No, this is not another thread asking about how to build a JR clone. It's more a clarification about how he did it, and a bit of reversed engineering.
In threads where people ask for schematics to a more powerful chip-based amplifier they are often referring to JR and they are often pointed to AN-1192 from National/Texas Instrument or to the website Shine7 Audio DIY Page where you can find 2 different ways on how to parallel more then 2 pcs of LM3886.
I don't want to discuss what method is best to use or is best sounding. I just want to show that Jeff Rowland did not use any of these configurations.
National/TI are using a bridge/parallel configuration with DC-servos to null the offset. 4 pcs of LM3886 that are paralleled 2 and 2 and then bridged together. One half of one channel is inverted, and the other half is non-inverting. This is a good sounding and reliable solution but it is expensive to build and frankly a bit to complex for many people.
Shine7 has taken a group of 3 paralleled LM3886 in an inverted configuration per PCB, where he null the DC-offset with a trim pot connected between the non inverted input and ground. He then bridges 2 identical PCB with a balanced line-driver or a trafo. This is working but, IMO, is not the preferred way to do it if you are going for a bridge/paralleled configuration, the way they do it in AN-1192 is more "correct".
Anyway. You can find more information about they do it in the links above.
Now, the Jeff Rowland amps differs a bit from AN-1192 and Shine7. JR uses 6 pcs of LM3886/channel but only paralleled, not bridged. Also he has connected the LM3886 in a non-inverted configuration and not inverted as many people claims on the Internet. At least at what you can tell from photos of his amps. I have not seen any of them in real life but there is some photos on the web like at the Shine7 website.
So how has he done it? Well, first and most, the LM3886 is nothing more than a basic Operational Amplifier with some extra addons. And if you know the basic of Operation Amplifiers then you know that you can easily trim the offset by just inserting a small amount of voltage to the inverted or non-inverted pin of the op-amp. Which one depends on how you have configured the Operational Amplifier.
I will not go in any deeper in the technical bits around this since how you do it is depending on your specs and requirements, but much information is to be found in MT-037: Op Amp Input Offset Voltage - Analog Devices and on page 6 and 7 in Application Note 31 Op Amp Circuit Collection among others.
Above is a photo from one of his amps. You can not see all whats going on there but you can tell that the LM3886 is non-inverted.
On the next image I have made a schematic from what I can tell from the picture above.
Im not sure of the value for R106 and C102. My guess is that C102 is a 100nF by it's small markings "A5". R106 should be set to give the gain in cooperation with R108. Below is a picture from where I have made a CAD of the schematic.
Pretty corresponding to the JR uh?
Well I don't know if this is for good use to anyone but I just wanted to share how you can make a more powerful amplifier similar to the JR-amps, or how to combine it with one half that is inverted and make a bridge/paralleled amp that is cheaper than using DC-servos, but more "correct" than the Shine7-version. I have tested both the inverted and non inverted and it works well and you can easy adjust the DC-offset. But to be honest there is room for tweeking if you read the technical documents above.
As always, to have the best result, use 0,1% tolerance resistors and always adjust the DC before soldering the output resistors.
Best regards.
In threads where people ask for schematics to a more powerful chip-based amplifier they are often referring to JR and they are often pointed to AN-1192 from National/Texas Instrument or to the website Shine7 Audio DIY Page where you can find 2 different ways on how to parallel more then 2 pcs of LM3886.
I don't want to discuss what method is best to use or is best sounding. I just want to show that Jeff Rowland did not use any of these configurations.
National/TI are using a bridge/parallel configuration with DC-servos to null the offset. 4 pcs of LM3886 that are paralleled 2 and 2 and then bridged together. One half of one channel is inverted, and the other half is non-inverting. This is a good sounding and reliable solution but it is expensive to build and frankly a bit to complex for many people.
Shine7 has taken a group of 3 paralleled LM3886 in an inverted configuration per PCB, where he null the DC-offset with a trim pot connected between the non inverted input and ground. He then bridges 2 identical PCB with a balanced line-driver or a trafo. This is working but, IMO, is not the preferred way to do it if you are going for a bridge/paralleled configuration, the way they do it in AN-1192 is more "correct".
Anyway. You can find more information about they do it in the links above.
Now, the Jeff Rowland amps differs a bit from AN-1192 and Shine7. JR uses 6 pcs of LM3886/channel but only paralleled, not bridged. Also he has connected the LM3886 in a non-inverted configuration and not inverted as many people claims on the Internet. At least at what you can tell from photos of his amps. I have not seen any of them in real life but there is some photos on the web like at the Shine7 website.
So how has he done it? Well, first and most, the LM3886 is nothing more than a basic Operational Amplifier with some extra addons. And if you know the basic of Operation Amplifiers then you know that you can easily trim the offset by just inserting a small amount of voltage to the inverted or non-inverted pin of the op-amp. Which one depends on how you have configured the Operational Amplifier.
I will not go in any deeper in the technical bits around this since how you do it is depending on your specs and requirements, but much information is to be found in MT-037: Op Amp Input Offset Voltage - Analog Devices and on page 6 and 7 in Application Note 31 Op Amp Circuit Collection among others.
An externally hosted image should be here but it was not working when we last tested it.
Above is a photo from one of his amps. You can not see all whats going on there but you can tell that the LM3886 is non-inverted.
On the next image I have made a schematic from what I can tell from the picture above.
An externally hosted image should be here but it was not working when we last tested it.
Im not sure of the value for R106 and C102. My guess is that C102 is a 100nF by it's small markings "A5". R106 should be set to give the gain in cooperation with R108. Below is a picture from where I have made a CAD of the schematic.
An externally hosted image should be here but it was not working when we last tested it.
Pretty corresponding to the JR uh?
Well I don't know if this is for good use to anyone but I just wanted to share how you can make a more powerful amplifier similar to the JR-amps, or how to combine it with one half that is inverted and make a bridge/paralleled amp that is cheaper than using DC-servos, but more "correct" than the Shine7-version. I have tested both the inverted and non inverted and it works well and you can easy adjust the DC-offset. But to be honest there is room for tweeking if you read the technical documents above.
As always, to have the best result, use 0,1% tolerance resistors and always adjust the DC before soldering the output resistors.
Best regards.
Yes I know 🙂
But since I don't know the marking on the R106 nor which gain JR uses I couldn't tell the value on R106. Maybe I was unclear but what I meant was that if anyone wanted to use the schematic then he/she has to calculate it him self to get the desired gain.
And, before any one else says it: this schematic is not finished and there should/could be more to it like zobel network, decoupling capacitors etc to ensure stability.
Last edited:
the manual states that absolute phase is inverting. however, since there is an input trafo the overall absolute phase may or may not be the same as the chipamp's circuit.
http://jeffrowlandgroup.com/Docs/M10Manual.pdf
to me the more interesting question is how it could be possible to deliver the claimed 150 w into 8 ohms without bridging given the LM3886 voltage ratings and the absence of an output trafo. pretty safe bet that it's a bridge-parallel design.
http://jeffrowlandgroup.com/Docs/M10Manual.pdf
to me the more interesting question is how it could be possible to deliver the claimed 150 w into 8 ohms without bridging given the LM3886 voltage ratings and the absence of an output trafo. pretty safe bet that it's a bridge-parallel design.
Last edited:
Well my thoughts is just a big guessing game (and some reverse engineering) since I have not found a schematic for his amplifiers based on the LM3886, nor any detailed photos of the input-board. The overall phase can certainly be inverting, but the I'm sure that the amplifier stage is built as a non-inverted stage only by looking at the photos of his amps.
Regarding the claimed power that the amp can deliver... well. I don't know from where he has gotten his numbers but if you enter 6 pcs of LM3886 paralleled into 8 ohms and with a power supply of +/- 24V into the Overture Design Guide then you get 160 W into 8 Ohm and 300 W into 4 Ohm. Paralleled should be able to drive more A than bridged? Or have I understood everything wrong?
And no offence, but if you study every photo of any LM3886-based JR amp you will see a "common rail" for the positive output to the speaker shared between all 6 pcs of LM3886. That would have been divided to 2 if the amps where in a B/P-config.
Best regards,
Carl
Regarding the claimed power that the amp can deliver... well. I don't know from where he has gotten his numbers but if you enter 6 pcs of LM3886 paralleled into 8 ohms and with a power supply of +/- 24V into the Overture Design Guide then you get 160 W into 8 Ohm and 300 W into 4 Ohm. Paralleled should be able to drive more A than bridged? Or have I understood everything wrong?
And no offence, but if you study every photo of any LM3886-based JR amp you will see a "common rail" for the positive output to the speaker shared between all 6 pcs of LM3886. That would have been divided to 2 if the amps where in a B/P-config.
Best regards,
Carl
The circuit look incomplete to me. I built a 4 channel 3886 poweramp after researching various 3886 circuits, and like it a lot. I don't recommend building poweramps for anybody who doesn't have a good understanding of phase margin, and a knowledge of how to verify it on the bench. If it blows up or only spuriously oscillates, there's a good chance it will destroy your speakers.
Rather than reverse engineering an existing circuit, why not just design a circuit that does what you want? If you want a bridged/parallel LM3886, National's BPA200 App Note (AN-1192) would be a good place to start. You'd probably also get quite a bit out of reading my Taming the LM3886 page.
Should you prefer to take a shortcut, you can assemble two of my Modulus-86 boards and connect them in a parallel or bridge configuration.
Not to knock Rowland's circuit, but I see several ways his layout could be optimized for better performance. At the very least, I'd move the local decoupling up to the IC, put the ground plane on the bottom of the board, and route the supply lines as pours rather than traces for lower supply inductance.
~Tom
Should you prefer to take a shortcut, you can assemble two of my Modulus-86 boards and connect them in a parallel or bridge configuration.
Not to knock Rowland's circuit, but I see several ways his layout could be optimized for better performance. At the very least, I'd move the local decoupling up to the IC, put the ground plane on the bottom of the board, and route the supply lines as pours rather than traces for lower supply inductance.
~Tom
Last edited:
the rowland power feed is through a tighly coupled parallel buss -- you can see a very small part of it and its shadow in the upper left of the pic (it's red-orange and it looks like a gigantic one-piece resistor network).
almost everything, yes.
see figure 5 in the application note in tomchr's post
parallel without bridging into 4 ohms the maximum output is about 120 watts with +/-35V -- at 24v and 8 ohms it would be much less.
Last edited:
In the photo's where you can see two or more chips you can see two traces that are alternately connecting to the input resistors. I would guess that one trace is inverted. That would also explain the use of input transformers.
AFAIK all versions included xlr inputs
An externally hosted image should be here but it was not working when we last tested it.
early rowland bpa models used jensentransformers later ones used lundahl
here is pic with a jensen
here is a multi-chip pic
as you can see, each chip has a trimpot
you can also see the parallel feed
Last edited:
You can see that R104 (middle chip)is connected to a different trace than R111 (left chip).
The absence of decoupling may also hint at bridged.
The orange bars look like they consist of two conductors. One bar for + - and one bar for the two outputs.
The absence of decoupling may also hint at bridged.
The orange bars look like they consist of two conductors. One bar for + - and one bar for the two outputs.
Last edited:
the shine7.com clone has the bypasses under the board but there is no comment AFAIK whether he copied this from the model 10 or just decided to add them.
looks like using smd caps would be easy here
looks like using smd caps would be easy here
Last edited:
there is another identical board in the shine7 chassis and the two boards form a bpa-300.
whether that's close enough to call it a clone, i'm not going to argue it one way or the other since we aren't entirely sure what the model 10 has for topology.
the board on the back panel is a drv-134 piece
get a look at the monster rifas!
whether that's close enough to call it a clone, i'm not going to argue it one way or the other since we aren't entirely sure what the model 10 has for topology.
the board on the back panel is a drv-134 piece
get a look at the monster rifas!
Last edited:
here's what the shine7 author says about it.
"After I built several LM3875 and LM3886 gainclone amplifiers, I was totally impressed by their audiophile sound quality. Then I saw the Jeff Rowland Power Amplifiers that use LM3886, I decided to create my own version - the BPA300. My design goal is to create a audio power amplifier that can deliver 300W into my 4-ohm DIY speaker with low distortion. I want it to produce deep, tight and punchy bass while keeping the excellent mids and highs from my other gainclones."
in a later version he used a lundahl instead of the drv-134.
http://www.shine7.com/audio/LL1545a.htm
"After I built several LM3875 and LM3886 gainclone amplifiers, I was totally impressed by their audiophile sound quality. Then I saw the Jeff Rowland Power Amplifiers that use LM3886, I decided to create my own version - the BPA300. My design goal is to create a audio power amplifier that can deliver 300W into my 4-ohm DIY speaker with low distortion. I want it to produce deep, tight and punchy bass while keeping the excellent mids and highs from my other gainclones."
in a later version he used a lundahl instead of the drv-134.
http://www.shine7.com/audio/LL1545a.htm
Last edited:
The supply rail current flow is completely different with this layout. A very good reason why he could not skip the standard HF and MF decoupling.
yes and there is a lot of trace inductance so i tend to think the layout calls for a 10-20uf oscon at the chip pins along with the low value caps a la tomchr.
even as is, though, shine7's measured results are respectable. perhaps he got lucky with the positive rail having the larger trace inductance but also the better positive side PSRR to offset it to some extent.
even as is, though, shine7's measured results are respectable. perhaps he got lucky with the positive rail having the larger trace inductance but also the better positive side PSRR to offset it to some extent.
Last edited:
- Status
- Not open for further replies.