Makes sense (although perhaps marginally better to dump the RFI to chassis at the board via one of the mounting fixtures rather than carrying it back alongside the incoming signal?). As I have not purchased boards I am not privy to the design documentation and it was something I was curious about having looked at the build photos.
Another turnkey solution might be to design the amp module with solder-to-board input connectors, but that of course would mean less flexibility as to case layouts. (Perhaps the design is compact enough to be configured/laid out so as to corner mount both to heat sink and rear panel.) Worth the hassle? Dunno.
Seems like a great little amp. Thanks.
When I first thought about my balanced to single-ended input board (a prerequisite for astx SA2014 build because I only have XLR sources) I was thinking of just using a THAT product. I was steered away from doing so by those helping me who viewed them as too noisy (plus Bruno's comment about the distortion performance).
Ps: I now understand why Mark designed the RingNot PSU and why he called it such.
One could certainly do that. Then you're requiring the end user to use metal standoffs.
Additional hassle. Additional expense. Another board to keep in inventory. I don't find it worth the hassle.
Sure. A lot of people certainly seem to like it. Thanks.
You can see the performance of the THAT products by looking at my measurements of my DIFF PRE, THAT Driver, and THAT Receiver. You can do a little bit better if you use one of the audio opamps that have been released since the THAT products came out. Whether the difference is meaningful in practice is another story. It is a measurable difference, as you can see in the measurements of my Modulus-286. The differences in noise and THD are partly due to the use of an OPA1612 rather than the THAT1200 and due to a tighter layout. Unfortunately the OPA1612 is SMD only and one of my requirements for the MOD86 was to use only thru-hole components for easy and beginner-friendly assembly.
Tom
Thanks. I will take a look. As a beginner myself I find SMD soldering soooo much easier than through-hole. Sure, the first time you look at a few 0805 parts it's a little intimidating, but grab a pair of fine-tipped tweezers, get going and before long you realize it just isn't hard at all and a darn sight easier than all that flipping and clipping for through-hole. Sometimes I think experts sharing their knowledge and boards with the DIY community miss a beat not encouraging newcomers to jump into SMD. Food for thought....
The Modulus-286 is all SMD except the two output inductors, the connectors, and the LM3886es. If you like SMD, that's the obvious option for you. It also provides slightly better performance than the Modulus-86.
I prefer SMD as well. No leads to clip. They also make for a tighter layout. However, SMDs do require a steadier hand and better eyesight for successful assembly than do the thru-hole parts.
For some insight into the building of a Modulus-286 from a beginner's perspective, I suggest having a look at the Modulus-286 Build Thread.
Tom
I prefer SMD as well. No leads to clip. They also make for a tighter layout. However, SMDs do require a steadier hand and better eyesight for successful assembly than do the thru-hole parts.
For some insight into the building of a Modulus-286 from a beginner's perspective, I suggest having a look at the Modulus-286 Build Thread.
Tom
stating "any length" is going too far.
Adding length and/or loop area will increase the impedance of the shunting leg of the filter.
H.Ott specifically refers to that.
I agree both with you and Mr. Ott about that. However, L = 0 m is not obtainable. So some common sense and perspective is useful.
Tom
Tom
how about 360 ring of smt C built into chassis connector
NC3FDX-EMC-Spec - Neutrik
looks like its even been talked up before in this thread http://www.diyaudio.com/forums/search.php?searchid=15971529
NC3FDX-EMC-Spec - Neutrik
looks like its even been talked up before in this thread http://www.diyaudio.com/forums/search.php?searchid=15971529
Last edited:
I'm in the process of building a 4-way speaker system. In your opinion, what kind of power supply would best suit two 286 amp boards (per speaker channel)?
Hi Tom
After finishing my Mod286 and while enjoying the wonderful performance I got inspired to read up on all the science behind it. Started with Cordell's book and this thread - phew tough read for a noob.
Anyway, I understood that part of the great overall performance is the exceptionally low THD+N and IMD, the very high CMRR and the very high PSRR. As you provided tons of data on the THD, IMD and CMRR, I could not find a figure on the PSRR. I think you mentioned, that you wanted to measure it and got some gear for it but apparently never got around to it. At least I couldn't find the PSRR figure anywhere. Please point me to it if I missed it.
So I thought maybe the PSRR is driven by the THAT1200 amp for the Mod86 or the OPA1612 for the Mod286. And indeed the OPA1612 datsheet says 0.1 uV/V typ PSRR. Is that also the PSRR of the Mod286?
Does this mean, that the approximate 1.3 to 1.5 V ripple voltage from the Power86 affect the Mod286 output by Gain (e.g. 10) x 1.3V x 0.1uV/V = 1.3uV only?
Could you please correct me with my assumption of the Mod286 PSRR or the estimated effect on the Mod286 output?
Thanks I really appreciate it.
After finishing my Mod286 and while enjoying the wonderful performance I got inspired to read up on all the science behind it. Started with Cordell's book and this thread - phew tough read for a noob.
Anyway, I understood that part of the great overall performance is the exceptionally low THD+N and IMD, the very high CMRR and the very high PSRR. As you provided tons of data on the THD, IMD and CMRR, I could not find a figure on the PSRR. I think you mentioned, that you wanted to measure it and got some gear for it but apparently never got around to it. At least I couldn't find the PSRR figure anywhere. Please point me to it if I missed it.
So I thought maybe the PSRR is driven by the THAT1200 amp for the Mod86 or the OPA1612 for the Mod286. And indeed the OPA1612 datsheet says 0.1 uV/V typ PSRR. Is that also the PSRR of the Mod286?
Does this mean, that the approximate 1.3 to 1.5 V ripple voltage from the Power86 affect the Mod286 output by Gain (e.g. 10) x 1.3V x 0.1uV/V = 1.3uV only?
Could you please correct me with my assumption of the Mod286 PSRR or the estimated effect on the Mod286 output?
Thanks I really appreciate it.
Last edited:
Pardon the belated response here. To reach me directly, it's better to send me an email, which you can do through the Contact Us page on my website.
I use my Power-86 with an Antek AS-3222 (2x22 VAC, 300 VA) transformer to power my 4xMOD86 amp for my LXmini speakers. The only reason I didn't use the MOD286 for that build is that I hadn't created the MOD286 at that time.
If you'd rather go with an SMPS, the Connex Electronic SMPS500R would be a good fit for a 4-channel amp. The SMPS300RE is a bit under-powered for that application. Get the ±30 V version of the SMPS.
Yeah. I recall reading Douglas Self's article series in Wireless World, which led to his Power Amp Design Handbook, back in 5th or 6th semester in college. It took quite a while before things started to settle for me even then.
Cordell's book is a little lighter - or perhaps just not quite as dense - but still a very heavy read for a newbie for sure.
Yeah. I picked up a transformer so that I can inject a signal onto the supply but never set up the experiment. Something about moving across an international boundary, setting up shop elsewhere, and starting life as a full-time student got in the way.
What matters in reality is the system performance. If the amp delivers good THD, IMD, etc. when supplied by a Power-86 and transformer, does it matter that the high performance is due to the high PSRR? It matters for your system-level understanding, of course, but I'm not sure the end user cares. So getting the measurement system set up for the PSRR measurement drifted down the priority ladder a bit. I'd still like to perform that measurement, but other projects are in the way.
The PSRR in the MOD86 is actually the PSRR of the THAT1200 combined with the PSRR of the on-board regulators. In typical numbers, the line regulation of the regulator is 0.01 %/V or 100 uV/V (= 80 dB). The total PSRR from the supply to the output of the THAT1200 should then be 82+80 = 162 dB = 8 nV/V.
For the MOD286, the PSRR should be that of the local regulators and the OPA1612 combined, so 140+80 = 220 dB or 10 pV/V.
Needless to say, I consider the MOD86 and MOD286 amps to be power supply agnostic.
10 * 1.3 * 8e-9 = 104 nV, actually. Or 130 pV for the MOD286. Both are well below the noise floor.
Now these numbers do decrease with frequency (on a 20 dB/dec slope), but for a quick back-of-the-envelope calculation the numbers above hold water. Also note that above math assumes typical silicon performance. If you use the worst case numbers the numbers get slightly ... worse. Still below the noise floor, though.
Tom
Hi Tom,
thanks for your insights…
To build on what you wrote in the above post and also in post 193 , I tried to guesstimate the SPL of the noise generated by supply ripple and came to the following result, which would (if correct) confirm the in-audiability of any ripple / ringing noise.
Ripple Voltage: Mod286 at +-35Vrail and 50Hz (mains freq. = 10ms cycle) with C=22.000uF per rail.
- 8ohms Speaker: I=4.4A -> Vripple = 0.01s x 4.4A / 0.022F = 2V
- 4ohms Speaker Imax = 7A (Mod86 Limit) -> Vripple = 3.2V
PSSR (worst case, upper end of amp bandwidth):
I was not sure for what “reference” frequency your above PSSRs (162dB Mod86, 220dB Mod286) stand for, so I assumed 50-60Hz, let’s say 10^1 Hz.
So at the upper end (ultrasound) of the Mod286 bandwidth ca. 10^5Hz (exactly at ca. 82kHz) the PSSR would have decreased by ca. 80dB (4 decades from 10^1 to 10^5Hz with -20dB/decade).
- PSSR Mod86 at 10^5Hz: ca. 80dB = ca. 10^-4 V/V
- PSSR Mod286 at 10^5Hz ca. 160dB = ca. 10^-8 V/V (wow … that’s small !!!)
Output Noise: with 20dB amp gain
- Mod86: ca. 10^-3 V => Output Power Noise: Pnoise = Unoise^2/2/Rspeaker = 10^-7 W
- Mod286: ca. 10^-7 V => Output Power Noise: Pnoise = Unoise^2/2/Rspeaker= 10^-15 W (wow … )
Plugging the Noise Power values into a SPL calc. , assuming 87dB/W speakers at 1ft listening distance, results in:
- Mod86: SPL Noise ca. 27 dB (well below the ambient noise of ca. 30-35dB), in addition that the majority is well into the ultrasonic range.
- Mod286: SPL Noise: simply not existent. (<<<<10dB).
Summary:
Are these back-of the envelope calcs correct?
If yes, it would show, that any ripple or ringing from supply voltage does not generate any audible noise whatsoever. Therefore, adding larger supply caps or adding snubbers is absolutely superfluous for the Mod86/Mod286 in terms of noise (except for psycho-acoustic reasons ;-)).
Please challenge and correct the above assumptions / calculations … it would help me personally to also better understand the subject in total.
Thx.
SH
thanks for your insights…
To build on what you wrote in the above post and also in post 193 , I tried to guesstimate the SPL of the noise generated by supply ripple and came to the following result, which would (if correct) confirm the in-audiability of any ripple / ringing noise.
Ripple Voltage: Mod286 at +-35Vrail and 50Hz (mains freq. = 10ms cycle) with C=22.000uF per rail.
- 8ohms Speaker: I=4.4A -> Vripple = 0.01s x 4.4A / 0.022F = 2V
- 4ohms Speaker Imax = 7A (Mod86 Limit) -> Vripple = 3.2V
PSSR (worst case, upper end of amp bandwidth):
I was not sure for what “reference” frequency your above PSSRs (162dB Mod86, 220dB Mod286) stand for, so I assumed 50-60Hz, let’s say 10^1 Hz.
So at the upper end (ultrasound) of the Mod286 bandwidth ca. 10^5Hz (exactly at ca. 82kHz) the PSSR would have decreased by ca. 80dB (4 decades from 10^1 to 10^5Hz with -20dB/decade).
- PSSR Mod86 at 10^5Hz: ca. 80dB = ca. 10^-4 V/V
- PSSR Mod286 at 10^5Hz ca. 160dB = ca. 10^-8 V/V (wow … that’s small !!!)
Output Noise: with 20dB amp gain
- Mod86: ca. 10^-3 V => Output Power Noise: Pnoise = Unoise^2/2/Rspeaker = 10^-7 W
- Mod286: ca. 10^-7 V => Output Power Noise: Pnoise = Unoise^2/2/Rspeaker= 10^-15 W (wow … )
Plugging the Noise Power values into a SPL calc. , assuming 87dB/W speakers at 1ft listening distance, results in:
- Mod86: SPL Noise ca. 27 dB (well below the ambient noise of ca. 30-35dB), in addition that the majority is well into the ultrasonic range.
- Mod286: SPL Noise: simply not existent. (<<<<10dB).
Summary:
Are these back-of the envelope calcs correct?
If yes, it would show, that any ripple or ringing from supply voltage does not generate any audible noise whatsoever. Therefore, adding larger supply caps or adding snubbers is absolutely superfluous for the Mod86/Mod286 in terms of noise (except for psycho-acoustic reasons ;-)).
Please challenge and correct the above assumptions / calculations … it would help me personally to also better understand the subject in total.
Thx.
SH
Last edited:
Those calculations actually assume a constant current draw of 4.4 A and 7.0 A respectively. A class AB output stage draws current in pulses. Half the cycle it's drawn from VCC the other half from VEE. For sine wave operation that turns above into:
8 Ω: I = 4.4/(2*sqrt(2)) = 1.55 A RMS
4 Ω: I = 7/(2*sqrt(2)) = 2.47 A RMS
You can work through the rest of your math to find that the ripple-induced output signal is even smaller than you thought. Note that the numbers assume that the amp is running at full blast with a sine wave, so the ripple-induced signal should be compared to the SPL of the sine wave in a traditional SNR or IMD fashion.
For music the ripple-induced signal on the output of the Modulus amps gets even lower as music tends to have a relatively high crest factor (ratio between peak and RMS voltage).
Some will rightfully argue that the math is rather simplistic. I'd agree with that. To get a better estimate of the ripple voltage, I suggest running a simulation of the power supply with a pulsed sine wave load at various frequencies in your favourite circuit simulator. You'll likely find that above numbers are reasonably close to reality - at least that's generally been my finding.
60 Hz.
Exactly.
Tom
Last edited:
Ok. Thanks....
Does that mean the rest of the math all the way to the estimated noise SPL is ok?
Even if to be corrected by a factor of 2-3 to a even lower value the estimates should represent a worst case scenario (higher than actual ripple) and higher decrease of PSSR towards higher frequencies (10Hz to 100kHz -> 80dB) than actual (60Hz-80kHz). Even if the math is off due to some some simplifications towards the conservative side (provided the logic is ok) the order of magnitude already shows that the resulting noise voltage, power and noise SPL are absolutely negligible. Is that right?
Last edited:
P = E^2/R. You have P = E^2/2R, which implies that you're working with peak voltages. That's a fair assumption, but it only works for sine waves (Vp = Vrms*sqrt(2)). The ripple voltage is not sinusoidal, so the math gets a bit more involved. Also the ripple voltage will have some higher frequency content (I see harmonics of the mains frequency reaching into the low kHz on the ripple voltage from common unregulated linear supplies) and the PSRR varies with frequency.
For a rough estimate, what you have is probably OK with the limitations and assumptions I've pointed out earlier. For higher precision, I suggest running a simulation. You can set up the power supply pretty easily in the simulator. In the past I've used an advanced current source that allowed for its current to be defined as an equation to create a pulsed sine source with an added DC component for the quiescent current. Two of those (one for each half of the period) model the Class AB output stage. You should be able to model the PSRR with a high-pass filter. Then get the simulator to calculate the RMS output voltage and work out the power dissipated in the speaker and the resulting SPL from there.
Tom