Sub' Amp - LM4780 questions!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
A while back I bought and built 2 x Audiosector LM4780 Gainclones - one for a friend and another for myself.

I completed the chassis for the first one and was really impressed with the result but as I have an NAD312 amp I didn't really get around to finishing the second amp which I had planned to use with my aging but enjoyable Infinity Ref10 (6Ω) speakers.

Now I'm thinking I'd like to get a bit more bass out of the NAD amp and speakers and figured I could rebuild the LM4780 as a bridged amp to power a DIY subwoofer. The NAD312 has a pre-amp in/out connector which baffles me slightly - but as it currently has a U joining the in/out signal feeds on theback of the amp I have the feeling I can route the currently playing signal to a subwoofer at linelevel.

I don't want sub' that gives the equivilent to a bass enema - just something that adds to the overall effect.

I guess the problem is that I know very little about subwoofer construction and how much power they need to drive them? Or how best to combine the L/R signals and remove the high-end frequencies.

I've looked through the this sub-forum, but I can't find anyone who has completed sub-amp build.

I've built the Gainclone, some CMOY headphone amps, 3 SOHA headphone amps, 3 Alien DACs and other bits and pieces, so I'm reasonably a reasonably competent builder.

Help? :)
 
Some figures and sums:

I have a Talema transformer that gives me 6.3Amps on each of the 18VAC secondaries.

Rectified that's about 25VDC

From the graphs on the datasheet - I might expect a maximum of 53W output power from 25VDC -/+ (with a high THD figure!).

Bridging theoretically gives me 4x this output? 212W? Egads! Or am I wrong?

The Overture Design Guide Excel Sheet from National says 106W and shows the load as 8ohms. Does this mean I can only use 8ohm, or that the amp sees the 4ohm I feed the sheet as 8ohm because of the bridge?
 
More thoughts:

ESP's Linkwitz transform looks excellent as a method for combining the L/R signals and getting the most from the sub's driver: http://sound.westhost.com/project71.htm
Though I think I would include the buffer section first with a filter as I can't see much need to expend energy amplifying signals the sub can't reproduce. Or is that nuts?

I'm still unsure about driving the sub in balanced or parallel mode and which is more suitable. Do I also need to buffer the GND signal for balanced mode? Or does the gainclone effectively to that?
 
raromachine said:
...ESP's Linkwitz transform looks excellent as a method for combining the L/R signals and getting the most from the sub's driver...

I'm sorry but since most subwoofer drivers are dual voice coil (i.e., accept two separate channels of input) why is combining channels important?

It sounds like you need to build two monoblock amps (LM3875) or one dual amp (LM4780)
 
I hadn't been looking at dual coil sub's - and figured that a parallel LM4780 would put more power into a single coil sub, than 2x single channels putting power into dual voice coils?

My amp has pre-out and power-in, i.e. it has no specific sub-out - so I would split the pre-out using a Y - feeding one back into the power-in, and another into the sub-in.

I don't know enough about the amp to know how much current the pre-amp can feed to the power amp - so wanted to mitigate this buy presenting a nice high impedance to the pre-amp.

I figured I woudl do this by buffering the R/L - and then I would combine them because one speaker = one channel. I wanted to use the linkwitz circuit to put the sub in a smaller box and get the most from it. So building a one channel linkwitz was less parts than a dual channel :)

But! If dual voice coils in the sub and 2 sep channels is easy - then that's pretty cool as it means no changes to my already assembled LM4780.
 
raromachine said:
More thoughts:

ESP's Linkwitz transform looks excellent as a method for combining the L/R signals and getting the most from the sub's driver: http://sound.westhost.com/project71.htm
Though I think I would include the buffer section first with a filter as I can't see much need to expend energy amplifying signals the sub can't reproduce. Or is that nuts?

I'm still unsure about driving the sub in balanced or parallel mode and which is more suitable. Do I also need to buffer the GND signal for balanced mode? Or does the gainclone effectively to that?


Hmmm....
linkwitz Railey designed an equaliser to correct the low-end frequency response of a totally enclosed loudspeaker in order to get the bass extension that the speaker designers intended.

The roll of at bass for a sealed box of correct dimensions is around 12 dB/oct and the equaliser will correct this by rolling in at 12dB/oct which effectively puls the frequency response of the speaker in line with what it should look like.

Now you have to match the equaliser with your speaker response so that the 3dB roll off is also the 3dB roll in of the equaliser.

In most case people make sub-woofers for effects like movies and these are single and dual cavity resonators that is already strongly band pass filtered by the mechanics of the design.

The roll-off is also quite servere sometimes as high as 24 - 30 dB/oct.

A linkwitz Raily equalisation for a bandpass box does not give good results becase it will boom and shake the house only which leads chesty voices and unnatural bass.

Also the combined phase shift of the bandpass box and the equalisation will make instruments not sound accurate.

Assuming a bass note of sorts is made of a fundamental and a harmonic, and in time the harmonic actually leads in the tone. After passing though both unmatched filters the leading harmonic may be delayed to a lagging harmonic and where there was good attack, a thump, there now becomes a decaying sound, a boom.

Be carful with bass boxes you can be very dissapointed.
 
Thanks Nico - that's very interesting.

From that I get that you would recommend not using the Linkwitz Riley circuit - but I'm not quite sure what you're recommending - I guess that it's to tread and think carefully about what I want.

My assumption so far as been that I should break down the amplifier into these steps:

----

1. Buffer the signal from my preamp

- to negate the impact on the power amp

2. Remove the frequencies the speaker driver cannot reproduce
filter

- it seems wasteful to spend energy amplifying frequencies I'm not going to use.

3. Do additional filtering to based on the requirements of the driver/enclosure.

- My choice in drivers doesn't seem to be huge, so I figured there would be some fiddling to get this right.

4. Feed the signal through to the LM4780.

- Amplifying it for the driver

----

Additionally (mostly to Carlos' comment): is it better to use a crossover? or to filter before amplifying? I would have thought the latter - but I'm easy either way.
 
Nico Ras said:
Hmmm....
linkwitz Railey designed an equaliser

raromachine said:
you would recommend not using the Linkwitz Riley circuit -

Firstly, gents, it's called a Linkwitz Transform after Siegfried Linkwitz (apologies for being pedantic).

This circuit can be very demanding power-wise, and was originally conceived for use with a sealed enclosure (power demands upwards of +6dB wrt the non-equalised part of the system). A useful tool for modelling the system power requirements as well as the actual circuit values can be found here FRD Group. See Unibox and Linkwitz Transform Design Program.
I believe that WinISD PRO may also have these tools.

I have adapted an alternative approach to the LT circuit, and that is a simple high-Q 4th order high-pass filter. Does the same job of LT, but includes a cut-off for the out-of-band lower frequencies*.

Good Luck with your design.

*IIRC the facility in Unibox has an option for adding a high-pass circuit.
 
Here is the schematic I build up to mix and buffer the input signals and then to strip out high freq's (-3db point is 337hz according to: http://ourworld.compuserve.com/homepages/Bill_Bowden/opamp.htm ).

buffer-and-lpfilter.png


For the time being the opamps are fed with a +/-4.5v supply from a 9v battery and a TEL2426 railsplitter.

I used the 100K resistor between the input and gnd as I'd seen it done on headphone amps, and without it the opamp spits out a few volts of DC that I didn't think the driver would like :)

It seems to work well - though I think even 337hz as a corner freq is too high. I also wonder if this should be a band pass filter, as when pushed the cone will (occasionally with very bassy music) meeps a little.

This is what it looks like:
subwoofer-buffer-and-lpfilter-pcb.jpg
 
Before I setup the power supply stage (2x vregs) I wondered if I needed to calculate how much power the 12 regulators might have to dissipate if they're fed from a 25v supplies.

I had hoped to use to-92 7xL12 regs as I have some spare.

Is it the voltage drop * output current? i.e. 13 * .1a = 1.3w?

Or does this not matter as long as the current drain of the opamps is kept within the reg's max output?
 
Here is the completed circuit layout - and I can say it DOES work, and works well (IMO).

The questions above are still open, but I have only 2mv of offset and the 12+/12- swing seems to avoid any clipping (which I did have with the 9v battery + rail-splitter). :)
 

Attachments

  • buffer-lpfilter-lwtransform.png
    buffer-lpfilter-lwtransform.png
    13 KB · Views: 303
CarlosT said:


I'm sorry but since most subwoofer drivers are dual voice coil (i.e., accept two separate channels of input) why is combining channels important?

It sounds like you need to build two monoblock amps (LM3875) or one dual amp (LM4780)

Umm. Feeding the same driver 2 potentially different signals sounds like a really bad idea to me. Your amps will fight over the speaker.

-Nick
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.