KSA-5 inspired amplifier builds

[kozard]

By faster do you mean lower capacitance? Or higher fT/Current Gain - Bandwidth Product?

I ask since at 10V I think the Cob for 2SC5171 is about 16pF versus 4pF for each MPSA06. So 4x MPSA06 would be about 16pF. And 2SC5171 has higher fT.

So does it mean in this case that one 2SC5171 would not be slower than four parallel MPSA06?

 

I could certainly put four parallel TO-92 but I would like to understand it better before I do the layout. I have a bunch of genuine 2SC5171/2SA1930 from dead receivers that I used the chassis/heatsink and transformer to make KSA-50 monoblocks. So I am looking for the best place to use them. Otherwise I can just get some MPSA06/MPSA56.

I found a couple Spice models for 2SC5171/2SA1930 but I don't know if I can trust them. One had quite a few default looking values.

 

I also have some leftover TTC004B (12pF Cob) and TTA004B. I have not found the BD139/BD140 Cob yet.

[AlexS]

Done the first power-on. No magic smoke escaped! 

But something's not right: Bench power supplies set to 21 Volts. Current draw is 161mA for both the positive and the negative rails.

Had to adjust the bias-pot all the way to one side to get a minimum current across the test points of 109mV. I thought that there should be more play here?

The power transistors / heatsinks hardly get even warm - after waiting for roughly 10 minutes.

DC voltage at the output (with input shorted to ground and no servo OP-amp installed) is at 70mV.

Any ideas, anyone?

Maybe I will stop for today and have a good look with fresh eyes tomorrow...

*frustrated*

?

[kevin gilmore]

there was a long explanation on this topic somewhere at diyaudio.  in general N x smaller transistors is going to have less total silicon area than 1 bigger transistor at the same power dissiption which probably means less cob. 

edit: without the servo 70mv on the output is about right. i don't remember what the output stage bias is supposed to be, too lazy to look it up.

hook up signal generator 16 ohm load and scope and you will know whether its working right or not.

edit: edit:  your desk is way to neat. nothing good can come from something this clean. look at pictures in the megatron thread for what your desk shoud look like.

Edited November 19, 2021 by kevin gilmore

[AlexS]

Right - I found my mistake: swapped the emitter and collectors on the two MJE15031 PNP output transistors...

Stupid me! No matter how often you check, re-check and double check something, good old murphy is always lurking somewhere!! 

It's an easy fix, optically not really pleasing. Lucky for me to have house guests over the weekend, so no time in the workshop. 🙄

@kevin gilmore nice one with the desk! 😂 ... Believe me, my workspace looks like yours from time to time!

[kevin gilmore]

if you use a circuit board system with integrated schematic, if the schematic is right, the layout has to be right. Believe me i have found out the hard way.

[AlexS]

@kevin gilmoreI use kicad and definitely have forward annotation in place. Funny coincidence, in this case I had two errors for the same part: I had the PNP symbol the wrong way round (Emitter and Collector swapped) AND I had a wrong assignment of the footprint (Q_PNP_BEC - should have been Q_PNP-BCE)... what are the odds??

Anyway - both errors cancel themselves out, leaving me with a layout that is correct - for that part anyway!

Then I inserted the OP amp and re-measured the DC Voltage at the output: 1,3mVDC - perfect!

Next up: signal generator, dummy load and scope. Set the generator to sine wave with 500mV pp and the load has 250 Ohms on the output.

• 20Hz ... 2,9V pp
• 200Hz ... 2,9V pp
• 2kHz ... 2,9V pp
• 20kHz ... 2,9V pp
• 50kHz ... 2,8V pp

Seems to be ok - amplification at 5,8. Is that what is to be expected? Removing the load didn't change anything.

I still can't figure out why I can't dial in under 110mV across the output resistors. And the heatsink barely goes over 28°C after 1 hour of operation. (19°C ambient)

Any ideas what might be wrong? From what I read in the forum posts, 400mA per power rail should be drawn. I have 165mA.

Here's a picture of the (still too tidy) workbench:

Edited November 24, 2021 by AlexS

[JoaMat]

On 11/20/2021 at 7:26 PM, kevin gilmore said:

if you use a circuit board system with integrated schematic, if the schematic is right, the layout has to be right. Believe me i have found out the hard way.

It really made my life a lot easier when I started to use integrated schematic. But if schematic is wrong, this might happen… 

Interesting project you have, AlexS.

[AlexS]

So, over in the diyAudio forum user Algar_emi from Canada shared a commented schematic of the original clone with a few measurements he made.

https://www.diyaudio.com/forums/headphone-systems/109618-krell-ksa5-9.html#post5189211

Turns out that a single channel will draw about 180mA per rail. That's with the voltage across the 2 ohms resistor at 125mV. That works for me. All values from his notes match with my measurements.

Strange though, this only seems to work with no heatsink attached! Having only the aluminium L-profile in place, the setup settles at 44°C (100mV). As soon as I connect the big heatsink, obviously the temperature drops and the voltage across the resistor increases to 137mV. (just under 200mA on the input rails). Temperature drops below 30°C.

After some ten minutes, temperature is up to 30°C and voltage down to 130mV.

So it seems that the thermal capacity of the heatsink has a big effect on the system. Any ideas, anyone? Do away with the big heatsink? Simply live with the increased current through the resistors / transistors?

 

Edited November 27, 2021 by AlexS

[kozard]

Perhaps the range of the Vbe multiplier (resistors and/or pot) needs to be adjusted if you want to use the larger heatsink and lower temperature?

Not sure what is best in this case but sometimes a certain temperature is targeted (such as 50C in another design).

[luvdunhill]

I would measure the temperature of the PCB and the free air around the output stage in both configurations…

[alexcp]

On 11/27/2021 at 5:16 PM, AlexS said:

Having only the aluminium L-profile in place, the setup settles at 44°C (100mV). As soon as I connect the big heatsink, obviously the temperature drops and the voltage across the resistor increases to 137mV. (just under 200mA on the input rails). Temperature drops below 30°C.

After some ten minutes, temperature is up to 30°C and voltage down to 130mV.

So it seems that the thermal capacity of the heatsink has a big effect on the system. Any ideas, anyone?  

This is quite normal. As the heatsink temperature rises, the Vbe of Q14 goes down (by about 2 mV/°C), the voltage between the bases of NPN and PNP output transistors decreases, the quiescent current of the output transistors goes down, which reduces the dissipation, causing the temperature to go down. In other words, the Vbe multiplier adds a negative temperature feedback loop, which maintains not the quiescent current (which it has no way of measuring) but the temperature of Q14. Eventually the system finds an equilibrium somewhere, and the temperature at equilibrium depends, among other things, on the thermal resistance of the heatsink.

In practical terms, you adjust the bias to your liking (e.g. by the lowest distortion, or by the desired heatsink temperature, or by the power dissipated by the output transistors) and be happy.

[kozard]

How is the project going? Have you listened to it yet?

[AlexS]

Yes - progress update: I have assembled the second channel and done some measurements. DC offset with OP amp at 1.2mV - happy with that. 

I didn't get much more done since everything went crazy at work... like every year in December. It's alway interesting when management tells you to use up your annual holiday within the year, you don't really get to take it during the year and then you have to be off for the better part of December. Top management happy, project management... not so much. Anyway...

I ordered the parts for the power-up and dc-protection circuits and want to CNC mill the traces of the board in the next couple of days. I want to have that done before I plug in any headphones... call me a chicken!

[AlexS]

Finally I got round to getting some time in the workshop: the DC protection circuit is up and running!

I decided to do a prototype on my CNC. Converted Gerber files to Gcode with a software called FlatCAM - free, full of features and easy to use!

Quite pleased with the result...

Assembled and tested:

- power on delay ist about 2.5 seconds. I might want to increase that a bit.

- The voltage divider for the reference voltages is a bit off: +410mV and -620mV. Not to worry...

- Applied a 1 kHz sine wave with 1 Vpp to the inputs and used the DC-offset adjustment on the signal-generator to tease the circuit. The relay clicks off at the exact values as set by the reference voltage divider. Both channels fine!

That's it for today - next step is to hook this up with the amplifiers and finally take a listen... looking forward to that!!!

[kozard]

That looks nice.

Sounds like I should look into FlatCAM.

How many different bit sizes did you need to use to make the board? If you don't mind me asking what are the sizes and types of bits that you used?

[AlexS]

It's alive 😊

How does it sound? Sounds great - to my ears... whatever thats worth! Seriously: after a short listening session, I am very pleased! Dead silence with no sound input and a very clear sound reproduction. In my opinion the HD-800 are very good in emphasizing any shortcomings in recordings or audio gear for that matter.  

Is it better or worse than others? I can't really judge. It is what it is and I am having fun during the process of building it - that's the main goal of the project!

I will be performing harmonic distortion measurements as soon as I have the PSUs built. That will be the next step! Probably a Goldenreference variant.

@kozard happy to share details - wasn't aware that it's of interest: three tools involved here: first is a 0.1mm V-shaped carbide engraving bit. Dirt cheap on eBay. That's for milling the isolation traces at a depth of 0.05mm. I used the same setting for the surround cutout. Only for marking, did the actual cutout with a hacksaw and a file. Then I used a 2mm endmill for clearing out the isolation areas around the 230V traces. Finally I drilled all holes with 0.6mm carbide twist drill. Didn't bother with tool changes for the different hole sizes. Did the ones that needed bigger holes later on manually. Thats really all the magic. I can maybe document the whole process in detail in the future if there is demand?

[JoaMat]

On 12/17/2021 at 9:12 PM, AlexS said:

Finally I got round to getting some time in the workshop: the DC protection circuit is up and running!

I decided to do a prototype on my CNC. Converted Gerber files to Gcode with a software called FlatCAM - free, full of features and easy to use!

 

That looks great! At picture above, are you removing copper with a 2mm end mill?

Edited December 19, 2021 by JoaMat

[audiostar]

Hey Alex, that looks really nice. Btw, why did you decided to have a separate trafo on the headphone protector board? Or is it temporary only? Cheers

Edited December 19, 2021 by audiostar

[AlexS]

@JoaMatYes - 2mm endmill... does not get into every little space, but saves a lot of time. If I had attempted to rub out the areas with the 0.1mm V-bit, that would have taken forever!

@audiostarThis is still to be debated on! The intention is to have a very weak power supply for the protection board, so that the voltage drops fast at mains power off and the relay disengages well before the HT rails of the amplifier come down. If I power the protection circuit off the same rails as the amp, that would obviously not work. So my initial thought was to have a completely separate supply with separate transformer.

However, this morning I had the following idea: why not tap the secondary AC from the amplifier transformer and go into a second rectifier? Well the positive part of it. Use the negative half of the existing rectifier for a common ground. That way the heavily buffered DC from the amp will not supply the DC of the protection (diodes block the current path) when the AC is removed. See attached pdf... drawn is a positive channel of a GRLV power supply as per @jamesmking/ @kevin gilmore. I added on a small +12V supply for the protection circuit.

Please kindly review this, anyone, and comment? Thanks a lot!!

 

[audiostar]

Hey Alex, I can't access the attachment. 

[AlexS]

@audiostarthat's odd... you can also try to directly access my files: https://51two.com/KSA5/PSU

[AlexS]

Merry Christmas to everyone here! I hope you had a few days in peace and quiet?!

Here is the first half of the power supply (a GRLV with +21V output and a 7812 with +12V output). 

The PCB area is what I can afford with the current plan for an enclosure - I wonder if it will be possible to place all components!! 🤪 Maybe standing resistors and diodes upright? The big input cap was substituted by 4ea 1000uF/50V. I only have 30mm height for the components.

[AlexS]

A happy new year to everyone! 

The positive part of the power supplies is done:

The required PCB dimensions were met, however I had to remove the fine-adjust capability. Everything rather tight!

Next step is the negative part and then it's time to order PCBs. These are double sided and I do not have the tools to insert vias underneath large components.

[audiostar]

Hey Alex, that looks fantastic ... and a Happy New as well, cheers!

[AlexS]

Thanks @audiostar! 

Here's a sneak peak preview of the mechanical design. This is the case that I ordered... not at all cheap, but made by a small company in Germany: https://19zoll.com/en/products/aluminum-cases/19-heatsink-enclosure-series-dsk/

And this will be the inside - Neutrik HP jack and RCAs. Volume pot in the middle. Power via a rotary switch a the rear with a long connecting rod to the front.

The toroidal transformer is 50VA and has 2x 22V output windings - that should do for the 21V outputs of the GRLVs.