L12-2 CFP Output amp 120W*2 8R

Member
Joined 2010
Paid Member
10-12mV is excellent for DC offset. Typically, anything below 40mV is fine, 25mV or below is plenty good enough as it becomes academic, like rows of zeros in a distortion measurement. When there is no provision to adjust offset (and for reliability, I don't think there should be either), the match of Q1:Q2 Vbe determines the current balance in the LTP (input pair) and hence the degree of offset voltage at the output.

Swap them or try other similar matched pairs and get a different offset or none at all, if you must have it.

The ratio of R10:R19 sets bias current, much as the pot usually placed there does in many other designs. I'd warn against experimenting with trying the next closest resistor you have to see what happens. The circuit will be more sensitive than you think and a toasted output stage of really good output transistors is not going to be nice to replace.

The double Szlikai OP stage is featured in Douglas Self's Amplifier Design Manual as the "Load Invariant Amplifier", meaning it performs almost as well on 4R as 8R loads.
 
Member
Joined 2010
Paid Member
Uhh..no... sorry...R13 and 20 are the "collector" resistors in the output stage, so you are measuring Bias current - I guess you have quoted for each channel and across both resistors?
The schematics show different resistors; 0.22R in your posting Evenharmonics and 0.1R in mine. If ohms law is still true, you need to calculate the current from the voltage across the actual resistor(s) you are measuring. Across both R13 & 20 is more accurate.

That calculated current should match the specification on the website documentation. Which, IIRC, is 20-40 mA
 
10-12mV is excellent for DC offset. Typically, anything below 40mV is fine, 25mV or below is plenty good enough as it becomes academic, like rows of zeros in a distortion measurement.
Does it matter whether it's positive or negative DC voltage? Mine shows negative when red (positive) lead of DMM is connected to + of speaker terminal and black lead to - of speaker terminal.
 
Member
Joined 2010
Paid Member
Does it matter whether it's positive or negative DC voltage? Mine shows negative when red (positive) lead of DMM is connected to + of speaker terminal and black lead to - of speaker terminal.
I thought 10 and 12 mV was what you measured across the output transistor collector resistors? :confused:

Well, if it was the voltage across the output terminals, that's fine - it's the offset voltage and it doesn't matter what polarity it is. In simple theory, if you swap Q1 and Q2 the polarity should simply reverse.

Re: the resistors: Yes the 0R1 is the same on the schematic I showed but for yours, these are all changed to 0R22 for Rev-2.
 
Last edited:
Bump.
10-12mV is excellent for DC offset. Typically, anything below 40mV is fine, 25mV or below is plenty good enough as it becomes academic, like rows of zeros in a distortion measurement. When there is no provision to adjust offset (and for reliability, I don't think there should be either), the match of Q1:Q2 Vbe determines the current balance in the LTP (input pair) and hence the degree of offset voltage at the output.

Swap them or try other similar matched pairs and get a different offset or none at all, if you must have it.
I'm guessing that you meant Q6 and Q8 (1015 transistors).
QQ20111218231516.jpg


I had little extra time and those transistors cost about 12 cents each at Mouser so I bought 6 and decided to play around. I took out the ones came with the board and there was a bit of mismatch when I checked with DMM. I measured the new ones and they are all very close (closer than what came out of the board). I soldered the matching new pair and DC offset is now 15mV (previously 24). Is the difference audible? I don't think so but it was a fun little project and I'm happy with the progress.
Thanks for the tips.
 
Member
Joined 2010
Paid Member
Great :), it's good fun tweaking but as you find, some notional ideals like zero DC offset are often a waste of time as they don't show any any obvious benefit apart from knowing you have....uh...zero DC offset.

The awkward truth is that the actual Vbe and Hfe matches are more important for the lowest amplifier THD condition than the Vbe alone is for the DC offset. That in turn results in max. cancellation of even order, mainly desirable 2nd harmonics. And so it goes on, lurching about from bad to worse when you start tweaking some parameters. FWIW, you then have to tweak the current mirrors, Q11,13 I think, to null the offset. The mirrors will fight any change the LTP, to maintain equal current in Q6,8 as they are designed to :xeye:

Yes, I did quote the wrong parts for the input pair as you guessed. :eek: I can't recall why other than I guessed because I couldn't read it at the time of typing and forgot to amend it later. That seems to happen a lot to me when I try to avoid losing the post completely trying to view earlier attachments if I opt to make an advanced reply. That is a right PITA and a good time to quit and go to bed. :mad:
 
Hi,

these small modules were one of the biggest surprises I came across yet.
I didn´t expect much, but the parts alone could have been worth a shot.
The Hitachi drivers and Sanken LAPTs belong to the best Sand You can use for audio amplification. So I ordered the completed modules, since the price difference renders self soldering obsolete.
As they came straight out of the box the modules looked fine. Ok, optically a few components could have been soldered in more carefully, but functionally everything was ok.
The schematics in #6 contain several failures. Besides the output transistors not beeing Darlingtons but Sanken LAPT, the emitter and output resistors are all 0R22. The input transistors, diodes and current mirrors are made from 1015/1815 not 5401/5551. Just the cascodes and current source are 5401/5551. Also the connection of current mirror Q11, R27 and R32 is wrong. At last the value of R10 in the biasing arrangement is 2k2 instead of 1k. The schematics in #12 are correct.
From first glance on the PCB I missed a Bias pot and indeed this is a tweak one should perform, since the Bias may vary wildly from module to module.
A 1k8 resistor, paralled with a 10k Pot (I always use 12- or 23turn types, e.g type 64W) instead of the 1k R19 allows to trim the Bias to the desired value.
Other Tweaks:
- small film caps in parallel to input and feedback lytics C5 and C8
- small film caps in parallel to power supply lytics C1,2,11 and C13
- increasing the high-current paths crossection by applying solder or copper. The Layout offers a good basis, as the high-current tracks are very short.
- adding a output inductance of a few µH, maybe paralleled with an 1-2.2R resistor.
- If any stabilty issues occur, one might add a small cap in parallel to feedback-R R15. I Haven´t had any problems so far, which is one of the points I really like about these modules..... stability into complex loads.

As long as output offset voltage remains within acceptable values I wouldn´t care about transistor matching, though it helps to minimize offset as well as THD.
The Bias value will also affect THD. Beeing CFPs the optimum Bias current may be very low, lower as needed by Emitter followers. 12-15mA is perfect, but higher values may be chosen as You wish. Bias remains very stable, due to the critical parts beeing the driver transistors (which work under more constant conditions as the usually critical output transistors). The very short feedback loop, encompassing the drivers and output transistors keeps control of the bias in the output transistors. Consequently Q5,7 and 9´s cases share a dedicated trace on the PCB to detect and level out temperature differences. They are not and must not be thermally connected to the main cooling fin. Due to the very low required Bias current, the modules run very cool. At idle You wouldn´t even require a cooling fin.

As I´m building and listening to Electrostats, amplification has always been a issue, as ESLs are often very difficult loads. In parts they exceed the -60° phase shift limit of the power cube diagrams You sometimes find with amp measurements. Even if they still work stable into such a loads, many amps suffer sonically. And it are not just cheap amps, but also some of the most renowned ones.
These small modules keep pushing the ELs as if there were no tomorrow and keep cool.
I can confirm ATAUDIOs statements in #15. Transparency, livelyness and control are words I´d use to describe the sonics and are words I got to hear from anybody who gave a listen. These modules are anything but boring.
Since they performed so well not just with dynamic speakers but also with ESLs, we made a small shootout between them and several commercial amps, among them a Accuphase A50.
Well, the King is dead, long live the King. ;) Poor Accu was dethroned within a few beats. Hard to believe but more on the listening panel preferred the small amp (tweaked) instead of the class-A monster. Differences were much smaller with dynamic speakers, but the fact that this amp could sonically cope with the Accuphase was sensational. Not taking build quality and finish into account, those preferring a more relaxed, laid-back performance would still be happier with the Accu. Those liking a lively performance would probabely prefer the small amp modules.

jauu
Calvin
 
Last edited:
  • Like
Reactions: 1 user
Member
Joined 2010
Paid Member
Any link where one can buy the boards?
L12 2 Power Amplifier 2 CH Assembled Boards Heatsinks | eBay
This guy is my go-to for genuine kits as advertised and I already asked if his were the genuine Sanken parts and not TTC5200/A1943 alternatives. Just enter "L-12 amplifier" in the Ebay search bar for other suppliers.

Hi Calvin
That's a very detailed and excellent review and discussion of the kit. Thanks indeed for your effort in informing us of what you found, it will inspire of lot of people to get into this neglected topology with new ideas and components.

Evenharmonics, the 2N5551/2N5401 complements are general purpose high voltage transistors that suit a lot of jobs in the low current areas of the amplifier but the 2SC1815/A1015 are lower noise, low voltage types for best performance of the input LTP and probably other low voltage tasks in the input stage. There will be overlaps of suitability but stick to the original types where fitted. Download the datasheets from say, Datasheetcatlog site and check them out for your interest. Check out the voltages present across C-E and C-B and draw your conclusions from that.
http://www.datasheetcatalog.org/datasheet/toshiba/964.pdf
http://www.datasheetcatalog.org/datasheet/toshiba/905.pdf
http://www.onsemi.com/pub_link/Collateral/2N5401-D.PDF
http://www.datasheetcatalog.org/datasheet2/0/00755x1fxd70zjeg12965dd3jffy.pdf
 
Member
Joined 2010
Paid Member
Sure, that's an option and thankfully, Fairchild cloned a lot of obsolete but excellent Japanese devices and changed the prefix with a K to identify their versions. Most are indistinguishable in performance to the OEM parts, though I thought the pinouts were still e-c-b. ??

There are other low noise parts such as 2SA970 that will be drop-ins that should perform well too. Take a look at this configurable chart of audio parts by member Greg Erskine as a good reckoner for most needs. You select the type of device you want from the top line and a pinout indication is included but you can also check pinouts by Googling the part numbers for a datasheet or just Google "KSC1815 pinout" for example. It's fast and easy.
VAS Transistors
 
Last edited:
Member
Joined 2010
Paid Member
post7:
the harmonics 5th to 13th "look" ugly.
Has any Member a view on this and whether it is important, or irrelevant?
You can be sure members Bob Cordell and Douglas Self have strong conflicting views on CFP output stages generally :D but as these kits by ljm-ljm are fairly new entries in the market and I haven't tried them yet, I'm taking note of Calvin's shoot-out comments and those of ATAUDIO.

The type of distortion spectrum you get with your own specific implementation of CFP stages is heavily, sharply dependent on bias and though lowest THD is achieved with just a trickle of of bias at the theoretical 12-15 mA level, great sound quality is possible with much higher current like 70-150 mA in the P3a, for example so perhaps we can take some expert analyses and generalised design comments with a grain of salt, in the sound quality areas at least. ;)