DIYHiFi.org

Folks,



I have been quite busy, including some more variations of the main amplifier circuit (more on these another time), but here one of the last pieces of the puzzle. Most Amp's have a fuse and a switch after the IEC Socket wired directly to the transformer. While this CAN be done and CAN work ok, I prefer normally a bit a more sophistication. In my case the supply is designed for 230V nominal, it will be running at appx. 240V though, as is the locally supplied voltage.

So in my case it will look like this:



Again, nothing revolutionary here really.

The inrush-current limiter is realised using a Timer relay. I got fond of them in my old days of constructing large scale stage sound systems. By setting the relays for varying time delays for both on after power is applied and for the delay before inrush current limiters are disengaged can help to power up large Amplifier racks, that otherwise can only be brought "up" by pumping the breakers... A NTC Thermistor could be used, but a big metal-clad resistor is more to my taste.

A big "canned" RFI filter is added to reduce the problems from all the SMPS and related RF crud that nowadays floats on the mains.

There is also a DC trap, the 16V rated "computer grade" low-Z cap can handle up around 3V reverse voltage, the big bridge clamps the voltage to significantly less than that for all expected conditions, so despite not using back to back "unipolar" capacitors this will not blow up.

A bunch of "X2" rated (means they can be connected across the mains safely and legally) are added to the transformer input (and will also help to lower the canned filters corner frequency) with a suitable snubber, to avoid creating too much of a resonant tank. I will not cover calculating the snubber in detail, the values shown are "standin" and must be optimised for each transformers.

Audio Star Ground and Earth are not linked directly, though for safety the voltage between audio ground and chassis is clamped by diodes, doing this right (make sure you can handle worst case fault currents etc.) creates an Amp that is both safe and free from earth loops, however unless you know how to do it right, maybe better avoid this and keep the normal safety earth.

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

What is the purpose of Bridge 25A?

Hi,



To clamp the voltage across the 68,000uF DC blocking Capacitor, which in turn keeps DC offsets away from the mains transformer.

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

Folks,

Before we come back to the Audio Circuit, a short note on protection circuitry.

I will keep the originally present one, it uses a single transistor SOAR detector, DC detection and power loss detection all feeding a monostable relay driver (that is any individual protection circuit that triggers will cause the mute relay to close and the output relays to open and retains this state until power is cycled).

This means the protection circuitry is non-invasive (that is it does not act on the audio circuit at all) and does not reset after triggering, something that would be unacceptable in a Pro-Audio Amp but is perfect for a High End Power Amplifier, where we are normally tend to not abuse the Amp. Most of the circuit may be found in the schematic for the Vincent SP-61 Amplifier, located at the usual places...

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

Absolutely, be careful here. Diodes seemingly have a high current rating, but you have to know how many cycles your circuit breaker or fuse will need to blow, and compare that to the diodes rating.
A very good thing would be if, in your house, there is a residual current circuit breaker installed: http://en.wikipedia.org/wiki/Residual-current_device

Your diodes may survive a fault easier....

Hi,



Actually, we do not really mind if the diodes fail "shorted" after a something put a lot of current into the chassis, what we must avoid is "failing open". Thankfully it is usually quite hard to make diodes fail "open" from over current.

And we must consider fusing. My design is intended for the UK, where mains connections are clearly polarised (so a fuse in the "Live" connection is correctly in the "Live" line), so it will work correctly and actually the fuse will fail before the diodes. In countries that lack proper polarisation it may be necessary to fuse both live and neutral or take other measures.

Note also that normally it is highly unlikely to have equipments mains wiring fail such that the chassis is fully earthed while the audio grounds have full live potential. Remember, the chassis is fully hard earthed in my depiction...

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

Folks,



Another time is now.

After going full circle I ended up trying also the original "two differential stages in series" circuit in the Simulator, but heavily degenerated. The results surprised me (though they should perhaps not have, given this topology is used in Kaneta's Preamp's and Amp's and some of the best sounding Op-Amp's). Not shown here, I again experimented with the open loop gain. It can be prodigious (this circuit with all degeneration and loading removed looks like > 130dB DC Gain), but so much loop gain does not really give us lower audible distortion, but generally tends to emphasis higher harmonics more while lowering lower order harmonics.

In the current circuit it is however kept down to around 47dB by degeneration and the loading on the VAS output from the inner feedback loop. The bandwidth incidentally is around 160Khz. The inner feedback loop uses up 9dB of this loop gain, leaving 12dB around the output stage. For what it is worth, the VAS uses around 42dB degeneration and the input stage (with the CFB pair) around 32dB. The result is very little upper harmonics (more later). Now to the actual circuit:



For one example, the harmonic distortion at 1KHz and 20KHz each for 100W/8 Ohm:



For 1W/8R very little distortion remains, in part because we are in Class A operation and in part because the BiMos - SEF (Source Emitter Follower) is really quite linear by itself.



For a 4 Ohm load distortion goes up a bit, a consequence of using only 3 output pairs. If we where to design this Amplifier not as 150W/8R but for 300W/4R doubling the number of output devices should return to the performance. Here the 4 Ohm distortion for 2W/200W 1KHz:



How about operation into capacitive loads? Here is what happens when 2uF/8R are connected and we use 10KHz/40V Peak-Peak. The ringing is strictly that of the output filter:



Overall the behaviour of this Amplifier circuit suggests it will suit my personal ideas of what produces "good sound" best, others may differ. As a bonus, it only needs minimal changes to the original circuit.

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

When are you anticipating building it?

Hi Thorsten, thanks for sharing this.

The input differential, i call it a CFP, although not done with two BJT but FET/BJT. Feedback to the emitter / source ("current feedback" as the OpAmp maufacturers are calling this feedback scheme), gain=1. Am I wrong?

I found the PSU primary quite interesting, especially that you want to get rid of DC component.
I can imagine that especially with all that switching mode psu's around, loading the grid asymmetrically, there can be a DC component.
Did you experience less hum when applying this cap?

You mentioned TINA-TI as a good program. I like the optical appearance of the schematic.
But the simulation tool (although i do not think i will use it often) and especially how external models can be "loaded" isn't clear at all to me.

Say I have this 2SJ74 JFET model:
.model J2sj74 PJF(Beta=39.21m Rs=0 Rd=0 Betatce=-.5 Lambda=4.338m Vto=-.5762
+ Vtotc=-2.5m Cgd=67.64p M=.2562 Pb=.3905 Fc=.5 Cgs=61.12p
+ Isr=158.7p Nr=2 Is=15.87p N=1 Xti=3 Alpha=10u Vk=100
+ Kf=109.9E-18 Af=1)

How can i implement it into TINA? Do i have to write a subcircuit model for this? How canthis be done? I din't find anything, just a video which shows how to import a macromodel or so.

Hi,



Well, in principle a compound feedback pair is just a combination of two devices. I have heard it calls "Inverted darlington", however I think this designation is not entirely correct.

Here the First device is a low Transconductance J-Fet (2SK246), the second a BJT.

As it is designed the J-Fet operates at around zero Tempco and has around 0.8mA/V Transconductance at that current. Compared to (say) 2SK170 it has however much lower input capacitance (and let's not forget that Zero Tempco thing - which means thermal distortion is defeated).

The BJT being a 2SA970 running at 0.8mA has around 30 Ohm Emitter impedance (or 33mA transconductance) and around 9K input impedance, so the effective load on the J-Fet is around 2.4k and the gain of the Fet around 2, so the combined transconductance becomes around 66mA/V.

In other words, we now have a J-Fet with 66mA/V transconductance, a 5pF reverse and 9pf input capacitance and a bandwidth that is more limited than either BJT or J-Fet alone, however, as the Ft for the 2SA970 is 100MHz the "slower" J-Fet is likely to still dominate.

How we connect this new compound device is up to us. It be a follower or even a common gate amplifier.

In the Amplifier Circuit is connected as differential pair with degeneration. As the effective source impedance is around 15Ohm, the 560 Ohm source resistors apply around 32dB degeneration, which helps to linearise the circuit, as does the cascode. Incidentally, the cascode incidentally is bootstrapped to the common mode signal (top of the tail), so common mode distortion is minimised

In addition, the low current in the J-Fet (> 2V) means we have (compared say to a BJT differential) an enormous dynamic range available to the differential circuit.

I might have applied a FET/BJT Compound to the output stage as well (then we can dump overall feedback), but the existing circuits are incredibly ill suited to that sort of modification and I do not fancy making new ones.



First, this circuit is quite old. You will find it (confirmed) in Bryson and Lamm Amplifiers (also almost all of my own designs). It helps with transformer hum if it is caused by DC (not always the case), but mostly it makes the transformer much more efficient and reduces temperatures.



You cannot "load" external models. You need to put the default component in (say P-Channel J-Fet) onto the schematic and then select the correct Model (Tina allows amny options) and then enter the parameters. You can then copy the component to all sorts of schematics. Each component has basically it's own local model in Tina.

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

Mike,



It will be build as modification of a Vincent/Sheng Ya SP-332 Amp I have on the shelf, that needs work anyway. Now that I have resolved all the questions in my mind to my personal satisfaction a few weeks i expect, i just need to order some more Tokyo Optics resistors to match what I am aiming at.

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

T, you are truly a sick individual. But, that is why I like you so much!!!!!!!!!

That.........and your, uh..........uh............contempt for the audio fascists.

Keep up the good work. On both fronts.

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

Hi T,
Can you explain how you end up with an input capacitance of ~10pF for the 2sk214? Looks like the datasheet specifies ciss as 90pF.
Thanks. I'm learning a lot from your analysis.

Hi,



Ciss is bootstrapped as we are using the 2SK214 as follower. The Output transistors have a beta of around 100 at the levels of interest. So the effective load on the follower with a 4 Ohm load is 400 Ohm. The output impedance of the 2SK214 & 2SJ77 together is around 5 Ohm. So 98.76% of Ciss is "bootstrapped"and hence the 90pf become around 1pF. So our real effecive input capacitance is around Crrs +1pF...

Ciao T

_________________
Theory is when you know everything but nothing works.
Practice is when everything works but you don't know why.
Guruhood is when you know everything and everything works accordingly.

2SK214??

Anyway... T, can you go into more detail about the thermal distortion in the JFET and how operating close to the zero temp coefficient point improves it?