Increasing track width, whilst increasing the current capacity of the design, also reduces the track impedance, which is also important - keep the track width when routing power to the op amps as thick as possible too. Looks like you should be able to use 80mil on the power boards.
I second the use of thicker tracks to help combat any etching/exposing problems, this can be a pain, but through a bit of trial and error this shouldn't be an issue.
Also Vikash, you could include snubbing caps over the diodes in the rectifier.
For making the PCBs, if using Fotoboard 2
http://www.megauk.com/pcb_laminates.php
as listed at the top of this page.
I just had great success using two, 100 watt equivalent, energy saving bulbs suspended about 20cm away from the board, with around a 6-7 minute exposure. I made 14 PCBs, 8 of which were double sided this way, with Zero failures. The bulbs have to be positioned horizontally over the boards for this to work. Remember to use half strength developer with Foto 2.
Using the largest surface mount caps (only for ease of use) you can find for decoupling seems like a good idea too.
Edit - I just had another look at the PCB - the area you are using for ground is split up quite considerably, this isn't a great idea.. What would work better possibly, reverse the smoothing caps, so they now take the +ve and -ve power towards the centre. Then rotate the diodes in the centre. This will increase the width of the copper in the upper and lower portion and increase the width of the copper running -> of the regulators in the middle. If you could liberate R1, R2, LED1 and LED2 out of the middle in the same way, then I would. This is of course the advantage of making a double sided PCB, you can go to town on the bottom with whatever track layout you want, and keep the copper top scotch free for the ground plane. Alternatively, you could leave the smoothing caps the way they are, and just clear D1, D2, R1, R2, LED1&2 out from the middle.
I second the use of thicker tracks to help combat any etching/exposing problems, this can be a pain, but through a bit of trial and error this shouldn't be an issue.
Also Vikash, you could include snubbing caps over the diodes in the rectifier.
For making the PCBs, if using Fotoboard 2
http://www.megauk.com/pcb_laminates.php
as listed at the top of this page.
I just had great success using two, 100 watt equivalent, energy saving bulbs suspended about 20cm away from the board, with around a 6-7 minute exposure. I made 14 PCBs, 8 of which were double sided this way, with Zero failures. The bulbs have to be positioned horizontally over the boards for this to work. Remember to use half strength developer with Foto 2.
Using the largest surface mount caps (only for ease of use) you can find for decoupling seems like a good idea too.
Edit - I just had another look at the PCB - the area you are using for ground is split up quite considerably, this isn't a great idea.. What would work better possibly, reverse the smoothing caps, so they now take the +ve and -ve power towards the centre. Then rotate the diodes in the centre. This will increase the width of the copper in the upper and lower portion and increase the width of the copper running -> of the regulators in the middle. If you could liberate R1, R2, LED1 and LED2 out of the middle in the same way, then I would. This is of course the advantage of making a double sided PCB, you can go to town on the bottom with whatever track layout you want, and keep the copper top scotch free for the ground plane. Alternatively, you could leave the smoothing caps the way they are, and just clear D1, D2, R1, R2, LED1&2 out from the middle.
Attachments
Thanks for the feedback guys. I'm hooked on this now.
Matt, BSC is accounted for in the broad notch. Plus I'd like to keep things as simple as possible to give me a fighting chance of finishing it in time.
I guess I had better get measuring so that I can complete the eq design and order everything in one go.
David, my parts bin is fairly large, but my experience with PCB layout is pi** poor. Hence the oversized caps chosen in haste. Al suggested using a paperclip as an inexpensive drill bit for PCB's. But I'm lucky to have an extensive drill bit collection. My uncle should never have left all his old tools in my garage
As for the PCB, I downloaded a 'hobby' DRC file which has already increased the track width and spacing etc. Track widths are 0.8mm at the moment.
Matt, I don't understand what yo mean by the ground being split up considerably. I've attached an updated layout which I thought looked neater.
edit: changed to 1.5mm traces.
edit: updated missing lines
edit: ok last time. 2mm traces.
Matt, BSC is accounted for in the broad notch. Plus I'd like to keep things as simple as possible to give me a fighting chance of finishing it in time.
I guess I had better get measuring so that I can complete the eq design and order everything in one go.
David, my parts bin is fairly large, but my experience with PCB layout is pi** poor. Hence the oversized caps chosen in haste. Al suggested using a paperclip as an inexpensive drill bit for PCB's. But I'm lucky to have an extensive drill bit collection. My uncle should never have left all his old tools in my garage
As for the PCB, I downloaded a 'hobby' DRC file which has already increased the track width and spacing etc. Track widths are 0.8mm at the moment.
Matt, I don't understand what yo mean by the ground being split up considerably. I've attached an updated layout which I thought looked neater.
edit: changed to 1.5mm traces.
edit: updated missing lines
edit: ok last time. 2mm traces.
Attachments
Yes indeed it is, I should have looked more closely
Ground. Basically Imagine where the current has to flow - From the component along the PCB, to the wire leading away from the board, that connects you to star ground inside the case.
Your ground point it seems, is the connector marked '0v'. In the first instance, imagine a current flowing from the smoothing caps to the star ground point, it has to travel through the middle of the PCB, which is fine, but has to pass through a tiny amount of copper along the way, if you can increase the size of this copper even better. Your second board layout is much better.
Also note ground paths. On IC2 From the middle pin, which is grounded, there are two paths for current to flow to '0v', either through the copper in the middle, or the copper over the top. This afaik is whats called a 'ground loop' and is undesirable. Either connect it to ground via the copper at the top, or the copper in the middle.
Adding in a highpass will take you all of 5 mins once you get used to the PCB program. If you have the time there is no reason not to add it in. You can always bypass the circuit if you like with a jumper.
Hope that helps clear some things up, Matt.
The almost final psu design attached. Its basically 5 planes now rather than many tracks.
I started on the eq circuit using dual opamps and it's not proving easy to do a single sided board. I wonder whether using single opamps rather than dual would make routing easier?. Also do you have any recommendations for a good but cheap opamp?
V
I started on the eq circuit using dual opamps and it's not proving easy to do a single sided board. I wonder whether using single opamps rather than dual would make routing easier?. Also do you have any recommendations for a good but cheap opamp?
V
Attachments
That layout is much better. My only comment is that the pads on the voltage regulators are almost touching! I doubt the toner transfer process or etching will work here. You only need to run a sharp knife between the pads (after etching) to break any remaining copper though.
As for layout problems, I put my op-amps in a row and run the power rails under the devices. Resistors and capacitors can then be placed so that they jump over the power rails. The attached pic shows the principle (blue is bottom copper, red is top with ground plane flood removed for clarity).
I'm with Richie on op-amp choices, and I'd add NE5532/5534 to the list too. Some people swear by them, others swear at them.
Nice one,
David
As for layout problems, I put my op-amps in a row and run the power rails under the devices. Resistors and capacitors can then be placed so that they jump over the power rails. The attached pic shows the principle (blue is bottom copper, red is top with ground plane flood removed for clarity).
I'm with Richie on op-amp choices, and I'd add NE5532/5534 to the list too. Some people swear by them, others swear at them.
Nice one,
David
Attachments
Vikash said:
All I can say is practise makes perfect. I have just designed a PCB for a four way speaker using about 20 opamp chips per mono board, all of it is routed on the copper bottom, apart from 2-3 tracks which found their way to the copper top. This of course is using the copper top as the ground plane, so I dont have to worry about cutting out ground on the bottom with the tracks. IME using duals or singles makes hardly any difference.
V [/B][/QUOTE]
richie00boy said:
I have breadboarded using OPA134/2134 with no bypassing to great effect. Of course it didnt sound as good as the PCB, but there were no issues with stability or massive 'buzzing' injected due to interference.
The price of a TL072 is 70p, OPA2134 is £2, at farnell. TL072 do sound rather good for the price I have to agree, but there is quite an obvious layer of dirt, or as if there is a veil over the sound, when using these. OPA2134 doesnt have this, I would recommend using these instead of TL072, especially as you dont need so many. There is always the samples program from TI.com
As you are using a PCB, thats likely going to end up reasonably well designed I see no problem in using OPAx134.I dont know what your PCB layout will be like, but you could use OPAx134 for the opamps in the signal path (dont forget to use an input buffer) and TL07x for those in the gyrator circuits,
Here's the filter board. I'm quite proud of it actually, but it's taken forever to do. Lousy autorouter.
I went with 15mm pitch for the caps which should allow me to fit any poly cap value which might be needed. Any parallel caps can be done under the board I guess (I reached the board length limit for Eagle Lite version).
Let me know if the design looks ok. Hopefully this w/e I'll get round to taking final FR's of the speakers and then finalising.
V
I went with 15mm pitch for the caps which should allow me to fit any poly cap value which might be needed. Any parallel caps can be done under the board I guess (I reached the board length limit for Eagle Lite version).
Let me know if the design looks ok. Hopefully this w/e I'll get round to taking final FR's of the speakers and then finalising.
V
Attachments
Here's another tip I forgot to give you: never use the autorouter.
The ground connection to R5, or R7 - I can make out the indent, is a little torturous. Rotating it 180 degrees would help.
Some decoupling caps on the op-amps would be a good idea. If they won't fit onto the PCB layout then just solder them directly onto the pins.
Other than that, it looks like a winner.
It appears you have the +ve input pin connected directly to the output, if these are singles, which the TL071 indicates, this is incorrect and should be changed to the -ve input pin and the output.
And yes decoupling caps, from what I can immediately see nothing else appears wrong nice layouts too.
And yes decoupling caps, from what I can immediately see nothing else appears wrong
nice layouts too.
I've taken some FR measurements today and I have two main questions. Firstly a subjective one - how much filtering is too much? In the attached I've modelled a response using four filters; three notches (1 for BSC) and another to fill in a peak. The resulting FR looks good but is it going too far?
Secondly, what (free) software can I use to model active filters against a frequency response? I'm strugling to find anything.
Vikash
Secondly, what (free) software can I use to model active filters against a frequency response? I'm strugling to find anything.
Davey, I played with LSPCad again but it's an early release of v6 and a little buggy at best . Anyway I created the four transfer functions (even though the fo entered didn't correlate to anything realistic) but are you saying you can then make it spit out a circuit from those transfer function blocks? I couldn't figure out how...Davey said:
Vikash
Attachments
With LspCAD there are two options, atleast in 5.25. One is called 'parameter mode' where it displays the filters as those 'transfer function blocks'. The other mode is component mode, where it will display the filter schematic.
May I recommend using a shelving network for BS compensation? It will/should allow for easier tweaking of the BSC applied, especially if you use trim pots for the resistors.
RE the peaks/troughs - take some off axis measurements. If they disappear and are diffraction related then you may not have to equalise them.
As a recommendation there are some equalisation plugins for winamp, that allow you to set the Q, centre frequency and depth in dB. I forget the name of the one I used, as I only toyed with it briefly. But perhaps you could use something like this, using the computer as a source to try out the notches.
- Found the plug in.
What I've got here are two dsp plugins. One allows you to use Winamp DSPs with foobar, the other is the EQ plugin.
From what I can see, you select two frequencies per notch. This I would imagine are the -3dB points either side of the notch/peak then you select the gain in dB.
May I recommend using a shelving network for BS compensation? It will/should allow for easier tweaking of the BSC applied, especially if you use trim pots for the resistors.
RE the peaks/troughs - take some off axis measurements. If they disappear and are diffraction related then you may not have to equalise them.
As a recommendation there are some equalisation plugins for winamp, that allow you to set the Q, centre frequency and depth in dB. I forget the name of the one I used, as I only toyed with it briefly. But perhaps you could use something like this, using the computer as a source to try out the notches.
- Found the plug in.
What I've got here are two dsp plugins. One allows you to use Winamp DSPs with foobar, the other is the EQ plugin.
From what I can see, you select two frequencies per notch. This I would imagine are the -3dB points either side of the notch/peak then you select the gain in dB.
If you want something perfect for a test
Download this..
http://tda.tention.org/index.php?cat=tdae&dest=eq2004p
Eq plug in. Its a VST plug in.
http://www.winamp.com/plugins/details.php?id=146317
is a winamp VST plug in.
This isnt complicated but long winded.
Chain =
Load foobar > Open the DSP > Load in the Winamp DSP bridge > Open the winampDSPbridge > Load the Winamp VST bridge> Open the VST plugin.
So essentially what you have running is thus.
A VST plugin 4 band parametric Equaliser. This is running in the Winamp VST plugin. Then the Winamp VST plug in, is running in the plugin that allows foobar to use winamp plugins.
Confusing but it all works, but only in 16 bit. If you manage to get it all going, I would imagine it should give you pretty accurate tests.
Download this..
http://tda.tention.org/index.php?cat=tdae&dest=eq2004p
Eq plug in. Its a VST plug in.
http://www.winamp.com/plugins/details.php?id=146317
is a winamp VST plug in.
This isnt complicated but long winded.
Chain =
Load foobar > Open the DSP > Load in the Winamp DSP bridge > Open the winampDSPbridge > Load the Winamp VST bridge> Open the VST plugin.
So essentially what you have running is thus.
A VST plugin 4 band parametric Equaliser. This is running in the Winamp VST plugin. Then the Winamp VST plug in, is running in the plugin that allows foobar to use winamp plugins.
Confusing but it all works, but only in 16 bit. If you manage to get it all going, I would imagine it should give you pretty accurate tests.
Using the bundled in v5.25, it's doing what I want perfectly now. It took but a few seconds to implement various eq filters and see the modeled FR, schematic, and component values required.5th element said:
I had a look at the shelving network but I can't see how it would better a notch in this instance considering the natural FR of this driver.
The 6.5/9.5/12.5kHz peak/dip/peak is inherent to the driver and apparent in the off-axis measurements too.
In the LSPCad calculated circuit there is an opamp followed by the input resistor R1 preceeding each new filter. Am I correct in assuming they're not required in implementation?
Attachments
Vikash,
I find myself using the v5.25 of LspCAD quite a bit. It doesn't have the freeform ability of the newer version, but I find it more comfortable to work with. (Creature of habit I guess.)
You should be able to achieve the results you're looking for with just a single serial resistor. You can't construct that in LspCAD 5.25, but it will do the component calculations and you can then transpose into LTSpice or LspCAD 6.0 to create a final topology.
Cheers,
Davey.
I find myself using the v5.25 of LspCAD quite a bit. It doesn't have the freeform ability of the newer version, but I find it more comfortable to work with.
(Creature of habit I guess.)You should be able to achieve the results you're looking for with just a single serial resistor. You can't construct that in LspCAD 5.25, but it will do the component calculations and you can then transpose into LTSpice or LspCAD 6.0 to create a final topology.
Cheers,
Davey.
Some of the centre frequencies are not modelleing as calculated in LTSpice. I can't even try the exact same schematic/values in LSPCad 6 either so I'm at a loss as to what to do.
I've attached the schematic which now includes 4 notches and 1 peak filter. The 4th notch was added just because there was an unused opamp channel left (did the pcb again using dual opamps).
I even tried Tina-TI but it will only export absolute values in FRD type format. Grrrr.
I've attached the schematic which now includes 4 notches and 1 peak filter. The 4th notch was added just because there was an unused opamp channel left (did the pcb again using dual opamps).
I even tried Tina-TI but it will only export absolute values in FRD type format. Grrrr.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.