• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

Update for Elektor Crescendo Power Amplifier 1982

I don't believe in ghosts I can't ear ;). I'm just chasing 50 or 100Hz hum.
Esperado, is the drawing below a good understanding of your last advises ? And what is better: W1 like in original, or W2.
PS: whatever the changes, R3 remains a complete mystery to me.
 

Attachments

  • wiring input.PNG
    wiring input.PNG
    19.9 KB · Views: 1,271
Last edited:
Esperado, is the drawing below a good understanding of your last advises ? And what is better: W1 like in original, or W2.
PS: whatever the changes, R3 remains a complete mystery to me.
Not complicated. In the source (preamp) ground and cold wire connected to the ground output plug. On power amp, shield not connected, and cold used as ground.
Just a question, if you short cut your inputs, with no preamp connected, did-you still have hum ?

R3 is just to allow ground reference of the input stage to be connected directly to the star ground reference, without creating loop (current across the 10 Ohm will be near null) while allowing a good enough reference when nothing is connected.
Is is because the ground connection to the power parts are high current, so, even a low z of cabling can create a voltage in it. Spécially at HF, where wires present an inductance.
 
Yes , i did build one a few months after publication and experimented oscillations with slightly reactive loads....rapidly fixed the thing this way.
My amp, before modified to CFB was flat, with no overshoots on 20Khz sqare waves after derating. Sure, decreasing the cascode pole cutoff will solve any oscillation problem, but don't you find it is sad to use a cascode to avoid capacitive load on the previous stage and increase bandwitch, then add a cap to decrease bandwitch and add a capacitive charge to the input ? With some expected TIM increase ?
Too, i REALY wonder why you have so different sim results than mine.
Attached my sim for your 22pf. (with no input filter)
Just for understand, HumNSmoke had yet solved his 1MHz oscillation. And is talking about hum, not the typical noise when some hf osc.
 

Attachments

  • vfb-cresc.gif
    vfb-cresc.gif
    30.8 KB · Views: 1,237
Last edited:
One thing is for sure, this amp can be marvelous or awfull. But because its poles are very near one from the other, it is has an open loop with a low pass filter of near 18dB/oct at his Hf limit. We flirt with Nyquist limits for stability, and it is not exactly the kind of amp for soldering and play. Need very often some derating in operating situation.
 
Sure, decreasing the cascode pole cutoff will solve any oscillation problem, but don't you find it is sad to use a cascode to avoid capacitive load on the previous stage and increase bandwitch, then add a cap to decrease bandwitch and add a capacitive charge to the input ? With some expected TIM increase ?

The miller compensation will decrease the Vas output Z , so
what is lost in gain is traded for better VAS linearity and driving capability, hence the distorsion will not rise , in the audio band at least.

Too, i REALY wonder why you have so different sim results than mine.
Attached my sim for your 22pf. (with no input filter)

Dont know , perhaps the fets spice models.

Just for understand, HumNSmoke had yet solved his 1MHz oscillation. And is talking about hum, not the typical noise when some hf osc.

Hum can be the result of oscillations or badly implemented ground circuit.
Often ground return using the main ground connection is the culprit ,
no wonder that symmetrical XLR connections are the rule in pro gear.

One thing is for sure, this amp can be marvelous or awfull.

Yes , at the time it was a breakthrough in amateur dedicated designs.
 
Last edited:
The miller compensation will decrease the Vas output Z
I was not talking about HD, but TIM, not on the cascode stage (with has enough current at 1Mhz anyway) but input stage, with has only ~300µa of collector current.

About the peak we get without miller comp, a little cap (around 3pf) can do the job in parallel with the FB resistance to flatten the bandwidth, and so, help to stability. Zobel is an other point where we can turn the open loop phase, with no negative effects. I would try those first, and use miller in last resort.
Of course, Miller cap would be a requisite for an industrially produced amp, where we want no trouble.
I believe, in a DIYer forum, each one prefer to tune his amp near the limit ?
 
Last edited:
...
Just a question, if you short cut your inputs, with no preamp connected, did-you still have hum ?
-1 °K :D hum with inputs shortened, this for each channel. When just both or only one unplugged, very very light hum and shuffle, but heavy hum when both-plugged-both-side (no matter the preamp is on/off/powercord removed). With no ground, earthquake :eek: but this is normal :p with such a powerfull amp isn't it?

R3 is just to allow ground reference of the input stage to be connected directly to the star ground reference, without creating loop (current across the 10 Ohm will be near null) while allowing a good enough reference when nothing is connected.
Is is because the ground connection to the power parts are high current, so, even a low z of cabling can create a voltage in it. Spécially at HF, where wires present an inductance.
Not sure I really understand as there is already a big gauge wire between ground board input and the star with nearly zero current inside. Do you mean this wire is say shortened by R3 at HF?
OK, I get the car now to Bricodépot to buy some copper bar for the supplies.
Added the earth link below.
 

Attachments

  • wiring input2.PNG
    wiring input2.PNG
    20.8 KB · Views: 1,170
Last edited:
Hi Esperado
We killed the hum :). Thank you for your help. Two no-concession stars did the job.
One last question please: I'd like now beginning to play with my DCX2496, plugging it between the preamp and the amps. All my home made is unbalanced, and when shorting inputs pins 1-3 does not frighten me at all, I wander if I can/must do the same for the outputs. This is not very clearly stated in Behringer's doc. Shorting the so called "cold" pin3 to ground permanently make me afraid to kill the output OP-amp. Should I instead leave the pin3 unconnected?
 

Attachments

  • wrong way.PNG
    wrong way.PNG
    13.8 KB · Views: 1,168
  • wiring input2.PNG
    wiring input2.PNG
    19 KB · Views: 1,167
  • PB300975.jpg
    PB300975.jpg
    576.7 KB · Views: 514
Shorting the so called "cold" pin3 to ground permanently make me afraid to kill the output OP-amp. Should I instead leave the pin3 unconnected?
Yes, you can. (No risk to short circuit neither: there is a 60 Ohm serial resistance )
But the best would be, or modify the DCX to bypass the output Opamp, or to charge the - output by the same resistance than the one of your input amp (31K).
nb: The audio stages of the DCX2496 are not top of the world.
 
Last edited:
...or to charge the - output by the same resistance than the one of your input amp (31K):

What an elegant idea man, great, this makes the difference between the centurion and the optione :wiz: . The charges will find room in XLRs until I decide to use balanced everywhere, because with all these stereo boxes it is difficult to remove all ground loops without using tweaked dedicated cables, ground here/not here.
BTW, no more need the 100/200Ω from ground(s) to earth to mute the hum.
Esperado said:
nb: The audio stages of the DCX2496 are not top of the world.
Yes, I read this... but shall I ear? I'll wait for the warranty to expire before I modify it ;)

Now next step, choosing amps for 8Ω/8Ω meddle/treble. Any advise about the power for them, suited with the power of Crescendo for bass (8Ω too)? (knowing I can't afford high-end and my -40dB 50 years old ears half broken by too many hard-rock concerts :rolleyes:... but I also must think to youngest at home).
What about this 50W or 100W or just LM3886 ?
 
Yes, I read this... but shall I ear?
If you run very dynamic and detailed speakers; OH YEAH !
Now next step, choosing amps for 8Ω/8Ω meddle/treble. Any advise about the power for them, suited with the power of Crescendo for bass (8Ω too)? (knowing I can't afford high-end and my -40dB 50 years old ears half broken by too many hard-rock concerts :rolleyes:... but I also must think to youngest at home).
What about this 50W or 100W or just LM3886 ?
Do yourself a favor: Run a Good (and cheap) Class D amp for basses (ex: Ncore400). And reserve your crescendo for trebles :)
 
I'm not an audiophile (read my signature).
I discovered the Otala's article near the 80s, while we where studying effects of speed on servos since 1970, and produced at this time an amp with one of the first 10Mhz Epitaxial power transistor for that reason.
I don't believe in words, or magazine papers. I believe in what i do have experienced and measured.
But, you are not about bring back this boring CFB/VFB controversy on this thread ? Do you ?
 
Last edited:
My point is that unless an amp enter slew rate limitation there will
be no other conditions under wich slew induced distorsion (SID/TIM
or in french "distorsion d intermodulation transitoire) could be measured.

As for an eventual CFB/VFB polemic about it, please no more prejudices
and other "proces d intentions" wich have no relevance with what i did
write earlier in this thread.....
 
My point is that unless an amp enter slew rate limitation there will be no other conditions under wich slew induced distorsion (SID/TIM
or in french "distorsion d intermodulation transitoire) could be measured.
My point was: Audible distortions (harmonic, intermodulation, phase modulation), in a non linear open loop amplifier -they are all because components non linearities- depend greatly on the ratio signal slew rate / amp slewrate in a closed loop.
It is some kind a exponential curve. So, faster the transmission in the loop, better it is.
There is no point where you can say: "that's enough".
You can verify this, comparing input/output of musical signals in any amplifying device with a very fast comparator. (i mean substracting the input signal from the output one). It is enough to look and listen to the results to be aware of this. We spend weeks on this kind of evaluations.
May-we stop arguing on this subject, far from the Crescendo point, it was just to explain why i preferred to use less miller cap as possible?
 
Last edited: