Even though you can't hear noise it masks definition. The quest for "no hiss" is a worthwhile pursuit!
One of the cheap as chips YJ amps might be at its noise floor - all linear power and a very special 3.3V regulator, the TPSA785 - an EVM that cost $20 - half the price of the amp. It has 8uV noise, wide bandwidth for digital, and low output impedance. The amp is fully silent (as expected) with no signal lock and I must have my ear right on the 86dB/W tweeter in a silent room to hear a very faint white noise - that's with signal lock and volume. I reckon this is the noise floor? A little better than -90dB?
The other YJ, the 2.1, is still standard and I can hear hiss 2 metres away with no signal lock. It's like they fitted a wind tunnel into that tiny case and I'm being tested for aerodynamic drag.
The saying goes that you can't turn a sow's ear into a silk purse, and perhaps that will apply, but I hope to enjoy trying.
One of the cheap as chips YJ amps might be at its noise floor - all linear power and a very special 3.3V regulator, the TPSA785 - an EVM that cost $20 - half the price of the amp. It has 8uV noise, wide bandwidth for digital, and low output impedance. The amp is fully silent (as expected) with no signal lock and I must have my ear right on the 86dB/W tweeter in a silent room to hear a very faint white noise - that's with signal lock and volume. I reckon this is the noise floor? A little better than -90dB?
The other YJ, the 2.1, is still standard and I can hear hiss 2 metres away with no signal lock. It's like they fitted a wind tunnel into that tiny case and I'm being tested for aerodynamic drag.
The saying goes that you can't turn a sow's ear into a silk purse, and perhaps that will apply, but I hope to enjoy trying.
Haha... but I can't add silk to this YJ. The pig's ear is too small to put more than 48kHz into it. The USB won't accept more than 16 bits either. It can't function as a purse. :-(
I tried the amp with more efficient speakers - Klispch 96dB/W and the noise is easier to hear - using the USB (16 bit) is the worst with Foobar playing at both zero volume and paused and the amp at max (30). When the volume is turned up full, the SPL is over 110dB. Oof. The room is very quiet otherwise - maybe ambient 15dB and the noise can't be detected 45cm from the speaker. So my back of a beer mat estimate says the noise floor is a little better than -90dB. It's not great but I really like the sound quality overall so I still think this amp is a total bargain.
My phone got an upgrade to Android Oreo so it now will allow APTX-HD and I have a receiver (CSR8675) that can do it too so the Bluetooth is now 48kHz and 24 bit - it's hard to tell the difference between this and the USB. In a blind test, I'm not sure I could, ... unless there were dropouts on the bluetooth ;-)
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Audeo, got a quick question need some clarification:
What determines 15uH's sat-current of 3A
Is it a relationship to the max rated current of MP2451?
Since MP2451 max out at 0.6A, do we need 3A of inductor?
That's just a very rough guesstimate from looking at the size of the trace that I had to cut to put the inductor over it. I suppose, the current there may have some peaks and transients dependent on the input voltage. This DC-DC converter powers up the entire circuitry except the power stage of STA326 and I did not measure the exact current and I'm not sure that this chip is exactly MP2451, although its application in D8 matches that in the datasheet and it indeed oscillates at about 2 MHz (that I did measure).
So, maybe a 1A inductor would be enough. Or maybe it would get too warm, and you don't want to have an additional heat source in that little box.
Last edited:
This was new to me. I was rather told that one can hear down into hiss/noise.
Is this your own observation and is it also acknowledged generally in "science"?
//
Observation from building dacs - being synthesized to this context - not true in every context of course. Improving power and lowering jitter always revealed more definition in the DACs I've played with, and this was related to working to lower noise. I'm sure there is a limit, like all things, so at some point in many contexts, there isn't any more definition to get from even the best of recordings, esp given most of us don't have ideal listening rooms.
However, these STA326 chips seem to be, by comparison, noisy, and the second YJ I'm currently playing with is the worst for noise I think I've ever heard. It's shockingly bad. So in this context, pursuing lower noise is worthwhile until I'm happy I've hit the noise floor Audeo mentioned. In fact, if I can just get the second YJ to match the first I'd be made up.
I'm afraid my humble YJ's are probably not the best test bed though, given they're sample rate limitations. I might hack the I2S lines with a 74HC157 (suitable switch?) to try a second input at higher rates.
Anyway, the second YJ is just a bit of fun while FDA tech matures - I'm convinced it has a premium future.
Thk you Audeo, for helping me understand the thought process behind inductor current. You have a point D8 has quite a number of signal + control + display circuitry to power & run. Taking your point of consideration, i would send high energy music into D8 audio signal input point (either toslink, Coax, USB), yet turning down the volume all the way to +1 level, and measure at DC supply socket, may give rough indication of the signal circuitry's current consumption (without power stage).
I try keep the inductor dimension within limited space - trying to max out inductor while sizing appropriate current rating.
for a start point, also found out that D8 product specs - the power-on without music play, consumes 48ma at DC supply socket of 28V.
I try keep the inductor dimension within limited space - trying to max out inductor while sizing appropriate current rating.
for a start point, also found out that D8 product specs - the power-on without music play, consumes 48ma at DC supply socket of 28V.
Wherever you are planning to insert the inductor, first insert a 1 ohm 1% 1/2 watt resistor and measure the voltage drop across it. V=IR. Then you can accurately work out what inductor to use to replace the resistor. Assuming a few hundred mA, you'll get a few hundred mV drop - check that doesn't mess up the dropout needed for any subsequent regulators in the circuit. You may need a 0.5 ohm.
Last edited:
I just "measured" noise using an allegedly calibrated app with the mic immediately next to the tweeter. Yes, I know, app, agc, etc so this isn't reliable data but does illustrate the issue. The modified YJ (2.0 version) is below the noise floor of the recorder at 36dB so no data, not even unreliable inaccurate data, but it does affirm what I already tested with a pro SPL meter years ago - my living room is very quiet. Regular listening volume gave a reading of between 86 and 89dB (a little over 1 watt - seems about right). The noise coming from the second YJ (2.1 version) gave a reading of 52dB. It's also worth pointing out that this noise is a white/pink noise character with a swirl of varying pitch noise mixed in. Could this be the crappest gear ever? It's a challenge! 
Thanks for your answer Nanoloop, and sorry for the late feedback.
So I understand from your message that from your perspective, it should not be a "generic flaw" from the STA326 (not delevering flesh / warmth).
I wanted to check if some have experienced similar feeling.
Best regards,
JMF
I've only tried one version of the STA326 and that with two different speakers so far - Klipsch and Kef - and what I hear is the colouration of the speakers/room more than anything else. Perhaps others who have tried more versions of this amp will comment.
Audeo point out another thread : Original Apogee datasheet design shows have 2 different inductor config cater to 4 & 8 Ohms, however D8 uses a castrated version of the filter intended for 8 Ohm speakers (D8 uses simliar inductor values from STA326/STA328 datasheet which only shows 8 ohm inductor config).
IF YJ uses similar inductor config as D8/STA326/328, then on contrary: 8 ohms speakers on YJ should be correct, as it is having 8 ohms inductor config.
IF YJ uses similar inductor config as D8/STA326/328, then on contrary: 8 ohms speakers on YJ should be correct, as it is having 8 ohms inductor config.
In my experiments with these amps the bass in particular is very sensitive to the power supply. The stock SMPSU sucks pretty bad for bass, it really needs a discretely regulated supply - I use an EF fed from a TL431 shunt reg.
The listening test verdict is out (non-qualitative):
Comparing (non A/B blind - placebo?) & referencing earlier comments:
1. D8 music instrument definition seems better to FX502Spro
(Optical digital source 16bit 44.1Khz -> Xmos U8 I2S-> STA326/8 D path amplification VS analog source-> ADC path -> TPA3250 D path amplification)
2. D8 much noiser than FX520Spro
D8 without mod noise-floor found at aux, but hiss could be hear for gain +1, ard 1 meter on my 6 ohms 86dB. Affirmed gain +0 speaker cut-off
However it is quiet on toslink digital-inputs; guess D8's ADC path was affect by smps V-reg noise? eg. hiss not started from STA326/8 amplifier.
FX502Spro impressive quiet featherlight hiss, maybe partly due to using LDO V-reg AMS1117 for 3.3V, LM317 for 24 V-reg.
Will guess have do the smps inductor noise removal mod.
Easiest and most accurate way to calculate and graph the effect of different output filter designs/values with different impedance speakers is to use TI's spreadsheet calculator. Free download. Free use. It rocks.
LCFILTER-CALC-TOOL Class-D LC Filter Designer | TI.com
What's an EF ?
I mostly use an LT1083 with the YJ 2.0 - sounds surprisingly good - ultrafast diodes with snubbers, 13,200uF pre regulator cap bank, zener at the base. I'm going to start using an AMB Sigma11 I made last night. I hooked it up and noise from the amp is the same as LT1083 despite S11 being a lower noise psu. I expect dynamics to be better - we'll see.
EE shorthand for 'emitter follower'. Sounds remarkably similar to your 'pre reg....zener at the base' except I use the TL431 as a programmable zener.
I've also tried the LM317 as CCS into a TL431 on a number of designs. Found its sometimes difficult to keep the LM from oscillating, depending on the voltage drop across it. I normally put resistors in series to limit the drop to 4V or so across the LM317. Doing this mod to the input stage PSU of my Xindak integrated amp made a very noticeable improvement to PRaT, previously the input opamp had been fed just from LM317/337s.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.