My_Ref Fremen Edition - Build thread and tutorial

Jump to Latest
S

Spartacus

Member
Joined 2002
lehmanhill said:
Last summer, I decided to learn more about resistors and started researching and collecting parts to try. About the same time, George pointed all of us to a thread where diyAudio member EVUL was playing with series/parallel resistors. I found that very interesting.

Let's talk about R3. It is in the main signal path and inside two feedback loops. It seems clear that an improvement in R3 would likely improve sound quality.

R3 is different from other signal path resistors because it needs to be a power resistor. Using Mauro's numbers, the My_Ref has 45W into 8 Ohms. For the standard build, R3=0R47, that's 2.6W in R3. For the evo A, 1.85W. We should be using resistors rated at 4 to 5W to have some safety margin.

In my opinion, power resistors used in audio, typically thick film and metal Oxide, are much weaker performing than the lower power metal film/foil resistors. They have 2 to 3 orders of magnitude worse current noise.

They are also subject to what is sometimes called “power modulation distortion.” Combine the change in resistance with temperature with how hot the resistor gets per watt and you have a PCR (power coefficient of resistance). A change in resistance in the circuit changes the output and that is distortion. More than that, music has loud and soft passages where the power dissipated by the resistor is changing with the music. You can see how this can negatively affect sound.

The Caddock has two orders of magnitude worse current noise than a typical metal film 1/4W resistor. It as a tempco of 200 ppm/C compared to 25 to 100 being typical for 1/4W metal film resistors. The Caddock does have the advantage of being designed for a heat sink. It would be great if we could find metal film resistors with better specs that could handle the power.

Quad resistors, also known as Quartet resistors or series/parallel resistors have some interesting properties. Starting with a description, a quad resistor replaces a single resistor with 4 resistors of the same value in a combined series/parallel package. In other words, 2 strings of 2 resistors in series and the 2 strings in parallel. That results in a resistor that has the same value as the single resistor, but 4x the power rating. Better than that, the theory says that the quad will have 12 dB less distortion than the single resistor.

Because the power rating is 4 times the single resistor, it is possible to use lower power rating resistors with better specs. In the case of R3, I have chosen to use Susumu KRL current sense resistors, specifically part number KRL2012E-M-R330-F-T5 for the 0R33 value used in evo A. This resistor is a foil resistor that is rated at 1W, 50 ppm/C, and should have very good current noise. Combining these into a Quad resistor will give us 4W power rating and, if we design the pcb properly, we should have a good opportunity for heat transfer.

In the picture, you can see my homemade version. A two sided copper clad board was cut into 15 x 30 mm parts. Near the bottom, I cut two narrow slices out of the copper and soldered the Susumu KRL over the gap. Note that this is a long side solder resistor. Two 1 mm holes were drilled and resistor lead were soldered in place. This was repeated on the back side of the board.

The resistors and leads are as close to the bottom as possible so that inductance is kept small. The height of the board is mainly to give surface area for heat transfer and it works pretty well. There is also a hole near the top so that a heat sink can be attached for additional heat transfer. As it sits here, the pcb runs at between 40 and 45 deg C without additional heat sinks. Of course, the ideal would be to have small pcb made professionally with a gold wash and no solder mask to get in the way of heat transfer.

Ok, so enough background, you want to know how they work. In my opinion, they are a very nice step forward. They improve the clarity of the ADA4627 even more and uncover instrument harmonics that were not heard before. For me, this warms the sound and makes it more realistic. But this is just one person's view. I would hope that a few builders, especially those using jfet opamps, will try this and report their impressions.

And one other thing. You need 8 resistors to make a stereo pair. In the US, Mouser is selling 10 of the 0R33 KRL for $3.70. Even if you throw away the extra two resistors, you are getting a quad resistor for about $1.85 each. Compare that with the current Mouser price for the Caddock at $5.47 and you have another reason to try the KRL.

Jac
Click to expand...

Jac, finally got around to implementing these output resistors. Results are good, better clarity and a more natural tone. Not a huge change, but useful. Thanks!

Next, on to replacing the pump resistors.
 
lehmanhill

lehmanhill

Member
Joined 2008
murgui said:
Could I have further explanation?

Regards
Click to expand...

Murgui,

Sorry for the lateness of my reply. Things got kind of quiet on this thread for a while.

The LM318 is an interesting and unusual opamp. Mauro chose it specifically for some of these characteristics. Among the features, pins 1, 5, and 8 are available as compensation connections that allow a designer to externally control the response of the opamp. If you look up the LM318 datasheet, you will find many examples of ways to use this compensation and a schematic of where those pins connect inside the chip. Some of these are changing slew rate, bandwidth, stability, and settling time.

Mauro experimented with compensation and came up with the compensation scheme, including C10, that you see on My_Ref. Over the years, others have played with the value of compensation components to tweak the subjective response of the amplifier. I am not one of those that has played in this area. Maybe someone else can tell you a general affect of C10.

The entire compensation network is specific to the LM318. For example, in George's experiments with jfet input opamps, the compensation components are removed from the circuit.

I hope that helps.

Jac
 
lehmanhill

lehmanhill

Member
Joined 2008
Telstar said:
I think you are uncorrect: the DC blocking cap is removed but the compensation network is only changed.
Click to expand...

You are correct. Not all of the compensation network is removed. From post #3085;

Change R3 to 0R33
Remove C10, R39, and C24 and/or R43
Keep C32 as a compensation cap.

Anyway, for murgui's sake, most of these compensation components are related to opamp choice. For example, the LM318 is a BJT input opamp which has much more input bias current than an FET input opamp like OPA827.

That said, I am NOT the deep knowledge base on this circuit. I would have to refer you to Dario, George, and others.

Have fun.

Jac
 
lehmanhill

lehmanhill

Member
Joined 2008
olivosunlight said:
I am wondering if RCA and Binding posts (mass) are connected chassis or not?
Click to expand...

I am not sure I understand your question completely. If you are asking if RCA and speaker output (binding posts) are electrically connected to the chassis, then the answer is no. You need insulation, usually plastic spacers that come with the RCA or binding post, to keep electricity from passing between the connector and the chassis.

If you are asking if RCA and binding posts are structurally attached to the chassis, then the answer is usually yes. It isn't a requirement, but it is often convenient. You can think of the chassis as a barrier between the outside world and the inside electrical components. The chassis is grounded to keep it safe and the RCA or binding post are a way to pass the signal through the barrier without connecting to it.

I hope that helps.
 
O

olivosunlight

Member
Joined 2013
lehmanhill said:
I am not sure I understand your question completely. If you are asking if RCA and speaker output (binding posts) are electrically connected to the chassis, then the answer is no. You need insulation, usually plastic spacers that come with the RCA or binding post, to keep electricity from passing between the connector and the chassis.
Click to expand...


Thank you for your advice,



I am working on chassis now and did not open RCA/binding posts completely, just measured and drilled holes. I checked and found spacers in RCA/speaker out as insulations which are insulated between chassis vs connectors.



Best regards
 
lehmanhill

lehmanhill

Member
Joined 2008
Mituisho said:
Hi Dario

I am finishing my build and wonder if it is ok for me to reduce the value of R23 in order to get a higher voltage (say 4v DC) for the external LED for cosmetic reasons?

The voltage I measured across R23 and LED are 15.7v and 1.7v respectively.

Thanks!
Click to expand...

Hi Mituisho,

I know you asked Dario, but I can answer as he is probably very busy with Group Buy 12.

It should be no problem to change R23 for a different color/voltage LED. The original value of R23 (6.8k) assumes a red LED at 1.7 V. That is about 2 mA which is safe for long life on a red LED. You will want to keep the current about the same so that you don't steal current from the speaker protection circuit, but it shouldn't be too sensitive.

Using your numbers and keeping 15.7V across the 4V LED and resistor, you will have 11.7V across the resistor. At 2 mA, the resistor should be about 5.8k. A 5.6k is close enough.

By the way, you may want to check the voltage drop across the 4V LED at about 2 mA because it may be rated 4V at a different current.

Jac
 
danny_66

danny_66

Member
Joined 2005
Luca72c said:
As already suggested, don't forget to try speakers crossovers resistors, like Mundorf Supremes, Duelund Graphite or Jantzen Superres. They are rated > 10W, non-inductive and are meant (on paper, at least) for keeping the signal clear, natural and uncolored.
They are expensive, but only 2 of them are needed...
Click to expand...

All those are beaten on paper and in measurements by the Isabellenhuette PBH
They were also tested in the doc: "Resistor Current Noise Measurements"
You have to zoom in on the graphs to better see the thickness of the different traces.

I use them in the crossover of my speakers and in my amplifiers (source, emitter, feedback resistors).
They are also available in different SMD packages, the SMR and SMV packages are easy to solder directly a wire to it.

I buy mine at buerklin, they have different series.

Definitely a resistor to put on the list !

Regards,
Danny
 
lehmanhill

lehmanhill

Member
Joined 2008
danny_66 said:
Isabellenhuette PBH
They were also tested in the doc: "Resistor Current Noise Measurements"

I use them in the crossover of my speakers and in my amplifiers (source, emitter, feedback resistors).
I buy mine at buerklin, they have different series.

Definitely a resistor to put on the list !

Regards,
Danny
Click to expand...

They clearly measure better than the Caddocks. Unfortunately, they are not very available in the US, but an interesting choice.

Jac
 
L

Luca72c

Member
Joined 2013
danny_66 said:
All those are beaten on paper and in measurements by the Isabellenhuette PBH
They were also tested in the doc: "Resistor Current Noise Measurements"
You have to zoom in on the graphs to better see the thickness of the different traces.

I use them in the crossover of my speakers and in my amplifiers (source, emitter, feedback resistors).
They are also available in different SMD packages, the SMR and SMV packages are easy to solder directly a wire to it.

I buy mine at buerklin, they have different series.

Definitely a resistor to put on the list !

Regards,
Danny
Click to expand...


Very interesting! I considered Isabellenhutte before, but i couldn't find the right wattage - rated resistance combo.
Surely i missed those PBHs. Did you try them on My_ref FE output?
 
G

giacintopapa

Member
Joined 2018
Hello everybody, I'm new in this thread and I'm starting to assemble the My_Ref_FE following the precious Tutorial by Dario Inserra. I apologize for my doubts as a beginner: I can not clearly understand what is meant by "thermal pads", those to be soldered at a later time, at temperatures above 350°C.
Every explanation will be welcome, thanks to everyone
Giacinto
 
You must log in or register to reply here.
Top Bottom