| soundNERD |
Hi, I finished designing some pcb patterns for the LM3886. This amp will be used to power a small 8-inch subwoofer speaker.
Please tell me if there is anything wrong with it.
Thank you!
-Mike |
|
|
| soundNERD |
Anybody?
Please tell me how this would work.
I based it off of the schematic on the first page of the 3886 datasheet
thanks, Mike |
|
|
| peranders |
I have just taken a very short look, but where are the deoupling caps? The trace between the LM3886 pins... not good. The trace width, consider this is a >10 A device! I think the pcb will act as fuses. Otherwise I think it's not so hard to succeed.
You can get some inspiration if you take a look at my LM3886 project. Download the pdf's for the pcb layout. Note also that I'm going to have 95 um copper (2.7 oz). |
|
|
| soundNERD |
The trace between the chips is the ground, so it doesn't matter, right? And the caps won't do anything, since the ps board is very close to the amp board.
I'll look at your project, though.
-Mike |
|
|
| soundNERD |
Actually, I changed my mind. I think I will just put the ps on the amp board. That will eliminate the need for the decoupling caps.
If I don't put the ps on the amp board, do I really need the 100uf decoupling cap? I would use the .1, though. I didn't put it on my pcb due to size. If I could just use the .1 cap, that would be easy to add.
Also, about the gnd. gnd carries no voltage, right? So does it really have to be thick?
If you didn't notice on the pcb, the pads on the top right are for input (this amp is for a sub, so connection to the input/crossover/volume board), the bottom center is for speaker, and the top left is power in. The pad connecting to the speaker pin is a jumper over the + voltage wires
I couldn't find your pdf file, but I only looked quickly.
Also, how much do traces limit power? The original amp I designed with the 3886 used all thin traces, not much wider than the pins themself. I designed that one manually using rub-on traces from radioshack.
Will using the thicker traces improve sound quality also?
Thanks, mike |
|
|
| crown300 |
Your PCB layout is a great first effort and you are applauded for your interest and work.
however............
IMO, the design will most likely oscillate at a very high frequency, maybe not at idle, but during power peaks. This will show itself as a lot of heat when running at a 'normal' level. I would not want to use my best speaker during your testing.
Please, please take the time to read the earlier GC threads. There is a lot of time saving information there.
Ground:
Note on page 1 of the LM3886 datasheet that the 8 ohm 'load' is connected to Ground. (please note that this is Power Supply ground - not earth ground.) On page three (near the bottom) there is a symbol "I out" which is "Output Current Limit". note the value is 7 Amps. This current has to flow through your speaker and "return" to the power supply (during audio peaks). Also the LM3886 needs to sense this current, hence it needs to be between the power supply and the load. This current path will cary high peak Amps and a small trace will have a (relatively speaking) resistance causing voltage drop. (Mr Ohm at work!).
Part deaux: if this 'ground' (with its varing output load voltage) has connected to it some of the input parts, they will see a voltage change where there is no change to the actual input signal. Instant oscillation. Meltdown.
So a GC PCB layout needs to account for this.
PS, IMO, most "automatic" PCB layout programs do not make power traces wide enough. |
|
|
| peranders |
| quote: | Originally posted by soundNERD
I couldn't find your pdf file, but I only looked quickly. |
You just have to click at "downloads" in the left frame or you can click here |
|
|
| soundNERD |
I looked at peranders pcb. I don't need anything that fancy, all I need is a small amp to power a small 8-inch subwoofer.
So, I would like to use the PCB I designed, but just modify it to not oscillate.
Lets say that I thickened up the PCB traces somehow (or left them as is, and covered them with a layer of solder, or solder a bare wire across all of them). Would that be enough for the speaker? If I made them as wide as the + and - supply wires, would that be enough?
What would be the cause of the HF oscillation? Is it the gnd trace? Or something else. If it isn't the ground, how could I fix it?
Thank you for the help, Mike |
|
|
| peranders |
| quote: | Originally posted by soundNERD
Actually, I changed my mind. I think I will just put the ps on the amp board. That will eliminate the need for the decoupling caps. |
One of the most importantthings are good ground and good power supply decoupling.
I think you should take a look at AN-1192 and the recommended design. Change then the design if you really know what you are doing.
If you remove the "fancy stuff" from my design see what's left and where it's located. Check also Thomas Madsen's LM3886 "Das Modul".
Consider AN-1192 as "the law" at the moment.
The casue of oscilllations is often not sufficiently low impedance (at the MHz region) between supply voltage and ground sensed at the LM3886.
BTW: I will order my LM3886 monster pcb's today. :) EDIT: Done! |
|
|
| skyraider |
quote:
'The casue of oscilllations is often not sufficiently low impedance (at the MHz region) between supply voltage and ground sensed at the LM3886.'
can you explain further? |
|
|
| soundNERD |
OK, I revised the PCB. Heres what I got.
There is a jumper from the gnd to pin 7 and from pin 7 to the resistor to its right.
Will this work fine?
Also, should i connect pins 2, 6 and 11 to gnd also? they don't connect to anything?
Thanks, Mike |
|
|
| soundNERD |
anybody?
I know i havent added the decoupling caps, but i am going 2 |
|
|
| soundNERD |
| Can somebody just tell me if the NC pins should be connected to GND or not? |
|
|
| bigparsnip |
the nc pins shouldn't be conected to ground or anything else if you can avoid it (just to be on the safe side), however, if you want to fix the chip onto the PCB more solidly, you can just have unconected pads for each of these pins, and solder the legs as normal (as they still won't be conected to anything, kind of like it looks like you have done.
Also, where will you be running the wire from the gnd to the resitor through? will it still be going between the chips legs, as if it is a bare wire it may touch one of them and cause a short or someting.
edit:
Just a thought, what design software are you using?
Also, could you possibly post a scheematic, if you have it, so that it is a bit easier for us to see where things should be going in your design. |
|
|
| peranders |
| You should always read datasheet if you have "NC" pins. Sometimes you can use them as guard (some opamps), sometimes the are not used normally but it can also mean "do not connect" meaning DO NOT connect. They can be connected to something internally for production test maybe. |
|
|
| soundNERD |
The schematic i used is from the first page of the datasheet. I am at a library computer now, so I don't have it.
I use the software called PCB Editor.
I will use insulated wires for jumpers.
I didn't think the NC pins had to be connected, but just wanted to make sure. |
|
|
| bigparsnip |
| Hi, I've just tried to have a real look at this and see where the components go and where the external connectors are, and I have realy confused my self, do you have a component overlay to go with this as (assuming it uses all of the smae components as on the national data sheet) it would prove very useful here I think. |
|
|
| bigparsnip |
Right, I have had a quick bash from scratch using the national reference design from the scheematic, and come up with the board below, which should also allow for reasonalbe sized local power supply decoupling caps to be fitted, as well as having mute circuitry (simply wire in the switch between pads 3 and 4) input conector (signal and ground, on pads 1 and 2) output pads (signal and ground, pads 5 and 6) and the power supply conectors (pads, 7 8 and 9).
If you want a copy of the eagle board files, or the schematic just let me now and I can post them up here for you.
Andrew. |
|
|
| bigparsnip |
Here is the board without the component overlay, to give you a clearer look at what is going on below.
edit: the red wire is a jumper link to simplfy the board layout a bit, but you could get away with un-insulated wire, as it will be no where near coming into contact with any other bare metal. |
|
|
| bbksv |
| Hey bigP...would you mind sending me that eagle file to bbksv@comcast.net........pretty please...:) |
|
|
| bigparsnip |
Hi, I have zipped the schematic and board files allong with the library file for the LM3886 chip and attached it below. You may find that you will have to add the library to your list in eagle before you can view the files properly, but I doubt that that will be too much of a problem (let me know if you get stuck, and I might be able to help out).
Otherwise, enjoy! |
|
|
| soundNERD |
I don't have fancy expensive software capable of displaying those files :(
Is there anyway you could post the Black and white copper bottom to scale or email it to me?
Thanks for the time you spent doing this for me!!
-Mike |
|
|
| bigparsnip |
Hi, the version of eagle that I used to do this is FREE! You can get a coppy from thie website below if you wnat one, but if you can't do that, I guess I could come up with some black and white gif's for you (just let me know if you need them).
http://www.cadsoftusa.com/ |
|
|
| bigparsnip |
Here is the board track layout that you can use straight off, without having the software (please remebmer, that this is a top down view, so you may need to mirro it depending on how you intend on transfering it tot he bottom layer of your board).
edit:
I forgot to mention that this should be printed out at 300 DPI without any resampling of the image ofr it to come out the correct size. |
|
|
| soundNERD |
WOW, thats amazing for a free program.
Is it easy to use? I am going to download it, but its a 7MB file, so I'll have to wait until later. I can only spend about 20 more mintutes on the internet.
I think I can print at 300x300dpi, but if not, by any chance would you know how much to scale it down?
Thanks again for designing this! |
|
|
| bigparsnip |
Eagle does have it's limits (at least in the free version, of board size, and the number of layers), but it is by far the easiest program I have used for trying to design PCB's (it makes trying to learn protel even worse). Just draw a schematic, press the CAD buttoin and it will create a set of components with all of the nets conected how you want them (providing you have teh pars int eh librbary). All you have to do is place them where you want, and join up the pins (there is an auto router, but I would advise against using this for the most part).
If you do need to try and scale the picture, just print it out on a piece of paper first, and try and match up the pin hole son your 3886 with the ones in the picture (when these match, you know you have it the right size). |
|
|
| soundNERD |
Yes, when I printed, it was way too big.
I then tried scaling it to 300 pixel width, and fireworks automatically adjusted the height.
Still too big. I guess this is a trial and error thing now, right?
Thanks :)
-Mike |
|
|
| bigparsnip |
the picture itself isn't supposed to be 300 pixels wide, but rather it is the corect size that when set to a resolution of 300 pixels (dot) per inch, it will be the corect size to print out.
So, if you can get fireworks to leave the picture with the same number of pixels (try selecing it so that it doesn't resample the image), and then change the resolution if the image, or i't s physical width (depending on what you have control of, I know you can do both in photoshop, as that's what I used to create it), so it is either 300 pixels per inch, or measures 2.573 inches wide by 1.487 inches high.
edit:
I have done the resizing for you and saved the image file as a tif, so it will remember the correct resolution and size of the image, so if you open this, and print it out without any resizing, it should be the correct size. |
|
|
| ukram |
| Pin 7, the signal ground, seems incorrectly connected to power ground, creating a ground loop. It should have its own wire to signal ground star point. |
|
|
| soundNERD |
| How else should it be connected? |
|
|
| theChris |
| odd Q, why not just make the ground traces as thick as possible? |
|
|
| bigparsnip |
| As far as the grounding goes, all of the signal grounds are conected together before going back to the main grounding point. However, as the current draw on the signal lines will at most be a few mA there shouldn't be any problems here (you would normaly only have this sort of problem if signal and power grounds share the same path I think). |
|
|
| ukram |
| I forgot that this is LM3886 with have separate pins for GND and vin-. Without schematic i'm not sure if there is a problem or not. I was worried that capasitor ripple current might flow in signal ground track. In my LM3886 amp i connected GND to signal ground. |
|
|
| saltnpeppah |
Hi.. This is my first post here in DIYAudio... if my questions seem dumb, thats because i am just overenthusiastic rather than stupid... please bear with me...
Anyway, i saw Parsnips pcb layout for the schematic on nat. semi's datasheet; and tried to come up with a design solution before i fabricated the boards...
please have a look at the component values on the below and tell me if they make a workable design.. if not, please suggest a suitable value...
Vcc/Vee : 35V
Rl : 8 Ohms
Rf : 22K Ohms
Ri : 1K Ohms
Ci : 10u Farads
Rb : 4.7K ohms
Rin : 47K ohms
Cin : 10u Farads
Thanks in advance...
Madhu. |
|
|
| bigparsnip |
| Which resistors in the schematic / layout have you listed there, as I think they are listed R1, R2 etc on the board. |
|
|
| saltnpeppah |
the schematic i am talking about is the one on the first page of the national semiconductors datasheet for the LM3886..
Andrew, i have downloaded the eagle schematics that you developed for soundnerd...
Madhu. |
|
|
| bigparsnip |
Right, I have had a look, and I think there are a few things you should note:
Firstly RL is the loudspeaker, so you don't need to include this in the circuit.
Secondly, you DO have to include Rm as this controls the mute function of the chip. This should be set to allow 0.5mA to flow from teh pin which I think is at ground, so for your rails (+/-35V) I think it should be about 70K, but I would check this yourself as I am probably wrong here.
Third, there is no provision for Cin, or Rin on the PCB, as Rin is the volume pot (which would be conected up to the corect pins in the PCB insted), and I didn't bother with in input cap as Soundnerd didn't want one (I think).
Fourth, I wouldn't bother changing the value of Rb, just keep it at 1K.
But appart from those few things, all of the rest of the components seem about right (although, I would be tempted to use biger caps for Cf and Cm as you can easily fit them on the baord, and the non-polar electrolytics aren't too expensive). |
|
|
| saltnpeppah |
| quote: | Originally posted by bigparsnip
Secondly, you DO have to include Rm as this controls the mute function of the chip. This should be set to allow 0.5mA to flow from teh pin which I think is at ground, so for your rails (+/-35V) I think it should be about 70K, but I would check this yourself as I am probably wrong here.
|
Rm on the spreadsheet is set to a default of 35K... would that be OK??? and what is Cf?? it is not in the schematic.... by the way parsnip, would you mind posting the component values you hav used for soundnerd's pcb??
Madhu. |
|
|
| bigparsnip |
Hi, sorry, that should be Ci, like you have listed above as being 10uF. As far as what components sound nerd used, I have no idea, you will have to hope he gives you a reply to your questions, as I have never tried to build up this PCB, only the LM3875 one for my little amp.
Oh, and 35k, would probably be fine, I doub't it has to be too close in this case, but perhaps someone with a bit more experience could give you a better idea about that. |
|
|
| tcpip |
Very interesting discussion. I saw it just now. And hi, Madhu. I too am from India. :)
I've been trying to finish working on a non-inv amp using the LM3875. Other than the muting function, the schematic should be identical to what I'd use for the LM3886. (The board layout will of course be different, because the pinouts of the two chips vary.)
I too started with the first schematic on the LM3875 datasheet, and then looked at some refinements in the single-supply version of the schematic in the same datasheet, and also looked around on the Net a bit, and played around with the Nat Semi spreadsheet which they provide to let you design non-inv amps, see the power dissipation, etc. I also ran the schematic through a friend much more experienced than I. I've attached it here. I just thought that since there are at least two of you on this thread who appear to be beginners like me, we could work on the schematic together. We can then do the PCB design in Eagle (I too use the Eagle freeware, Bigparsnip, but on Linux. I like it a lot.)
The schematic I've shown here is exactly what I'll put on my PCB, once I get an okay from you guys that it'll work. The only parts external to the PCB will be the transformer and bridge rectifier. (I was thinking of using mass-market parts to build it, and not opt for the hi-tech diodes that some others seem to prefer for Gainclone bridge rects.) I want an input cap, and I omitted the RL circuit at the speaker output. Many others seem to have omitted it too, in their gainclones. From what I've understood, C8+R5 form one first-order high-pass filter and affect the lower frequency limit, and R2+C2 together act as a second first-order filter. With my values, C8+R5 will give me 0.16/(1uF * 56K), which gives me 2.86Hz. And R2+C2 gives me 0.16(10K * 4.7uF), which is 3.4Hz. So the -3dB point, as far as I've understood, will be 3.4Hz for my schematic. C1+R4 reduce the gain of the amp at high frequencies, reducing the possibility of high frequency oscillations. All this is what I've acquired in bits and pieces by pestering others (something I'm good at :D ) --- my own understanding of these things is minimal.
As my schematic shows, there's no connection between signal ground and power ground in this schematic, and there will not be any in the PCB. I intend to take leads from both terminals to a star ground off-board somewhere. And the return path from the speaker will also go directly to that star-ground, not to this PCB. In fact, I think I'll take the lead from the power-ground (marked GND) of this PCB back to the centre tap of the power transformer, and take a separate lead from that centre tap to the star-ground. That way, the heavy charge/discharge currents between the smoothing caps (C4 and C5 in my schematic) will not flow through the star ground.
Any comments? Please? |
|
|
| tcpip |
I also have another schematic for a finished non-inv amp for the LM3886, from Elektor. In that schematic, they've handled the mute pin too, so one can simply pick up the resistor values based on what they've put in it. (They've put in the option for a mute switch, which I don't intend to use.... I'll just keep the circuit permanently "un-muted" when I design the PCB.)
I have this article in PDF format, in case you want me to mail it to you. It's a one-page thing with schematic and PCB layout. |
|
|
| theChris |
the circuit as shown has a gain of about 3. it needs to be boosted by a bit.
why 2 42k resistors in parallel. also the input capacitor may now be needed. |
|
|
| tcpip |
| quote: | Originally posted by theChris
the circuit as shown has a gain of about 3. it needs to be boosted by a bit. |
You're right! I'll have to make R2 into 1K, and probably make C2 to 47uF to compensate and keep the same roll-off frequency.
| quote: | | why 2 42k resistors in parallel. also the input capacitor may now be needed. |
The two resistors in parallel are because some people on the chip amps forum say that a higher power rating for the feedback resistor seems to improve the sound a bit. I thought this would be a good way to reach the 0.5W rating using two 0.25W MFR.
And can you please explain the comment about the input cap? You think it may now be needed? I already have one. |
|
|
| roadkill |
Hi Tarun
I don't get how increasing the power rating of the f/b resistors can improve the sound. Anyway, electrically it should be the same. Some may say that heating effects in smaller resistors can raise the temperature enough to change the resistor value, leading to distortion. Could be, but IMO, the effect won't be audible except to the golden eared. Esp. if you're using metal film, which has a good tempco.
Please verify (using SPICE and an ordinary opamp model in place of the 3875) the effect of the 50pF cap in the feedback loop. Or else some simple math will tell you the 3dB point of the ckt (no, I'm recovering from flu, and you can't tempt me even with an Uncle Harry knob to do the math for you! :) ). If there's a rolloff of more than 1dB in the audio band, it's gonna be a lot more apparent than the effect of resistor heating.
On splitting the grounds apart, you could do the following: drill many holes on the PCB along the edge where the ground planes are split. You can solder short bits of wire across these, which you can later clip off if the hum gets annoying. You can also use a 50 - 100 ohm resistor here, because in case the input connection to the amp is removed (remember Murphy's law: sh*t happens!), you can say goodbye to whatever load is connected. The resistor will provide an "emergency" ground connection. In normal use (when there is an input), the ground return path will have a lower impedance than the resistor, so it shouldn't matter.
Madhu, welcome to DIYAudio. Using a supply of 35 volts with this chip is a bit risky. I once blew out a pair of 3886s when I used a 30-0-30 trafo. Of course, no-load DC is 42V here, but under load it should be lower than the chip's rating. Well, I guess turn-on transients did the 3886s in. Using a 25-0-25 (to get a nominal of 32V) is probably a safer idea. Cin of 10uF means using an electrolytic (err... BAD!) or a Z5U ceramic (also bad). You probably wont need LF response to go down to 0.72 Hz anyway. Use a 1uF film capacitor (get them at Cee Pee electronics. Ask for "box type" capacitors, he'll know). Perhaps you can use two of them in parallel. |
|
|
| roadkill |
Looks like Tarun already did the math... Maybe I should read a bit more before I post? :smash:
One more thing: you mentioned high tech diodes. These are probably Schottky or fast recovery types. I can't get why they're used in an audio amp supply. They're meant for switchers. In a linear supply, their fast turnoff will inject harmonics of 50 Hz into the power lines, which if not adequately filtered, can cause problems. Look at the PSU to any low-noise measurement equipment from a quality firm, for instance an HP spectrum analyzer. The psu diodes will have capacitors across them (these must be able to withstand the AC voltage being applied) to slow down the diode switching. Thus, turn-off transients (which can sometimes go up to several megahertz!) are avoided at the source. The downside to all this is power loss in the diodes... small price to pay IMO.
Aside: ultra-fast switching diodes (backward diodes, some schottkys, etc) are sometimes used as "harmonic generators" to multiply the frequency of an RF signal. Introduce switching harmonics, then filter out the one you need. |
|
|
| roadkill |
| Here is my gainclone amp schematic. Note J1, a wire link between grounds. There will be an inductor wound around R7, which is not showin in the schematic or board. |
|
|
| roadkill |
| And here is my LM3886 PCB. Of special interest (maybe?) is the board size. Note that this PNG was exported at 150 dpi from Eagle. |
|
|
| Aud_Mot |
I am a little confused.
Output connection X2-1 and X2-2. If the load (speaker) goes between these 2 points, you will not get any potential. Your speaker will see the equivelant of a 12.7 ohm resistor in series with a 100nF cap accorss it's terminals. The amp output just flapping in the wind, with no path to ground.
Maybe there is supposed to be a ground/common connection between the C8 and terminal point X2-1?
A quick look at your PCB seems to refelct what I think is an error in the schematic.
Aud_Mot |
|
|
| theChris |
roadkill, you left off the inductor on the output of the amp. as it is you have just the 10ohm resistor
\ |
|
|
| tcpip |
| quote: | Originally posted by roadkill
I don't get how increasing the power rating of the f/b resistors can improve the sound. Anyway, electrically it should be the same. Some may say that heating effects in smaller resistors can raise the temperature enough to change the resistor value, leading to distortion. Could be, but IMO, the effect won't be audible except to the golden eared. |
See the schematic here, taken from a common Gainclone reference source, the Decibel Dungeon. The author uses two feedback resistors in parallel. I agree with you 100% that electrically, a higher-wattage resistor will not have an audible effect, but I decided to make provisions for it on the PCB (hence the schematic) anyway, in case someone sometime wants to use my PCB to experiment with high-wattage. And if ears could be categorised, I'm probably steel-eared rather than golden. :D
| quote: | | Please verify (using SPICE and an ordinary opamp model in place of the 3875) the effect of the 50pF cap in the feedback loop. |
I haven't done the math myself, but I've picked up those values from the LM3875 datasheet. I also looked at the LM3876 amp published by Elektor just now. They have 1K resistor to ground, 18K feedback resistor, and another 18K+47pF in parallel to this feedback resistor. So I think I'm in the safe league here.
| quote: | | On splitting the grounds apart, you could do the following: drill many holes on the PCB along the edge where the ground planes are split. You can solder short bits of wire across these, which you can later clip off if the hum gets annoying. |
This is interesting. May try it.
| quote: | | You can also use a 50 - 100 ohm resistor here, because in case the input connection to the amp is removed (remember Murphy's law: sh*t happens!), you can say goodbye to whatever load is connected. The resistor will provide an "emergency" ground connection. In normal use (when there is an input), the ground return path will have a lower impedance than the resistor, so it shouldn't matter. |
Couldn't quite understand what you have described here. (Haven't had my morning coffee yet, otherwise I'm usually a brilliant guy.) Can you please explain exactly between which two points I should connect the 100R?
| quote: | | Cin of 10uF means using an electrolytic (err... BAD!) or a Z5U ceramic (also bad). |
Or you can use metallised polyester or something. I have axial-lead metallised polyester, 160V, of that value, that I picked up from L.Road. Quite expensive, but available. |
|
|
| tcpip |
| quote: | Originally posted by roadkill
One more thing: you mentioned high tech diodes. These are probably Schottky or fast recovery types. I can't get why they're used in an audio amp supply. They're meant for switchers. In a linear supply, their fast turnoff will inject harmonics of 50 Hz into the power lines, which if not adequately filtered, can cause problems. |
I have very limited understanding of the subject, but I have heard someone else exactly echo your opinions. I once asked Randy Slone about this question, and he said that he can't figure out why people are looking for faster and faster diodes when he uses snubber caps to slow them down.
| quote: | | The psu diodes will have capacitors across them ... to slow down the diode switching. |
Yes, snubber caps are a common refinement. I'm using them in the (fairly low-tech) regulated symmetric PSU I've built for my preamp. In fact, in high-quality rectifiers designed by Randy Slone, he uses separate diodes instead of bridges, and puts a small-value resistor in series with each diode, and puts the snubber cap across the diode+resistor combo. He doesn't have anything against integrated bridges, but if he wants to put resistors, he can't afford to use bridges.
There are plenty of people in the DIY audio crowd who keep logic aside and just take decisions based on what sounds better. Very often, they don't attempt to explain even to themselves why something sounds better. If such people prefer high-speed diodes, you can't get much analysis out of them. On the other hand, I've also seen some very interesting analyses (with scopes, meters, and semiconductor physics) on the effect of PSU noise generated by high-speed diodes. See this thread. I'm not competent to understand everything discussed there, so I can't comment. |
|
|
| tcpip |
Roadkill,
Thought I'd actually list out all the differences between our schematics, so that we can see which if any are significant:- I have an input cap, you don't.
- Your input impedance is 22K, mine is 56K. Yours gives you slightly less Johnson noise, I get the advantage of using a smaller input cap if I want an input cap at all.
- Your time constant of R4+c7 is at a frequency about double of mine, because I use 47uF and you use 22uF. I personally like the NatSemi datasheet recommendation about the upper and lower cutoffs: they recommend that the lower one should be one-fifth of 20Hz and the upper should be five times 20KHz. Your time constant places you a little closer to 20Hz than what I'd like.
- You have an extra RC combo for the mute pin. I will need to add that when I shift from 3875 to 3886. BTW, someone had recommended something like 70K for the mute resistor... are you sure that value can be changed as widely as 70K to 22K? How does one determine that one's value?
- Your PSU filtering caps are one-fourth the value of mine. Are you sure such small values are preferable? I have no clue other than what the NatSemi datasheet and other diy builders are using. They seem to talk of at least 1000uF per rail.
- You've put the so-called Boucherot network (that R6+C8 combo) and the RL filter at the output of your amp, and I haven't. I'm still undecided whether I should. I'm told by a few experienced constructors that I should put them only if I see/hear any problems in their absence.
What do you think, about any or all these diffs? :) |
|
|
| theChris |
| tcpip. i actually meant "not" instead of "now". basically that cap on the input is to remove DC from previous stages. if you have a small DC offset from the source/signal processing, then there is no need to block it out. the cap in the feedback, along with the bias resistor, are designed to reduce DC offset from the IC. |
|
|
| bigparsnip |
| quote: | Originally posted by Aud_Mot
I am a little confused.
Output connection X2-1 and X2-2. If the load (speaker) goes between these 2 points, you will not get any potential. Your speaker will see the equivelant of a 12.7 ohm resistor in series with a 100nF cap accorss it's terminals. The amp output just flapping in the wind, with no path to ground.
Maybe there is supposed to be a ground/common connection between the C8 and terminal point X2-1?
A quick look at your PCB seems to refelct what I think is an error in the schematic.
Aud_Mot |
I would have to agree here, you appear to have your speaker conected accross your output ressitor and zobel network rather than from your output resistor to ground.
This may simply be because you have forgotten to ti the bottom of the zobel to ground in your schematic, but as it is the circuit shouldn't work very well at all and will need to have this change made before you can use it. |
|
|
| saltnpeppah |
Hi all,
Came back from a very refreshing weekly off and am charged up....
first of all, good to see that the discussion has improved...
now for my questions...
1. are you guys using unregulated power supplies or smps??
2. how do you post images on this forum??
3. and would using a toroidal transformer instead of a E-I core improve the sound quality?? ( why i am asking is because the cost factor is almost 10 times)
Madhu. |
|
|
| roadkill |
In reverse order now... :D
Madhu, - Unregulated supplies for me.
- ****! Then use the "Attach file" option below the reply composer.
- Toroid cores have the benefit of reducing fringe fields (the fields are kept within the core much better than EI), so there's less chance of a ground loop or signal loop picking up the field. Another benefit: you can use a slimmer case.
bigparsnip and Aud_Mot,
No, this is not a mistake in the circuit, it's a mistake in my post. I should have explained more clearly (and not type these replies at 2 in the morning!). I return the speaker back to the power supply star ground (midpoint of the capacitors). I also seperately return the zobel network's ground end to the PSU star point through X2-1.
Tarun,| quote: | - I have an input cap, you don't.
- Your input impedance is 22K, mine is 56K. Yours gives you slightly less Johnson noise, I get the advantage of using a smaller input cap if I want an input cap at all.
- Your time constant of R4+c7 is at a frequency about double of mine, because I use 47uF and you use 22uF. I personally like the NatSemi datasheet recommendation about the upper and lower cutoffs: they recommend that the lower one should be one-fifth of 20Hz and the upper should be five times 20KHz. Your time constant places you a little closer to 20Hz than what I'd like.
- You have an extra RC combo for the mute pin. I will need to add that when I shift from 3875 to 3886. BTW, someone had recommended something like 70K for the mute resistor... are you sure that value can be changed as widely as 70K to 22K? How does one determine that one's value?
- Your PSU filtering caps are one-fourth the value of mine. Are you sure such small values are preferable? I have no clue other than what the NatSemi datasheet and other diy builders are using. They seem to talk of at least 1000uF per rail.
- You've put the so-called Boucherot network (that R6+C8 combo) and the RL filter at the output of your amp, and I haven't. I'm still undecided whether I should. I'm told by a few experienced constructors that I should put them only if I see/hear any problems in their absence.
|
- My input cap is at the preamp board's output (1+1uF plastic film). Once again, sorry this wasn't mentioned before. You should use an input cap, otherwise the LM3886 bias currents will flow in/out of your preamp.
- This is the problem with dealing with RF designs by day and audio by night. I prefer lower impedances for the lower noise. Not that I can hear it, though!
- The lower cutoff frequency is quite high since most of the bass will end up going into a sub. The amplifier rolling off the lower end will help protect any vented speakers connected to the amp if the crossover (main to sub) is switched off.
- The mute pin stuff was straight out of the LM3886 datasheet. From what I can see, as the cap charges up, it mutes the amp for a while, removing any possible thump which the preamp may have generated.
- No no no!!! These are not PSU filtering caps, these are merely local decoupling to offset the effect of a long cable delivering power to the amp from the main PSU. The PSU itself has a C-R-C structure (4700uF, 0.1 ohm, 4700uF) per channel.
- As for the Zobel, I prefer the opposite. Include it, remove if there are any problems :D In my limited understanding of this subject, I think they will act like a minimum load for the amp, in addition to compensating for the inductive nature of the speaker cable and voice coil.
- The badly written part on the 50-100 ohm resistor should go something like this: if the input connector to the PA falls off, then the amplifier has no ground reference against which it compares the input voltage. Thus, the output can swing to one of the rails, welding tight any speaker connected to the output. If you connect the grounds together, there *may* be hum (depending on your layout and wiring), so the compromise is to use a resistor between the two grounds (input and power). This way, there is a ground reference even if the input connector is removed. When there is a connector in place, though, the ground path through the connector, to the preamp and back to the PSU star point is much lower in resistance compared to the resistor between the planes on your amp board. Thus, no "ground loop" is formed.
Some random ramblings on bridge rectifier switching noise: in a radar system, the master oscillator's power supply must be absolutely free of noise. Typically, they are powered from 400Hz gensets, not 50 Hz mains. If 400 Hz ripple gets through, then this becomes phase noise on the oscillator's output. Phase noise 400 Hz away from the carrier in a typical radar oscillator is around -130 to -140 dBc/Hz, so the rejection had better be good. These PSUs usually use large geometry diodes so that their junction capacitance itself slows down the switching action, and use extra 0.1uF and 0.01uF caps across each diode. I've not seen a resistor used in series with the diodes yet. Of course, MASSIVE filter caps follow (47000uF + :bigeyes: ) along with a (usually discrete) linear regulator. Audio stuff probably won't need this, though. |
|
|
| roadkill |
| Here's a pic of the amp board. That's a 1 rupee coin, for size comparison. The black connector at the bottom is the input connector, the green screw terminals are for speaker and supply voltage. Note the inductor (rather small :) ) wound around R7. I've used the isolated version of the LM3886 here, since good sil-pads of the right size for a 3886 are hard to come by here. |
|
|
| tcpip |
| quote: | Originally posted by theChris
i actually meant "not" instead of "now". basically that cap on the input is to remove DC from previous stages. if you have a small DC offset from the source/signal processing, then there is no need to block it out. the cap in the feedback, along with the bias resistor, are designed to reduce DC offset from the IC. |
Oh, okay. I understand clearly now. Thanks for the patience. :)
But I think a lot of people are fanatical about keeping DC offset at the output to single-digit millivolts. I'm not debating whether that fanatical approach is necessary... I was just wondering whether your approach, of cutting DC gain, will be adequate to protect speakers? Just curious. |
|
|
| theChris |
to protect? well the cases where DC damages speakers is in cases where something has gone horribly wrong. so it's unlikely that anything not specificially designed to protect the speaker will help. this means if you want to ensure the speaker is protected from DC, all you can do is use a fuse, large capacitor, or some other form or protection circuit.
basically, if you look at what causes the DC offset problem, you'll notice that there are 2 places for the DC offset to originate: from the source, and from the amplifer. a large value cap on the input will correct the source issue, and a cap in the feedback path should help the DC offset of the amplifer. without it, any DC offset gets multiplied by the gain of the amplifier. |
|
|
| tcpip |
| quote: | Originally posted by roadkill
Tarun, |
Thanks for all your patient answers. I now know why you have higher frequency for your low roll-off... you have a separate sub. Kewl!| quote: | - The badly written part on the 50-100 ohm resistor should go something like this: if the input connector to the PA falls off, then the amplifier has no ground reference against which it compares the input voltage. Thus, the output can swing to one of the rails, welding tight any speaker connected to the output. If you connect the grounds together, there *may* be hum (depending on your layout and wiring), so the compromise is to use a resistor between the two grounds (input and power). This way, there is a ground reference even if the input connector is removed. When there is a connector in place, though, the ground path through the connector, to the preamp and back to the PSU star point is much lower in resistance compared to the resistor between the planes on your amp board. Thus, no "ground loop" is formed.
|
Understood. What your low-value resistor does is basically act as a ground-lift resistor, used in many professional power amps. (Randy Slone uses values about one-tenth of what you've suggested, but the principle is the same.) But in my case, I am thinking of connecting the IN-GND to a common star ground point using a separate wire anyway. That way, even when there's no preamp connected at the input, the IN-GND will have a ground reference. Should be okay, don't you think? |
|
|
| tcpip |
| quote: | Originally posted by theChris
basically, if you look at what causes the DC offset problem, you'll notice that there are 2 places for the DC offset to originate: from the source, and from the amplifer. a large value cap on the input will correct the source issue, and a cap in the feedback path should help the DC offset of the amplifer. without it, any DC offset gets multiplied by the gain of the amplifier. |
Got it. I had never thought of these two as alternatives till you pointed them out. I'll think some more, and when I've thunk, I'll pester my other friends and you some more, maybe. :)
Going by your reasoning, all well-designed amps (i.e. those which, left to themselves have very small output DC offset) can omit their input caps. The only time that input cap will be useful will be when I accidentally pump in a huge DC offset at the input itself, which can then travel through the power amp with a gain of 1 and still hit the speaker. But such situations will only happen due to very unusual freak accidents, right, and one can't really protect well against such accidents.... I mean, I wouldn't know how to get a high DC voltage into my amp's input even if I deliberately wanted to, other than connect some dry cells to it.
Interesting. Thanks a lot. :) |
|
|
| tcpip |
| quote: | Originally posted by theChris
basically, if you look at what causes the DC offset problem, you'll notice that there are 2 places for the DC offset to originate: from the source, and from the amplifer. a large value cap on the input will correct the source issue, and a cap in the feedback path should help the DC offset of the amplifer. without it, any DC offset gets multiplied by the gain of the amplifier. [/B] |
Sudden insight: does this mean that I can omit the (usually expensive) input cap in all non-inv amps? With these amps, I now have a means of bringing down DC gain to 0 (or is it 1?) without putting any cap in the signal path. Does this mean that I can eliminate the input cap completely in all such amps? If yes, then it's one big reason to opt for non-inv configuration, and save money and improve sonics (they say sonically the best cap is no cap at all).
For protecting the input transistors against accidental high voltages, I can always use clamping diodes between IN and IN-GND, to limit input voltage. I don't need caps for that.
What do you think, theChris? And if I'm right, then why does anyone ever use input caps on non-inv amps? (Or they actually don't, and I'm slow on the uptake? :D ) |
|
|
| tcpip |
| quote: | Originally posted by saltnpeppah
1. are you guys using unregulated power supplies or smps?? |
Class B power amps (this includes all chip amps I've heard of, other than the recent spate of digital amp chips) usually do not behave well with regulated power supplies of any kind, unless you have a regulated PSU which is rated for at least 10 times the current rating of the peak current drawn by the amp. This is because once the regulation circuitry begins to respond to the ebb and flow of current drawn by the power amp, it introduces one more element of non-linearity into the amp itself. This may actually degrade the stability of the amp in serious cases.
| quote: | | 3. and would using a toroidal transformer instead of a E-I core improve the sound quality?? ( why i am asking is because the cost factor is almost 10 times) |
There is no reason to use a toroidal transformer over an E+I in audio, other than (i) space constraints and (ii) minimising the electromagnetic field around the transformer. The former may or may not be an issue with you, and the latter can be addressed either with more space around the transfy, or by metal enclosures around it.
I have decided to opt for inexpensive 15-0-15V 500mA toroidals in my preamps just to keep the EMI around low-voltage high-impedance circuits as low as possible. These transfies are Rs.280 each, all inclusive, delivered from B'lore to Bombay. (An equivalent E+I in Bombay is Rs.50 each.). For power amps, I refuse to use toroids... they're just not worth the money. Constructors from other countries often use them, to be able to keep everything in small chassis, but their budgets are often higher than ours, and I wouldn't pay that kind of premium for cuteness and a slightly smaller size. Moreover, I often prefer to keep the (large) E+I power transformer in a separate metal box on the floor, so that EMI issues are pretty much eliminated anyway.... I'm sure my approach gives me lower EMI than the best toroid housed in the amp chassis. The cost of a 8"x6"x6" sheet steel powder-coated box in L.Road is Rs.80. Add some high-current connectors to wire it to the power amp chassis, and it still remains much lower than the price difference between a power toroid and E+I.
And if you still feel like opting for toroidal, I think the price differential will be much less than 10 times, if you source both transfies from the same country. A ratio of 1:2 is more likely. How did you get the 1:10 figure? |
|
|
| tcpip |
Here is my non-inv LM3875 PCB, which I've yet to freeze on. Compared to Roadkill's PCB, this one is larger: it's about 3.5" by 1.5". The red rectangle around the chip is an area marked in the "tKeepout" layer of Eagle for the chip. I use that area to keep other components out of that restricted area, because I want to use a clamping bar on the chip to hold it to the heatsink, and put two bolts on two sides for the bar. Also, my PCB contains the smoothing caps (1000uF to 2200uF per rail); I don't have any off-board smoothing caps. The rectified power supply will come straight from a bridge (yes, I intend using a plain vanilla bridge of 25-35A for a set of amps) to the board.
I intend to use L-clamps to attach the PCB to the heatsink, like I've done here. In that case, only the two mounting holes at the rear edge (i.e. top of picture) will be used, the other two are there "just in case."
I have provided space for a large input cap, with lots of alternate pads to accommodate various sizes. Now, with the inputs that theChris has given, I may be able to eliminate the input cap altogether. That'll save me some money... I was planning on a good MKT cap there.
I've also put the smoothing caps to the sides of the chip, not in front, to allow me to use a screwdriver easily to tighten the bolts of the clamping bar holding the chip down. I intend to use non-isolated "T" type chips, incidentally, with mica sheets and heatsink grease.
Comments please? And the Eagle files are available in case any of you want them. |
|
|
| BrianGT |
| quote: | Originally posted by tcpip
Comments please? And the Eagle files are available in case any of you want them. |
I would put the LM3875 closer to the edge of the board. I placed mine .1" from the back of the board, which seems to work good for mounting the board directly to the heatsink.
--
Brian |
|
|
| tcpip |
| quote: | Originally posted by BrianGT
I would put the LM3875 closer to the edge of the board. I placed mine .1" from the back of the board, which seems to work good for mounting the board directly to the heatsink.
|
I too had the fear that the chip shouldn't be any further forward than I could avoid, even jutting out a bit if needed, but of course, my beloved Eagle DRC would begin hollering if I did that. :)
However, there seems to be some confusion somewhere... I've certainly put my chip closer than 100mil from the edge of the board. If you want, I can send you the .BRD file and you can take a look. From the picture, does it appear that my chip is far from the edge? Here's another picture, much magnified. See if this makes things clearer? As a data point for comparison, the centre-to-centre distances between two adjacent front-row pins, or two adjacent back-row pins for that matter, is 0.134". Visually, it seems that my chip is way closer than 0.1" from the rear edge, isn't it?
One more reason to keep the chip where it is, and not shift it any further back, was to keep the supply rail tracks at their current widths. I don't know how much I need to worry about that if I move the chip back by a few mil, but I thought this was a nice compromise.
And since I'm not using the "TF" type chip, I'll get a few mils from the mica sheet and heatsink grease. |
|
|
| BrianGT |
| quote: | Originally posted by tcpip
I too had the fear that the chip shouldn't be any further forward than I could avoid, even jutting out a bit if needed, but of course, my beloved Eagle DRC would begin hollering if I did that. :)
However, there seems to be some confusion somewhere... I've certainly put my chip closer than 100mil from the edge of the board. If you want, I can send you the .BRD file and you can take a look. From the picture, does it appear that my chip is far from the edge? Here's another picture, much magnified. See if this makes things clearer? As a data point for comparison, the centre-to-centre distances between two adjacent front-row pins, or two adjacent back-row pins for that matter, is 0.134". Visually, it seems that my chip is way closer than 0.1" from the rear edge, isn't it? |
I was referring to putting the top row of the pins on the LM3875 header 100mil from the edge of the board. The rows are 200 mil apart, which is enough to put the V- trace between them.
If you get DRC errors, just modify the silkscreen part of the chip.
--
Brian |
|
|
| tcpip |
| quote: | Originally posted by BrianGT
I was referring to putting the top row of the pins on the LM3875 header 100mil from the edge of the board. The rows are 200 mil apart, which is enough to put the V- trace between them.
If you get DRC errors, just modify the silkscreen part of the chip. |
Okay. Will see how I can get this effect.
But regarding the space between the rows of pins, I seem to see a lot of the space getting eaten up by the pads. Will the gap still be enough? Can you post a magnified image of just this portion of your PCB layout, like I did? Also, there's a track going from front to back to pin 8. Since mine is a single-layer PCB, I'll have to put a jumper there. I'll find it difficult to put my feedback path components on the underside, since I have four components in the feedback path (okay, I can reduce it to three) to your one.
Thanks a lot. |
|
|
| BrianGT |
Here it is.
If you are curious, I decided to solder the NFB resistor directly to the IC pins instead of putting it on the pcb. This cleans up the board a bit, and provides the shortest possible feedback path, which is very important to the design.
--
Brian |
|
|
| tcpip |
| quote: | Originally posted by BrianGT
If you are curious, I decided to solder the NFB resistor directly to the IC pins instead of putting it on the pcb. This cleans up the board a bit, and provides the shortest possible feedback path, which is very important to the design. |
Actually, if I had to choose between moving the chip closer to the edge and switching to a double-sided PCB, versus keeping the board and FB resistors as they are but retaining a single-sided PCB, I think I'll choose the latter, just to keep costs low. Please try to accept my different perspective. :) I am not as advanced an audio system builder as you (and many others) are, and I'm not at all sure I'll be able to hear the sonic improvement that comes out of shortening the FB path by a few millimetres. On the other hand, I want to retain the other FB components as I've shown them, i.e. the cap and extra resistor, because others on the forum have reported instability with this amp at 300KHz and above, and I don't want to suffer that if possible.
Perhaps the only reason why I didn't opt for your excellent group-buy opportunity was because I could see that your design was geared for very high-end (and expensive) passive components, whose benefits would be lost on me. I'm trying to design this PCB to let me sample the Gainclone's sound as a first attempt with off the shelf MFR and caps.
But I'll see if I can move the chip further back anyway. |
|
|
| BrianGT |
| quote: | Originally posted by tcpip
Actually, if I had to choose between moving the chip closer to the edge and switching to a double-sided PCB, versus keeping the board and FB resistors as they are but retaining a single-sided PCB, I think I'll choose the latter, just to keep costs low. Please try to accept my different perspective. :) I am not as advanced an audio system builder as you (and many others) are, and I'm not at all sure I'll be able to hear the sonic improvement that comes out of shortening the FB path by a few millimetres. On the other hand, I want to retain the other FB components as I've shown them, i.e. the cap and extra resistor, because others on the forum have reported instability with this amp at 300KHz and above, and I don't want to suffer that if possible.
Perhaps the only reason why I didn't opt for your excellent group-buy opportunity was because I could see that your design was geared for very high-end (and expensive) passive components, whose benefits would be lost on me. I'm trying to design this PCB to let me sample the Gainclone's sound as a first attempt with off the shelf MFR and caps.
But I'll see if I can move the chip further back anyway. |
I could have easily made my board single sided, but it is the same cost here in the states to get the board done double sided as single sided, as most all of the board manufacturing places use a double sided process more often.
My pcb also works with cheaper components. I am looking to offer a more budget orientated kit, in April, with cheaper resistors, knocking $20 or so off the price of a kit.
Your layout looks good so far, and it certainly is fun making your own pcb layout if you are able to do so. I hope your boards work out well for you. You could even start your own group order there in India for all your friends :)
--
Brian |
|
|
| BrianGT |
Here is a picture with the more budget components available from most vendors. 0.5w compact metal film resistors, and 0.5w carbon film.
--
Brian |
|
|
| peranders |
| I like red pcb's and also black with yellow printing :nod: Blue is also OK. It's a shame though that it's a little bit harder to order other colours than green. My pcb's supplier uses only green solder mask. |
|
|
| theChris |
to TCPIP:
the locations of c6 and c7 vs c5 and c4 are switched. the smaller caps should be located closest to the ICs, ideally next to the power pins, but next to the IC is fine. |
|
|
| tcpip |
| quote: | Originally posted by BrianGT
I could have easily made my board single sided, but it is the same cost here in the states to get the board done double sided as single sided, as most all of the board manufacturing places use a double sided process more often. |
Wow... I didn't know single-sided had gone so out of fashion with guys at your end of the world. :) BTW, what kind of pricing are you getting for the board manufacture, if, say, you order a few hundred boards? And are you doing the panelising or will the board maker do it himself?
| quote: | | My pcb also works with cheaper components. I am looking to offer a more budget orientated kit, in April, with cheaper resistors, knocking $20 or so off the price of a kit. |
The price I'm hoping to do for very small quantities of my amp will be:- PCB: Rs.50 each, single sided, 2ounce Cu
- Smoothing caps: Rs.40 for a pair
- other passive components: Rs.10
That's a total of Rs.100 (a shade over USD 2.00) without the chip amp and with no input cap. If I buy the chip amp from Digikey, and use the GE metallised polyprop caps on offer at low rates from Madisound, I'll add another USD 8.00 to the total. That makes it USD 10.00. And I'll get 1% MFRs at this price, plus ordinary electrolytic Al can-type caps. Of course, I'm not putting any power resistor at the speaker output. That would have added another few cents to the total, if I'd gone for the wirewound white-coffin types available everywhere.
I don't know how this figure compares with your budget kit. I know that if I forego solder masking and legend printing, I can get the PCBs (my size, 3.5" x 1.5") done on 2-ounce Cu for Rs.20 each. And this fab-house makes good PCBs, if you see all the details I'd posted here and in subsequent posts. Of course, I'm omitting the cost of the films, which will be a one-time cost.
I want to make ten amps straight away, because I'm working on a two-way five-driver speaker system, which will be driven by one amp per driver, with active xo. I'll see whether your budget group-buy looks easy for me, or else I'll try local sources. :)
But I'd be nowhere without the help of friends on the forum, including you. You guys are the only reason I have the confidence to wing it on my own, because I'm actually not on my own. Thanks. :) |
|
|
| tcpip |
| quote: | Originally posted by theChris
the locations of c6 and c7 vs c5 and c4 are switched. the smaller caps should be located closest to the ICs, ideally next to the power pins, but next to the IC is fine. |
Thanks. Will do. Seems like it will be easy to do this change. |
|
|
| saltnpeppah |
| quote: | Originally posted by tcpip
And if you still feel like opting for toroidal, I think the price differential will be much less than 10 times, if you source both transfies from the same country. A ratio of 1:2 is more likely. How did you get the 1:10 figure? |
Well, I got a quote of Rs. 2300.00 ( :bigeyes: ) for a 750VA toroidal tranformer; but on the other hand, a local manufacturer is quoting Rs 250.00 for an E-I core... its not exactly 1:10; but close to it.
By the way, what method do you guys use to fabricate PCBs?? i tried to use the toner transfer method with limited results..
I tried to use copier paper (UGH!! what a mess!!) and then with label backing sheet..I removed the labels from the sheet and printed the pattern on the glossy side; but the toner wouldn't get into the material. I tried to iron this onto the board, but again with limited results.
Please help!!!
Madhu. |
|
|
| soundNERD |
If you can get those pcbs for only .20 each, thats extremly cheap. thats much cheaper than making them at home with the toner tranfer method and photo paper.
can you post a link to the place here?
How much do they charge for shipping?
Also, Brian. those red pcbs look awesome, same questions to you.
thanks!
-Mike |
|
|
| roadkill |
Madhu,
Go to Om Electronics (SP Road, third shop), give him a laser print of your layout (on A4 copier paper), he will quote a price (for PCBs of the size we're talking about, it's usually less than Rs. 100), pay him, get an order slip. Come back after a few days, show him the slip, collect your PCBs. Easy :) I tried the toner transfer method, results were quite shabby. Make sure you ask for a Glass Epoxy PCB. Phenolic is not worth the 10-15 rupee savings.
Also, the price you've mentioned is normal for a 750VA toroid. A 750VA E-I will cost a LOT more than Rs. 250 (close to 1k, I guess). For 250, you'd probably get a 250-300VA transformer if you buy it readymade. A custom wound transformer (try Universal Transformers, opposite Vishal Electronics), this price will get you a 225VA transformer.
Mike,
I don't think they will ship to the US, they're a relatively small-time operation. If you like, I could try and help you out with this, although I haven't got a clue about the shipping process. The courier charges would probably be many times the cost of the board.
Tarun,
The reason I put the ground lift resistor is because a 25-paisa el-cheapo carbon film resistor can save the life of an expensive loudspeaker in the event that the wire/connector between the star ground point and the amp's input ground ever comes loose.
Nice layout, here are some points:- Either manually insert some space between the polygons for the ground plane and power planes, or use Eagle's "Net Classes" command to assign a clearance of more than 15 mils. The planes, if they're so close, will cause problems due to solder bridging, etc.
- How are you going to connect the speaker wire? If you just solder it to a pad, the joint will eventually either break or cause the pad to lift off the board, since the wire used will be heavy gauge.
- The pad dimensions you have used are more suited to a PTH board. If you use similar pads for a single sided board, the adhesion will be very poor, and the pads will tend to lift off easily. Usually, octagonal, elongated pads are used for single-sided boards.
- I'd advise you to use large filter caps close to the bridge rectifier. It may sound like paranoia, but the charging/discharging current into the filter caps will cause EMI, esp. for bridge rectifiers without snubber caps. Keep the smaller caps (220u or 470u) on the board, for local decoupling.
- Disable thermals for the various planes being used. They will limit the current flow very badly.
Brian,
What PCB software are you using? I hope there's a freeware version available, since it can do rounded bends and polygon edges!
I asked Om Electronics' proprietor about the PCB process. He said the laser prints are used to make films, which are then used to make a silkscreen. He then squeezes etch-resist ink onto a bare PCB and etches it. Naturally, the quality is nowhere nearly as good as a true photo-etched PCB, but it's good enough ;) |
|
|
| saltnpeppah |
| quote: | Originally posted by roadkill
Madhu,
Go to Om Electronics (SP Road, third shop), give him a laser print of your layout (on A4 copier paper), he will quote a price (for PCBs of the size we're talking about, it's usually less than Rs. 100), pay him, get an order slip. Come back after a few days, show him the slip, collect your PCBs. Easy :) I tried the toner transfer method, results were quite shabby. Make sure you ask for a Glass Epoxy PCB. Phenolic is not worth the 10-15 rupee savings.
Also, the price you've mentioned is normal for a 750VA toroid. A 750VA E-I will cost a LOT more than Rs. 250 (close to 1k, I guess). For 250, you'd probably get a 250-300VA transformer if you buy it readymade. A custom wound transformer (try Universal Transformers, opposite Vishal Electronics), this price will get you a 225VA transformer.
I asked Om Electronics' proprietor about the PCB process. He said the laser prints are used to make films, which are then used to make a silkscreen. He then squeezes etch-resist ink onto a bare PCB and etches it. Naturally, the quality is nowhere nearly as good as a true photo-etched PCB, but it's good enough ;) |
Hi all,
roadkill, i went to om electronics and talked to suraj, the propreitor, he said that he would take 3 days for delivery. i didnt bother to ask him about the pcb process though.
anyway, the transformer price; it was quoted by a friend of mine. he has a coil winding setup and he said he would be able to do it for me. however, he asked me for the design and i am left in the dark. could you people point me out on how to design E-I core transformers??
so you say making pcbs at home is not worth the effort?? if it is, then it hurts my DIY spirit a great deal.
thanks for the suggestions and hoping for more....
Cheers,
Madhu. |
|
|
| saltnpeppah |
| quote: | Originally posted by tcpip
I also have another schematic for a finished non-inv amp for the LM3886, from Elektor. In that schematic, they've handled the mute pin too, so one can simply pick up the resistor values based on what they've put in it. (They've put in the option for a mute switch, which I don't intend to use.... I'll just keep the circuit permanently "un-muted" when I design the PCB.)
I have this article in PDF format, in case you want me to mail it to you. It's a one-page thing with schematic and PCB layout. |
Hi all,
I digged around on the internet some more after Tarun mailed me the above pdf; and i found one more elektor article, dated july-august 1998, FOR the 3886... however, the schematic and the pcb layout resemble (to a Mechanical engg. eye... ) the one in the above article... dunno the difference... you guys have a look at it..
just send across a mail to me.. ill mail it to you... by the way, is it ok if i just post the link to the file here??
Cheers,
Madhu. |
|
|
| tcpip |
| quote: | Originally posted by saltnpeppah
Well, I got a quote of Rs. 2300.00 ( :bigeyes: ) for a 750VA toroidal tranformer; but on the other hand, a local manufacturer is quoting Rs 250.00 for an E-I core... its not exactly 1:10; but close to it. |
I got a custom-wound 600VA E+I done for Rs.2750. I don't think these guys (Disco Winding Works on L.Road here) are inexpensive, but I think his transfies are up to spec (very low voltage drop at heavy loads, etc). This was for 25-0-25V, 12A per rail. Guru in Delhi sells ready-made E+I transformers. They sell a 26-0-26V, 5A transfy for Rs.550. That's about 250VA, and they're much cheaper than Disco of L.Road, but I suspect their load regulation will be worse.
BTW, my 600VA transformer will power a parallel array of 10 Gainclones. What kind of humungous setup are you planning to build with a 750VA transfy?
| quote: | | By the way, what method do you guys use to fabricate PCBs?? i tried to use the toner transfer method with limited results.. |
I don't know why Indian diyers in big cities make their own PCBs, unless (i) they're in a tearing hurry for some emergency work, or (ii) they enjoy the process. Neither of these applies to me. And making my own PCBs will usually result in much worse quality of work than what these small professional workshops do, and of course I save time, huge amounts of capital investment, and I can do things like solder masking which no home-brew PCB maker seems to do. So why don't you take Roadkill's advice and try getting the PCB made from a professional source?
If you want to get them made from Bombay, there are places which will take all your datafiles by email, and courier the finished PCB to you. Please start reading from here and read through all the posts related to PCB making. There's more information here than any diyer needs to know, about Indian PCB making options. :) |
|
|
| tcpip |
| quote: | Originally posted by roadkill
Nice layout, here are some points:- Either manually insert some space between the polygons for the ground plane and power planes, or use Eagle's "Net Classes" command to assign a clearance of more than 15 mils. The planes, if they're so close, will cause problems due to solder bridging, etc.
|
In Eagle, you can set the "isolate" parameter of each polygon separately. I always set mine to 16mil, so I guess I'm already doing what you've pointed out. Just use "change isolate", select the value you want, and click on any of the wires of the polygon. You don't need to go into "Net Classes."
And solder bridging doesn't pose a problem for me because I get my PCBs solder masked. :)
| quote: | - How are you going to connect the speaker wire? If you just solder it to a pad, the joint will eventually either break or cause the pad to lift off the board, since the wire used will be heavy gauge.
|
The internal hookup wire for the speaker connections doesn't need to be any heavier gauge than the supply rail wires, and one can even use solid core if one wants. So, within the chassis, I intend to use any suitable thick stranded/solid wire for speaker connections, and one wire would come from the banana socket to the power-star-ground, and another piece would connect the PCB's power ground (marked GND) to the same star ground. I'm not sure they'll lift off the board due to their weight... they're short pieces of thick copper wire. They don't need to be 300 strands. :)
| quote: | - The pad dimensions you have used are more suited to a PTH board. If you use similar pads for a single sided board, the adhesion will be very poor, and the pads will tend to lift off easily. Usually, octagonal, elongated pads are used for single-sided boards.
|
I've not used the default pad sizes which come in Eagle. I have standardised on larger pad sizes, by going into Eagle's "DRC" function, and changing the minimum restring radius. Eagle comes with a default minimum restring radius of 10mil for pads and 8mil for vias. I use at least 18mil. I've been using it regularly, and my PCB making workshop too thinks that they're large enough. I've faced no problems.
And I certainly don't see any need to use elongated pads for all components in general. Are you specifically referring to the pads for the amp chip? Well, in that case, I started out with elongated, but then changed to circular (maybe I'll make them octagonal, I don't know) because elongated was taking up too much space between the rows of pins.
| quote: | - I'd advise you to use large filter caps close to the bridge rectifier. It may sound like paranoia, but the charging/discharging current into the filter caps will cause EMI, esp. for bridge rectifiers without snubber caps. Keep the smaller caps (220u or 470u) on the board, for local decoupling.
|
Interesting. I can see see your point. None of the gainclone PCBs I've seen on this forum have done what you're suggesting, but at least I can see the logic behind it. Let me see what I do. Your suggestion seems particularly appealing because there will be a few metres of distance between my bridge and the amp PCB, and putting some filter caps near the rectifier seems like a good thing to do.
| quote: | - Disable thermals for the various planes being used. They will limit the current flow very badly.
|
Interesting. When I'd made my earlier layout, I was told by friends that I should put thermals because otherwise they make soldering difficult. But for very high-current pads (like the chip amp here), I think your suggestion makes sense. I'll probably revert to "change thermals off" in Eagle. :)
| quote: | | Brian, What PCB software are you using? I hope there's a freeware version available, since it can do rounded bends and polygon edges! |
Not to speak on behalf of Brian, but my Eagle 4.11 does rounded bends for tracks and polygon edges. The earlier Eagle 4.09 didn't do this.
| quote: | | I asked Om Electronics' proprietor about the PCB process. He said the laser prints are used to make films, which are then used to make a silkscreen. He then squeezes etch-resist ink onto a bare PCB and etches it. Naturally, the quality is nowhere nearly as good as a true photo-etched PCB, but it's good enough ;) |
This is how Balaji Hertz (Mr.Subramaniam of Bombay) does it too. Like Om, he too prints using etch-resist ink through his silkscreens onto the copper-clad board. You can get 12mil tracks and you can easily pass one track between two legs of a DIP IC. You can't easily get 8mil tracks, and you can't usually pass two tracks between DIP IC legs. On the whole, for audio, I think one doesn't need anything better. Of course, if you get your films made directly on a Gerber photoplotter instead of via laserprint->photofilm, you'll see that finer details are possible using the same silk screen process. The limitation is in the initial film-making more than the silk screen, I think. |
|
|
| tcpip |
| quote: | Originally posted by saltnpeppah
however, he asked me for the design and i am left in the dark. could you people point me out on how to design E-I core transformers?? |
Forget your friend, because he appears inexperienced in the specific area of transfies. Go to a shop which either sells ready-made transfies, or go to a winding company which regularly makes custom-wound transfies. There are many all over the place, and you will not need to pay prices as high as I paid to Disco Winding here. There are some design parameters to be taken into account when deciding the size of the E+I core, depending on the VA rating, current flow, and voltages used. But any transfy winding shop will do it without problems in about one minute. When I place an order with Disco, they go back to their PC, feed in my VA ratings and voltages into their software, and come back a minute later with the size of the core, overall dimensions, and therefore, pricing. It's that simple for a shop which regularly makes transformers. If your friend can't do this, go somewhere else.
| quote: | | so you say making pcbs at home is not worth the effort?? if it is, then it hurts my DIY spirit a great deal. |
I guess you must've noticed that this is a "diy audio" site, not a "diy pcb" site. We use a lot of things which we don't create ourselves. For instance, we all use eCAD software which we haven't written ourselves. We also don't hand-craft our own capacitors. :D No one will stop you, but just don't be very surprised if not too many others show much enthusiasm for your line of interest. :)
| quote: | Originally posted by saltnpeppah
I digged around on the internet some more after Tarun mailed me the above pdf; and i found one more elektor article, dated july-august 1998, FOR the 3886... however, the schematic and the pcb layout resemble (to a Mechanical engg. eye... ) the one in the above article... dunno the difference... you guys have a look at it.. |
There is no difference between the LM3886 and LM3876, other than an extra connection in the LM3886 for -Vcc, I think. The article I'd sent you caters to both the LM3876 and LM3886, where one jumper for this -Vcc is optional for the '76, but is needed for the '86. |
|
|
| tcpip |
| quote: | Originally posted by roadkill - How are you going to connect the speaker wire? If you just solder it to a pad, the joint will eventually either break or cause the pad to lift off the board, since the wire used will be heavy gauge.
|
I got some interesting data points about what others are talking about regarding wire gauge sizes, when they're building no-compromise Gainclones. See this post where a reader asks what gauge of internal hookup wire to use, and Brian says 18AWG should be fine for everything. This includes supply rails and speaker output lead. I don't think 18AWG wire will rip itself off a reasonable-sized solder pad easily, do you?
And what's interesting is that this crowd is going for a no-compromise high-end build project, where they'd go for the thickest, lowest-impedance wire that they'd think would make a difference. |
|
|
| roadkill |
Madhu,
If you want a good source for E-I transformers, it's Universal Transformers on SP Road. There's also Miracle Transformers (www.toridal.com), who make toroid core trannies. Use their website to choose a model, then email them for a quote. They charged me ~1200 for a 225VA (dual 25 volt windings @ 4.5A each), couriered it to me within a week. Good stuff!
I echo Tarun's comment, it's really not worth making your own PCBs. If we could get presensitized boards, that's a different story.
Tarun,
The Disco Winding Works quotes seem a bit high. Then again, the transformers from Universal sag by about a volt at full load.
Yeah, the isolate option's better, coz then the DRC wont complain as much.| quote: | | And solder bridging doesn't pose a problem for me because I get my PCBs solder masked. |
I'm not sure, but the solder mask alignment is very imp. for this, if it's even slightly off, you can get bridges.
| quote: | | The internal hookup wire for the speaker connections doesn't need to be any heavier gauge than the supply rail wires, and one can even use solid core if one wants. So, within the chassis, I intend to use any suitable thick stranded/solid wire for |
| | | |
