TDA8932 has been my workhorse amplifier for a number of years now and its great for a desktop setup. But what about when I need a bit more power, like for a living room or an even larger space for demonstration purposes? I have played around on paper (and some experiments) with getting more juice out of the chip by paralleling and using a trafo on the output. But nothing has really grabbed me as a way to boost the output power into the above 100W region.
For higher power amps, I've gone in the past for IRS2092-based designs and these can sound pretty good with the signal power supplies optimized - the same technique I've used to get the best from TDA8932. So that is one route that looks worth following. But I would like to understand classD amps a bit better and nothing beats trying to design one for getting on a steep learning curve. IRS2092's main drawback is it needs a split supply, it would be really excellent if I could come up with an amp that just needed a single 48V supply and dished out 100W comfortably. A monoAMP on steroids so to speak.
Not using IRS2092 means going to opamps and comparators and H-bridge drivers as the next level components down. The first puzzle for me is - IRS2092 uses an OTA internally and those are definitely not too common as separate components. The venerable old CA3080 (beloved of analog synthesizer makers) has been obsolete for years and LM13600 can't be too far behind. I downloaded the long list of IRAUDAMP datasheets and discovered that, while many used IRS2092, no.1 did not. So I figured this must be a good place to make a start - instead of an OTA it has an LT1220 which is a fairly mid-range fast opamp. Its LTP input stage is heavily degenerated which leads to bags of slew rate and less-than-stellar noise numbers. The LF noise corner also seems rather high (around 1kHz) so I went in search of an alternative with lower noise, or at least a lower LF noise corner.
AD817 turns out to have an LF noise corner around 10Hz so altogether much more promising. Its less DC-capable but I doubt that matters too much for an audio amp. Second-hand on Taobao AD817 is affordable enough so I ordered some. In the meantime I wanted to build something. I had a bag of AD744 to hand so that went in (in place of LT1220) to my first classD prototype built from the ground up (rather than from using a dedicated classD chip). For the output stage I decided to use CD4049 inverters in parallel as they can run up to 18V supply. The idea was to make an amp just powerful enough to drive my HD6XXs (300 ohm).
On first firing this amp up, the oscillation frequency was wildly too high, well over 1MHz. The variable R which controls the idle freq I had as an SoT fixed R and I had to reduce this to about 15ohm to get into the right ballpark, by which I mean under 500kHz. Apart from this the thing did work and amplified my DAC's output by 6dB. The idle current though was a bit higher than I was hoping for, around 40mA for both channels. In the flush of a successful build, at first I didn't really notice any shortcomings but on quick listening my wife said 'too noisy!'. How much of the noise comes down to the AD744 I wondered?
(to be continued)
For higher power amps, I've gone in the past for IRS2092-based designs and these can sound pretty good with the signal power supplies optimized - the same technique I've used to get the best from TDA8932. So that is one route that looks worth following. But I would like to understand classD amps a bit better and nothing beats trying to design one for getting on a steep learning curve. IRS2092's main drawback is it needs a split supply, it would be really excellent if I could come up with an amp that just needed a single 48V supply and dished out 100W comfortably. A monoAMP on steroids so to speak.
Not using IRS2092 means going to opamps and comparators and H-bridge drivers as the next level components down. The first puzzle for me is - IRS2092 uses an OTA internally and those are definitely not too common as separate components. The venerable old CA3080 (beloved of analog synthesizer makers) has been obsolete for years and LM13600 can't be too far behind. I downloaded the long list of IRAUDAMP datasheets and discovered that, while many used IRS2092, no.1 did not. So I figured this must be a good place to make a start - instead of an OTA it has an LT1220 which is a fairly mid-range fast opamp. Its LTP input stage is heavily degenerated which leads to bags of slew rate and less-than-stellar noise numbers. The LF noise corner also seems rather high (around 1kHz) so I went in search of an alternative with lower noise, or at least a lower LF noise corner.
AD817 turns out to have an LF noise corner around 10Hz so altogether much more promising. Its less DC-capable but I doubt that matters too much for an audio amp. Second-hand on Taobao AD817 is affordable enough so I ordered some. In the meantime I wanted to build something. I had a bag of AD744 to hand so that went in (in place of LT1220) to my first classD prototype built from the ground up (rather than from using a dedicated classD chip). For the output stage I decided to use CD4049 inverters in parallel as they can run up to 18V supply. The idea was to make an amp just powerful enough to drive my HD6XXs (300 ohm).
On first firing this amp up, the oscillation frequency was wildly too high, well over 1MHz. The variable R which controls the idle freq I had as an SoT fixed R and I had to reduce this to about 15ohm to get into the right ballpark, by which I mean under 500kHz. Apart from this the thing did work and amplified my DAC's output by 6dB. The idle current though was a bit higher than I was hoping for, around 40mA for both channels. In the flush of a successful build, at first I didn't really notice any shortcomings but on quick listening my wife said 'too noisy!'. How much of the noise comes down to the AD744 I wondered?
(to be continued)
