It is not small problem for a follower. 6dB down (meaning half of voltage) from 200 to ~1.3k and another 6dB down (meaning only a quarter) at 3k. Need to trace where the voltage drop develops.
Last edited:
Thats a very nice post, and exactly relevant to the scenario too, maybe i should have seen it sooner
Thanks
Wow thanks, i will take you up on the offer, but i will pay you something (shipping+extra)
Maybe that would fix it, and i dont know what else to do more now, but when i was making all modifications to other channel, i changed LD just to see if it was bad luck or what..
Now we have 3 completely identical FR's from 3 different LD's...even if they werent good units, what are the odds of having the same exact failiure mode?
I checked with phone now as signal generator, same curve, so its not the interface for sure
Yeah, not a small problem, probably in some active device
not that it'll help , but try ( in future ) to switch to greater resolution in graphs
who care for zillion db on ordinate, when we want curve in window of db or 2
test of your pcbs: remove source follower/level shifter, remove Lu + Cascode
place plain N Channel (IRFP240) up, accordingly connected and try it as simple complementary follower OS
I mean, circuit is proven, it really is just matter of your particular build, finding where is actual ookoop
who care for zillion db on ordinate, when we want curve in window of db or 2
test of your pcbs: remove source follower/level shifter, remove Lu + Cascode
place plain N Channel (IRFP240) up, accordingly connected and try it as simple complementary follower OS
I mean, circuit is proven, it really is just matter of your particular build, finding where is actual ookoop
Good point, noted
I am a bit layman at this stuff, but ok removed irf510 level shifter, remove LU (but irfp240 is its cascode, if i remove that then i lose complementary follower in the end), place irfp240 up not sure what is meant. Probably whole layer of my knowledge missing in order for it to "click" what is intended
This is only thing i could get that runs in the sim
removed bd139 and bc550 not sure if needed
remove everything needed to try it as Babelfish M25 R.2
that will exclude level shifter and LU outa game
that will exclude level shifter and LU outa game
Huh.
Even 75$ (DDP) oscilloscope would reveal in minutes where problem is, without any component removal or circuit provisional changes. Consider this or something better for all future projects: entry level oscilloscope
Yupp...dammit, why God give the most esoteric problems to his least bright soldiers
I was thinking, ok circuit is proven, circuit works 100%. I copied schematic, from what i checked 10 times now, correct.
So it "must" be fault of specific implementation itself, either pcb (unlikely? would have to be a massive ookup to have this result) or components.
So tried now changing the dc block caps on input, thinking, maybe they have some weird transfer characteristic (curve looks kind of analog), but it wasnt it either.. Only notable thing i can think of here, the bigboy "steel rod" cap is 4700u instead of 3300u, because AC analysis said fine and mentally i thought fine, but now god knows.
Resistor values i checked, LT3092 cant be because it biases fine, optos cant be because offset sets fine. IRF510 and ludef isnt because we just tested that. Basically tore down, changed or whatever'd 90% of amp and its still there.
So either its something incredibly silly and unrelated to amp, God hates me and cant have nice things, power mosfets, or some combination of above. I just realised i spent like at least 3 days continously doing little else but this, embarassing. I remembered mosfets measured fine before installing so probably i think this is end of the road, but i think i have some irfp250/9240 laying around....
I kind of have oneandonlyitis for "like SIT but cheap" thing now, alephs feel too normal and ok (i had AJ and mini A3 already)...from here, i dont know. Maybe some schade like steepleguy or something
Wanted to get one of those USB hantek's but somehow always delayed it and said ok dont need now...problem is when you need it (like now) it's too late then. Should have been smarter..
Still, problem seems pretty absurd, im not sure where i'd begin even if i had scope, unless just by poking randomly something obviously sticks out as wrong
It’s an exemplary absurd situation.
As both channels measure exactly the same, and changing JFETs or simplifying circuit brings no change whatsoever, it is not likely that any semiconductor is defective. Babelfish output stage is so simple that I don’t see what could you have done wrong if biasing works.
When I receive absurd measurement results, after checking circuit, next step is to verify again measurement equipment and setup.
Check again Motu loopback response. I don’t know if Motu has input overvoltage protection and could line input be damaged.
With oscilloscope (use Motu + REW) check signal generator level at 5 kHz and then connect signal to amplifier input. If signal level drops, amplifier input is heavy loading Motu output. Check if there is no drop at 200 Hz. If so, Motu is unable to drive this follower.
Is there a problem with Motu output or amplifier input impedance above 400 Hz, is another question.
As both channels measure exactly the same, and changing JFETs or simplifying circuit brings no change whatsoever, it is not likely that any semiconductor is defective. Babelfish output stage is so simple that I don’t see what could you have done wrong if biasing works.
When I receive absurd measurement results, after checking circuit, next step is to verify again measurement equipment and setup.
Check again Motu loopback response. I don’t know if Motu has input overvoltage protection and could line input be damaged.
With oscilloscope (use Motu + REW) check signal generator level at 5 kHz and then connect signal to amplifier input. If signal level drops, amplifier input is heavy loading Motu output. Check if there is no drop at 200 Hz. If so, Motu is unable to drive this follower.
Is there a problem with Motu output or amplifier input impedance above 400 Hz, is another question.
I agree with thinking here...at this point definitely, likely nothing is actually physically wrong and just chasing ghosts..must be something with setup as a whole.
That would make sense, but i tried driving amp with both aikido and phone, as well as motu's headphone-out and line-out. That's 4 different outputs giving identical result, if it was loading issue maybe at least they wouldnt all be 0.01dB matching curve i figure..
I tried your idea now too, play say 10khz, and measure output unloaded, then plug in and see if it drops (because its loading the output).
But for all frequencies i tried it remained same to within 0.1mV.
An interesting thing though i found out thoug is even when signal is flat, it shows lower and lower rms number with higher frequencies.
At one point out of frustration I checked Motu output with AC mode of multimeter (no scope sadly) unloaded and varied frequency, and AC level shows decreasing with frequency, in my head seems starts dropping at similar place like with amp measurements. So then i think ok must be interface, but when i actually make loopback of the interface, perfectly flat, measurement of aikido preamp, flat, its headphone output, flat. And then i take my phone jack output (which when plugged into motu and sweeping, stays flat) and do the same multimeter AC mode and it shows same behaviour of dropping the higher the frequency (or motu and xiaomi are conspiring). So, perhaps maybe its just a red herring and how true RMS measurement works with multimeters (or at least this one), maybe there's less energy in higher frequencies so it shows as lower number, but peak is actually of the same amplitude. So i dont know in the end.
Average multimeters can’t measure correct RMS value past 200 Hz and would show measured value constantly drop with frequency increase. Better true RMS multimeters have specified upper frequency for correct AC voltage measurement.
Now, we have confirmed that signal voltage at 10 kHz doesn’t drop at amplifier input. Next is to verify AC signal level at the LD1014 and IRFP9240 gates. Is it the same or there is a drop?
You can use REW with FFT measurement and dBV scale. Set level to -10 dBV using unloaded generator output, and then connect signal to amplifier input and check level for any level difference at the at mentioned points.
Be careful not to connect Motu input directly to rails voltage. It could be fine as there is a capacitor at input but I don’t know how good voltage spikes clamping Motu has, if any.
Now, we have confirmed that signal voltage at 10 kHz doesn’t drop at amplifier input. Next is to verify AC signal level at the LD1014 and IRFP9240 gates. Is it the same or there is a drop?
You can use REW with FFT measurement and dBV scale. Set level to -10 dBV using unloaded generator output, and then connect signal to amplifier input and check level for any level difference at the at mentioned points.
Be careful not to connect Motu input directly to rails voltage. It could be fine as there is a capacitor at input but I don’t know how good voltage spikes clamping Motu has, if any.
I'm slowly getting caught up on 2 year absence and found this thread, very cool as I have what I think are the perfect P chan FETs. 2SJ115 What do you think? I bought these probably 20 years ago, and I can't remember why but I ended up with 2SJ115 O Vgs range and 2SK415 Y Vgs range.
Check out the data sheet, I think it should pair well with the LU1014D.
Cheers
PS I forgot I also have 2SJ49 and 56, similar Vgs. Looks good although TO-3, not 247.
Last edited:
always good to have various parts on disposal, but I decided to abandon future Schade-ing of part in negative rail of amp, simply because in already established source follower OS arrangement, it brings 2nd as dominant, but positive phase
so, depending of need, when follower/level shifter is needed - doing that in positive side, all scenarios are covered with plain simple IRF510 or DN2540 (DN2530) ....... 2nd dominant intended result but also negative (more liked) phase of same
where TF! I can buy 2SJ115 and 2SK415 these days, for peanut money?
(quoted in your post was exact chase for small mosfet , serving as source follower/voltage shifter/driver for output mosfet in negative rail ....... and question irrelevant for this thread, but relevant for Babelfish M25 R.2 ; I'm perfectly happy with IRFP9140 as "complementary" partner for cascoded LU)
so, depending of need, when follower/level shifter is needed - doing that in positive side, all scenarios are covered with plain simple IRF510 or DN2540 (DN2530) ....... 2nd dominant intended result but also negative (more liked) phase of same
where TF! I can buy 2SJ115 and 2SK415 these days, for peanut money?
(quoted in your post was exact chase for small mosfet , serving as source follower/voltage shifter/driver for output mosfet in negative rail ....... and question irrelevant for this thread, but relevant for Babelfish M25 R.2 ; I'm perfectly happy with IRFP9140 as "complementary" partner for cascoded LU)
where TF! I can buy 2SJ115 and 2SK415 these days, for peanut money?
Perhaps kind person from Vernon BC might send care package to behind iron curtain.
Maybe I'm confusing the requirements with DEF thread, I thought the P channel FET Vgs match and tempco was critical for this design??...so much to keep straight in old brain...it pays to keep up daily with threads and not let 2 years pass before reading again.
Cheers
tnx, no tnx - don't need more unobtanium parts ......... my fun is to play with toyz all Greedy Boys are able to get
my personal needs for gadgets of my use are long ago fulfilled .......
so, if you don't need them, make happy some Needy Boy in neighborhood, greater good than sending them to (Old) Geezer
........that post (inquiry for P channel TO220 mosfets) was really related only with Babelfish M25 R.2; this thread was (most likely) frequent in that moment, and I simply cried where crowd was .......
my personal needs for gadgets of my use are long ago fulfilled .......
so, if you don't need them, make happy some Needy Boy in neighborhood, greater good than sending them to (Old) Geezer
........that post (inquiry for P channel TO220 mosfets) was really related only with Babelfish M25 R.2; this thread was (most likely) frequent in that moment, and I simply cried where crowd was .......
Ha!!, the ideal part combination for F5 tuned for neg H2, intoxicating highs and bandwith of a video amplifier.
@Dendrobium
It is almost certain that problem is your measurement setup and REW settings.
FR response you get is exactly (within fraction of dB) what looks REW output of a perfectly flat amplifier or it’s loopback, if 1/48 smoothing + Flat-Top window are used. Smoothing reduces measured level with frequency increase.
However, automatic distortion and FR measurement using stepped sine option, ignores smoothing options from the standard FFT measurement setup in my REW version, so question remains how is that happening in your case.
Do you have latest REW version 5.20.13?
P.S.
Listening test should immediately make clear does that -15 dB at high frequencies really exist.
It is almost certain that problem is your measurement setup and REW settings.
FR response you get is exactly (within fraction of dB) what looks REW output of a perfectly flat amplifier or it’s loopback, if 1/48 smoothing + Flat-Top window are used. Smoothing reduces measured level with frequency increase.
However, automatic distortion and FR measurement using stepped sine option, ignores smoothing options from the standard FFT measurement setup in my REW version, so question remains how is that happening in your case.
Do you have latest REW version 5.20.13?
P.S.
Listening test should immediately make clear does that -15 dB at high frequencies really exist.
Last edited: