Andy and Sonny, thanks for your emails. Andy, 2 uF without overshoot or ringing, how many amps of ANY kind will manage this? Rather far from a difficult speaker.
I don't expect that CFA is substantially better than a VFA but it caught my interest. The headphone amp I designed was a complete "desktop product" and bam it worked. No instability or other bad behaviour.
I gather that much that a high power amp, CFA style, is much harder to make good.
BTW: What do you call a product which is only designed in a computer or by pen and paper and no part is tested before the pcb is made? The pcb is made and bam it works. In swedish desk = "skrivbord" =>"skrivbords produkt"
Sometimes "skrivbordprodukt" is a negative word. Things that are produced but they have not been developed too much....
I don't expect that CFA is substantially better than a VFA but it caught my interest. The headphone amp I designed was a complete "desktop product" and bam it worked. No instability or other bad behaviour.
I gather that much that a high power amp, CFA style, is much harder to make good.
BTW: What do you call a product which is only designed in a computer or by pen and paper and no part is tested before the pcb is made? The pcb is made and bam it works. In swedish desk = "skrivbord" =>"skrivbords produkt"
Sometimes "skrivbordprodukt" is a negative word. Things that are produced but they have not been developed too much....
peranders said:
I see that you and I have the same tendency toward overkill 😉. I believe it can be done.
I don't know of any English phrase that corresponds directly to that idea. There is a phrase that describes the approach of going bit overboard on the analysis side of things. It is "analysis paralysis". I've sometimes heard this phrase used by management types who don't completely appreciate what's required to get a product right the first time, especially when the requirements of the product are difficult. But sometimes this phrase can be right on target too...
He may not know you........
But we do!
My experience with them for power amps is that eventually it became necessary to isolate the output with a series inductor.....a few nH worked. I think.....
Nothing we did in the output stage seemed to effect the way it measured.......distortion, bandwidth, slew rate, etc......blah, blah.
CFAs for amp or low-level stuff sounded the same, and while very good.....there was something odd about them. We abandoned them.
Have fun anyway.
Jocko
But we do!
My experience with them for power amps is that eventually it became necessary to isolate the output with a series inductor.....a few nH worked. I think.....
Nothing we did in the output stage seemed to effect the way it measured.......distortion, bandwidth, slew rate, etc......blah, blah.
CFAs for amp or low-level stuff sounded the same, and while very good.....there was something odd about them. We abandoned them.
Have fun anyway.
Jocko
I like also nice step response and I will give it a try to achieve that but I also give it up when it's a clear overkill and I don't succeed...andy_c said:
Help... simulation guys!
How do I get a frequency sweep as plot with SwitcherCAD? I have messed around with a file which had it in the beginning. I get only time at the X-axis instead of frequency.
I notice also that the high impedance node is critcal and also the feedback resistor (all accordning to theory!)
How do I get a frequency sweep as plot with SwitcherCAD? I have messed around with a file which had it in the beginning. I get only time at the X-axis instead of frequency.
I notice also that the high impedance node is critcal and also the feedback resistor (all accordning to theory!)
you have done a transient response. You need to do an .AC like
.AC DEC [step] [start] [stop]
.AC DEC 100 100 100meg
for 100step/decade start=100Hz, Stop = 100MHz
Set "AC" voltage on your source to 1mV
.AC DEC [step] [start] [stop]
.AC DEC 100 100 100meg
for 100step/decade start=100Hz, Stop = 100MHz
Set "AC" voltage on your source to 1mV
Does the SPICE model consider the variations of fT with collector current?
Using a Wilson current mirror doesn't add a thing compared to using a simple one. Is this really true in the real world?
Now I have located the choice of step analizes and sinus sweep. The program is not a wonder of user interface 🙄
Using a Wilson current mirror doesn't add a thing compared to using a simple one. Is this really true in the real world?
Now I have located the choice of step analizes and sinus sweep. The program is not a wonder of user interface 🙄
Re: He may not know you........
Jocko, can you develop the "something odd", please?Jocko Homo said:
He's back!!!!
Good.....I can send the requests I get for simulation help to you, ol' buddy.
Per:
Not sure that any of the variations in current mirrors/sources made much difference. Except maybe in the realm of how likely they were to drift.
About the sound...........long story short......the dealers really hated the first product that we made with CFB. Nearly caused a full-scale revolt.
They complained about dynamics. Lifeless, unemotional, all that subjective stuff that drives us engineers crazy. One of them insisted that I buy a particular CD to hear for myself why he didn't like it.
I'll spare everyone the name of the CD, but let us say that I didn't like his CD and we parted ways.
Later, we developed a power amp with CFB, in conjunction with another company. I sent protos out for evaluation. Same complaints. Some were astute enough to figure out that it sounded exactly like that previous product that we made. And none had any idea either used CFB. Of course, they wouldn't know what CFB was anyway.
My buddy did market his version. I received calls from people who heard its debut at CES, and suggested that I talk him into not relaesing it. Same complaints.......
unemotional......uninvolving.....lifeless. Again, from people who knew nothing of its topology.
Any help? No......probably not. Best that I can do.
Jocko
Good.....I can send the requests I get for simulation help to you, ol' buddy.
Per:
Not sure that any of the variations in current mirrors/sources made much difference. Except maybe in the realm of how likely they were to drift.
About the sound...........long story short......the dealers really hated the first product that we made with CFB. Nearly caused a full-scale revolt.
They complained about dynamics. Lifeless, unemotional, all that subjective stuff that drives us engineers crazy. One of them insisted that I buy a particular CD to hear for myself why he didn't like it.
I'll spare everyone the name of the CD, but let us say that I didn't like his CD and we parted ways.
Later, we developed a power amp with CFB, in conjunction with another company. I sent protos out for evaluation. Same complaints. Some were astute enough to figure out that it sounded exactly like that previous product that we made. And none had any idea either used CFB. Of course, they wouldn't know what CFB was anyway.
My buddy did market his version. I received calls from people who heard its debut at CES, and suggested that I talk him into not relaesing it. Same complaints.......
unemotional......uninvolving.....lifeless. Again, from people who knew nothing of its topology.
Any help? No......probably not. Best that I can do.
Jocko
So what you are saying is that if they knew it was CFB they were very unhappy and if they only listened not knowing it was alright?
Jocko, did you agree with them? Is CFB useless according to you?
Jocko, did you agree with them? Is CFB useless according to you?
Oki doki, seems that they didn't like it regardless but maybe the CD was the cause of it?
But what had happened if they didn't know at all what they were lsiten to, a real blind test?
But what had happened if they didn't know at all what they were lsiten to, a real blind test?
Jocko,
You make some good points. Would you say that if you kept the number of stages in a CFB amp to a minimum the situation would improve? How about a similar topology with drastically reduced feedback or none at all?
Regards,
Jam
You make some good points. Would you say that if you kept the number of stages in a CFB amp to a minimum the situation would improve? How about a similar topology with drastically reduced feedback or none at all?
Regards,
Jam
I would not say that it is useless, just that its key point is that bandwidth remains (relatively) constant as a function of gain, is not all that important here. So.....in retrospect, there may not have been the need to try it. But I did. You don't learn unless you try.
So....Per.....please try it. Tell us what you think. Sonics, as well as how it functions.
The people who squawked about it knew absolutely nothing about how anything works. They were dealers, that should explain it all. But they all said pretty much the same thing, they didn't know each other, and that so many others were complaining. If they all perceived it in pretty much the same way, then there must have been some merit to their beefs.
I solved the problem in the first case by taking pretty much the same circuit......building it discretely.....and eliminating the feedback loop. Problem solved. Many said it was the best thing I had ever made.
(I have similar theories on VFB, but let's leave that to another time.)
In the case of the power amps, I was surprised to discover that:
A. It sounded just like the first product, which used an AD846 (they were new, and easier to get and cheaper then.)
B. Nothing that I did......current in the front-end.....number of output devices......you name it.......seemed to have any effect at all on how it sounded.
So Per.......please press on, so we can mull it over with you. We all might learn something.
Just don't expect it to supplant all your previous efforts.
Jocko
So....Per.....please try it. Tell us what you think. Sonics, as well as how it functions.
The people who squawked about it knew absolutely nothing about how anything works. They were dealers, that should explain it all. But they all said pretty much the same thing, they didn't know each other, and that so many others were complaining. If they all perceived it in pretty much the same way, then there must have been some merit to their beefs.
I solved the problem in the first case by taking pretty much the same circuit......building it discretely.....and eliminating the feedback loop. Problem solved. Many said it was the best thing I had ever made.
(I have similar theories on VFB, but let's leave that to another time.)
In the case of the power amps, I was surprised to discover that:
A. It sounded just like the first product, which used an AD846 (they were new, and easier to get and cheaper then.)
B. Nothing that I did......current in the front-end.....number of output devices......you name it.......seemed to have any effect at all on how it sounded.
So Per.......please press on, so we can mull it over with you. We all might learn something.
Just don't expect it to supplant all your previous efforts.
Jocko
Muzak Sonics ?....
I have heard amplifiers that fit your above description, and more recently one in particular that measures pretty well, and does not care a hoot about what speaker or cable you hang off it.
My evaluation after a few hours listening on all kinds of music was that many minor nuance details in music that I know well were missing or at least diminished, and that it was overall sonically uninteresting.
My violinist mother's comment on the way home was that it was 'dead' and 'boring' and had no 'life'.
Jocko, what was your sonic evaluation of the prototypes ?.
Eric.
Was that a product of being too good and too clean - by this I mean dealers, users etc often like amplifiers that are strongly load dependent and the dynamic sonic artifacts that are produced.
I have heard amplifiers that fit your above description, and more recently one in particular that measures pretty well, and does not care a hoot about what speaker or cable you hang off it.
My evaluation after a few hours listening on all kinds of music was that many minor nuance details in music that I know well were missing or at least diminished, and that it was overall sonically uninteresting.
My violinist mother's comment on the way home was that it was 'dead' and 'boring' and had no 'life'.
Jocko, what was your sonic evaluation of the prototypes ?.
Eric.
The first product used an AD846 as the I/V in a CD player. I don't think that load had anything to do with it.
When I changed over to the next version of the I/V, I didn't feel that the previous version should have been bashed so severely. But then, I didn't have, or care to have, a dozen different brands to compare it too.
As for the amps.....I felt that unless it was clearly superior to what we were building at the time....it wasn't....that there was no reason to pursue it.
In all fairness, in both cases, CFB did a lot of things right. Its downfall was the one thing it was perceived to do poorly was one that no one was able to live with.
I hope that Per...or someone.....will make some of these things. I think we are all interested in how it turns out.
So.....Per.....put away your simulator that you probably are about as confused with as I would be.....(no, wait.....he seems to understand it better that me).......and just build one!
Jocko
When I changed over to the next version of the I/V, I didn't feel that the previous version should have been bashed so severely. But then, I didn't have, or care to have, a dozen different brands to compare it too.
As for the amps.....I felt that unless it was clearly superior to what we were building at the time....it wasn't....that there was no reason to pursue it.
In all fairness, in both cases, CFB did a lot of things right. Its downfall was the one thing it was perceived to do poorly was one that no one was able to live with.
I hope that Per...or someone.....will make some of these things. I think we are all interested in how it turns out.
So.....Per.....put away your simulator that you probably are about as confused with as I would be.....(no, wait.....he seems to understand it better that me).......and just build one!
Jocko
but some CFB power amps seem to do well in the marketplace
There are several commercial CFB designs which seem to be selling well and get good reviews in "golden ear" magazines like the German magazine Stereoplay. Maybe it's because the marketing people told the reviewers that their amps had this revolutionary topology. I believe Accuphase and either Electrocompagniet or Thule are among those using CFB.
There are several commercial CFB designs which seem to be selling well and get good reviews in "golden ear" magazines like the German magazine Stereoplay. Maybe it's because the marketing people told the reviewers that their amps had this revolutionary topology. I believe Accuphase and either Electrocompagniet or Thule are among those using CFB.
Accuphase indeed use CFB and is considered as a pioneer in such topology. Early Electrocompaniet surely did not use it, do not know what have happened later though.
Pedja
Pedja
> Noone knows, not even the experts here?
No. It seems obvious to me, but I wanted to see if anybody raised any point I missed.
Refering to the schematic at the top of the thread, with Q3 Q4 on the Vp input and Q1 Q2 on the Vn input:
Per, I know your work and I think you can design Q3 Q4 just like any other amplifier. They should carry enough current to drive Q1 Q2 with authority, yet not so much current that your input bias current is too high (complementary means you only need to keep the difference in bias current low). Then you consider noise: high current low Rb for low-Z sources, low current high Beta for hi-Z sources.
Speed will suffer if the Q3 Q4 current is not high enough to smack the Q1 Q2 bases around. If Q1 Q2 flow 1mA, then 0.1mA is probably plenty in Q3 Q4 (assuming Beta of 100 or more).
Noise may be high in any case, because you have four transistors in the input comparator. OTOH a typical voltage-feedback amp may need some compensation that will degrade noise, so a CFB amp is not always bad.
> a long print which came out from a VAX computer. But I suppose the SPICE models have been developed and also the computers....
Print-out? Original SPICE tended to use stacks of punch-cards.... what fun!
The actual SPICE module still works on punch-cards emulated by a text-file.
> The program is not a wonder of user interface
"Better than punchcards" seems to be the motto. But they just hide the punchcards deep in dialog boxes-- it was almost easier to have a bunch of cards punched with standard analyses and shuffle them in and out of the deck than to go through a GUI. I do think it helps a lot to have used the punch-card or text-file version before you use a GUI SPICE.
> Does the SPICE model consider the variations of fT with collector current?
Yes, incidentially. It models voltage variable junction capacitance, Gm, Hoe. However the Rb emulation is usually pretty poor. If you drive from a real zero-ohm source it will be wrong. In most real cases, the driver impedance swamps the device Rb.
> Using a Wilson current mirror doesn't add a thing compared to using a simple one. Is this really true in the real world?
I have an opposite trouble with the usual models: the collector impedance emulates much lower than I see with real devices. My 1-transistor current sources emulate much "softer" than I measure, and I often have to cascode to get performance in SPICE that I know can be done without cascoding with real transistors. That's with the rather old models that come with free SPICEs, I have not really checked new models such as Zetex's.
No. It seems obvious to me, but I wanted to see if anybody raised any point I missed.
Refering to the schematic at the top of the thread, with Q3 Q4 on the Vp input and Q1 Q2 on the Vn input:
Per, I know your work and I think you can design Q3 Q4 just like any other amplifier. They should carry enough current to drive Q1 Q2 with authority, yet not so much current that your input bias current is too high (complementary means you only need to keep the difference in bias current low). Then you consider noise: high current low Rb for low-Z sources, low current high Beta for hi-Z sources.
Speed will suffer if the Q3 Q4 current is not high enough to smack the Q1 Q2 bases around. If Q1 Q2 flow 1mA, then 0.1mA is probably plenty in Q3 Q4 (assuming Beta of 100 or more).
Noise may be high in any case, because you have four transistors in the input comparator. OTOH a typical voltage-feedback amp may need some compensation that will degrade noise, so a CFB amp is not always bad.
> a long print which came out from a VAX computer. But I suppose the SPICE models have been developed and also the computers....
Print-out? Original SPICE tended to use stacks of punch-cards.... what fun!
The actual SPICE module still works on punch-cards emulated by a text-file.
> The program is not a wonder of user interface
"Better than punchcards" seems to be the motto. But they just hide the punchcards deep in dialog boxes-- it was almost easier to have a bunch of cards punched with standard analyses and shuffle them in and out of the deck than to go through a GUI. I do think it helps a lot to have used the punch-card or text-file version before you use a GUI SPICE.
> Does the SPICE model consider the variations of fT with collector current?
Yes, incidentially. It models voltage variable junction capacitance, Gm, Hoe. However the Rb emulation is usually pretty poor. If you drive from a real zero-ohm source it will be wrong. In most real cases, the driver impedance swamps the device Rb.
> Using a Wilson current mirror doesn't add a thing compared to using a simple one. Is this really true in the real world?
I have an opposite trouble with the usual models: the collector impedance emulates much lower than I see with real devices. My 1-transistor current sources emulate much "softer" than I measure, and I often have to cascode to get performance in SPICE that I know can be done without cascoding with real transistors. That's with the rather old models that come with free SPICEs, I have not really checked new models such as Zetex's.
- Status
- Not open for further replies.