Yeah, looks like a clickbait but I had such a design target.
Analog Processing Unit contains two devices in the same housing. THAT1510 based preamp-LNA 34/60db gain, 10Vrms max output level, and 48VDC phantom power, just in case if need to work with a condenser mic. The input impedance is 47kOhm 100pF in case it needs to use as a MM phono-preamp without RIAA(Cosmos ADC has internal DSP and may handle RIAA EQ online accurately). The input noise of the LNA is about 130nVrms(A) i.e.
5-6x times fewer than APx555 which, for instance, is not capable to measure the dynamic range of $13 Meizu HiFi DAC, APx555 result is 123-124db(A). Cirrus promises AES17 DR 130db(A) for CS43131, Cosmos APU+Cosmos ADC result is 130.5db(A) with 12db room to APU's residual AES17 DR for 2mV. The LNA's balanced inputs are clamped at 5-6V by TVS thru the 10ohm 0603 resistors in serial to each input i.e. an overvoltage with a good enough current rather may burn the resistors than TVS but that's safe for THAT1510.
The second and biggest half of Cosmos APU is the 1/10kHz active notch filter with -30db ratio(Q about 50) at the fundamental frequency and nearly 0db at 2nd and 3rd harmonics i.e. you can simply divide by 30(or subtract 30db) the THD/THD+N result to get a normal value. Also, with REW or Arta, you can use a calibration file to normalize the notch's frequency response and observe FFT results directly. The max balanced input level is 10Vrms, it is also clamped but by the active clamp with opamp's rail voltage-1V, hence, 10ohm 0603 again may be burned like a fuse. The residual harmonics of the distortions are <-150db@1kHz or <-130db@10kHz, that's probably isn't too spectacular but still better than APx555. The residual THD+N, practically achieved one, -132db@1kHz@10Vrms(DAC+LPF approach), calculations based one is -134db but I can't confirm that so far because of can't find a perfect-zero-noise sine for that
It is clear that APU is kinda a microscope for your ADC, even any laptop audio input with APU gets the ability to measure a tiny noise of DACs or LDOs, and distortion levels are perhaps lower than AP SYS2722 or even Cosmos ADC.
The best way to power Cosmos APU with a power-bank(or a good smartphone charger), is to minimize GND loops issues.
The device is quite simple and inexpensive, strange that's not on the market already. Especially inexpensive(~$70) would be a version without the case as a PCBA which any DIYer can adjust by 4pcs trim-pots(2pcs for the 1kHz Twin-T -30db or -40db, another pair for 10kHz -30db).
REW with the calibration file DAC+LPF->CosmosAPU->CosmosADC
The residual THD+N of 5 preproduction samples, -101db of reading need to subtract 30db i.e. THD+N -131db@9.5Vrms the same about harmonics -127db means -157db
APU Notch frequency responses:
AES17 Dynamic Range test performs at -60dbfs level by measuring the THD+N A-weighted. That plot shows the residual THD+N (A) of the LNA at 1-10mV level sweep. 2mV corresponds 2V 0dbfs DAC like CS43131, and -82db needs to subtract 60db to get AES17 DR = 142db. A typical high-performance DAC with 5Vrms 0dbfs could be measured down to 150db and so on.
The residual noise of 5 preproduction samples(the S2 using SSM2019 instead of THAT1510), the reading need to divide by 1000 due to 60db of LNA i.e. a uV is actually nV.
The gain of Cosmos APU preamp is +60db(1000x times), hence the scale units are nVrms(A). After the APU unit turns On, you can see 5s of settling,
from 6s the input noise reaches 130nV, about 9s 48V power was turned On, and after 1s reached 142nS. Hence, APU's phantom power adds just 12nV to the preamp noise, that's a nearly ideal result as I think The test was performed with a shorted preamp's inputs(In+ to In-) and biased to the GND by 2x6.8kOhm resistor as a dummy condenser mic.
and the spectrum 20Hz-130kHz of the same test, units need to divide by 1000 as well, so uV becomes nV. This is a good illustration to killing the myth that SMPS is always dirty Cosmos APU contains no linear voltage regulators at all, only SMPS with Fswitching > 1MHz.
Finally, I got working Cosmos ADC with internal DSP and RIAA EQ implemented for all Fs from 44.1 to 384. A pair of Cosmos APU was used as a dual-mono phono preamp +34db gain with no analog RIAA, which was applied in Cosmos ADC digitally. The video is a comparison of LP vs CD sound, recorded this way.
The formal specs of such a combo:
THD+N@5mV-40ohm(AP output) -83db, and suddenly it places Cosmos APU to the 2nd position of the phono-preamps ASR rating(and yes, I know his tests are silly)
I remeasured Cosmos APU+Cosmos ADC+RIAA max accurate, AC line hum virtually zero.
5mV 10ohm source APU gain 34db Cosmos ADC 1.7V sensitivity, THD+N -89db or -97.7db(A). Due to H2 and H3 being very low, SNR is also 97db(A).
250uV 10ohm source APU gain 60db Cosmos ADC 1.7V sensitivity, THD+N -66db or -78db(A). Due to H2 and H3 being lower than the noise floor, SNR is also 78db(A).
The real SNR i.e. APU source is a cheap MM cartridge AT3600L, the LP record is digitized by Cosmos ADC@32/384, normalized to fit max peak to the -6dbfs, next, the stylus is up to the air, Pioneer PL-300(made in Japan 1980) turntable keeps 33.33rpm. This way the SNR = 91,2dbfs-6db=85.2db(A), that's looks decent if the LP "silence" between the tracks shows SNR = 65db(A) and less.
Of course, both these functions, mic and phono preamp, are features just for the feature list, but I think both are decently implemented and virtually for free so I couldn't call that a marketing
https://e1dashz.wixsite.com/index/cosmos-apu
https://archimago.blogspot.com/2022/05/early-look-cosmos-apu-high-performance.html#more
https://www.l7audiolab.com/f/e1da-cosmos-apu-2/ APx555b measurements
Analog Processing Unit contains two devices in the same housing. THAT1510 based preamp-LNA 34/60db gain, 10Vrms max output level, and 48VDC phantom power, just in case if need to work with a condenser mic. The input impedance is 47kOhm 100pF in case it needs to use as a MM phono-preamp without RIAA(Cosmos ADC has internal DSP and may handle RIAA EQ online accurately). The input noise of the LNA is about 130nVrms(A) i.e.
5-6x times fewer than APx555 which, for instance, is not capable to measure the dynamic range of $13 Meizu HiFi DAC, APx555 result is 123-124db(A). Cirrus promises AES17 DR 130db(A) for CS43131, Cosmos APU+Cosmos ADC result is 130.5db(A) with 12db room to APU's residual AES17 DR for 2mV. The LNA's balanced inputs are clamped at 5-6V by TVS thru the 10ohm 0603 resistors in serial to each input i.e. an overvoltage with a good enough current rather may burn the resistors than TVS but that's safe for THAT1510.
The second and biggest half of Cosmos APU is the 1/10kHz active notch filter with -30db ratio(Q about 50) at the fundamental frequency and nearly 0db at 2nd and 3rd harmonics i.e. you can simply divide by 30(or subtract 30db) the THD/THD+N result to get a normal value. Also, with REW or Arta, you can use a calibration file to normalize the notch's frequency response and observe FFT results directly. The max balanced input level is 10Vrms, it is also clamped but by the active clamp with opamp's rail voltage-1V, hence, 10ohm 0603 again may be burned like a fuse. The residual harmonics of the distortions are <-150db@1kHz or <-130db@10kHz, that's probably isn't too spectacular but still better than APx555. The residual THD+N, practically achieved one, -132db@1kHz@10Vrms(DAC+LPF approach), calculations based one is -134db but I can't confirm that so far because of can't find a perfect-zero-noise sine for that
It is clear that APU is kinda a microscope for your ADC, even any laptop audio input with APU gets the ability to measure a tiny noise of DACs or LDOs, and distortion levels are perhaps lower than AP SYS2722 or even Cosmos ADC.The best way to power Cosmos APU with a power-bank(or a good smartphone charger), is to minimize GND loops issues.
The device is quite simple and inexpensive, strange that's not on the market already. Especially inexpensive(~$70) would be a version without the case as a PCBA which any DIYer can adjust by 4pcs trim-pots(2pcs for the 1kHz Twin-T -30db or -40db, another pair for 10kHz -30db).
REW with the calibration file DAC+LPF->CosmosAPU->CosmosADC
The residual THD+N of 5 preproduction samples, -101db of reading need to subtract 30db i.e. THD+N -131db@9.5Vrms the same about harmonics -127db means -157db
APU Notch frequency responses:
AES17 Dynamic Range test performs at -60dbfs level by measuring the THD+N A-weighted. That plot shows the residual THD+N (A) of the LNA at 1-10mV level sweep. 2mV corresponds 2V 0dbfs DAC like CS43131, and -82db needs to subtract 60db to get AES17 DR = 142db. A typical high-performance DAC with 5Vrms 0dbfs could be measured down to 150db and so on.
The residual noise of 5 preproduction samples(the S2 using SSM2019 instead of THAT1510), the reading need to divide by 1000 due to 60db of LNA i.e. a uV is actually nV.
The gain of Cosmos APU preamp is +60db(1000x times), hence the scale units are nVrms(A). After the APU unit turns On, you can see 5s of settling,
from 6s the input noise reaches 130nV, about 9s 48V power was turned On, and after 1s reached 142nS. Hence, APU's phantom power adds just 12nV to the preamp noise, that's a nearly ideal result as I think
The test was performed with a shorted preamp's inputs(In+ to In-) and biased to the GND by 2x6.8kOhm resistor as a dummy condenser mic.and the spectrum 20Hz-130kHz of the same test, units need to divide by 1000 as well, so uV becomes nV. This is a good illustration to killing the myth that SMPS is always dirty
Cosmos APU contains no linear voltage regulators at all, only SMPS with Fswitching > 1MHz.Finally, I got working Cosmos ADC with internal DSP and RIAA EQ implemented for all Fs from 44.1 to 384. A pair of Cosmos APU was used as a dual-mono phono preamp +34db gain with no analog RIAA, which was applied in Cosmos ADC digitally. The video is a comparison of LP vs CD sound, recorded this way.
The formal specs of such a combo:
THD+N@5mV-40ohm(AP output) -83db, and suddenly it places Cosmos APU to the 2nd position of the phono-preamps ASR rating(and yes, I know his tests are silly)
I remeasured Cosmos APU+Cosmos ADC+RIAA max accurate, AC line hum virtually zero.
5mV 10ohm source APU gain 34db Cosmos ADC 1.7V sensitivity, THD+N -89db or -97.7db(A). Due to H2 and H3 being very low, SNR is also 97db(A).
250uV 10ohm source APU gain 60db Cosmos ADC 1.7V sensitivity, THD+N -66db or -78db(A). Due to H2 and H3 being lower than the noise floor, SNR is also 78db(A).
The real SNR i.e. APU source is a cheap MM cartridge AT3600L, the LP record is digitized by Cosmos ADC@32/384, normalized to fit max peak to the -6dbfs, next, the stylus is up to the air, Pioneer PL-300(made in Japan 1980) turntable keeps 33.33rpm. This way the SNR = 91,2dbfs-6db=85.2db(A), that's looks decent if the LP "silence" between the tracks shows SNR = 65db(A) and less.
Of course, both these functions, mic and phono preamp, are features just for the feature list, but I think both are decently implemented and virtually for free so I couldn't call that a marketing
https://e1dashz.wixsite.com/index/cosmos-apu
https://archimago.blogspot.com/2022/05/early-look-cosmos-apu-high-performance.html#more
https://www.l7audiolab.com/f/e1da-cosmos-apu-2/ APx555b measurements
Attachments
Last edited:
Yes, I plan that DAC with the model name Cosmos DAC. So far I can't get .5% Viking resistor for some value, only 1%(actually, even 1% we got by spots 2k here, 6k there) or waiting up to the Jul. So, the CMRR wouldn't be too impressive but, anyway I recommend using APU with power-bank battery power, hence, CMRR isn't too critical. If we have those resistors in the hands that day +1 week you can find the PCBA in our Aliexpress for $70.
Regarding the cost, my Chinese wife pushed me to set that higher, and L7, the guy who did the review in China, said to her that "so high-performance stuff you can sell for $1000" )) But I'm pretty much DIYer as well, just living in China now and can order 1k pcs assembly, so I understand the real level of my products, almost any DIYer here can make the same for the weekend and free "if have time"
So, I insisted on the $70 tag for the PCBA but the assembled/adjusted version could be >$100(i.e. who can't make the metal-case for the PCBA will pay more)
Regarding the cost, my Chinese wife pushed me to set that higher, and L7, the guy who did the review in China, said to her that "so high-performance stuff you can sell for $1000" )) But I'm pretty much DIYer as well, just living in China now and can order 1k pcs assembly, so I understand the real level of my products, almost any DIYer here can make the same for the weekend and free "if have time"
So, I insisted on the $70 tag for the PCBA but the assembled/adjusted version could be >$100(i.e. who can't make the metal-case for the PCBA will pay more)
FWIW I would order the version in the case. The illustration looks well executed. I believe you are using minijacks (3.5mm) and I understand that the small form factor really leaves few choices. Getting well made shielded 3.5mm cables is not easy but you are better located to solve that issue as well. I would pay a premium for a collection of 3.5mm to various audio connection cables.
The Viking resistors look pretty good on the datasheet. They indicate they have higher power MELF resistors with close tolerance. For the critical locations larger resistors are good if possible. For very high broadband CMRR you will probably need trimming. How much CMRR is needed in this application?
CMMR for the LNA is huge by design, THAT1510 has about 120-130db@1kHz so there is no problem at all but the notch contains OPA1612 buffers, next differential amp OPA1612x2 composite. That diffamp will use 620ohm 1%, and this is the reason why I recommend using APU only with isolated 5VDC USB power i.e. power-bank or a charger but a good one with not too high Y-capacitance.
Not sure if I mentioned this before or not, but I find that the AutoRanger product works very well with the Cosmos ADC. It gives appropriate balanced output with a choice of balanced and unbalanced input.
Of course, it performs a somewhat different function than the unreleased APU. I guess it's good to have both. Plus a Cosmos DAC. Plus a Cosmos LPF. Plus... Still way more hobbyist friendly than even the annual software subscription for AP products. (Yes, I know that you're not a hobbyist, at least by day...)
hah, it seems I bought 6000pcs fake Viking 0207 MELF resistors, but fortunately, I have 2000pcs original from the distributor )) Look isn't exactly the same as the original, the reel is also not exactly the same, no QC stamp, UV marks are different. Even 1% 620ohm Viking has resistance difference within 1ohm, at least neighbors in the reel's cells, fake has 3x time worse. Well, I hope I'll find some good MELF to replace Viking then because they have a lead time up to the end of Jul(or Jun, I'm not sure).
Interesting. Digikey has no stock on Viking. For 1% they do have options but you will need to qualify them if they are of interest. Mouser also has some in stock. Vishay seems to be the most popular. Mouser lists tolerances to .01% but that may be wishful thinking in practice.
Let us know if we can help from this side of the "pond".
Let us know if we can help from this side of the "pond".
Well, I did test the alternative MELF 0207 from Chinese supplier NYS .25%. I tested that in the differential stage only, the most critical notch filter resistors are Viking. The H3 is about -155db i.e. about the same as Viking. We sent 6000pcs NYS 0207 to the SMD production, they promise to finish that in 2 days(Monday).
@IVX , just ordered your E1DA Cosmos ADC at Audiophonics France. I am unable to measure some of my designs like
https://www.diyaudio.com/community/threads/small-class-a-amplifier-with-thd-below-0-0001.385676/
properly with my equipment and I have high hopes in your product.
https://www.diyaudio.com/community/threads/small-class-a-amplifier-with-thd-below-0-0001.385676/
properly with my equipment and I have high hopes in your product.
actually, I mark Cosmos ADC units on the bottom side like "THD+" or sometimes "THD++" when the unit exhibits a spectacular THD performance. Whatever is grade-A or grade-B, so, you could buy a very good THD performance unit for a minimal price. In the case of Cosmos APU, the distortions are anyhow lower than -150db@1kHz but you have to use some R/R divider to read >10Vrms.
