
However, if something looks like a duck, walks like duck, and quacks like a duck, it is most likely a duck (to paraphrase the old saying).
Regardless, a dual bipolar in that location for the circuit only makes sense, and fits the pinout for the board layout.
@@ronschauer839 However, if something looks like a duck, walks like duck, and quacks like a duck, it is most likely QUACKERS (to paraphrase the old saying). I wonder how many viewers get the joke Ron???
Hello michael, I just want to ask as well if I could use non-geniune or un-original LM7915CV as well as some LM7815CV regulators for my preamp build similar to this. Would that be an issue? I also would want to ask how can I improve and/or reduce unwanted noise other than adding some ceramic capacitors as well as polymer caps (if I remember that correctly.) Thanks
Fake anything are usually of lesser performance. In my limited experience of fake LMs, they do tend to work but don't make the current output. Possibly not an issue on low current stuff like a preamp. But you simply don't know what's really inside. These chips already have good noise reduction, doubt changing caps will help much, if at all.
I would have increased C5 from 100uF to at least 1000uF now that the capacitive multiplier around Q9 is defunct.
You could, but they won't fit. It also will make no difference to stability. That was our only real goal. If your main rectifier and cap board is placed as near to the amplifier as possible, that will be fine. In fact, exactly as tested.
@@MichaelBeeny *Yes indeed.*
If we could have fitted larger caps for C5 and C8 (don't forget about the negative rail) onto the board we certainly would have.
However as Mike says, if you have the supply filter caps close by (and preferably they are large in capacitance with stout connecting wires), it doesn't matter nearly as much as does adding the 1uF ceramic bypass caps across C5 and C8.
In my testing I accidentally left them off of one board but had them added onto the other.
They board without them still had traces of instability under certain conditions and yet the one with them was clean.
It took a bit of head scratching to realize why this could be.
When I realized my mistake and added them, those issues went away.
Now the two boards behave *exactly* the same.
So they really do matter..!
2.2uF or 4.7uF ceramics would be even better, but be careful of their rated voltages.
I am using a rectifier/filter board similar to the one Mike is going to use in his build.
The only difference is that my board has four 10,000uF 50V filter caps and his are rated at 63V.
Because I am going to run my boards at +/-37.5VDC, the 50V ratings on my caps are sufficient.
Mike is planning to run his closer to +/-50V, as I recall.
FWIW... 🙂
Where can we buy
the t-shirt😂
I had the shirts made by TEMU, very low cost even for a one off. Send me your e mail and I will return with the artwork. Send these details to TEMU with your choice of colour and sizes, two weeks later you will have them. (sadly, no commission for me lol) I'm not sure I understand why anyone, but me, would like my name displayed over there chest? I do have some strange viewers. But I still value everyone.
Q9/Q10 are very-very-very strange solution, cause a signal amplification stage voltage must be 2-5 volt higher than an output stage voltage (even with a Sziklai pair). I understand these are capacitance multipliers (or transistor equivalent chokes), but one additional low power transformer these days is $15-20 as a max.
A differential stage 4mA is a bit higher than an optimal current for a higher S/N, and 4mA for the next stage is a bit low for my taste.
Sziklais everywhere is good. Zobel network inside a feedback loop is good.
Yes, the "upgrade" LJM made to this board in that regard actually caused the vast majority of the issues with the board. The other major problem is that he tripled the amount of current required in the driver stage, and that causes stability problems with the QI adjustment among other things. Once returned to 100 ohms (or thereabouts) the circuit behaves itself. Of note- 100 ohms happens to be the value used for those two resistor positions in the earlier versions of L12 and they worked nicely. Why LJM changed this is a mystery that only he can resolve. But don't hold your breath waiting for him to do so.
@@ronschauer839 I see. I would expect yet more driver stage before the output Sziklai. For the stability and better bias control. Cause here we have a common emitter cascade without a feedback resistor (a very low input impedance for a previous stage). I'd also try to connect this emitter to the 0.1Ohm (0.22 or 0.47 would be better imho) rather then to an output directly. Because Sziklai and Darlington have directly connected emitters in order to work like a single high b transistor, and so a feedback would work properly.
Yet more thing is that there is no a tiny capacitor shunting R17, and correcting a feedback loop pole. This schematic must be modelled with a stability analysis first.
We tried capacitance across R17. It did nothing to resolve the original problem unfortunately.
You had a tiny problem which could be solved by a tiny 10pf capacitor across R17. You destroyed the PSRR, the slew rate, the NFB.
For sure that would have been the first thing I tried. Limiting the bandwidth to about 30 khz with higher values might have also been a good idea.
We tried something like that as a first pass solution. It did NOT solve the problem.
@@ronschauer839
Okay .... so you had the same idea as me... try it and see what happens. Good stuff.
I still think part of the instability is the board itself, not a great layout for sure. Our aim was to get the board AS SUPPLIED working as best we could. It is not for us to design a better amplifier. You now have a working amplifier with good performance, I think we have done this, and NO COST to you guys. The alternative was e waist.
@@MichaelBeeny
Yep. No argument there.
I am gonna buy this kit and do what LJM says to do
I'm not really sure I understand your post!
@@MichaelBeeny he said on the forum to replace the bad tr's that came with the kits,
@@RoverWaters I learn this after testing them with my curb tracer, but this can't justify the bad board layout and the instability of the design so go ahead through the dice but don't test them without speaker protection.
@@fredcailloux2526 What's all this about bad transistors. Is someone saying these boards were supplied with faulty transistors? Any more info you have would be very interesting.
I begin to worry my L15 & L20.🤣
I have built and tested the L15. I had no stability issues but did not really like it's sound personally. No experience of the L20.
@@MichaelBeeny Thanks for you kindly reply. In fact, I have no impression of L15 and L20 at all.😅 Due to space considerations, I use tpa3255 module.
Not far behind
Grato!
bridging out the regulators that's a poor regulator no Zener diode
Actually, they weren't really "regulated" to begin with.
The two transistors were configured as capacitance multipliers, not regulators.
However the components that were chosen are grossly too large for the current that they actually carried.
Far worse, they induced non-linear elements into the +/- power rails at very critical points.
This in turn caused instabililties whenever the waveform peaks even approached the incoming rail voltages.
Shorting across the transistors is admittedly a band-aid fix, there is no argument about that.
However, we were trying to find the simplest possible DIY fixes for what is a really poor design (short of tossing them into the nearest dumpster).
The board layout itself is so convoluted that if you remove the two transistors and short across the pads of what would have been their collectors and emitters, it oscillates horribly (believe me, we tried this)..!
So the transistors and the "filter" devices (R20, R29, C6, C9) must remain in place to help stabilize things even though the transistors themselves are no longer acting as amplification devices (or filters, as you prefer).
Bad... but in the end- salvageable. 🤨
Buy revision 3 or 4 L12 boards instead IF you can find them,
Just watched part 3 video, seems like you are having big problems with instability.
Your zobel network (R and C in series across the speaker terminals) looks WAY too small. Increase the C to 4.7uF. This ADDS resistance at high frequences (where instability is occurring). Use high quality mylar or polypropylene cap and resitor and solder on the speaker connectors.
For your additional bypass cap, ceramic type is a BIG no. Use mylar or polypropylene types, they are large physically but worth jamming on your pcb. Ceramics have 'peaky' capacitance, at some frequencies they dont capacitor.
I know you didnt design the pcb but consider a redesign. Mounting npn and pnp outputs at opposite ends is frankly bizarre. No thermal stability and your drivers wont be thermally matched either. Bet you have bias problems with this design.