Spinoff from Motorola some years ago. Moto stayed for a while in the cellphone business. The discrete semis went to On Semi. Freescale makes uC chips and does research.
See for example: About Freescale
Hi,
It is strictly concept.
You need J-Fets with a high enough pinch-off voltage to bias the Bipolars, unless you explicitly bias them.
In this case you see two daisychained rush cascodes.
In this case it is a non-inverting gainstage with a Fet follower driving it.
A 2SK246BL/2SJ103BL will have around 4mA current with the BJT Bias and hence around 3mA/V transconductance which is passed on to the BJT's collector. Several common BJT and FET distortion problems are removed, no doubt we are adding some different distortions somewhere else, so that the eternal balance of the universe is preserved.
The concept also shows some freaky feedback approaches, for fun.
I probably would not even remotely attempt to build this, however it can serve to show some possible alternatives to the common way things are done.
BTW, adding one resistor and a choke to the output stage turns it into a quad style current dumper, which is way more effective for optimum bias or overbiased class AB than for pure class B...
Ciao T
It is strictly concept.
You need J-Fets with a high enough pinch-off voltage to bias the Bipolars, unless you explicitly bias them.
In this case you see two daisychained rush cascodes.
In this case it is a non-inverting gainstage with a Fet follower driving it.
A 2SK246BL/2SJ103BL will have around 4mA current with the BJT Bias and hence around 3mA/V transconductance which is passed on to the BJT's collector. Several common BJT and FET distortion problems are removed, no doubt we are adding some different distortions somewhere else, so that the eternal balance of the universe is preserved.
The concept also shows some freaky feedback approaches, for fun.
I probably would not even remotely attempt to build this, however it can serve to show some possible alternatives to the common way things are done.
BTW, adding one resistor and a choke to the output stage turns it into a quad style current dumper, which is way more effective for optimum bias or overbiased class AB than for pure class B...
Ciao T
No note like that. Spice shows added capacitance which by my interpretation ( potentially flawed understanding), makes the CCS not as effective. This is a small Naim. Are LED's much slower than small switching diodes?
Staying with just the ccs. Baby steps.
I also notice on just about every schematic the IPS ccs is a single transistor where most of the VAS ccs are a feedback pair. Because of the higher gain in the VAS?
Comments say green led's are more stable than red, but red non-linearities track opposite and compensate for the ccs BJT. Both comments sound reasonable. What are the thoughts here?
Some authors suggest to isolate the IPS and VAS ccs references. Some use the same reverence. Cost savings, but would not they have negative influence on each other? Or is the influence an advantage?
Not an issue of LED or diode "speed". They are providing a reference voltage of a cetain temperature coefficient. The lower Z the better. It is not tracking of "nonlinearites" that is of importance, it's the temp coefficient ~match. If one has a stage that is intentionally modulating the LED current, that's another matter altogether --- but that would be peculiar, and rely on even more non-guaranteed parameters.
Paralleling an LED of any visible color with two standard junction diodes is of no imaginable advantage it would seem --- unless you anticipate the LED failing open-circuit, in which case it's like wearing suspenders anticipating that your belt is going to fail. They lost the footnote, which would have added that for the two diodes the emitter resistor would be correspondingly smaller than for the LED, for the same collector current.
I doubt that sharing the reference has any advantage. Again, it's a voltage reference of modest capability.
Hi,
For this type of CCS more capacitance on the base is actually a "good thing". As these are open loop the amount of local feedback and thus the collector impedance are largest if and when the emitter resistance is lowest and thus the base resistance is lowest.
A very worthwhile improvement is to keep one of the Diodes and add an LM329 instead of the other and a re-adjusted emitter resistor with a big elcap (maybe 10uF/10V?) across the whole shooting match on the base. The 329 gives around 10 times the emitter resistance of the single diode and thus 10 times the effective CCS Impedance.
In the input stage we have no issue with the extra voltage dropped out. It's not as usable for the VAS Load, lest we use boosted rails.
For a Killer CCS take an additional J-Fet with at least 1V GS voltage at the CCS and enough voltage/power handling and cascode this suckah (gate to BJT base, source to BJT Collector, drain is new CCS Output. This trick is really cool for modding existing stuff...
It remains debatable how much of impact an improved CCS has though.
LM329 and cascoding can easily get you 50o times the impedance of the original though and a flat impedance to a much higher frequency.
Hard to say. Maybe because of dropout voltage or greater collector current?
I would avoid LED's, they are fashionable, but I find them too inconsistent. The days when you could just buy HP Red Ultrabright and have a really good and consistent Voltage source are over...
I think DS showed that especially the VAS CCS can influence the IPS one adding distortion. If anyone would spend a few cent on a small 1,000uF/10V (< 2 Ohm @ 100Hz) cap most of that would go away. So in a new design I would seek to optimise the CCS's individually while for an existing one with shared reference I'd probably just put the "magic cap" across the diodes and be done...
BTW, in Samuel Groner's commentry on Selfs APADH he proposes a small modification that turns the VAS current source into a Push-Pull VAS at high frequencies (only really usable with a buffered VAS though), which will help the slew rate quite a bit. I first saw this clever bit of trickery in an all Fet design by Jim Strickland (Hafler 915/100SE Preamp linestage), very clever.
If one was interested in polishing a t... ähhhhmmm standard "Lin" type amplifier rather than designing a better topology there are many of the "a resistor here and a cap there and a J-Fet buffer elsewhere" tricks that can dramatically change the circuits performance.
If you like, open a thread strictly on that subject invite me and elect a subject Amp and we can have fun...
Ciao T
For this type of CCS more capacitance on the base is actually a "good thing". As these are open loop the amount of local feedback and thus the collector impedance are largest if and when the emitter resistance is lowest and thus the base resistance is lowest.
A very worthwhile improvement is to keep one of the Diodes and add an LM329 instead of the other and a re-adjusted emitter resistor with a big elcap (maybe 10uF/10V?) across the whole shooting match on the base. The 329 gives around 10 times the emitter resistance of the single diode and thus 10 times the effective CCS Impedance.
In the input stage we have no issue with the extra voltage dropped out. It's not as usable for the VAS Load, lest we use boosted rails.
For a Killer CCS take an additional J-Fet with at least 1V GS voltage at the CCS and enough voltage/power handling and cascode this suckah (gate to BJT base, source to BJT Collector, drain is new CCS Output. This trick is really cool for modding existing stuff...
It remains debatable how much of impact an improved CCS has though.
LM329 and cascoding can easily get you 50o times the impedance of the original though and a flat impedance to a much higher frequency.
Hard to say. Maybe because of dropout voltage or greater collector current?
I would avoid LED's, they are fashionable, but I find them too inconsistent. The days when you could just buy HP Red Ultrabright and have a really good and consistent Voltage source are over...
I think DS showed that especially the VAS CCS can influence the IPS one adding distortion. If anyone would spend a few cent on a small 1,000uF/10V (< 2 Ohm @ 100Hz) cap most of that would go away. So in a new design I would seek to optimise the CCS's individually while for an existing one with shared reference I'd probably just put the "magic cap" across the diodes and be done...
BTW, in Samuel Groner's commentry on Selfs APADH he proposes a small modification that turns the VAS current source into a Push-Pull VAS at high frequencies (only really usable with a buffered VAS though), which will help the slew rate quite a bit. I first saw this clever bit of trickery in an all Fet design by Jim Strickland (Hafler 915/100SE Preamp linestage), very clever.
If one was interested in polishing a t... ähhhhmmm standard "Lin" type amplifier rather than designing a better topology there are many of the "a resistor here and a cap there and a J-Fet buffer elsewhere" tricks that can dramatically change the circuits performance.
If you like, open a thread strictly on that subject invite me and elect a subject Amp and we can have fun...
Ciao T
IMHO, the main trick is not to let the IPS current be affected by the VAS output voltage.
The pic below shows how to completely screw it up. If the output ever clips to the negative rail, Q3 saturates, the voltage across the diodes drops, the IPS is starved of current, and the output remains stuck to the negative rail indefinitely.
Attachments
Hi,
Correct. How do we fix it?
1K in line with the based of Q3.
And then maybe 1nF from the base of Q3 to that of Q5...
Ciao T
Correct. How do we fix it?
1K in line with the based of Q3.
And then maybe 1nF from the base of Q3 to that of Q5...
Ciao T
I probably won't be trying it myself, but it will teach me some.
As you have probably guessed, I am not too comfortable with FETs, I have done very little work with them. Perhaps this is an opportunity to catch up a little. Lord be praised, my good friend Oliver, who does my PCB artwork (I'm pretty good at it, but he's way better), loves FETs, MOSFETs and tubes, so I do have someone to check up on me.
One thing I think I see as definitely coming is use of FETs in the input stage as shown on my previous schematic, and in version "B", the one which has cascodes for their second stage of voltage amplification.
FETs make a lot of sense, and there's nothing to be gained by shying away from them, especially now that I have been told that those which are available to me will do quite nicely.
This also serves to remind me of my own view that after having the initial idea, the most important link in the development chain of anything is the feedback one receives if one asks for it. It's so easy to get locked up in one's own little world and miss out a lot that's going on and that is quite possible.
If I may ...
Sharing of whatever is, in my experience, most often (though not always) a matter of economy and ease of build than anything else. The industry is literally looking out for the number of resistors used, despite their ridiculously low per piece price, but multiply that infinitesimal by a sufficient series number and it starts to make a difference.
Another problem is stratifying your product line. If you make a cheaper model too good, your more expensive model won't sell. I am a beneficiary of just such a situation. In their day, Dual made their TT model CS604 so well, that their top model dropped in sales very significantly, and, unseen in Germany for decades before that, an order list was formed, you had to wait for your unit. People are not fools, it was soon discovered that the difference in sound between the two was say 2 or 3%, but for double the price. All you needed was to remove Dual's 204 cartridge and put in something decent, in my case Ortofon, and you had a TT worth every penny it cost, and then some.
It appeared in July 1977 (or 1978, not sure) and was withdrawn by November of the same year, meaning it was replaced by model CS 606 - which was not half the model it replaced, but it restored the sales of their biggest model. I bought mine in September. I still have it. Never even a hiccup.
These things happen from time to time.
My view is that each function inside a circuit should have its own electronics and should share as little as possible. Obviously, this will make for a more complex and costly circuit. On rare occasions, it does make sense to share something; look back at my schematic and you'll see the same zener diode supply the required 12V to both the cascode stage and the constant current transistor, the catch being that this is the input stage where there are least problems with things like loads, etc, assuming you got it right and in view of the fact that I have split power supplies.
Sure, you can replace a 0.6W zener with a 1.5W one and then share, but if anything goes wrong, it will go wrong in two places rather than just one. Not a good idea.
Last edited:
Here's a not totally-irrelevant discussion:
http://www.diyaudio.com/forums/pc-based/208898-ripping-cds-safemode-sounds-much-better.html
A respected audiophile swears he can hear large differences in "detail, tone, depth and clarity" (very familiar phrases in all listening tests, it seems to me) between 'rips' of audio files depending on whether the PC is in safe mode or not when the 'rip' is performed. The files appear to be identical, yet blind tests apparently show that other people can hear the difference too.
I feel that the bluff of the listening-tests-are-all-that-matters fraternity has been called. Do you go along with this, or do you dismiss it out of hand because, in this case at least, you 'know' that the listening tests must be flawed? Would you be anywhere near as sceptical if listening tests had 'proved' something more obscure and less contentious?
http://www.diyaudio.com/forums/pc-based/208898-ripping-cds-safemode-sounds-much-better.html
A respected audiophile swears he can hear large differences in "detail, tone, depth and clarity" (very familiar phrases in all listening tests, it seems to me) between 'rips' of audio files depending on whether the PC is in safe mode or not when the 'rip' is performed. The files appear to be identical, yet blind tests apparently show that other people can hear the difference too.
I feel that the bluff of the listening-tests-are-all-that-matters fraternity has been called. Do you go along with this, or do you dismiss it out of hand because, in this case at least, you 'know' that the listening tests must be flawed? Would you be anywhere near as sceptical if listening tests had 'proved' something more obscure and less contentious?
I'd want to read his article first.
If he describes his test setup and procedures in enough detail to determine that there's no obvious flaws (as reported) and to allow opportunities to replicate, then I'd sure be interested in trying to do so. Colloms is an odd duck. He'll say sensible things, then turn around and say stuff that's so stupid you just know he doesn't believe it and just wants to continue his career selling books and articles to the audio-porn audience. Sometimes it's hard to tell which Colloms you're dealing with.
Well, a DVD should behave the same as an audio CD in theory.
Test that hypothesis by making a copy of the original. Then copy the copy to end up with 5 copies. You will easily see that the 5th copy is considerably less sharp and defined than the first copy.
This proves the point that despite what we are told, a copy is NOT exactly like the original. If you accept that, then it stands to reason that there may be people who actually can hear the difference between the original and a copy.
But because it's just some people and not everybody, the logic is made relative. What do I care if it's a copy, so long as I can't hear the difference?
If the files are identical, then the sound is identical. Period. Now, what are we doing when we rip? We are reading less than perfect media and we are changing the data to a different encoding scheme. Bad ripping software has a possibility of making incorrect calculations in encoding or error corrections. Could how busy the computer is effect this? It should not, but it is not impossible within the laws of physics.
Simple test. Rib one cut with a busy computer. Rip it again in safe mode. Do a binary compare. If they are different, then the listening camp has made a valid point that they need to get better software and a better computer. If they are identical, they are blowing it........
I use EAC to copy CD's.
Simple test. Rib one cut with a busy computer. Rip it again in safe mode. Do a binary compare. If they are different, then the listening camp has made a valid point that they need to get better software and a better computer. If they are identical, they are blowing it........
I use EAC to copy CD's.
Was there a preference? Wait, I see there was.
Well, sounds like "next stop homeopathy!"
Yes Colloms is an odd one. I'd say I raised my eyebrows more when he got affiliated with NXT than when he went on, with detailed numerical ratings and expanded vocabularies, about component sound.
Wait, I see there was. Well, sounds like "next stop homeopathy!"
Yes Colloms is an odd one. I'd say I raised my eyebrows more when he got affiliated with NXT than when he went on, with detailed numerical ratings and expanded vocabularies, about component sound.
Last edited:
@dvv
"Test that hypothesis by making a copy of the original. Then copy the copy to end up with 5 copies. You will easily see that the 5th copy is considerably less sharp and defined than the first copy"
I am curious to know how the DVD/copy apparatus knows it is dealing with a "music file"?
When data copies (eg binaries) are made one can go copying ad finitem without error, unless the copy process is faulty. If this wasn't true, the whole computer industry would grind to a halt.
"Test that hypothesis by making a copy of the original. Then copy the copy to end up with 5 copies. You will easily see that the 5th copy is considerably less sharp and defined than the first copy"
I am curious to know how the DVD/copy apparatus knows it is dealing with a "music file"?
When data copies (eg binaries) are made one can go copying ad finitem without error, unless the copy process is faulty. If this wasn't true, the whole computer industry would grind to a halt.
- Status
- Not open for further replies.