I recently repaired an amp with zeners on the LTP collectors to lower the voltage over the input transistors. No bypass caps or anything, just plain 35V zeners in series. I found it a bit weird, but the amp sounds nice, and no noise problems even if its pretty high power with 80V rails. Just an example of a simple way to overcome the voltage limitation of the input transistors, not complex 🙂
Hi Rally,
I would not expect zeners used that way to add any significant noise. Their noise is voltage noise and the output of the differential stage is current. Kirchoff says the current into and out of the zener is equal, so no noise added, at least to a 1st or 2nd order approximation.
The benefits of a floating cascode on the inputs are
1) Protecting input transistors from over voltage
2) Increasing the BW of the input stage
3) Increasing the gain of the input stage
4) Increasing the linearity of the input stage
The zener solution only provides the first benefit. So, simple, but not elegant.
I can add the floating cascode with a resistor on a bias leg that is already there. So it only adds three components, including the cascode transistors.
I would not expect zeners used that way to add any significant noise. Their noise is voltage noise and the output of the differential stage is current. Kirchoff says the current into and out of the zener is equal, so no noise added, at least to a 1st or 2nd order approximation.
The benefits of a floating cascode on the inputs are
1) Protecting input transistors from over voltage
2) Increasing the BW of the input stage
3) Increasing the gain of the input stage
4) Increasing the linearity of the input stage
The zener solution only provides the first benefit. So, simple, but not elegant.
I can add the floating cascode with a resistor on a bias leg that is already there. So it only adds three components, including the cascode transistors.
Cascoding the input differential pair does not improve linearity, increase bandwidth or increase gain.
Good input.
But all of three are partly (or conditionally) based on input common mode range, used rail voltage, tail current and amp topology.
The advancements made to MOSFETs over thae last 30 years have greatly reduced RDS on and ciss vs size and voltage but at the expense of (Linear Region) linearity and unwanted oscillation.
niether of these problems are seen if the device is switched like class D.
I am going to try irf610 , 9610 drivers on 2SC5200 , 2SA1943 as these lower power FETs have really nice datasheet output curves.
niether of these problems are seen if the device is switched like class D.
I am going to try irf610 , 9610 drivers on 2SC5200 , 2SA1943 as these lower power FETs have really nice datasheet output curves.
P-channel IRF from Vishay, IR, Infineon and Siliconix has linearity issues. Unless you can find IRF9610 from Harris, Intersil, Samsung, On-semi or Fairchild, better use FQP3N30 & FQP3P20 instead,
https://www.firstwatt.com/pdf/art_mos_test.pdf
https://www.firstwatt.com/pdf/art_mos_test.pdf
Incorrect.
Gain: Gain = gm*Ro, and Ro is hugely multiplied by the cascoding. Other impedances on the diff stage output will also limit gain, but in my circuits, the Ro of the input pairs is the limit on impedance at that node.
Linearity: Ro varies with Vce, so the gain changes with signal swing. That's nonlinearity that is corrected by cascoding. Rout of the cascode has almost no change with signal swing.
Bandwidth: Any gain on the input stage Miller multiplies Ccb at the input nodes. Cascoding kills this Miller effect for the input nodes. Obviously, this is not the dominant pole, but any high frequency pole that you can move to an even higher frequency enhances the stability of the amplifier.
All of these are small, low order effects, but they are real.
No, usually my designs end up in racks that never move.
Where I've seen this happen before is in equipment that lives in mobile racks or carts. I've seen it a few times in test equipment that lives on a cart, and more than a few times in amps / outboard equipment that have been toured for a long time.
A closely related issue (that is a big problem in Crest CA series amps) is stress put on solder joints and connectors by a flexing chassis. The CA9 is really heavy, but it doesn't have the most rigid chassis, especially when some idiot comes along and thinks "Nah, I don't need no stinkin' rear support on these amps!".
On CA series amps it's connectors that fail, but on other equipment I've seen it shear leads on voltage regulators or TO-220 transistors that are screwed to the chassis.
Speaking of the Crest CA series, they used kind of a neat little package for the input stage- the MPQ6700. From what I remember the individual transistors in the device are similar to a 2N3904 / 2N3906, but they're all on the same die. There are other devices in the same series that are probably better suited.
IMO, these sort of packages are pretty pricey for what they are, but then again I also *hate* matching transistors. It's one of my least favorite activities.
Another interesting dual device is the JNSVJ6904DSB6T1G (what ever happened to sensible part numbers?). Of course, being that the two JFETs share a common source, that precludes the use of degeneration resistors. Anybody played with these things?
IMO, these sort of packages are pretty pricey for what they are, but then again I also *hate* matching transistors. It's one of my least favorite activities.
Another interesting dual device is the JNSVJ6904DSB6T1G (what ever happened to sensible part numbers?). Of course, being that the two JFETs share a common source, that precludes the use of degeneration resistors. Anybody played with these things?
There are lots of matched devices available these days if you don't mind using SMD. Looking at my library of devices in my layout app, I've got:
BCM847BS NPN $0.16
BCM857BS PNP $0.16
DMMT5401 PNP $0.15
DMMT5551 NPN $0.15
SSM2212 NPN $5.08
I put in the digikey price/device if you buy 100 pieces at a time. Only the SSM2212 is special, but it's a supermatched extra low-noise part for input pairs, like for a phono or mic preamp. I don't think I'd buy a 100 of those at a time. Ten of them cost $61.
There's tons more devices available, but those are just the ones I use.
I remember finding those double BC847/857 about 20 years ago when the company was still called Philips. I think they didn't have the M preceding the 8 back then.
Two things I didn't like:
- they were not 849 = low noise, so good for current mirror but not necessarily for LTP
- they were not monolithic but just roughly matched
The lot I had back then was pretty poorly matched if you tested them at a different temperature or current.
Matching on the DMMTs is much poorer but they are high voltage. Not sure how to compare noise figure as they don't give bandwidth.
Two things I didn't like:
- they were not 849 = low noise, so good for current mirror but not necessarily for LTP
- they were not monolithic but just roughly matched
The lot I had back then was pretty poorly matched if you tested them at a different temperature or current.
Matching on the DMMTs is much poorer but they are high voltage. Not sure how to compare noise figure as they don't give bandwidth.
Last edited:
I agree with that, capslock. I don't use them for input pairs. Some of the grades of the part have really high beta, so those might not be too bad for inputs. Datasheet gives the noise figure as 2 dB typ and 10 dB max. I'm guessing the high beta units have the 2 dB NF.
Looking around, some of the manucturers use the "M" and others don't. I don't know what's inside the plastic, regarding single or dual die.
--Russell
Looking around, some of the manucturers use the "M" and others don't. I don't know what's inside the plastic, regarding single or dual die.
--Russell
Last edited:
I was looking at the THAT 300 series of matched devices the other day. Several configurations available. I have no experience with them.
THAT Corporation 300-series Low-Noise Matched Transistor Array ICs
THAT Corporation 300-series Low-Noise Matched Transistor Array ICs
Pretty hard to beat,
(2N)MMBT5401 and MMBT 5551 for high volt (IPS), support circuitry. KSA1381 and KSC3503 for VAS, high voltage cascoding, and predrivers. MJE15032/33 for drivers.
Just today I was working on an amp using these parts.
TI developed the 2N5401/5551 a long time ago and it is still relevant in todays amplifier designs.
Digi-key sells a OnSemi 2N5401Y that is hfe graded.
I measure these Rectron 2n5401 coming off the ammo pack paper tape, Vbe is within a 1mV or two side by side, which is what the 2n2920 was spec'd at ? 🙂
I you use a DC servo and it is basically a non-issue IPS bias offset. You eliminate the DC FB ecap too, if you plan to use one.
(2N)MMBT5401 and MMBT 5551 for high volt (IPS), support circuitry. KSA1381 and KSC3503 for VAS, high voltage cascoding, and predrivers. MJE15032/33 for drivers.
Just today I was working on an amp using these parts.
TI developed the 2N5401/5551 a long time ago and it is still relevant in todays amplifier designs.
Digi-key sells a OnSemi 2N5401Y that is hfe graded.
I measure these Rectron 2n5401 coming off the ammo pack paper tape, Vbe is within a 1mV or two side by side, which is what the 2n2920 was spec'd at ? 🙂
I you use a DC servo and it is basically a non-issue IPS bias offset. You eliminate the DC FB ecap too, if you plan to use one.
Last edited:
BC550C/560C for input, but as they looks missing in the last few months (i use Onsemi), changed to BC549C/559C. 2N5551/5401 if cascode is necessary.
KSC1845/KSA992 if want to use same type for input+cascode+VAS for lower power amplifiers.
KSC3503/KSA1381 for VAS, and for predriver. They are pretty good devices. 2SC4883/2SA1859 for driver, Toshiba TTC004B/TTA004B or TTC011/TTA006 are useful up to +/-35-40V.
For output I use mainly bipolars, and my favourites are 2SC2837/2SA1186 or 2SC3519/2SA1386 from Sanken.
Sajti
KSC1845/KSA992 if want to use same type for input+cascode+VAS for lower power amplifiers.
KSC3503/KSA1381 for VAS, and for predriver. They are pretty good devices. 2SC4883/2SA1859 for driver, Toshiba TTC004B/TTA004B or TTC011/TTA006 are useful up to +/-35-40V.
For output I use mainly bipolars, and my favourites are 2SC2837/2SA1186 or 2SC3519/2SA1386 from Sanken.
Sajti
