New mixed-device-type symmetrical folded cascode amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This is something I've been slowly piecing together since I started coming to these forums. It makes use of all three common types of transistor (BJT, JFET and MOSFET), making use of the good points of each. It incorporates many features inspired by threads here, such as the symmetrical folded cascode topology from the thread of the same name, current mirrors with gain from a distortion cancellation thread, and it is through this forum that I finally managed to find somewhere in the UK that sells the Toshiba dual JFETs that I've used..

I've posted information and pictures here.

It sounds pretty good to me (as if anything I made could possibly not sound good!), and it has 10 - yes, ten - LEDs, which guarantees it must be good:D
 

Attachments

  • amp_photo.jpg
    amp_photo.jpg
    61.3 KB · Views: 1,407
jacco vermeulen said:
Love the trace job, no idea how you do it...
I might put an article up on my PCB creation method one day. In short, I print out from the PCB layout software onto a piece of paper and use that as a template to drill holes (for SMT devices I use a small hammer and a nail to put an indent where the pads will be). Then it's just a matter of joining the dots with an etch-resist pen. It's a mix of art and science: I try to bear in mind what form might be best for a track and also draw them in a way that pleases my eye.


jacco vermeulen said:
..(Are the Cow speakers redone in Conolly by now, or still in the furr ? :clown: )
Fur forever!


Charles Hansen said:
...What are the driver MOSFETs (X1 and X2)?
I was supposed to explain that on my site, but I seem to have forgotten. They are ZVN2106A and ZVP2106A from Zetex. They seem pretty good, although not my ideal choice. What I really wanted were some rated for 2-3W so I could run them at about 50mA rather than the 20mA they are at now to better drive the input capacitance of the output FETs. However there seems to be a bit of a gap in what's available, with plenty of low-power devices and then a jump to power FETs of 20W+, which are a bit too big.
 
VERY nice, mr evil :)

looks like a neat project and great to see someone using all the free expertise that's been shared over the years here!

your board making technique brings back good memories. I used to used staedtler (sp?) markers, the black Lumicolor #317. Bought 'em buy the box at art supply stores. Those pens did a GREAT job on copper. The lumicolor pens and a couple of bubble stones on an aquarium pump in a pyrex dish made some fantastic boards very quickly!

drilling however, was ALWAYS painfully slow to me.

i used a spring load tapping device i purchased from Sears probably 30 years ago(!) to mark the holes.

heck, i even used to buy jfets from Dalbani when they were in the states. :)

mlloyd1

Originally posted by jacco vermeulen
Love the trace job, no idea how you do it.
Like the Roedersteins and the leds too.
Thanks for posting the Dalbani link, they do sell nice devices.
 
Mr Evil said:
I was supposed to explain that on my site, but I seem to have forgotten. They are ZVN2106A and ZVP2106A from Zetex. They seem pretty good, although not my ideal choice. What I really wanted were some rated for 2-3W so I could run them at about 50mA rather than the 20mA they are at now to better drive the input capacitance of the output FETs. However there seems to be a bit of a gap in what's available, with plenty of low-power devices and then a jump to power FETs of 20W+, which are a bit too big.

Did you look at the 2SJ76 + 2SK213 from Renesas (nee Hitachi)? These lateral parts are in a TO220 package and are very easy to heatsink.
 
Mr Evil,

Is there any significant DC offset at the output in your design? Do you think thermal drift would still generate any significant DC offset?

I ended up putting a servo on a recent symmetrical folded cascode design to make sure that 100+ db/watt speakers would not see any physical coil bias that the typical 50mv offset can cause. Naturally, there's a long story behind this........ Like you, I found the symmetrical folded cascode topology attractive for low voltage output applications.

The Ayre V3 amplifier topology includes a second symmetrical output stage to create a balanced output. It seems like a cousin to your cascode.
 
Charles Hansen said:


Did you look at the 2SJ76 + 2SK213 from Renesas (nee Hitachi)? These lateral parts are in a TO220 package and are very easy to heatsink.
They look quite interesting. Lower capacitance than other similarly rated devices is good. That was a particular consideration, since any increase in the speed of driving the output stage by choosing a power FET run at high current tends to be offset by them being slower devices. It's too late to change now, but I'll keep them in mind for future projects.



LineSource said:
...Is there any significant DC offset at the output in your design? Do you think thermal drift would still generate any significant DC offset?..
Offset is about 3.5mV (which implies that it would be ~38mV without the DC feedback cap) and it's the same immediately after being turned on as after heavy use, so it doesn't seem to change much with temperature. Offset depends a lot on the current mirrors being well matched in gain.



LineSource said:
...The Ayre V3 amplifier topology includes a second symmetrical output stage to create a balanced output. It seems like a cousin to your cascode.
Indeed I looked at the V3, and one of my earlier designs was very close to it. I did want to use a balanced output like that at first (for insanely high PSRR), but in the end I couldn't because both sides of the LTP are used just for one.



AndrewT said:
...I note your use of high value resistors in the NFB loop and at the input. Why?..
I used high value resistors so that I could use small capacitors. That allowed me to use the polypropylene ones I had spare, otherwise I would have had to use electrolytics.



AndrewT said:
...Should r3=r11? to minimise offset?
Matching the impedance at both inputs is not necessary, since a JFET input stage has extremely low input bias current.
 
And another in the daytime, complete with 'scope probes. As you can probably tell from all the extra holes, fans, VU meters (PCB in the lower left corner), this case has been used for several previous projects. It's not exactly a compact amp, taking a whole 4U rack case for just one channel.
 

Attachments

  • case_photo_light.jpg
    case_photo_light.jpg
    98.7 KB · Views: 747
Hi,
just gone back for another look at your circuit.
What effect does Q5 wrapped around Q2 have on mirror action and what improvement does it bring?
You have matched the gains of the 3 NPN and 3 PNP transistors. How important is this?
D1 to D10 are guaranteed to improve the sound (hah hah) but why?
 
AndrewT said:
...What effect does Q5 wrapped around Q2 have on mirror action and what improvement does it bring?..
Q2 + Q5 form a Sziklai pair. By sizing R6 appropriately this causes the current through Q2 to be approximately equal to that through Q1, helping to keep the two sides more evenly balanced.


AndrewT said:
...You have matched the gains of the 3 NPN and 3 PNP transistors. How important is this?..
It's quite important to ensure that the gains of the two mirrors are fairly close, which keeps DC offset to a minimum. Matching of Q1 to Q2 and Q3 to Q4 is what matters most, Q5 and Q6 don't have much effect.


AndrewT said:
...D1 to D10 are guaranteed to improve the sound (hah hah) but why?
The practical reasons to use LEDs is that they have lower noise than e.g. Zener diodes, and slightly lower impedance at low currents (1mA). A further benefit is that with diodes in all four legs, it allows a quick visual check that the circuit is functioning properly (all four strings lit with equal brightness).

However, the real reason I chose LEDs over Zeners (or any other voltage reference), which would have been much cheaper, is that they look nice, which definitely helps me enjoy the amp, and therefore the sound it makes. :)


Giaime said:
My best compliments! And how do those leds influence the sound? I have heard that red ones give a "warmer sound".... :D
I'm not sure if it would be possible to tell the difference between colours. I have been led to believe that green ones have the lowest noise, which is why I chose green. The cascode stage has high voltage gain, so it is desirable to keep noise from its bias voltages low.
 
Is it really folded cascode?
I thought folded cascode is nothing but common base amplifier.
Here is common base mixed with common emitter. (let's assume all transistors are bjt :D )
I don't mean to argue about semantics- my point is rather that this doesn't eliminate miller effect, does it?
 
Giaime said:
...just kidding about that "warmer" sound... :D

Really Green Leds are less noisy than others? I didn't know...
Yeah, the smily sort of gave it away, but I thought it deserved some serious explanation anyway. There is a thread somewhere on this forum where someone did some noise measurements of various types of LED and other diodes, and green was slightly lower than other colours, except blue which was very poor. They were tests on a small selection of devices only, but I suspect that the same holds true for LEDs generally. How significant a difference it makes here I don't know, but I didn't want to take any chances.



darkfenriz said:
Is it really folded cascode?
I thought folded cascode is nothing but common base amplifier.
Here is common base mixed with common emitter. (let's assume all transistors are bjt :D )
I don't mean to argue about semantics- my point is rather that this doesn't eliminate miller effect, does it?
A folded cascode is a common emitter stage followed by a common-base stage, with current being reflected into the CB stage by a current mirror or other means. The LTP is the CE stage here, then the MOSFETs are the CB stage (or common-gate rather). At first glance it may not seem like it, but the gates are held at a constant voltage and the input is applied to the source, so calling it a folded cascode is correct. If the three-LED strings were taken out of the LTP and connected directly to the rails it would still work exactly the same and would look more obviously a folded cascode.
 
Hi MrEvil,

I would appreciate any Green LED noise and voltage stability information you could share. I want to get educated on:

What is the voltage drop range of Green LEDs?
What is the noise range of Green LEDs?
Which vendor's Green LED has the best voltage stability? the lowest noise?

What are the issues with putting Green LEDs in series in terms of voltage stability and noise? When does a single Zener LED have better specs?

Thanks.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.