A rotary encoder controlled integrated chip amp LM3875

[justblair]

I am trying to come up with a pcb design for a very compact integrated amp based on the LM3875 chip, I have posted a couple of threads already.

My basic premise is that the whole amp will fit in a hammond 1455 case 220mm*100mm*55mm. That is pretty small when you consider I wish to use a regulated power supply.



I had looked before at using Pedja Rogic's Buffer. However the problem of volume control becomes an issue. The torroid really needs to be at the front of the case. That way signalling in and out is as far as possible from it.

If I use a stereo pot, then the signal wires must pass very close.

A digital volume pot seems the easiest solution. I have a DS1802, and I certainly could use it. However on searching the Maxim site, I noticed the MAX5440.

Call it silly, but the idea of an encoder controlled volume pot seems good. The added advantage of a 5 LED readout kinda got me interested. The chip also includes buffers at the wiper out stages, which makes me think it would suit an Inverted amp. So I thought I would give a try at designing one.

So far I have sketched out the schematic on Dip trace. I would appreciate comments.



The caps obviously are wrong in the schematic, I will of course be using bipolar caps in the signal paths, bias to ground etc.

I opted for using dual +/-2.75V suplies for the chip. It will allow single voltage operation, but the datasheet indicates higher levels of THD+N in this configeration. I have also opted to bypass the -V supply to ground.

I have included a resitor between the source and the chip. This is to bring down the value of the peak voltage to within the tolerances of the chip (Vmax 2.75v peak) as my calculations say that the max rms the chip can handle is roughly 1.9vrms.

Maxims datasheet and description

Essentially I see this as being a similay implementation to the way Pedja ran the DS1802, but with a differnt chip.

I have yet to decide the values for the rest of the components, I will aim to have pretty high gain on the lm3875 to counter the lower input voltage.

I am also concerned by the datasheet. It shows the frequency response of the chip tailing sharply at 100khz is this too low?

Your comment please. Try not to be too negative about the choice of chip. I know it's only got 32 wiper positions etc

[justblair]

Silence was the loud reply...

I have been busy putting together a pcb for the above schematic. Here it is

Top



Bottom



I have been reading the datasheets and MAxim have another chip the MAX 5480 which is a push button version of the 5440. My design is such that if I widen the solder pads, either chip should be compatible. The 5480 requiring a trace to be cut.

There is still some tidying up to do.

I have to produce a second board for power supplies as well. The aim is that the boards will stack.

Comments please. I hope to get the boards etched this week. So if I have made a horrible mistake, it would be good to know this weekend.

Thanks in advance

Blair

ps if theere are any diptrace users out there. I am having trouble getting the traces to align with the pads of the chip. I have decreased the gridlines to their lowest setting. What have I missed?

[justblair]

I am going to bed now after a marathon19 hour session in front of the PC.....

I have updated my design to include a low pass filter before the LM3875, I aim to add a slope off at around 20Khz. I wont need to worry about the chip at higher frequencies as the max5440 filters frequencies above 100Khz quite sharply.

Have narrowed the width taken up by the components to allow more heatsink, and have put together the bulk of a schematic for a regulated power supply.

I solved the issue of the uneven tracks on the chip's pads, discovered some issues with ground loops and resolved them and picked, designed (in pattern editor) the sizes for all of my caps. I just have to place 2 of the filter caps. I'll do this tomorrow morning.

I haven't decided yet on all the resitor values, this is where I could do with some help.

I also have some questions...
[list=1][*]Is my idea to place a resistor in series with the input of the pot going to reduce the voltage correctly? Value wise the pot is 40K so a 40K resistor would halve the voltage at the pot?[*]What is a typical RMS voltage from a source. The GF currently is taking sound from a soundblaster live 24bit soundcard. What sort of attenuation will I require pre MAX5440? It will accept a peak voltage of 2.75v.[*]What do you think my chances of success are? [/list=1]

I realise that this is not everyones cup of tea, but if anyone is interested, I would be happy to make up boards at cost/trade of parts etc.

[BWRX]

Hi Blair. Based on what you have planned so far it looks like this will turn out to be a real nice project. Now to address some of your questions.


The frequency response of the chip is perfectly fine for audio use. If you wish to add a low pass filter before the LM3875 choose a higher cutoff frequency. Move the cutoff frequency up to at least 40kHz to minimize its effect on the frequency and phase response in the audible frequency range.

When choosing the gain you should consider the whole picture from source to speakers. Normally I prefer to set the gain so that the output is on the verge of clipping with a maximum source voltage. This prevents me (or anyone else) from ever horribly overdriving the amp. You mentioned that the source will be a sound card. The maximum output voltage for these can vary, so it's best to just measure it yourself. Generate a 1kHz full output test tone, turn up the volume the whole way, and measure it with a true RMS multimeter. That's the max RMS output voltage of the sound card. Generally this will be in the range of 1-2Vrms. Now you know the max input voltage to the LM3875 and can calculate the necessary gain based on the rail voltages you plan on using.

The resistor in series with the pot intput will work as you are thinking it will. You won't need to use the resistor unless the sound card has a very strong output, but it is a good idead to have provisions for it just in case.

One thing I noticed is that you have the midbias pin connected to ground on your schematic. The dual supply schematic on page 6 of the MAX5440 datasheet shows that the midbias pin is not connected to anything.

What resistors values do you need help selecting? There aren't that many you need to choose!

Also, don't forget that you will need a rotary encoder with a grey code output to work with the MAX5440.

Last but not least, I'd say your chance for success is very good

[justblair]

Thanks for the reply... I was begining to feel kinda lonely!

Originally posted by BWRX

The frequency response of the chip is perfectly fine for audio use. If you wish to add a low pass filter before the LM3875 choose a higher cutoff frequency. Move the cutoff frequency up to at least 40kHz to minimize its effect on the frequency and phase response in the audible frequency range.

The values I have gone for were pretty much on the advice at Peter Rogic's site, and orgionated from Joe Rassumen I believe.

I have opted for 1k/4.7nf/9K before the amps (lm3875) neg input.

This should give a drop off of -1.1db at 20Khz according to Pedja (I haven't worked out how the calculation is made) On Joes site he shows similar values in his ideal low pass figure. I dont have enough experiance to say wether this is to my tastes. (I gues that if the thing works I will be happy!)

Originally posted by BWRX

When choosing the gain you should consider the whole picture from source to speakers. Normally I prefer to set the gain so that the output is on the verge of clipping with a maximum source voltage. This prevents me (or anyone else) from ever horribly overdriving the amp. You mentioned that the source will be a sound card. The maximum output voltage for these can vary, so it's best to just measure it yourself. Generate a 1kHz full output test tone, turn up the volume the whole way, and measure it with a true RMS multimeter. That's the max RMS output voltage of the sound card. Generally this will be in the range of 1-2Vrms. Now you know the max input voltage to the LM3875 and can calculate the necessary gain based on the rail voltages you plan on using.

That is good advice. I will order up the parts for 26db gain. If its a little quite the GF wont mind. The 760i's she has (on loan from me) are pretty sensitive, and she does not play music too loud.

I also dont want to amplfy any noise from the max chip too much. I am making several boards anyway, so I can experiment with others.

Originally posted by BWRX

The resistor in series with the pot intput will work as you are thinking it will. You won't need to use the resistor unless the sound card has a very strong output, but it is a good idead to have provisions for it just in case.

Thanks for reassuring me. I was not sure wether to use a resitor or a divider, the maths told me a resistor would be ok. However with unknown values for the source, I am guessing. The datasheet says maximum peak of 2.75v This is 1.9rms. Wherever I read, this is a low figure for a source. My guess is that 20K is around the right value for this resitor, but maths is not my strong point! 40K will have the voltage. Anyone like to give me a clue here?

Originally posted by BWRX
One thing I noticed is that you have the midbias pin connected to ground on your schematic. The dual supply schematic on page 6 of the MAX5440 datasheet shows that the midbias pin is not connected to anything.[/B]

The datasheet I downloaded has the bias attached to system ground (I hope that means signal ground.) The first schematic on the sheet shows it and the attched low outputs floating. This is setting the input voltage at 5.5v and 0v. I am balancing the chip in between voltages. At least I hope I am reading the sheet correctly?... The little triangle means ground in this application?

Originally posted by BWRX
What resistors values do you need help selecting? There aren't that many you need to choose![/B]

First the input resistors. I guess another way of asking is what is a normal peak signal from a source (or the ranges that sources may reach)

Secondly the LED resistors. Now my understanding is that the LED resistors are switched to ground. This poses a problem. I want to use blue LED's. The forward voltage on these is higher than the 2.75v on the logic circuit. I chip shorts these to ground.

Will I be able to just feed them from a seperate (higher Votage supply?) Damn I wish I thought about this before!

Originally posted by BWRX
Also, don't forget that you will need a rotary encoder with a grey code output to work with the MAX5440.

I found one on flea bay. 20positions gray code continuous rotation with a push button. £1.99. The push button I can rig to either mute or balance.

I have finished the PCB I think.





Though it could do with a real going over with a toothcombe. I literally worked through Fri-Sat without sleep, And hit some more problems/mistakes today. There comes a point where I cant see the mistakes anymore.

I have also produced this Scematic for the PSU.



I would appreciate comments on this as I am about to start the PCB design on this one....

My Etchant, boards, developer, case and torroid arrive tomorrow. I want to get the parts ordered this week and start building on Friday (Day's holiday)

ANY help appreciated!!!

edit

A bigger version of the schematic

And I just noticed, i have 24v in for the outputs. It shoudl read 20V. I can only fit a 18-0 18-0 transformer in the case...

[BWRX]

If you used 1k/4.7n/9k you would end up with a cutoff frequency of about 33kHz. I would suggest getting some 3.3n and 2.2n caps just to try. Since the 4.7n only introduces a small droop at 20kHz you probably won't even be able to tell the difference.

26dB (20V/V) is a good choice for the gain. Keep the input resistor value at 9k or higher. If you lower it's value you will increase the droop in the frequency response due to the low pass filter.

If the source is capable of 2Vrms output a 20k resistor in series with the pot will drop the source voltage to 2/3 (about 1.89Vpeak) when the pot is at the max of 40k. If the source is only capable of 1Vrms output then you won't need to use any series resistor.

Look at figure 6 on page 13 of the datasheet. That schematic is for a dual supply implementation. From what I can see, the midbias is not connected to anything. The midbias pin should only needed when using a single supply because you can't connect the low side of each pot to ground like you can when using a dual supply.

The datasheet is not very clear about what the LED are switched to. It appears as though they are switched to ground, unfortunately. What is the voltage drop across the blue LEDs at 10mA? If you apply a lower voltage across them they will still light up but will drop less voltage and not be as bright. A simple way around this is to use the LED open drain to ground outputs to control PNP transistors tied to the Vdd rail. When the LED output is open drain to ground it will pull the base of the PNP transistor to ground (through a resistor) and turn on the transistor. You can then wire the LEDs to the Vss rail and use the higher voltage drop LEDs. You can get 2N3906's for less than 10 cents a piece. Heck you may even have some laying around.

Your PSU schematic looks fine.

[wxn]

SBLive 24bit doesn't fit there. I mean, pre EMU/XFI Creative products are simply not up to hi-fi standards, not to mention upper/hi end.
I'd recommend getting at least a prosumer grade sound card like EMU 1212, ESI Juli@. There are many choices.

[justblair]

Originally posted by BWRX
If you used 1k/4.7n/9k you would end up with a cutoff frequency of about 33kHz. I would suggest getting some 3.3n and 2.2n caps just to try. Since the 4.7n only introduces a small droop at 20kHz you probably won't even be able to tell the difference.

I will order up several values and try them.. Thanks

Originally posted by BWRX
26dB (20V/V) is a good choice for the gain. Keep the input resistor value at 9k or higher. If you lower it's value you will increase the droop in the frequency response due to the low pass filter.

Will do..

Originally posted by BWRX
If the source is capable of 2Vrms output a 20k resistor in series with the pot will drop the source voltage to 2/3 (about 1.89Vpeak) when the pot is at the max of 40k. If the source is only capable of 1Vrms output then you won't need to use any series resistor.

I will order a couple of values here. 5k seems more reasonable by the looks of things

Originally posted by BWRX
Look at figure 6 on page 13 of the datasheet. That schematic is for a dual supply implementation. From what I can see, the midbias is not connected to anything. The midbias pin should only needed when using a single supply because you can't connect the low side of each pot to ground like you can when using a dual supply.

I was clearly looking at the sheet through very tired eyes.... Of course you are right. Thanks for pointing that out. I decided to take a break from the design side for a couple of days after that.. I think you can look at something for too long and become blind. Wood and trees come to mind

Originally posted by BWRX
The datasheet is not very clear about what the LED are switched to. It appears as though they are switched to ground, unfortunately. What is the voltage drop across the blue LEDs at 10mA? If you apply a lower voltage across them they will still light up but will drop less voltage and not be as bright. A simple way around this is to use the LED open drain to ground outputs to control PNP transistors tied to the Vdd rail. When the LED output is open drain to ground it will pull the base of the PNP transistor to ground (through a resistor) and turn on the transistor. You can then wire the LEDs to the Vss rail and use the higher voltage drop LEDs. You can get 2N3906's for less than 10 cents a piece. Heck you may even have some laying around.

The Leds in my box right now are a bit unknown. However I looked up the datasheets for a few blues. Big problem there. Most will pass little to no current at 2.75v. The current they will pass I dont think will light them up. Red or yellow on the other hand pass current at much lower voltages, so will probably be fine. I guess that red might be OK. Blue is cooler... I will see if I have time to rework the circuit.

Another issue that crops up there is the LED indicator from the MAX works in increments. The lowest LED's will light up to full brightness, the final one dims bit by bit before going out. This allows for more increments to be represented. I think I could bias the transistors to give this effect, but I will have to look into it.

Originally posted by BWRX
Your PSU schematic looks fine.

Thats a relief!

Originally posted by wxn
SBLive 24bit doesn't fit there. I mean, pre EMU/XFI Creative products are simply not up to hi-fi standards, not to mention upper/hi end.
I'd recommend getting at least a prosumer grade sound card like EMU 1212, ESI Juli@. There are many choices.

Hi WXN. The soundblaster is not an ideal choice. It will do for now. The Mission speakers that this amp will be used with are not high end. A pair of Metronomes are next on the building list for the GF. After that I will look at a USB DAC. I am currently really enjoying the DDDAC1453MKII in my system. The sound from it is magical. I think I will probably go this route for the GF's system in the end. I may even adapt the design to produce a more compact version.

I have spent the last two nights producing a case for my DDDAC and its power supply. I am happy with the end result. Its based on the same HAmmond case that the amp design above is going to fit in. I really like the idea of producing a series of boards for this case. Amp, DAC, etc Watch this space...

[justblair]

An update.

I forced myself to take a break from this project. The rising panic of a deadline was forcing errors.

However the other day I got my order from Rapid.



I also received a couple of Rotary Encoders from Flea bay as well as some other Bits and Bobs.

Here is the case that I will be using...



Looks pretty big till you put the torroid in...




I really like the hammond 1455 case. It comes with black plastic surrounds to the end pieces. However it is well enough produced to do away with them..

I have spent the last couple of nights adapting the same case to house my DDDAC (Click to jump) . SO when I say a couple of nights off, i dont mean in front of the telly with a glass of red wine.

The GF comes back off a family holiday tomorrow. So no work will be done tomorrow . But I have a day off on Friday. Back to the computer for another serious CAD session...

[banana]

Is it the MAX5440 will be at -12dB position upon every powerup? That seems too loud.

Anyway, I like the idea of graycode encoder control without MCU. If there's a way to access the resistor ladder directly without using the build-in opamp, I'll like to try it too.

Interersting project.
Please show us the LED display looks like