LuDEF

[Zen Mod]

if you have missing 6K8, know that it's all me, and you have my apologies

error I made on packing list was fixed, eventually

regular 0207 resistor size is perfect

just ignore wattage - some are calling them 1/4W, some are calling them 500mW and some are calling them 600mW

as I'm using those in my schematics, even 250mW will suffice

 

[pfdavid]

if you have missing 6K8, know that it's all me, and you have my apologies

error I made on packing list was fixed, eventually

regular 0207 resistor size is perfect

just ignore wattage - some are calling them 1/4W, some are calling them 500mW and some are calling them 600mW

as I'm using those in my schematics, even 250mW will suffice

I was hoping for a simple oversight, whew! Easily fixed. Thanks! And having fun. Such a fantastic hobby and community.

 

[Dendrobium]

I tried my luck making own LuDEF for my own heatsink format.
For shi$% n giggles decided no front end, so like current buffer (thus NuDEF...cause its bare). Idea is i will keep it modular and not commit to one front end baked into the amp, but just put my own preamp or whatever in front (was thinking Aikido...mmm toobz). A lot of time i listen under 2W, so maybe even line-out from DAC can work.
Anyway here is what i came up with, after few layout attempts tried to keep things logical and also have continous ground plane with no traces interrupted

There's 2 LU's on the sch because you only mount one, one is for SMD under pcb, other for off-board ( i just wanted to be flexible, used offboard in the end).
And here is how it came out (ok, looks like s..., moving on ):

Only substituion is KSC1845 instead of BC550, being mindful of different pinout, everything else should be as perscribed by Dr.ZM

Here is finished thing.. i use SMPS for cheap power, then just filter out everything with juma style cap multiplier (the purple boards)...but capmult does not approximate a good DC source, i'll switch to superreg based thing soon. Anyway, it works fine for all amps i had in here so far ( like, mini Aleph 3, AJ, etc)
Everything biased up fine, went to 150mV (Iq - 1.38A) because smps starts being uncomfortably hot past that..DC offset few mV, LU VGS about 2.4V

 

[Dendrobium]

Went to do some measurements and fine tune LU VGS...idea was plug it in REW, play sine wave, then tweedle deedle the pot until it is happiest?
But i come across very unexpected problem.
Frequency response not come close to flat whaaat?

Any guess as to how this could be? I've never seen something like it, loopback of just the audio interface (MOTU M2) that has the dac and adc shows flat as a plank.
Is this what you get when you buy aliexpress LU?

 

[Zen Mod]

ali LU measuring as real deal?

I mean - for prescribed Iq and Vds, what is LU Ugs?

not everything of needed facts, but first necessary one

 

[tombo56]

Frequency response not come close to flat whaaat?

Could be an effect of MOSFETs input capacitance on MOTU output, as there is no high impedance input buffer. I have no experience with MOTU, but using headphones output to drive amplifier would be a way to check if capacitance load is the culprit.

 

[Dendrobium]

Sorry i just realised i kept writing VGS above, i meant Vds whole time.
Ok, so for 1.38A Iq and 2.44 Vds, LU Vgs is 1.466V (not sure how to get Ugs now)
One suspicious thing i tested is LU's performance (i.e distortion and FR) are basically identical from 2.3 to 3.1 Vds (this are the extremes pot will allow me to set apparently).

I did not have the opportunity to curve trace them beforehand (no equipment), frankly i picked random 2 from the pile because all i have is one of those crappy Ali small signal transistor tester which tests far from real operating condition, but it did confirm at least these are n-jfets. Oh, i made my own LU/LD mounting thingies too..

They are actually LD1014, i call them LU just to not confuse as it should be same thing different package, but since now there is problem i have to disclose that just in case.
If anyone is willing to match them i will give all away in return for 2-3 matched pairs.

Could be an effect of MOSFETs input capacitance on MOTU output, as there is no high impedance input buffer. I have no experience with MOTU, but using headphones output to drive amplifier would be a way to check if capacitance load is the culprit.

Ooh, good idea! I checked it now but result is the same, same delta of about 15dB, but was worth a shot...

EDIT:
One thing i noticed is...the amp itself seems to very faintly make the same sound as what is being played..i can only hear it when everything is dead quiet and my ear is next to it, but it's definitely physically making the noise its playing electrically (so, sine waves, multitones, etc)...
Basically, internally oscillating somwhere maybe?
But FR stays identical even at lowest volumes, and distortion stays good even at highest volumes, so i dont know..

 

[Zen Mod]

have a scope?

 

[Dendrobium]

Hmm, no unfortunately.
Sometimes i use scope in REW, but with 192khz sample rate of soundcard, bandwith is 98khz...hardly a scope
Also the thing i mentioned with it making a faint noise...upon really sniffing with my ears it actually comes from the SMPS of all things, not the amp boards..even though there is capmult between them. I cant even think how that could be, that class A amp makes PSU sing... i wonder if its even related to the FR problem, or just sideproblem..

 

[Zen Mod]

have any buffer-like stage to put temporary on input , and check freq. that way?

 

[Dendrobium]

I have Aikido preamp...i did add cathode follower mode on it (untested, if you dont hear from me call fire department) that is unity gain buffer.
I will test it with that now then (dont you love English language).

 

[tombo56]

They are actually LD1014, i call them LU just to not confuse as it should be same thing different package, but since now there is problem i have to disclose that just in case.

Those LD1014D are real deal. I bought a batch of 50, as some other members did.

However, several were nonfunctional, so I have a suspicion that those LD1014D are some sort of factory rejects or leftovers.

For curve tracing, I used programmable voltage supply, setting fixed Vgs and varying Vds from 1.5V to 3.0V. Curves were all around the place and it was possible to get only 3 tightly matched pairs. In the end, I’ve assembled LuDEF using Papa’s gifted LU1014, obtained through group buy.

 

[Dendrobium]

Looks like its really something with amp... this FR behaviour remains no matter what is changed
One bruteforce ""fix"" is to just EQ inverse of this, then you get something approximating flat but...somehow im not happy underlying problem is still there
As a bonus you get this bizzaro distortion profile

Those LD1014D are real deal. I bought a batch of 50, as some other members did.

However, several were nonfunctional, so I have a suspicion that those LD1014D are some sort of factory rejects or leftovers.

For curve tracing, I used programmable voltage supply, setting fixed Vgs and varying Vds from 1.5V to 3.0V. Curves were all around the place and it was possible to get only 3 tightly matched pairs. In the end, I’ve assembled LuDEF using Papa’s gifted LU1014, obtained through group buy.

Hmm interesting, good to know they're at least possibly ok.
I think sounds about right, these arent that well matched either, for same Vds i have about 100mv different Vgs roughly.
But im guessing these 2 in particular arent broken or anything, since their behaviour is identical and in ballpark of real part. Beggars cant be choosers i figure, so i'll try to live with it..still, i cant even begin to think how FR would end up like this, what can even cause it other than some analog filter (not present) or something.

 

[Zen Mod]

well, considering that I personally measured zero specimen LuDEF (see inserted image), only conclusion is that there is some omission on your schm , or just pcb

I remember tombo56 did his own measurements

 

[tombo56]

Were other amplifiers, that were using the very same power supply (SMPS + Cap Mx) measured in the same way for their frequency response? Question is could PS somehow affect frequency response?

I’m not impressed with that PS:

• Small SMPS close to max. load
• Cap Mx seems to have no capacitors at the output and +/- rails don’t have the same capacitors??
• SMPS transformer ‘singing’ audibly in tune with amplifier output

It is worth to mention that LuDEF has very asymmetric AC load current, where positive rail will have about 2.6 times larger current than negative one.

 

[tombo56]

well, considering that I personally measured zero specimen LuDEF (see inserted image), only conclusion is that there is some omission on your schm , or just pcb
I remember tombo56 did his own measurements

As is usual case, production specimens measure better than T-bed and LuDEF has -3 dB at 38 – 40 kHz.

 

[Zen Mod]

Were other amplifiers, that were using the very same power supply (SMPS + Cap Mx) measured in the same way for their frequency response? Question is could PS somehow affect frequency response?

I’m not impressed with that PS:

• Small SMPS close to max. load
• Cap Mx seems to have no capacitors at the output and +/- rails don’t have the same capacitors??
• SMPS transformer ‘singing’ audibly in tune with amplifier output

It is worth to mention that LuDEF has very asymmetric AC load current, where positive rail will have about 2.6 times larger current than negative one.

well ZM Hawkeye didn't even got to spot those caps lacking

another case of (Casey🙂 - Your ambition outweighs your talent.......... , when speaking of over-enhusiastic made CapMulti, without final C Bank

not blaming, have my share of same mistakes

As is usual case, production specimens measure better than T-bed and LuDEF has -3 dB at 38 – 40 kHz.

yup, exact purpose of T-Bed confirmed - approximation of pretty much lousiest possible build, and it must behave

 

[Dendrobium]

Were other amplifiers, that were using the very same power supply (SMPS + Cap Mx) measured in the same way for their frequency response? Question is could PS somehow affect frequency response?

I’m not impressed with that PS:

• Small SMPS close to max. load
• Cap Mx seems to have no capacitors at the output and +/- rails don’t have the same capacitors??
• SMPS transformer ‘singing’ audibly in tune with amplifier output

It is worth to mention that LuDEF has very asymmetric AC load current, where positive rail will have about 2.6 times larger current than negative one.

well ZM Hawkeye didn't even got to spot those caps lacking

another case of (Casey🙂 - Your ambition outweighs your talent.......... , when speaking of over-enhusiastic made CapMulti, without final C Bank

not blaming, have my share of same mistakes

Thank you for roasting me (reality check) and well deserved
Yes used the same SMPS and capmult for aleph J (ZM's version), my_ref (ok this is class AB) and Mini aleph 3 (single output stage) had same smps but no capmult, all were flat..but i agree and also keen observation of everything. This little bugger has had a hard life at this point because the A3 and AJ also ran it close to its limit, its case getting warm like..65C ish. Though i must say, the singing thing is new and didnt happen before. I was quite surprised it was the smps singing and not the amp board, makes no sense to me since class A means it should see constant load more or less.
I'll populate some more caps (especially on the output after capmult stage, there are 2-3 positions), i was originally lazy with this because i fast realised i wanted superreg or something like this feeding amp but here we are almost 2023 with this still. Cant find high current superreg for both pos and negative rails (though @grunf says something is in oven).

One thing i will also test tommorow, running just one channel at a time. It barely gets warm with one (i originally set operating points one by one), it can handle one no problem, then we'll see.

[...] only conclusion is that there is some omission on your schm , or just pcb

I keep thinking must be too, did triple check sch in #963 to yours in post one, and seems ok...i remember i did ton of transient and AC analysis in ltspice, i guess that's all for nothing (ok, squiggly lines pretty to look at).

 

[Zen Mod]

as tombo reminded us, LuDEF is special case of DEF OS stage

characteristic of proper DEF(like) stage (absolutely basic one is pure Square Law) is one part having much lesser (or much higher) xconductance, which results in wonder antiphase function of two output parts ...... where speaker is getting difference in their current, not sum of it

while usual specimens are having depletion mode part/SIT in upper part of OS, this one is opposite - cascoded/not degenerated Lu is pretty much super-mosfet behaving, having ridiculous xconductance, thus mosfet in lower rail is actually having significantly lesser xconductance

eyeball technique (read - LTSpice) is saying that Lu is having approx triple xconductance vs. IRFP9xxx ; was too lazy to confirm the same observing with my venerable scope (did that while prototyping SissySIT), too much of cheapskate to invest in modern smart ones, having all Gizmo as bal probes, floating as option etc. ....... even old equipment is smarter and more precise than I am

and, that's result with any OS having non-equal xconductance parts, side with greater xcon will have greater current swing; for DEF logically antiphase behavior starting when diff. is double

what you're hearing as singing in switcher is exactly that - output xformer singing with difference of its two winding currents

not saying lack of cap bank after cap multi is having anything with discrepancy in amp freq. characteristic ..... that's certainly another issue

 

[Dendrobium]

Thank you ZM and Tombo for that, i just checked it in ltspice too and checks out (not that i doubted, but i did miss this first time), gives some food for thought, especially this what you say transformer singing with difference of currents... that really does make sense now, thank you making me at least slightly less confused and dispelling at least a little ignorance from my mind...food for thought..even if i am never amp wizzard like you