Rich,
Thanks for the enlightening review and detailing your trials and tribulations of implementing an active crossover for your Kairos/Trairos speakers...
Anand.
Sorry to start out with a downer, but that's really bad advice. The input resistance of an opamp is usually not very well controlled as "high" is the target spec. Some bipolar opamps may have biasing resistors that could work if they are tightly controlled, which they most likely won't be. Even analog-optimized semiconductor processes tend to have rather poor absolute tolerances (say ±10 % resistor tolerances to give you a general idea) but stellar matching between components. Unless the input resistance of your opamp of choice is specified with a min, typical, and max spec in the spec table, do not rely on it for anything that requires precision.
Sweet! Thank you for letting your friend test the amp and letting us know of the results. I greatly appreciate it.
Taming those idiosyncrasies was interesting. I got to know the LM3886 rather intimately during the development, and in particular in the development of Rev. 2.0. I should get some of those learnings onto my Taming the LM3886 pages.
The vast majority of the development effort for Rev. 2.0 went into addressing the performance at clipping. Thank you for recognizing my efforts.
I'm glad you like the Modulus-86. That's good to hear.
I went from my DG300B to the MOD86 in my setup. I hear ya. I like the sweet colouration of the 300B, but nothing - and I mean nothing - beats the MOD86 if you want precision.
You're welcome. I am fully committed to the Modulus-86 amplifier. It is by far the best performing amplifier I've ever designed and I've designed quite a few.
Tom
Not trying to stir things up, but JonMarsh has started posting some measurements over at http://www.htguide.com/forum/forumdisplay.php?58-On-the-Bench . Based on what he has posted, the Benchmark AHB2 ($3k) looks to be slightly ahead of the Modulus in THD+N and probably at least comparable in THD while providing more power. OTOH, the Modulus performs better than the still-very-good Cambridge 651W. The performance of the Benchmark seems to be corroborated by the Stereophile measurements here. Benchmark Media Systems AHB2 power amplifier Measurements | Stereophile.com
I'm slowly starting to actively plan around building a Modulus-86. I'm running an active wideband desktop setup kinda-sorta simlar to the Lx-Mini, but using the Scan 10f in the Ikea salad-bowl spheres on top of the same Seas L16 used in the Lx-Mini. Crossed at 300-400. I'm still noodling on whether to build a 4-channel Modulus or whether to just build a stereo amp for the Scan 10f's and use an available amp for the Seas.
My initial question is that I had the idea to try to fit the amp into an old Alesis RA-150 chassis. It looks to be about 13x8x3.25 or so, with heat-sinks along either side. I'm not sure on the dimensions of the cases that folks have used. If this doesn't look like it will be suitable, it may push me towards a 4-channel setup like Tom built since I'd have to buy a new case anyway.
I'm glad the new Rev 2.1 design is available. I move rather slowly, and I was concerned that my procrastination was going to be costly when the parts availability was looking questionable. Now I can just claim that I was waiting for the new version all along
The HT Guide link didn't seem to work. It got me to HT Guide alright, but to the main page rather than a post. A search for AHB2 turned up blank. Thank you for the link to Stereophile's measurements though. It looks like the AHB2 ($3k) has a lower noise floor than the MOD86 ($0.4k in parts, boards, chassis, transformer, etc.). I suspect the THD is likely about the same but with the lower noise floor, the THD+N gets lower.
It is worth noting that the IMD of the MOD86 is considerably lower (looks like 6-8 dB lower for the 18+19 kHz @ 1:1 case).
Very cool. It's good to see I'm not operating in a vacuum.
I have plenty of boards in stock and intend to keep the design alive for a long time to come.
The eBay BZ4309 chassis I used measures 330 x 274 x 84 mm internally. The heat sinks protrude out to the sides.
These days I'm directing folks to ModuShop for the chassis. Their products are solid and the price including shipping to North America is less than a similar sized eBay chassis. I hear great things about them and their products.
I'm glad the new Rev 2.1 design is available. I move rather slowly, and I was concerned that my procrastination was going to be costly when the parts availability was looking questionable. Now I can just claim that I was waiting for the new version all along
Just executing according to the plan, Sir!
Tom
Last edited:
Benchmark in my book are good guys. They aren't cheap, but neither are they 100k of flooby dust and shiny shiny. They are also honest and, for their DAC-2 point out that, except for 2 edge cases it should sound exactly the same as their original DAC-1!
"...includes a driver amplifier stage and a main output stage, with a plurality of impedance networks, providing among other things, feedback paths..." I find it interesting that I can match their THD performance and beat their IMD performance with only a multitude of impedance networks rather than a plurality. Now I just need to lower the noise floor a bit.
Tom
Tom
"............Now I just need to lower the noise floor a bit.
Tom
Apart from the question if there is any need for it, just put in a lower noise op-amp, say a LME49990, and lower the overall feedback impedance. A 1K resistor already has app. 4nV/sqrt(Hz).
I doubt it will make any difference sonically.
It's desirable, actually, to look at the full noise gain structure much like one does when budgeting signal gain structure. Current noise from BJT op amps is seldom negligible in audio circuits, meaning selection of low voltage noise parts---such as the LME49990---often results in greater noise than parts---such as the LME49710---with somewhat higher voltage noise but lower current noise. The Mod is a typical audio circuit in this regard.
You're correct reducing feedback impedances reduces current noise. Doing so, however, risks increasing other error terms. At error levels like the -124dB THD of the Mod noise optimization becomes not a simple exercise. This is particularly the case when desiring to preserve other attributes of good amplifier design, such as high CMRR and appropriate signal gain.
How much this matters sonically mostly depends on how one feels about the ear in the speaker test. Which, in objective terms, is largely a function of efficiency; whilst something of a niche application, 105dB SPL/W horns are fairly demanding in this regard.
You're correct reducing feedback impedances reduces current noise. Doing so, however, risks increasing other error terms. At error levels like the -124dB THD of the Mod noise optimization becomes not a simple exercise. This is particularly the case when desiring to preserve other attributes of good amplifier design, such as high CMRR and appropriate signal gain.
How much this matters sonically mostly depends on how one feels about the ear in the speaker test. Which, in objective terms, is largely a function of efficiency; whilst something of a niche application, 105dB SPL/W horns are fairly demanding in this regard.
Last edited:
That's right Twest, but 105dB efficient speakers require their own dedicated amps anyway. The noise story at THX is a bit like phono pre amps are advertised. Some claim huge SNR, but only when the input is shorted. When driven from the impedance of a cart those figures become meaningless.
Yeah, that's a significant part of it. This is the usual precision analogue whack-a-mole, so numerous other restrictions, conditions, and limitations apply. Move one thing out of the way and four others become problems.
That's the tradeoff right there. Do you want 90 dB CMRR or do you want 3-4 dB lower noise?
If you can live with 60 dB CMRR - and I think most people can - you can build your own front end that's about 3-4 times the cost of the THAT1200 but would give you that 3-4 dB improvement in noise floor. For a DIY project, that's certainly possible. For a commercial product, one would have to do a little research to ensure that one isn't stepping on THAT's patents.
To get to the 90 dB CMRR of the THAT1200, you'd need a good semiconductor process or a handful of laser trimmed resistors carved from a solid block of grain-oriented unobtanium by Ipswich virgins.
Yep. Pretty much.
Tom
Is there a matching/mating THAT device to go from unbalanced to balanced without sacrificing performance
The idea is to see about the viability of avoiding 7 metre unbalanced interconnects (that's pretty much guaranteed to degrade any audio signal), or 7 metre speaker wires that possibly going to do the same - good quality unbalanced to balanced signal transformers aren't the cheapest item in the catalogue!
The idea is to see about the viability of avoiding 7 metre unbalanced interconnects (that's pretty much guaranteed to degrade any audio signal), or 7 metre speaker wires that possibly going to do the same - good quality unbalanced to balanced signal transformers aren't the cheapest item in the catalogue!
My THAT Driver does exactly that. It takes a single-ended (unbalanced) input and provides a differential (balanced) output. Uses the THAT1646.
True. In my DG300B I use the Jensen JT-11P-1HPC. $70/each as I recall. They're nice transformers but distort significantly more than the THAT chips, in particular at 20 Hz. We're talking 0.04 %, so nothing earth shattering but approaching audible.
Tom
They're in Newmarket rather than Ipswich, but yes
Or were. Impressed the heck out of me aged 16 seeing laser trimming in action.