MyRef RevD - successor to RevC

[mcspack]

I am going to be building one of these in parallel configuration using the lm4780, I will be using the amp to drive 4 ohm speakers eventually. I have a few questions. Will I need to change compensation values for the lm318, if so how should I start that process? Also, I have found metal foil vishay resistors that are 0.01 tolerance, how will metal foils sound in the R3 spot? The closest I cound find at this spec to 0.47 is 1ohm will this be ok for the circuit? They are rated to 8w which sould be ok especilly if I find a good heatsink. I have been able to find other vishay metal film in values 0.02 and 0.05 tolerances for the R5,6,8,9 spots.

[linuxguru]

Originally Posted by mcspack

I am going to be building one of these in parallel configuration using the lm4780, I will be using the amp to drive 4 ohm speakers eventually. I have a few questions. Will I need to change compensation values for the lm318, if so how should I start that process? Also, I have found metal foil vishay resistors that are 0.01 tolerance, how will metal foils sound in the R3 spot? The closest I cound find at this spec to 0.47 is 1ohm will this be ok for the circuit? They are rated to 8w which sould be ok especilly if I find a good heatsink. I have been able to find other vishay metal film in values 0.02 and 0.05 tolerances for the R5,6,8,9 spots.

If you haven't already built an LM3886-based Rev C, don't try a parallel version as your first attempt! Lots of things can go wrong.

Compensation for the LM318: As long as your overall voltage gain is the same, i.e. (1+12k/390), you can probably get by with the same compensation as the RevC. YMMV, because Mauro's Evolution apparently uses slightly different compensation (and also a DC Servo) in addition to the dual LM3886 per channel topology.

Vishay Metal Foils for R3: No idea, but stock Futaba and Caddock non-inductive power resistors are easily available with ratings from 3W-25W, value 0.47R, and work well at R3. No real point in changing this unless you're big on experimenting. 1 ohm impoves the phase margin (stability) but wastes more power.

Precision resistors for R5, R6, R8, R9: No need to buy the very tight-tolerance Vishay S102 or similar resistors. Just get a 100 or so regular 1% tolerance non-magnetic MFRs of high quality, e.g. Holco precision or similar, and match them with a 4.5/5.5 digit DMM. You'll find at least 10-20 pairs matched to 0.01% or better, and at least a few pairs matched to 0.001% or better. You can use the unmatched resistors at other locations or in other projects.

[mcspack]

Thanks guru for the reply,

I decided on the caddocks (MP930 0.5R) for R3, They are already matched at 1/100th per their value right? the others I will use PRP as they are reportedly very tight already. I will eventually build two of these amps so I bought plenty of these to match. 1% audio grade carbon films for the rest of the signal path, except r11 where I have an old resista. Wima fkp2 and silver mica for almost all of the caps. No output safety circuitry.
I"m using a BG for the C9, I know Mauro used the DC servo but I have to idea how to compensate for that in this circuit so? I will be powering this amp with either smps or batteries with TL431s as shunt regs for the lm318 ps. When I get this first amp working I'll report back with my impressions. Currently this amp is built as a parallel gainclone. I will be rebuilding this all on ratshack boards PTP wiring.

MC

[mrsavage]

Is anyone thinking about another group buy for BOTH the C and D Version of the My_Ref
amplifier? I'm new and would like to build several of these amps. Please count me in.
Take care.

[bcmbob]

If a group buy doesn't materialize, linuxguru has everything you might want in both bare PCB and kit form. Send a PM by right mouse click on name two posts back

[danielwritesbac]

Not practical:
LM3886 run at full blast with a 4 ohm speaker, 35vdc~37vdc rails made pops and cut-outs due to overrun and thus longevity problems are expected.

Practical:
LM3886 run at full blast with 4 ohm speakers Works with amp power from 29.5vdc~31.1vdc rails, approximately a 20,0,20 transformer--No pops/cut-outs detected. LM3886 behaved very well!

Not practical:
Under-run with 18,0,18 transformer may be inappropriate due to exacerbated clipping of the spike system not conducive to useful output at under-run voltages.

Test environment:
Generous heatsink, cool air intake vent underneath and hot air output vent above, were used during this testing.
However, bad enclosure design like an oven wasn't tested.

Not yet measured:
There wasn't test with 4 ohm speakers while LM3886 was powered in-between 32vdc~34vdc, because I didn't have a 22,0,22 transformer available.
Question: In real life, does LM3886 do mute pop/cutout if run at full blast with a 4 ohm speaker? Perhaps someone who owns the 22,0,22 transformer, LM3886 and 4 ohm speakers could contribute that practical information?

[AndrewT]

Spike is about protecting the chipamp from abuse.
That abuse could be the result of asking for too much output voltage, or could come from asking for too much output current, or a combination of both.

Setting up a chipamp using a PSU voltage to suit the 8ohms or 4ohms or other speaker impedances available makes good scientific sense.

Voltage clipping resulting from asking for too much output voltage is a user error, not a Spike error.
Current clipping resulting from asking for too much current is an amplifier assembly/design error (not a chipamp designer error). It is not a Spike error.

The builder should look to end use and try to eliminate the possibility of designer error and user error.

Choosing a lower voltage transformer than the maxium for 4r0 loading is a good designer choice. Chipamps in general are current crippled into low impedance loads.

I strongly suggest that aiming for 60 to 68W into 8r0 is preferable to the same power target into 4r0.
I further suggest that the power target for a 4r0 and 4ohms loading be significantly reduced below that 60W resistor capability.
Aim for 40W into 4r0 and that should leave some margin for extreme current peaks into reactive 4ohm speakers.

This is in complete disagreement with

Practical:
LM3886 run at full blast with 4 ohm speakers Works with amp power from 29.5vdc~31.1vdc rails, approximately a 20,0,20 transformer--No pops/cut-outs detected. LM3886 behaved very well!

Not practical:
Under-run with 18,0,18 transformer may be inappropriate due to exacerbated clipping of the spike system not conducive to useful output at under-run voltages.

[udailey]

There should be a many 22 volt experiences. When I sold the kit I used 22 volt transformers .
I use 32 volt rails and 6 ohm speakers with no pop not at all. I did have a pop until I used a better on off switch .

[AndrewT]

Hi U,
I cannot tell if the 22V is AC or DC.
The context of 32 seems to confirm you mean 32Vdc.

[udailey]

For my post I mean both.
I sold 22vac transformers with my kit so there should be lots of guys with experience with 22vac and right now I use 32vdc rails. I get no pop. I used to get pop with 32vdc into 6R but I bought a broken Sonance Sonamp and removed the guts. I use their on/off switch and no pop. I attribute it to the switch. They have a cap across it.
If we muted the amp by taking signal + to Signal GND until the voltage on the PSU came up fully would we remove the POP? I always wondered if the pop was the result of the quick increase in voltage as the PSU caps charged or is it just the arcing across the on/off switch creating a high voltage for a short time before the switch closes?.
Uriah