More transparency in active speakers - part II
Posted 15th November 2010 at 01:59 PM by abraxalito
Updated 5th March 2012 at 03:58 AM by abraxalito
Updated 5th March 2012 at 03:58 AM by abraxalito
Its not something I claim originality for, this use of Cat6 cable for running to speakers. What though I believe is original is combining the function of speaker cable with the sharing resistors - I've seen that nowhere else. Remember guys, you saw it here first 
Not only does this save on having low value fairly high power resistors but it comes with another benefit which suits diyers. That's the fact that the amps can be easily tested separately - they're not going to be in parallel until the drive unit is connected to the output, since the cables join together only at the far end, at the terminals of the driver. I've certainly found this an advantage in development - the individual gains I've checked with a multimeter between the outputs when playing a 400Hz sinewave. This frequency was chosen because I'm too lazy to check the frequency response of my multimeter and I wanted a mid-range test.
The gains of the individual chip amps do need matching very closely across the whole frequency band - for the gain setting resistors I used normal 1% parts which I hand-selected with my 4 3/4 digit DMM to better than 0.1%. It will be interesting to see how stable the gains are over time. I'm planning to build the next prototype using some old (at least 25years) Holco H2s from a box I have of them. When I measured these I found some of them had drifted beyond their 0.1% tolerance so that their mean value was around 0.06% high - they had drifted pretty much together which bodes very well in terms of how stable the gain will be over time.
Below is the schematic for the bass/mid driver amp - all DC coupled (no servos, no caps) but with low enough offset because of the low voltage gain. So no trimmers required - I haven't measured the output offsets but on connecting the speaker unit didn't notice a change in quiescent current on the PSU meters (bench supply, 3A). The quiescent current cold is about 32mA per chip, rising to 40mA when warmed up. The photo shows the stereo pair of prototype amps - the tweeters are fed from a halved variant (2+2) as on the original crossover there's approximately 6dB more gain to the bass/mid than the tweeter. A careful comparison of the two channels will reveal differences in the zobel networks - on the second build I split the zobel in two so that I could experiment with returning them to both - and + supply rather than to ground. Another step in the direction of going truly "ground free".
Notice that the way of getting the TDAs stable with such a low gain is to increase the noise gain once beyond the audio band. This is achieved with the 2n/1700R networks between the summing node and ground. This way the chip sees its normal 24dB or so of gain at HF, but down in-band, its a measly 6dB. A small amount (under 1dB) of peaking beyond the audio band is apparent with this method of compensation. In order to develop this amp I created a Spice model for the LM1876 (which I originally planned to use until I found just how cheap the TDA2030a was). This model is based on the Bowers/Alexander app note from ADI. If anyone's interested in the LTSpice file to sim this design for themselves, please drop me a PM with your email address and I'll forward it to you.
The next stage in development is to change the nice clean linear bench PSU for a cheap and cheerful pair of SMPSUs. With the original LM3886s (not bridged) when I tried this, the sibilance increased markedly and no amount of snubbing or filtering ameliorated this condition. The hope is that this bridged design will be more tolerant of rubbish on the supplies - we shall see.
In my next entry, I'll talk a bit about the active crossovers originally in the speakers and the changes I've made to them to feed the BDUB.
Not only does this save on having low value fairly high power resistors but it comes with another benefit which suits diyers. That's the fact that the amps can be easily tested separately - they're not going to be in parallel until the drive unit is connected to the output, since the cables join together only at the far end, at the terminals of the driver. I've certainly found this an advantage in development - the individual gains I've checked with a multimeter between the outputs when playing a 400Hz sinewave. This frequency was chosen because I'm too lazy to check the frequency response of my multimeter and I wanted a mid-range test.
The gains of the individual chip amps do need matching very closely across the whole frequency band - for the gain setting resistors I used normal 1% parts which I hand-selected with my 4 3/4 digit DMM to better than 0.1%. It will be interesting to see how stable the gains are over time. I'm planning to build the next prototype using some old (at least 25years) Holco H2s from a box I have of them. When I measured these I found some of them had drifted beyond their 0.1% tolerance so that their mean value was around 0.06% high - they had drifted pretty much together which bodes very well in terms of how stable the gain will be over time.
Below is the schematic for the bass/mid driver amp - all DC coupled (no servos, no caps) but with low enough offset because of the low voltage gain. So no trimmers required - I haven't measured the output offsets but on connecting the speaker unit didn't notice a change in quiescent current on the PSU meters (bench supply, 3A). The quiescent current cold is about 32mA per chip, rising to 40mA when warmed up. The photo shows the stereo pair of prototype amps - the tweeters are fed from a halved variant (2+2) as on the original crossover there's approximately 6dB more gain to the bass/mid than the tweeter. A careful comparison of the two channels will reveal differences in the zobel networks - on the second build I split the zobel in two so that I could experiment with returning them to both - and + supply rather than to ground. Another step in the direction of going truly "ground free".
Notice that the way of getting the TDAs stable with such a low gain is to increase the noise gain once beyond the audio band. This is achieved with the 2n/1700R networks between the summing node and ground. This way the chip sees its normal 24dB or so of gain at HF, but down in-band, its a measly 6dB. A small amount (under 1dB) of peaking beyond the audio band is apparent with this method of compensation. In order to develop this amp I created a Spice model for the LM1876 (which I originally planned to use until I found just how cheap the TDA2030a was). This model is based on the Bowers/Alexander app note from ADI. If anyone's interested in the LTSpice file to sim this design for themselves, please drop me a PM with your email address and I'll forward it to you.
The next stage in development is to change the nice clean linear bench PSU for a cheap and cheerful pair of SMPSUs. With the original LM3886s (not bridged) when I tried this, the sibilance increased markedly and no amount of snubbing or filtering ameliorated this condition. The hope is that this bridged design will be more tolerant of rubbish on the supplies - we shall see.
In my next entry, I'll talk a bit about the active crossovers originally in the speakers and the changes I've made to them to feed the BDUB.
Total Comments 1
Comments
-
Hi abraxalito, I have a LM1876, and I am mounting a board to be a simple amplifier, but I wanted too much have a Spice model of this component, I use Multisim 11, and it only have LM1875, I have much headache trying to make a LM1876 model, I am tired. I would be very happy if you could send me your model to test. Thanks in advance.Posted 7th November 2011 at 01:05 PM by Phistar
Updated 7th November 2011 at 01:07 PM by Phistar (Multissim = wrong, Multisim = correct)
