I hope so, for the listening pleasure of the DIYers.
We are still missing some symmetrical middle power versions of CFAs to can bridge them easy for high power and double slew rate.
If i can risk a point on our actual situation, highly subjective, but based on my experience:
On its simplest variations Single INPUT, VAS OPS, CFAs are able to give more than decent distortions results, very musical, plus killing results on both bandwidths and slew rates.
Most of the attempts to improve distortion numbers, by adding stages *inside the loop* in order to cancel local distortions, like in some of the very instructive Dadod studies, will be paid by instability (High Q peaks at the upper limit) or noticeable slew-rate limitations or/and tricky compensations. And, i more than believe, some signal definition lost.
As simplest designs can give, omho, distortions far from the point were we are able to notice them, i think it is the best way to begin with.
The quality of CFAs are lightened with listening, not by distortion numbers.
As far as CFAs and I are in concern, the *simple is beautiful* (in signal paths) law is gold.
While VFA designers are used to care first about distortion, when we come to VFA designs, we have to change our minds, and *care at square waves* first in CFAs, both little signals and saturated (slew rate).
Caution that, in real world, some very high frequencies oscillations can occur, due to the high bandwidths: you sometimes need a 100mHz oscillo to notice them.
Notice too that CFAs are often more requiring about layout good practices.
And quality of some important parts: Low noise supplies of input+VAS, high quality resistances in feedback network...
Agree ?
Last edited:
Esperado:
You mentioned this demanded good layout practices, can you suggest a layout to be followed to ensure that is followed? Any layouts on the forum that would be a good starting point of design?
You mentioned this demanded good layout practices, can you suggest a layout to be followed to ensure that is followed? Any layouts on the forum that would be a good starting point of design?
to begin:
Star for grounds. At least, take the input +VAS grounds from the common ground point where the ground wire come, free from any high current, or use a separate one going to the general star point of the amp.
Same-thing for feedback networks. They have to be taken *at* the loudspeaker wiring point.
Take care of very symmetrical tracks, in OPS (same lengths). etc..
And, for everithing, never forget that any track is an inductance+resistance, and build a capacitance with the tracks aside.
This as a practical example: http://www.diyaudio.com/forums/vend...lateral-mosfet-amplifier-128.html#post3655825
Look at the images at the bottom.
Last edited:
It looks really good. Esperado.
What voltage coefficient do you expect from the BAT54???
- Sonny
Not the slightest idea. ;-)
Just enough to compensate exactly the input stages ?
(Where is the clown icon ?)
I ask because i have only seen about 1mV.
But an improvement could be exchanging the the position of D1 and D2 so that they are soldered close to base pin of the KSC3503 and KSA1381. It would give you a good thermal coupling as the wire from the base would heat the solderjoint of the diode. Just an idear.
See the attachment.
I like to run the frontend with an lower voltage and cascode the diamond second stage to be able to reach the higher working voltage.
It allows the use of jfet like J111 - J113 and J174 - J176 for cascoding in a simple way. or the use of small signal bipolar.
But an improvement could be exchanging the the position of D1 and D2 so that they are soldered close to base pin of the KSC3503 and KSA1381. It would give you a good thermal coupling as the wire from the base would heat the solderjoint of the diode. Just an idear.
See the attachment.
I like to run the frontend with an lower voltage and cascode the diamond second stage to be able to reach the higher working voltage.
It allows the use of jfet like J111 - J113 and J174 - J176 for cascoding in a simple way. or the use of small signal bipolar.
No no.... dont!! 😀
Actually. +/-45VDC is a good compromise.
I have been using my amps at a +/-35VDC. But it has not enough dynamic range at that voltage.
It just start to get muddy very fast. But at +/-45Vdc (80 - 90Watt) it starts to fit better at normal (okay a little bit above) listening levels.
But my Raidho C1.1 just need some juice!
Actually. +/-45VDC is a good compromise.
I have been using my amps at a +/-35VDC. But it has not enough dynamic range at that voltage.
It just start to get muddy very fast. But at +/-45Vdc (80 - 90Watt) it starts to fit better at normal (okay a little bit above) listening levels.
But my Raidho C1.1 just need some juice!
By the way, i am doing a cormercial highend CFA amp. In scary contrast to VSSA, i am hitting 280parts per channel + power supply.
I know it is a lot, but it has only a trimmer for bias. The rest is kept in place.
I know it is a lot, but it has only a trimmer for bias. The rest is kept in place.
We share the desire for simple circuitry and very fast - CMA domain -- Would you like to try to reduce the circuit by two more transistors?
Remove the input pair (Q7,10). Tie bases of Q13,14 together for input.
Use CCs to bias the Q13,14... to do this you need to add a small amount of Re in each emitter of Q13 and Q14. (maybe then increase R4, to get back gain). If chosen well, distortion null can occure (2H).
Apply CCS current to Q13 and Q14 emitter. The CCS thru the added re will develope the .6v needed to bias the new input stage.
Massage it and you will have a very good performer with 2 fewer transistors.
Thx-RNMarsh
It is time to compare the simple cfa to real amplifier implementation. What is the differents of sound character if one is optimise by slew rate and the other optimise by THD....
Is the preservation/alteration of waveforms not the performance summarizing both slew-rate and THD and directly in relation with the sonic character ?
Seems that some dont realize that speed doesnt equate precision,
although , quite ironicaly , the CFA contenders are now relying
on THD numbers to brand such amps as good , i guess it will take
a month a two before one starts to ask about IMD and so on.
(underlining is by me)
And as IMD and THD have the same origin, the loop will be closed.
I *started* from this, Richard.
The problem is the CCS+ fixed gain. Reason why L.C. used the big caps. You can do-it by removing the CSS too, saving 6 tranies, but will be not so good.
I don't care the added 2tranies+2res. They are not in the loop, they bring isolation between the input and the loop, they are transparent, they provide higher impedance, they do not add distortion (< -150dB), they are not expensive. At least, less than the big caps.
On a finished prototype, i would add at least two more tranies to provide a cap multiplier to the input+VAS stages.
+1bimo said:
While i'm quite confident with the answer to had done previously this kind of study...
Opening some hidden posts, I thought irresistibly of "Statler and Waldorf" from The Muppet show.
Ok Thats fine. Buffered input outside of loop. You got the big caps out. That is a Big plus. Was looking for direct comparison of all the numbers with same transistors, currents etc. Because i have been able to get (measured - not sim) results without buffer of well under -100dB (<.001%) in real world. I always compare with real world eventually. However, if you can do that with listening tests, then Ok also. Unfortunately, I can't say one is better based on SIM work along. But the SIM has shown what might be possible to achieve. It is the only topology I have used since introducing it to audio and the DC servo several decades ago. The only competing design would be my friend John Curl's topology which has been refined and refined in real (not SIM) world development. This is where the other topologies have a head start.
For myself, these are all old topologies now. But I went thru the same process to get here. My goal is only to have many people try and compare against the prevailing VFA topology/design which has been the dominant circuitry and most books cover it and not the less well know CFA. Now, I think there is enough going on here at DIYAudio to show everyone that the CMAmp is extremely competitive for audio use.
Now to the slew rate -- it needs some listening to -- to see if it is a 'marker' for better sound. Need ideas on how to do this is an apples to apples way. perhaps same CMA with adjustable low - mid-high SR?? or what can we come up with? NOT a different topology like VFA vs CMA. That has too many variables.
Thx-RNMarsh
For myself, these are all old topologies now. But I went thru the same process to get here. My goal is only to have many people try and compare against the prevailing VFA topology/design which has been the dominant circuitry and most books cover it and not the less well know CFA. Now, I think there is enough going on here at DIYAudio to show everyone that the CMAmp is extremely competitive for audio use.
Now to the slew rate -- it needs some listening to -- to see if it is a 'marker' for better sound. Need ideas on how to do this is an apples to apples way. perhaps same CMA with adjustable low - mid-high SR?? or what can we come up with? NOT a different topology like VFA vs CMA. That has too many variables.
Thx-RNMarsh
Last edited:
PS - Note to Esperado: Recall you always thought fewer transistors/simpler circuits sounded best. That's the only reason i suggested trying it without buffered input regardless of the SIM numbers. Is the buffer really transparent? You could build and listen to just the buffer portion and tell if that is true. Maybe listen to amp both ways to decide before accepting an assumption.
-RM
-RM
Last edited:
Some shunt variable capacitance somewhere in the loop, taking care it does-not introduce distortion by running the previous stage out of current ?
[edit] I will do-it soon, with my own amps witch is not enough destroyed, when we will receive our VSSA to keep some music safe during those surgeries...
Last edited: