and now for something completely different part 3

[JoaMat]

On 7/17/2024 at 12:13 AM, justin said:

Last year I ordered a lot of different boards of KG's amps/PSUs and had someone here put them together as a project. One of them was this 6DJ8 hybrid that uses the triple output buffer of the CFA3. Single-ended output only. There was 1 missing trace on the board that prevented it from working, found and fixed that, so possibly nobody had assembled one of these before. Problem is there's a substantial hiss (very easily audible with even sensitive planars, something like 0.3mV rms of 'white noise'. The +100V is coming from a TL783 regulator, tried AC, DC, batteries for heater, different grounding, gate stopper resistors, feedback caps, w/out servo..exhausted a lot of stuff but couldn't get more than a few % improvement on the noise. Anyone have ideas?

 

th.PDF 51.92 kB · 23 downloads

What dual op amp are you using?

[justin]

have to look, probably either a OPA2134 or 2132

But I also tried it without the servo and using a pot instead

[JoaMat]

Have had a new bias servo working with my CFA3smd something for a few days. Schematically it looks like this,

and the new thing is in red square. Now I get a good reference point to the servos. Don’t understand why I didn’t figure this out earlier – slow brain perhaps. All four servos are fed from the same point. So, one trimmer and full control over the BIAS of all channels.

Diagram for new servo - 6 minutes from cold.

and diagram for old servo (in mA for the first 40 minutes from power on)

Please have in mind that the graph for no servo is for my CFAsmd something. I guess the original CFA3 have a different and better behavior – where all tripple Darlington and vbe transistors are mounted on heat sink.

 

Edited July 27 by JoaMat

[kevin gilmore]

would like to see what effect this has on thd. especially large voltage swings. still think an optoisolator drive to the vbe multiplier.

[JoaMat]

Like this? Think I got inputs to op amp right.

Edited July 28 by JoaMat

[kevin gilmore]

if it works right then yeah.

[kevin gilmore]

so this is the way to do it correctly. its the only way not to seriously effect the thd.

unfortunately both opamps have to take the voltages of the output devices. which means they have to  be opa445

you set rv2 to the maximum amount of current. Then B1 is a 10k pot to V+ and ground.

r11 is actually 100k.

Edited August 4 by kevin gilmore

[JoaMat]

So, I got myself QuantAsylum QA-403 Audio Analyzer. Now I can make some simple analyses of amplifiers I build.

Screenshot – my CFA3smd something with HD800 headphone,

2Vrms into headphone. Probably room for improvements. But I’m satisfied and I’m not competent enough to make it better. 

[arcanaking]

Hi all, I am following the schematic from the document "cfa3productionss.pdf" to build the cfa3 zero feedback version. I make use of the verion 1.35 board as a reference for the later building. After checking all the components and soldering to the pcb, I found that there is no mv when I short the input and measure across the 1 ohm.

The 2sa1860 is replaced by mje15033 and the 2sc4886 is replaced by mje15032. The 47uf cap is smd. After I switch the amp for 30 mins, the smd bjt is hot but the 15032 & 15033 bjt is cold. Is there any wrong implementation for the cfa_amp board that lead to this situation?

 

[audiostar]

Are you building a headphone amp and isn't this the board for the power amp version? 

[kevin gilmore]

what is the name of the gerber file for the output board

 

[arcanaking]

11 hours ago, kevin gilmore said:

what is the name of the gerber file for the output board

 

The output board name is cfp3largeT.zip which should be posted on 11  Jan 2023.

  

On 1/11/2023 at 10:20 PM, kevin gilmore said:

here you go

 

cfp3largeT.ZIP 42.53 kB · 164 downloads

 

I try to measure and record the voltage between each pin and the gnd.

 

[kevin gilmore]

minimum power supplies +/-24

might have to change a couple of resistors in the bias generator for this output transistor.

[arcanaking]

23 hours ago, kevin gilmore said:

minimum power supplies +/-24

might have to change a couple of resistors in the bias generator for this output transistor.

Currently, the power supplies is +/-30v.

I have ordered some mje15030/15031 for further test. Also, If I have to change the resistor in the bias generator for the output transistor, which resistors should I change? Does the 500ohm and 110ohm in last output stage? or more?

[kevin gilmore]

you need to measure the voltage across q36  should be about 5 volts. if not adjust rv2 and if that is not enough change r36

[arcanaking]

On 10/7/2024 at 5:20 PM, kevin gilmore said:

you need to measure the voltage across q36  should be about 5 volts. if not adjust rv2 and if that is not enough change r36

For the MJE15032 and MJE15033 version, I try to rotate the 5k trim pot at least 10 rounds. Until the the voltage across the Q36 (also call Q10 in my schematic) greater than 3.3v, the voltage across the 1 ohm start changing. When the voltage across Q36 comes to 3.6V, all the voltage across 1 ohm comes to 150mv to 200mv range.

This measurement only rotate the 5k trim pot, and the 10k trim pot does not rotate (just for a little record) and stay in the middle point. Only one side of the CFA3 amp board (L+ channel) has already generate a great heat to the heatsink after 5 minutes in a fanless environment which suprise me a lot (the heatsink is around 20mm).

[justin]

I cant even tell if this is a stupid question but where are your heatsinks? if you try to power it even for a short time without heatsinks the bias is going to be very unstable

On 10/7/2024 at 5:20 AM, kevin gilmore said:

you need to measure the voltage across q36  should be about 5 volts. if not adjust rv2 and if that is not enough change r36

3.6V sounds right to me, 5V would probably result in about 700mA bias

Edited October 9 by justin

[kevin gilmore]

with the sanyo transistors 5v results in about 500ma bias. which was the original design spec for this board.

[arcanaking]

9 hours ago, justin said:

I cant even tell if this is a stupid question but where are your heatsinks? if you try to power it even for a short time without heatsinks the bias is going to be very unstable

3.6V sounds right to me, 5V would probably result in about 700mA bias

Hi Justin, sorry for the misleading of the picture. It is because the picture I post is only for a test version for myself to validate if the schematic I draw has any problem. 

After building the Dynalo, it is easy to measure the voltage across 20 ohm. Yet, I cannot measure the voltage across 1 ohm after building the cfa3. Therefore, I make use of the previous printed original PCB from Dr. Gilmore to test if there is the same situation. Therefore, I don’t install the heatsink and test it with minimal components.

Actually, this is my CFA version (with a 20mm width heatsink). At the beginning, I would like to create a standard 1U size, therefore, I am fine with the original size (the 3u height version). However, I would like to minimise the space after watching your design to 32mm width. Therefore, I start drawing the new board by referring to the published schematic and the board measurement to ensure the circuit is correct.

[arcanaking]

6 hours ago, kevin gilmore said:

with the sanyo transistors 5v results in about 500ma bias. which was the original design spec for this board.

Thanks for your information. I am ordering an Infrared thermometer for the further temperature measurement of the heatsink to see if it is suitable to increase the bias. 

Long time ago, I have pmed you for a simple preamp setting in the cfa3 design. Yet, the setting is not suitable to the CFA3. Therefore, I would like to integrate Dynalo and CFA3 together (+-30v input in a separate case) to archive both hp amp and preamp function in one case.

I would like to ask how large of the transformer’s VA should be used when supply both circuit at the same time? Currently, I am using two 50VA 2*25V transformer for either one part to switch on. According to Headamp information, the Dynalo seems 40w power consumption and CFA3 seems at least 75w in one pair output BJT situation. I am curious if two 80VA 2*25v is enough for my situation.