Trouble with thick film is that purportedly the noise performance is even worse than carbon film TH resistors, along with even a worse thermal stability coefficient. Although having said that I haven't had that much of an issue with carbon film (still trying to deplete my stocks on all my trial projects, haha)
RS components carries MELF thin film/metal film resistors, which are 'only' twice as expensive as TH metal films that they sell (though compared to local prices both options are far from bad). I'm looking at DIY quantities only though.
I had a re-browse and it seem Digi-Key carries some Susumu 0805 thin films which aren't priced that badly in quantities of 100, but shipping fees have to be considered.
I've built a few crossovers with thick film resistors, no noticeable noise problems on them so far. Quoted tempcos aren't too hot admittedly, perhaps 200ppm. But if they were really poor in thermal stability then they'd not make them in close tolerances, and they do (0.1% is available). I've never heard of carbon film in 0.1%
Ah, but there's a catch - the 200ppm thick films are not the ones boasting 0.1% precision. The 0.x% thick films will have a better tempco rating (for e.g Rohm's MCR10 datasheet that I'm looking at now has 50ppm for 0.5% vs 100ppm for 1%, and 200ppm for 5%)
ED- datasheets are all good and that, but only if you can buy the product advertised :/
ED- datasheets are all good and that, but only if you can buy the product advertised :/
Last edited:
Its a good point, though Ralec do 100ppm right across the board, they don't do 0.1% in 200ppm. I'm not sure whether the 1% ones I have are 100 or 200ppm.
http://www.ralec.com/download/products/1_Thick_film_chip_resistor.pdf
http://www.ralec.com/download/products/1_Thick_film_chip_resistor.pdf
As for the tweaking et cetera - variable resistors Vs board space:
You can buy a bunch of bourns-like trimmers off fleabay,
20k ohm 3296 Variable Resistor Precision ,40 pcs - eBay (item 310285395860 end time Feb-03-11 18:26:18 PST)
short the center pin of the trimmer to one of the adjacent pins,
bend the 1 and 3 pins so they would fit the holes in the board and solder them slightly,
and viola - you got a variable resistor, which doesn't takes much space.
When you have settled on specific resistor values, you just buy the required values and replace the trimmer.
Still the board space is an issue - these trimmers wouldn't fit in standart configuration - you'll need to place the caps on the opposite side of the board (if you use full-sized caps).
Or quite the opposite - temporarily mount the trimmers underneath the board, tweak them, order specific values.
I'll make some program (or even web service) for finding a "composite" resistor by series-paralel connection of lower grade (E24) resistors. Say you need 6.31oHm resistor? Take 27ohm and 8.2ohm (27|8.2) and you get 6.29, pretty close. Or (100|(6.8|68)), and take 6.31.
Here are some calculations i've made for 10kOhm stepped attenuator in E24-series (bottom tables):
http://www.vegalab.ru/forum/attachment.php?attachmentid=89503&d=1277238017
I'll do the silkscreen printing on both sides of the board.
You can buy a bunch of bourns-like trimmers off fleabay,
20k ohm 3296 Variable Resistor Precision ,40 pcs - eBay (item 310285395860 end time Feb-03-11 18:26:18 PST)
short the center pin of the trimmer to one of the adjacent pins,
bend the 1 and 3 pins so they would fit the holes in the board and solder them slightly,
and viola - you got a variable resistor, which doesn't takes much space.
When you have settled on specific resistor values, you just buy the required values and replace the trimmer.
Still the board space is an issue - these trimmers wouldn't fit in standart configuration - you'll need to place the caps on the opposite side of the board (if you use full-sized caps).
Or quite the opposite - temporarily mount the trimmers underneath the board, tweak them, order specific values.
I'll make some program (or even web service) for finding a "composite" resistor by series-paralel connection of lower grade (E24) resistors. Say you need 6.31oHm resistor? Take 27ohm and 8.2ohm (27|8.2) and you get 6.29, pretty close. Or (100|(6.8|68)), and take 6.31.
Here are some calculations i've made for 10kOhm stepped attenuator in E24-series (bottom tables):
http://www.vegalab.ru/forum/attachment.php?attachmentid=89503&d=1277238017
I'll do the silkscreen printing on both sides of the board.
I already have an early Excel sheet with this on it.
You can usually find any value to closer than 0.1% to what is required. Then it's down to resistor tolerance for absolute value.
Email & I'll send it.
AndrewT,
Thanks for the excel sheet
I'm doing some DSP based xover too... and thought of implementing the scheme of this analog filter in biquads. Should be simple... As long as i'll study the DSP biquads voodoo, and take the values from s-plane multipliers of the filter's transfer functions (i've already calculated them for the xover blocks in the software).
Then, having the biquad values, doing basic multiplication in computer should be easy = you get computer based simulator of analog circuit... I've already tried the directsound library awhile ago - yet just in stereo mode. Now it's 5.1 time
Looks way too ambitious... I'll need to split the development steps, so i could pass the waypoints during development with working device in hands.
I've setup a web page for this project
s3t.it> Active crossover: AXO3
and for the digital one
s3t.it> Digital crossover: DXO4
Thanks for the excel sheet
I'm doing some DSP based xover too... and thought of implementing the scheme of this analog filter in biquads. Should be simple... As long as i'll study the DSP biquads voodoo, and take the values from s-plane multipliers of the filter's transfer functions (i've already calculated them for the xover blocks in the software).
Then, having the biquad values, doing basic multiplication in computer should be easy = you get computer based simulator of analog circuit... I've already tried the directsound library awhile ago - yet just in stereo mode. Now it's 5.1 time
Looks way too ambitious... I'll need to split the development steps, so i could pass the waypoints during development with working device in hands.
I've setup a web page for this project
s3t.it> Active crossover: AXO3
and for the digital one
s3t.it> Digital crossover: DXO4
Last edited:
The TI DSP you've chosen looks to be good bang for the buck at $5 or so but I'd be a bit concerned about its lack of flexibility. If it doesn't sound much good, what is there to tinker with? In particular, the dither feature is a little worrying - it says that an LFSR "can be used to dither the audio". Doesn't sound like much attention's been paid there.
As long as there are 24-bit DACs on the outputs, the dither shouldn't make difference - the lower bits will be lost in DAC's noise floor anyway. At least i hope so
16bit DAC's performance will suffer from the dither tho... Still, most of 16bit DACs require upsampling DF on their input - maybe their DF will do the dither job...
Ahh, that means i couldn't use TDA1541 in NOS for midrange output
This device uses the TAS:
Ground Sound
16bit DAC's performance will suffer from the dither tho... Still, most of 16bit DACs require upsampling DF on their input - maybe their DF will do the dither job...
Ahh, that means i couldn't use TDA1541 in NOS for midrange output
This device uses the TAS:
Ground Sound
Seems a little odd to me that a $4k box would use a $5 DSP. They've paid a fair bit of attention to various implementation details but not to dither. Many years back, when I was working amongst a team of pro audio guys designing a digital console, they were concerned about dither and devoted a significant area of silicon in their custom processor to getting it right. I doubt one feedback shift register really cuts the mustard. For a start, TPDF dither (the kind that doesn't introduce noise modulation) is created by summing together two noise sources.
The notion that the dither is beneath the noise floor might have some merit, but I'd not want to bank on it.
I guess I'm just conservative in my engineering instincts, I wouldn't feel comfortable investing development effort into something based on just hoping it will sound fine
The notion that the dither is beneath the noise floor might have some merit, but I'd not want to bank on it.
I guess I'm just conservative in my engineering instincts, I wouldn't feel comfortable investing development effort into something based on just hoping it will sound fine
Yes, they use the TAS3108:
http://www.groundsound.com/dcn24.php
The $4k price tag is made from nice case, ADCs, DACs,PSUs, volume reg, propertiary software, pro assembly and "High-End" thing...
Well, one of the I2S inputs could be used as dithering noise input And then mixed with a processed signal...
L/R input and bitclock will be sourced from the WM8804, and data signal could be generated externally and then processed in the DSP to become 1-bit "noise"...
The problem as i see it is to make this noise uncorrelated with audio signal and other noise sources - we are going "analog" on this input, and proper function should be applied to normalize the noise source...
I've choose this chip just because it is easy to implement, cheap, and wouldn't require much development efforts/time/costs. In addition, it's clocking scheme seems nice and flexible (could work as slave for the DACs, could take it's clock from the I2S input, could be master for all other components...).
WM8804+TAS+EEPROM+4xWM8804 - and viola, you've got the cheapest DSP board ever
By the way, it could do FIR too - going down to ~400Hz of filtration at 44.1kHz.
Too bad it doesn't have sample rate converter inside - then i could use it as DF for the TDA1541/PCM63/58/etc. I could use some ASRC on it's input, and then do all the filtration at 4xFs rate... Seems pretty flexible for it's 5$. No analog inputs - doing ADC is always wrong.
Still, the project is "for fun", fully opensource e.t.c.
http://www.groundsound.com/dcn24.php
The $4k price tag is made from nice case, ADCs, DACs,PSUs, volume reg, propertiary software, pro assembly and "High-End" thing...
Well, one of the I2S inputs could be used as dithering noise input
And then mixed with a processed signal...L/R input and bitclock will be sourced from the WM8804, and data signal could be generated externally and then processed in the DSP to become 1-bit "noise"...
The problem as i see it is to make this noise uncorrelated with audio signal and other noise sources - we are going "analog" on this input, and proper function should be applied to normalize the noise source...
I've choose this chip just because it is easy to implement, cheap, and wouldn't require much development efforts/time/costs. In addition, it's clocking scheme seems nice and flexible (could work as slave for the DACs, could take it's clock from the I2S input, could be master for all other components...).
WM8804+TAS+EEPROM+4xWM8804 - and viola, you've got the cheapest DSP board ever
By the way, it could do FIR too - going down to ~400Hz of filtration at 44.1kHz.
Too bad it doesn't have sample rate converter inside - then i could use it as DF for the TDA1541/PCM63/58/etc. I could use some ASRC on it's input, and then do all the filtration at 4xFs rate... Seems pretty flexible for it's 5$. No analog inputs - doing ADC is always wrong.
Still, the project is "for fun", fully opensource e.t.c.
Last edited:
Well, one of the I2S inputs could be used as dithering noise input
Yes
Good thinking!Its about twice the price but given that it includes 8 channels of ASRC, this one might be better if you're really hankering after SRC:
Analog Devices ADAU1445
I'm planning to do something similar to your idea, but neither of those chips appeals to me owing to proprietary lock-in. So I'm eagerly awaiting the LPC4000 series from NXP to kick start my ideas. In the meantime I do have an LPC1768 board to get me some way along the learning curve with the tools.
Wow, LPC's are nice - ethernet/usb stack + real MCU + DSP in single chip - looks like pretty sophisticated network/usb transport/crossover/room EQ could be made.
Lots of coding tho... Not that i'm afraid of coding, but i hate C language
Another interesting option for DSP xover/room correction is a plain FPGA with full-sized FIRs.
Lots of coding tho... Not that i'm afraid of coding, but i hate C language
Another interesting option for DSP xover/room correction is a plain FPGA with full-sized FIRs.
Yes, and remarkably low power too. Oh, you forgot the LCD
Lots of coding tho... Not that i'm afraid of coding, but i hate C language
Me too, so there's only one solution - Thumb-2 assembler
Assembly coding is so much more fun even though it takes longer. If I used C I'd be wanting to find out why it ran so slowly anyway, so why not miss out that step? Yes, pretty much subject to proprietary lock-in there too. So ARM's the way to go for me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.