Winston OBoogie
Well-known member
...
Folded-Cascode Amplifier (Dead Thread)
- Thread starter Winston OBoogie
- Start date
Help Support GroupDIY Audio Forum:
G
Guest
Guest
Thanks a lot for the circuit and explanation. It helps to clear up some questionsI had, especially on biasing. I will read and ponder now.
tk@halmi
Well-known member
Thanks John! I just got up and seeing this is like starting the day with eating chocolate in the morning.
> not intended for driving extremely heavy loads due to the somewhat lean 6mA biasing of Q09 & Q10.
That... and the open-loop gain is something like Rl/82, more like Rl/100 counting JFET Gm.
With IHF 10K load, 10,000/100= gain of 100, feedback gain is 3, 30dB NFB, fine.
With 600 ohm load, 600/100= gain of 3, feedback network is 3, almost no feedback. And it will need it because it will have to strain to the limit to make 6V peak.
As a hi-fi line amp, or maybe to buffer 10K faders, great.
That... and the open-loop gain is something like Rl/82, more like Rl/100 counting JFET Gm.
With IHF 10K load, 10,000/100= gain of 100, feedback gain is 3, 30dB NFB, fine.
With 600 ohm load, 600/100= gain of 3, feedback network is 3, almost no feedback. And it will need it because it will have to strain to the limit to make 6V peak.
As a hi-fi line amp, or maybe to buffer 10K faders, great.
bcarso
Well-known member
[quote author="Winston O'Boogie"]
Q05, Q06, Q07 & Q08 are current mirrors and any decent low noise transistor will work for these. Ideally, Q05 should be matched with Q06 and Q07 with Q08. The amount of current reflected from the left hand side transistors (Q05 and Q07) to the right (Q06 & Q08) is, in this instance, not 1. We have a 1:3 ratio due to the values of emitter resistors R04/R05 & R06/R07.
[/quote]
Because of this current gain (which means the bipolar stages could be called current amps rather than current mirrors) the open loop gain (i.e., transconductance) is multiplied by that gain amount.
One caveat: The 2SJ74 on the output is being run near its specified abs max drain gate voltage at zero signal, and outside of the rating when the output goes negative a bit. I mentioned this to Fred F. in connection with his all-FET op amp, and he had noticed it and found that in practice things worked o.k. So as usual Toshiba is conservative in their specifications. I would screen the candidate parts for breakdown before installation.
The SK170 is challenged a bit but not nearly as much as the P device.
EDIT: Sims show the transconductance to be about 90mA/V. The analysis is a bit trickier than first meets the eye.
Q05, Q06, Q07 & Q08 are current mirrors and any decent low noise transistor will work for these. Ideally, Q05 should be matched with Q06 and Q07 with Q08. The amount of current reflected from the left hand side transistors (Q05 and Q07) to the right (Q06 & Q08) is, in this instance, not 1. We have a 1:3 ratio due to the values of emitter resistors R04/R05 & R06/R07.
[/quote]
Because of this current gain (which means the bipolar stages could be called current amps rather than current mirrors) the open loop gain (i.e., transconductance) is multiplied by that gain amount.
One caveat: The 2SJ74 on the output is being run near its specified abs max drain gate voltage at zero signal, and outside of the rating when the output goes negative a bit. I mentioned this to Fred F. in connection with his all-FET op amp, and he had noticed it and found that in practice things worked o.k. So as usual Toshiba is conservative in their specifications. I would screen the candidate parts for breakdown before installation.
The SK170 is challenged a bit but not nearly as much as the P device.
EDIT: Sims show the transconductance to be about 90mA/V. The analysis is a bit trickier than first meets the eye.
Winston OBoogie
Well-known member
...
Winston OBoogie
Well-known member
...
Winston OBoogie
Well-known member
...
Svart
Well-known member
i do have to add one thing, nothing electrical, but rather applauding your explaination of the circuit! to often around here we get schematics and everyone adds their .02$ but rarely do we get an in depth and practical explaination of a circuit as you have done. this is MUCH easier for me to understand than reading 10 books from cover to cover.
many thanks!
:guinness:
many thanks!
:guinness:
bcarso
Well-known member
Wow it is hot here too. My a/c went out in most of the place last week, so I am holed up in the bedroom with a portable one. The max temp predicted today was 95F and right now it is 97F in the shade. Still moderate humidity though, so I am really not complaining.
The caveat about the output FETs (and even the inputs for small closed loop gains, thus potentially large input swings even before overload) is that when the output swings, the voltage across one of the FETs gets larger---in the extreme case the P part for example could see almost the rail-to-rail voltage (source at slightly less than 25 - Vled, drain at as low as -25 + Vled + a bit).
Fortunately the current is limited by the other FET, so breakdown, if it occurs, is likely nondestructive.
The caveat about the output FETs (and even the inputs for small closed loop gains, thus potentially large input swings even before overload) is that when the output swings, the voltage across one of the FETs gets larger---in the extreme case the P part for example could see almost the rail-to-rail voltage (source at slightly less than 25 - Vled, drain at as low as -25 + Vled + a bit).
Fortunately the current is limited by the other FET, so breakdown, if it occurs, is likely nondestructive.
> about 90mA/V
OK, 100 ohms, 11 ohms, same within an order-of-magnitude.
Actually, I totally screwed-up. I saw the 82R5 and lept to a conclusion. Yeah, it is something like K170BL's GM. And I missed the X3 factor. I didn't fool with the X2 and /2 factors, though they do seem to cancel out.
So taking Gm of one JFET as 1/33Ω gives your result and seems plausible for the 170.
> For all I know, it's DEEPLY flawed in some way
It has way too many costly single-source parts.
The ratings may be insufficient.
No DC stability at all.
Otherwise, it seems to do the job.
> Mr. Thegard relied, maybe rightly so, on the transcendental equation for determining the mirror ratio with a single emitter resistor in the left hand transistor.
For 2:1 ratio you want 20mV offset; so for 3:1 you need around 28mV offset... ah, but at what current? Your Q01-Q04 current is ill-defined. Your resistors hold a 3:1 mirror ratio over a wider range of current. I don't know which of Norm's plans you looked at, but he may have used actual known-current sources.
OK, 100 ohms, 11 ohms, same within an order-of-magnitude.
Actually, I totally screwed-up. I saw the 82R5 and lept to a conclusion. Yeah, it is something like K170BL's GM. And I missed the X3 factor. I didn't fool with the X2 and /2 factors, though they do seem to cancel out.
So taking Gm of one JFET as 1/33Ω gives your result and seems plausible for the 170.
> For all I know, it's DEEPLY flawed in some way
It has way too many costly single-source parts.
The ratings may be insufficient.
No DC stability at all.
Otherwise, it seems to do the job.
> Mr. Thegard relied, maybe rightly so, on the transcendental equation for determining the mirror ratio with a single emitter resistor in the left hand transistor.
For 2:1 ratio you want 20mV offset; so for 3:1 you need around 28mV offset... ah, but at what current? Your Q01-Q04 current is ill-defined. Your resistors hold a 3:1 mirror ratio over a wider range of current. I don't know which of Norm's plans you looked at, but he may have used actual known-current sources.
Winston OBoogie
Well-known member
...
Marik
Well-known member
Anybody ever saw a tube implementation of the folded cascode? I have been looking for one for awhile.
Winston OBoogie
Well-known member
...
Winston OBoogie
Well-known member
...
bcarso
Well-known member
[quote author="Winston O'Boogie"]
I have about 4 or 5 batches of the J74, dates are spread out quite a bit. When I can, I think I'll take a few from each and see what it takes on average to cause a catastrophic breakdown. It's worth the sacrifice :twisted:
[/quote]
Don't sacrifice them! Put a megohm or thereabouts in series with the gate and ramp away on the volts. You'll see the breakdown and roughly how sharp it is without destruction (you can tie drain and source together for the other end---it isn't exactly the same as the normally biased conditions but it's a good indicator).
I screened some 74's myself when I was frustrated about how restrictive the 25V rating is. If memory serves they managed about 40V without incident, although I was not directly measuring gate leakage which can climb a lot for both N and P devices in this family. Your mileage may vary.
I have about 4 or 5 batches of the J74, dates are spread out quite a bit. When I can, I think I'll take a few from each and see what it takes on average to cause a catastrophic breakdown. It's worth the sacrifice :twisted:
[/quote]
Don't sacrifice them! Put a megohm or thereabouts in series with the gate and ramp away on the volts. You'll see the breakdown and roughly how sharp it is without destruction (you can tie drain and source together for the other end---it isn't exactly the same as the normally biased conditions but it's a good indicator).
I screened some 74's myself when I was frustrated about how restrictive the 25V rating is. If memory serves they managed about 40V without incident, although I was not directly measuring gate leakage which can climb a lot for both N and P devices in this family. Your mileage may vary.
bcarso
Well-known member
[quote author="Winston O'Boogie"]
I was cruising Erno Borbely's site for any info and I saw this newer article I hadn't read before:
http://www.borbelyaudio.com/pics/305borbely-new.pdf
It's a similar circuit to the one I posted above and he states that it is a folded-cascode. However, on fig. 4 from the article it seems to me that the gates are being driven by signal from the diff. amp. as are the sources. He claims a similarity to the JC-2 by John Curl. The Curl circuit is indeed similar (BJT's are used) but the bases on that are not fixed and it's not a folded-cascode according to Mr. Curl himself. Either in reference or in execution, is Erno actually wrong? :shock: What do you think? PRR, Brad, any of the learned ones out there[/quote]
Erno is speaking somewhat loosely---it's more of a single-Q differential amp, and tricky in that regard for achieving balance of current in the input devices.
Many are the ways...
I was cruising Erno Borbely's site for any info and I saw this newer article I hadn't read before:
http://www.borbelyaudio.com/pics/305borbely-new.pdf
It's a similar circuit to the one I posted above and he states that it is a folded-cascode. However, on fig. 4 from the article it seems to me that the gates are being driven by signal from the diff. amp. as are the sources. He claims a similarity to the JC-2 by John Curl. The Curl circuit is indeed similar (BJT's are used) but the bases on that are not fixed and it's not a folded-cascode according to Mr. Curl himself. Either in reference or in execution, is Erno actually wrong? :shock: What do you think? PRR, Brad, any of the learned ones out there[/quote]
Erno is speaking somewhat loosely---it's more of a single-Q differential amp, and tricky in that regard for achieving balance of current in the input devices.
Many are the ways...
Winston OBoogie
Well-known member
...
> I was being a bit tongue-in-cheek
Me too. "Too many ... parts" is a no-problem for one-off work. "over ratings" can be demonstrated (or not) with bench-abuse. And you did mention a servo (more parts!!).
> current is ill-defined. .... Bit of a loosey-goosey arrangement isn't it?
Works in production for several companies. Sure can work in one-off projects.
Me too. "Too many ... parts" is a no-problem for one-off work. "over ratings" can be demonstrated (or not) with bench-abuse. And you did mention a servo (more parts!!).
> current is ill-defined. .... Bit of a loosey-goosey arrangement isn't it?
Works in production for several companies. Sure can work in one-off projects.
Winston OBoogie
Well-known member
...
Similar threads
Latest posts
-
-
-
-
Germanium circuit: 500 or 19" rack? For dirt/drive or colouration.
- Latest: stylesforfree
-
-
