UGS MUSES Scion Preamplifier

It has been decided that the MCU Controller card will be designed around the ATmega 2560 (ATMEGA2560-15AU) from Microschip.

Summary:
The high-performance, low-power Microchip 8-bit AVR RISC-based microcontroller combines 256KB ISP flash memory, 8KB SRAM, 4KB EEPROM, 86 general purpose I/O lines, 32 general purpose working registers, real time counter, six flexible timer/counters with compare modes, PWM, 4 USARTs, byte oriented 2-wire serial interface, 16-channel 10-bit A/D converter, and a JTAG interface for on-chip debugging. The device achieves a throughput of 16 MIPS at 16 MHz and operates between 4.5-5.5 volts.

medium-ATmega2560-TQFP-100.png


This chip will have enough I/O pins to be able to support multiple UI inputs and commands, more than I will use, but customisable for does who wish to have an extended UI front. This will allow users to use breakout boards to communicate with the main MCU controllercard. The MCU controllercard will have a USB interface communication for firmware updates among other.

Oneminde




 
I guess this is it for now. The investigative period is now over and I can start looking at redesigning things. This will ofc take a while since it includes learning a PCB CAD software, learning about PCB rules, selecting components, do some math and so far. So updates will not be that frequent for a while.


 
What about control interface, are you going to do that as well ?
I mean write software as well ?
Yes because I have to. But I am covered, a friend of me who also want to build one for himself is a programmer. So he has some nice skills. It will be based upon Arduino, so open source. This will allow anyone with enough skill to add their own strings of code for added or changed UI. Communication will mostly be done over I2C. This will allow users to design their own shields / breakout boards and functionality if they so desire, but is nothing I will provide in terms of service.

That is why the ATMega chip is the core of the MCU Controller. While it supersede the pin-count (input) I need, it leaves room for added front-end functionality. Or in my case, I will use a remote with some added functionality which is not featured on the front of the amp, just to keep it clean.

Oneminde




 
I will. Just so that you are all in the loop of what's what. The next following months will be reworking the schematics and PCB layout as well as going over the components used and perform a healthy overhaul there as well.

As an example and as you might know, the JFet's and BJT's will be replaced with newer models, but don't worry, in the end, it will be better. I will do sporadic updates of the development :)



 
Member
Joined 2006
Paid Member
unfortunately the original 2SKXX 2SJXX s dont have better and newer replacements...that is why NP went another way with newer UGS versions....

But i do take my hat off to your thorough and systematic approach...looking forward to following installment.

;)
 
SOT-23 package.

unfortunately the original 2SKXX 2SJXX s dont have better and newer replacements...that is why NP went another way with newer UGS versions....

But i do take my hat off to your thorough and systematic approach...looking forward to following installment.

;)
A note on original 2SK*** series JFet's. While they are available through different collectors on Ebay as an example, since this is a "new" preamplifier, my goal is to make sure that does who want one in the years to come can secure themselves high quality transistors without paying for them with a leg and and arm. One of does JFet's can run as high as $10 a piece or even higher compared to the $0.50 for the replacement JFet.

Therefore the Toshiba 2SK209GR (monolithic) is an excellent replacement. The noise figure @ 1kH is 1dB compared to the 2SK170 @ 1kH and 0.8dB. The 2SK209 exist as a dual-Jfet via the 2SK2145. But I will go for monolithic to enhance the matching. the 2SK2145 has a typical 10% in the same package. Using monolithic will allow us to match to less than a fraction of a percentage.

(GR: 2.6~6.5 mA)

The BJT's will also be replaced. The original UGS call for ZTX550/450 and aren't that special in the first place. The replacement BJT will perform much better.

I am struggling between:

On Semiconductor BC850/860 - (is the SMD equivalent to BC550/560)
Toshiba 2SA1312 and 2SC3324.

But the nice thing here is that I selected components that come in the SOT-23 size package, that way, they are low profile and will allow for one heatsink attached to all of the components on the Buffer card(s) :D

In the end, there will be 3 Buffer cards available.

Buffer 1 (stage 1): The original Flatlabs UGS.
Buffer 2 (stage 1): CUGS wich is a paralleled version of the UGS. That is to say there are 4 JFet's in parallel / side compared to just one. This reduced the overall noise figure and THD and possibly increases the dynamic performance.
Buffer 3 (stage 2): Is the additional output buffer for the MUSES chips.

--- OBS ---
The original UGS call for complimentary JFet pairs (N-chn and P-chn), this will not be the case for this amplifier. I will only use N-chn. The P-chn will basically be an "inverted" N-chn. That together with stupidly narrow matching is a recipe for great success.

The original UGS wich use BJTs from ZTX have a CMRR (Common Mode Rejection Ratio) of more than 100dB - PASS Laboratories XP-30 has an (Stereophile measurement) CMRR of 60dB. This I suspect will improve even further with the better BJT's and the use of paralleled JFet buffer stage.

On top of that, I will more than likely end up with WIMA Foil Capacitors which is among the best available. They perform better than C0G (NP0) capacitors.

:D


 
Member
Joined 2006
Paid Member
a big thank you for your approach, dedication and thoroughness.

that being said ...we all fell back from 2sk389 and complementary, to 2sk170 and complementary and now to 2sk209....2sk2145....but always downward...i still have 4 modules with originals to compare so it's going to be interesting.

2SK2145 is good but not as good as original and not truly matched, unlike 2sk389 and 2SJ109 these were made exactly for this job!, true that the 2sk2145 and 2sk209 are probably the best you have right now so by using them and supporting them with the best the result will be brilliant .
The cascodes are no so hard to replace ...the core , that is another matter, not possible, but i am being too picky, you are doing all that is possible and doing it brilliantly DIYers take note !!!.
For example NP from UGS 3 to 6 changed all but the core and it is still used today so carefull design arround core will yield results to surpass UGS3.....and it seems Muse volume control is a plus as well.

By the way i aplaude your option of buffer before and buffer after.....gain stage.
 
Last edited:
. For example NP from UGS 3 to 6 changed all but the core and it is still used today so carefull design arround core will yield results to surpass UGS3.....and it seems Muse volume control is a plus as well.

By the way i aplaude your option of buffer before and buffer after.....gain stage.

An externally hosted image should be here but it was not working when we last tested it.


UGS6 use:
Diodes Incorporated ZVN2110 ... which is a MOSFET
Nexperia BCP53 .. which is a BJT.
Ofc, I can't speak for N.Pass, he knows why he made does choices. But UGS3 and UGS6 is wastly different.

When you say core, what exactly do you mean ?

And I work with what I've got. But do notice that the pcb size etc will be different for this version, so you can't drop in the older UGS. Basically, the only reason to build the UGS MUSES Scion preamplifier is that you have neither of the other. If you have any of does, I see no need for an upgrade, unless say you are using a resistor based volume attenuator and want to jump to a MUSES version, then yes, switching amplifier makes sense.

Oneminde


 
Member
Joined 2006
Paid Member
the core is 2K398-2SJ109 or s2k170-2SJ74.....that is where the magic starts.
as for working with what you have ..yes i said as much and you are doing a brilliant job.

bettering UGS3 is going to be tough...saying parts are better...what parts are better than 2SK389 or 2SK170 for that job ?

What i am trying to say is that the weak spot is the input transistors and those cannot be bettered....

but to your credit, the others and topology may be as NP has shown in UGS 6 .....and others , many others.

not criticizing, just stating facts.
You have all my support.
 
Got it. Btw, NAD is using the 2Sk209 in their pre-amplifier / stage and NAD are not bad products :) -

My 2 cents goes as follow: The reason for the 2SK398-2SJ109 or 2SK170-2SJ74 being so loved is their natural low NF - Noise figure. However, with that being said, paralleling the 2Sk209 which will happen in the Buffer 2- Stage 1 will surpass the performance of a single 2Sk170 and in the end, that is what matters. (yes, paralleling 2SK170 will surpass paralleled 2SK209).

In the end, when all is said and done, what matters is the whole circuit and the preamplifier as a whole. It is the sum that define if it is a good preamplifier or not :)




 
I need to go over the schematic in more details, because there is another JFet that if - IF - Drain-to-source-voltage is bellow 15V, One Semiconductor 2SK3557-6-TB-E might be a better option than Toshiba 2SK209GR. BF862 is not available. Believe I have tried to track it down. They stopped manufacturing it back in 2017.


 

Attachments

  • IFD2542_FO.jpg
    IFD2542_FO.jpg
    163.5 KB · Views: 483
  • IFD2542_Table.gif
    IFD2542_Table.gif
    43.8 KB · Views: 486
Last edited:
Voltage, [hfe] etc.

Okay, time for another update, this one should be interesting. I looked over the notes I have for the original UGS Buffer which is the one we will use primarily, this one as you know, I call Buffer 1, stage 1. In addition to this, there is a version which use paralleled JFets, this one is called Buffer 2, stage 1. And for the followup gain stage which is an additional stage used in the UGS MUSES preamplifier is called Buffer 3, stage 2. All of these buffer/gain stages will in the end use the same JFet and BJT. Meaning that the selection of replacement JFets and BJTs affect all gain stages.
A side not to that. I am going to use SOT-23 package, so all of the FETs of interest comes in this package, therefore, changing between different FETs should not be an issue.

***

No capacitor in the signal path.

This preamp is powered by -/+ 24V.

The JFets are biased around 3.5mA, and the BJTs around 7mA, and the assembly in the order of 25mA /channel.

Pairing is done with the same mA and similar transconductance curves.

Idss must be larger than 4mA for the JFets and 7mA for the BJTs.

24V * 25mA (0,025A) = 0.6VA/Chn.

The JFet bias current is roughly 3.5 mA/JFet, and it is set by the JFets' source resistors. It obviously depends on the Id=f(Vgs) curve, so Idss matters.

The mirror resistance is here to provide some gain, helping the first stage in order to get a higher open loop gain (around 35dB in simulations). Their values are on par with the drain resistors, with a rough current gain of 2 here. And if the resistor is too low, current will increase, and the dissipation in the output transistors will become too high, so I had to make some compromises between gain and thermal equilibrium. With values here, the current through the output BJTs is around 7mA.

The max output swing, and simulation gave me (Flat) a rough 25V peak to peak (12.5/-12.5V).

Question; The Idss of K389 and J109 BL grade is from about 6mA and up to 12mA, since the bias is at about 3.5mA, can I use JFet of about 6.5mA Idss? What is the performance difference (gain, distortion, sound etc.) for 6.5mA and 10mA Idss parts?
Answer; Bias will depend on the Idss ratings, since it impacts on the Id=f(vgs) curve. I quickly simulated using GR parts, with an Idss of about 4mA, and as awaited the bias is lower (2.3mA). The whole pre will work, but expect lower gain, both open loop and with CR. So, to answer your question, it will work for a 6.5mA part, but the input range before clipping will be reduced, and hence the output swing. But in "normal" applications, the clipping levels will seldom be reached... So if you find BL parts, just try them, listen carefully on your setup, and try to go ahead by matching Idss, pinch-off, transconductance, Vbe, Hfe, etc
The source resistors on the input FETs are another of the biggest contributors to 1/f in this circuit, so using low noise resistors here really pays off - bulk metal foil or RN60 types (maybe PTF series, some of the Caddock parts, or CMF60's at the very least) will work best. Carbon, SMD, or even cheap metal films here are likely to degrade the DC stability - IMO it's not worth the cost savings to go with anything less than the RN60's.

Paralleled JFets: 1/f noise of the output transistors is further reduced about 6dB by using 4 devices in parallel.

Question: Could I get by with BC550/560 replacing the zetex parts?
Answer: The "historical" reason for using zetex parts was their form factor which allowed a more efficient thermal coupling with my "heatsinks", and their thickness that is close to the Toshiba's
***
That effectively ends the notes aspect of things. The next section will cover some interesting data on the JFet and BJT runner ups.
***

Original JFet
Toshiba 2SK389-BL / 2Sk170-BL - Idss; 6-12mA - [Yfs] typ 22mS - Pd: 400mW - NF: typ 0,5dB


Replacement JFet
1) Toshiba 2SK209-BL - Idss: 6-12mA - [Yfs] typ 15 mS - Pd: 150mW - NF: typ 1dB.
2)
On Semiconductor 2SK3557-6-TB-E: 10-20mA - [Yfs] typ 35 mS - Pd: 200mW - NF: typ 1dB.


In order to keep things fresh, here is the graph displaying the performance between 2SK209 and 2SK3557. Notice that the 2SK3557 and Linear Systems LSK170 is the same.

IFD2542_FO.jpg
IFD2542_Table.gif

The voltage swing of 25V peak-to-peak (+12.5/-12.5V) is equivalent to:


83.3%
from maximum for On Semiconductor 2SK3557-6-TB-E which has a Gate-to-Drain Voltage (VGDS) maximum of 15V.
25%
from maximum for Toshiba 2SK209-BL which has a Gate-to-Drain Voltage (VGDS) maximum of 50V.

Original BJT.
Diodes Incorporated ZTX550/450 - [hfe] 100-300 - General voltage is 45-60V. No info on NF - Noise Figure. - Pd: 1W - No capacitance info.

Replacement BJT.
1) On Semiconductor BC850/860-C (SMD equivalent to BC550/560)
- [hfe] 420-800 - General voltage is 45-50V - NF: typ 1.2dB - Pd: 310mW - Input Capacitance: typ 9pF - Output Capacitance: typ 3.5pF.
2) Toshiba 2SC3324-L/2SA1312-L
- [hfe] 350-700 - General voltage is 120V - NF: typ 0.2(0.5)dB :D - Pd: 150mW - Output Capacitance; typ 3pF.

***
The CUGS (Buffer 2, stage 1) used Toshiba 2SC2240BL/
2SA970BL BJT (discontinued a long time ago) and the report is that the paralleled version decreased the noise and increased the dynamic performance. Keep that in mind because the NF - Noise Figure of this BJT is typ 4dB. Meaning many times noisier than the replacement BJTs I have in mind :cool:
***

Are you excited ? I know I am. I will eventually run simulations (LTSpice) with the replacement FETs and check what software tells me in the form of voltages, current and wattage. If the software clares the On Semiconductor, then the
2SK3557-6-TB-E is cleared for bench-top trial. This trial will be an endurance one to see how it behaves over time operating at the full voltage swing of 25V.

Oneminde saying goodnight.



 
The Pd for the Toshiba BJT is "only" 150mW, that can be solved by running them in parallel, which also in turn reduces the amount of base current that flows, and therefore the portion of the shared current that the affected transistor carries. This technique is commonly known as “Emitter Degeneration.”

Running them in parallel and they share the load. The wattage is shared among them

Sources:
Bipolar Transistors connected in parallel - Tech Tips - Engineering and Component Solution Forum - TechForum │ Digi-Key
heat sink calculation for parallel bjt - Electrical Engineering Stack Exchange