Hi,
There are many boards from Tinysine that use I2S input to ADAU1701 and then use the inbuilt DACs of the chip. It is comforting that the ADCs are not used and one can feed I2S from Toslink, Coaxial, bluetooth, and of course one could take HDMI ARC converter and feed Toslink to these amp boards from TV too.
The board I have in mind is this https://www.tinysineaudio.com/products/tsa1702c-spdif-coaxial-dsp-audio-receiver-board
Does the DAC quality matter on real speakers?
My speakers are https://jblpro.com/en/products/jbl-nano-k8
And I am building 12 inch stereo subs using TPA3255 for a 2.2 system.
The reason why I ask this question is I have a feeling that one could make out the deficiency of the ADAU170 DACs using headphone based listening but probably not speakers.
Warm Regards,
WonderfuAudio
There are many boards from Tinysine that use I2S input to ADAU1701 and then use the inbuilt DACs of the chip. It is comforting that the ADCs are not used and one can feed I2S from Toslink, Coaxial, bluetooth, and of course one could take HDMI ARC converter and feed Toslink to these amp boards from TV too.
The board I have in mind is this https://www.tinysineaudio.com/products/tsa1702c-spdif-coaxial-dsp-audio-receiver-board
Does the DAC quality matter on real speakers?
My speakers are https://jblpro.com/en/products/jbl-nano-k8
And I am building 12 inch stereo subs using TPA3255 for a 2.2 system.
The reason why I ask this question is I have a feeling that one could make out the deficiency of the ADAU170 DACs using headphone based listening but probably not speakers.
Warm Regards,
WonderfuAudio
Yes. ADC and or DAC quality can reasonably be expected to have some effect on resultant sound. Whether or not there is an objectionable effect is likely going to be up to the person doing the listening.
To put it in somewhat blunt terms, by some measures the TinySine board could be considered a toy for tinkerers to have some fun with.
Also, the JBL speakers you link to are clearly built to a low-ish price point. Probably they are not intended for serious hi-fi; maybe more for low cost gaming, or maybe for simple monitoring when working out in the field. Maybe some things like that.
Last edited:
So Mark, you are implying the DACs (hence DSP) need upgrade (probably with ADAU1452 with better external DAC chips) and also the speakers ?
ADAU devices are more than enough for most applications as loudspeakers are the weak link in the signal chain when it comes to quality / fidelity / distortion. And with speakers, bigger is always better.
One thing many people would like to avoid in many of the ADAU range of devices is the sigma-delta ADC / DACs, but some have S/PDIF Tx/ Rx which is handy.
One thing many people would like to avoid in many of the ADAU range of devices is the sigma-delta ADC / DACs, but some have S/PDIF Tx/ Rx which is handy.
Think it is best to measure the specific board. I have mixed experiences with the implementation of those cheap boards.
I have for 15 years run a miniDSP with ADAU1701. No external ADC/DAC. Home use
In my setup it is silent and has a good gainstaging between the components. No hiss in my 90db+ sensitivity high end speaker elements.
Extremly transparent sound. Thats why I stick with it, even with its 90dB THD/N
The good thing with 1701 only is the common masterclock on input/output conversion.
Think the main reason to go external ADC/DAC is balanced system. Then there is a real benefit in system noise suppression.
But if noise is no problem, why go balanced?
I have for 15 years run a miniDSP with ADAU1701. No external ADC/DAC. Home use
In my setup it is silent and has a good gainstaging between the components. No hiss in my 90db+ sensitivity high end speaker elements.
Extremly transparent sound. Thats why I stick with it, even with its 90dB THD/N
The good thing with 1701 only is the common masterclock on input/output conversion.
Think the main reason to go external ADC/DAC is balanced system. Then there is a real benefit in system noise suppression.
But if noise is no problem, why go balanced?
And of course the ADAUs has limited drive capability. So there has to be attention on the loading. Eg 10k+ like olden day 47k input poweramps
The problem with some implementations is the inferior analog components in signal chain in input/output. Where the resistors or capacitors distorts more than the ADC/DACS. Sometime horrible amounts more. It is analog filters that are supposed to filter out HF to avoid aliasing, but destroys the performance itself with excesive distortion. Good intentions, but bad implementation.
It kind of looks like to me that everyone is trying to tell you more or less the same things in their own way. Basically, the ADAU line of DSP processors is good enough for many purposes. Sometimes people need more than that in which case they may use things like SHARC chips and or FPGAs. But for the most part ADAU are pretty common.
The problems are with analog aspects of ADC and DAC circuitry, including electronic noise coupling when the data converter function is located inside the DSP chip. SPDIF has its issues too, mostly with clock jitter, although their are ways to make that less bad. That said, jitter is something that only affects mixed signal functions where analog is a part of it: DACs, ADCs, and ASRCs.
Speakers are another matter. Its not low-cost to make good ones. You might look around at semi-professional monitors from Mackie or better manufacturers. JBL knows how to make good speakers too of course, but the ones you are looking at are an attempt to break into the very low-cost entry-level part of the market. Not the same as the professional or semi-professional studio part of the market, and certainly not anywhere near the professional mastering room part of the market.
In case its still not clear about the speakers, there is a lot more to it than just specified frequency response. It can be almost meaningless without a response graph and without knowing how the graph was taken. Sometimes there are tricks used to make a FR graph look smoother than it really is (such as by sweeping the signal generator faster than the filters can settle). Distortion is another big factor in speakers, as is dispersion as a function of frequency.
Last edited:
Well, I agree that it can be a problem, but not with the 1701. It is just engeenered low cost, good enough.
To put it in perspective the ESS9039 has DSP inside the DAC and measures incredible. Better than -120dB thd+noise even in pretty basic, but good engeneering, circuits implementations.
But we agree the 1701 is not state of the art and the difficult part is making good filters.
But using it with a standard 24 dB LR filter for SUB, main filtering is good enough for most applications.
Then the user can spend a lifetime to get the bassresponce right. That is an art.
But using it with a standard 24 dB LR filter for SUB, main filtering is good enough for most applications.
Then the user can spend a lifetime to get the bassresponce right. That is an art.
Before I forget. One main quality with the 1701 and other integrated chips is that the outputs can be in phase. No clock drift between the DACs. For filtering it is the most important aspect in my view.
True, however we are talking about rather different extremes between that and ADAU parts.
For ES9039Q2M, I count 5-different power rails, each of which should be on its own voltage regulator. Plus, it needs a good local crystal clock for good performance (again, ideally on its own voltage regulator), and not a PLL for lowest-cost clock recovery from a SPDIF stream. So with lots of internal design tricks and good surrounding circuitry, yes, it is possible to get good measurements from what amounts to several ICs in total.
Since I have extensive usage experience with a variety of the SigmaDSP IC lines and the programming thereof as well as in-circuit integration, I'll wager in my two cents on the ADAU1401/ADAU1701 (same chip really).
Some things I really like about this arguably outdated chip/platform:
Is it sufficient for real speakers (whatever that's supposed to imply)?
Oh yeah. Absolutely. Depending on your acceptable input-to-output delay, you can squeeze a lot out of this thing.
On the other hand, there's some other things to keep in mind that may unexpectedly mean the ADAU1401/ADAU1701 is no longer suitable:
More acceptable delay = larger chunks of data = more efficient instruction scheduling == more performance
That goes for sample rate as well. The higher the sample rate, the higher the delay (buffering of chunks); keeping the delay the same, raising the sample rate will reduce headroom in processing power.
The register-written output muting value or DAC/ADC or DSP core disabling/enabling is NOT inherently free of pops or clicks. Changing the values of these registers in the GUI is for debugging only.
They're intended to be programmatically written as part of a custom SigmaDSP or C/C++ sketch or program that also controls GPIO's to trigger relays or the likes.
The mutes are there for preventing crosstalk across neighbouring circuits when one's disabled and one's not, or turning off outputs if you don't need them click-free. The power-off is there for power saving purposes. Reducing the click on power-on/off of cores and/or outputs ("mutes") requires DC-blocking capacitors or other circuitry external to the chip, which can impact the signal if chosen improperly.
One quirk to note: changing the instructions per cycle register setting in the register window effectively means all your frequency points should be shifted by the same multiple. The software, however, doesn't do this; in other words, it doesn't reflect this real-world frequency shift in the displayed values.
The outdated version of SigmaStudio you'd use for the ADAU1701 and similar older chips is no longer being updated. It's... not a dealbreaker.
And last but not least: If you're debugging/experimenting/testing with a physical eval kit board, be careful with static and shorting things. The outputs do not have short circuit protection, current sinking protection or overvoltage protection; you will kill them if you do this to them, or if any of your components do when they die.
I have since moved on to AD SHARC and TI DSP chips, but that's simply because I wanted something more challenging. My main rig still runs a modified ADAU1701-powered dev kit board in a little box. I have to admit adding this note was a last-minute thought, and that's frankly because it's been just on and running 24/7 for years, and I forget it's there. it just works.
Feel free to ask any other questions, even if they're specifically for me. I'd be happy to respond, as would lots of others here.
Regards,
HumbleDeer
Some things I really like about this arguably outdated chip/platform:
- It's perfectly adequate for many jobs, purely processing power wise
- It's very easy to get it sounding really good with minimal componentry
- It's not hard to add external DAC's or ADC's, granted you add a manual compensation delay on the built-in ones
- Circuit integration is a treat. Multiple ways exist of programming or booting a program;
Two common ones being start-up bootstrapping from an external µC, and programs programmed into external EEPROM. - The realtime "tuning"/debugging can be done over I2C "just like that". It just recompiles on the fly and programs it into the registers.
Is it sufficient for real speakers (whatever that's supposed to imply)?
Oh yeah. Absolutely. Depending on your acceptable input-to-output delay, you can squeeze a lot out of this thing.
On the other hand, there's some other things to keep in mind that may unexpectedly mean the ADAU1401/ADAU1701 is no longer suitable:
More acceptable delay = larger chunks of data = more efficient instruction scheduling == more performance
That goes for sample rate as well. The higher the sample rate, the higher the delay (buffering of chunks); keeping the delay the same, raising the sample rate will reduce headroom in processing power.
The register-written output muting value or DAC/ADC or DSP core disabling/enabling is NOT inherently free of pops or clicks. Changing the values of these registers in the GUI is for debugging only.
They're intended to be programmatically written as part of a custom SigmaDSP or C/C++ sketch or program that also controls GPIO's to trigger relays or the likes.
The mutes are there for preventing crosstalk across neighbouring circuits when one's disabled and one's not, or turning off outputs if you don't need them click-free. The power-off is there for power saving purposes. Reducing the click on power-on/off of cores and/or outputs ("mutes") requires DC-blocking capacitors or other circuitry external to the chip, which can impact the signal if chosen improperly.
One quirk to note: changing the instructions per cycle register setting in the register window effectively means all your frequency points should be shifted by the same multiple. The software, however, doesn't do this; in other words, it doesn't reflect this real-world frequency shift in the displayed values.
The outdated version of SigmaStudio you'd use for the ADAU1701 and similar older chips is no longer being updated. It's... not a dealbreaker.
And last but not least: If you're debugging/experimenting/testing with a physical eval kit board, be careful with static and shorting things. The outputs do not have short circuit protection, current sinking protection or overvoltage protection; you will kill them if you do this to them, or if any of your components do when they die.
I have since moved on to AD SHARC and TI DSP chips, but that's simply because I wanted something more challenging. My main rig still runs a modified ADAU1701-powered dev kit board in a little box. I have to admit adding this note was a last-minute thought, and that's frankly because it's been just on and running 24/7 for years, and I forget it's there. it just works.
Feel free to ask any other questions, even if they're specifically for me. I'd be happy to respond, as would lots of others here.
Regards,
HumbleDeer
Bypassing the 1701's ADCs is a good idea to improve the 1701's performance. To me, the biggest limitation of the 1701 DACs is the limited 0.5 V max output (part of the reason the DAC's THD + Noise is 83 dB). Consequently you have to be vigilant and properly match the amplifier gain. Ideally you want to be able to adjust the amp's gain so at the max 0.5 V input you have enough output, but aren't overdriving the amp.
Here's an example: Let's say you use one of those cheap TPA3116 boards that come with 36 dB gain. The max output, with a 4 ohm load and 10% THD (!) is 50 watts. This corresponds to an output voltage of about 14 Vrms. With 36 dB of gain gain (a gain factor of 63) the max input would be 222 mV. So you're not even using 1/2 of of the already limited 500 mV DAC output. So matching the amp gain is important if you want to minimize problems like hiss.
I would still recommend buying the board and get experience using SigmaStudio. The 1701 is by far the easiest Analog Device's DSP to program. It's a big step using the more capable devices like the 145x/6x series. That's my experience.
Here's an example: Let's say you use one of those cheap TPA3116 boards that come with 36 dB gain. The max output, with a 4 ohm load and 10% THD (!) is 50 watts. This corresponds to an output voltage of about 14 Vrms. With 36 dB of gain gain (a gain factor of 63) the max input would be 222 mV. So you're not even using 1/2 of of the already limited 500 mV DAC output. So matching the amp gain is important if you want to minimize problems like hiss.
I would still recommend buying the board and get experience using SigmaStudio. The 1701 is by far the easiest Analog Device's DSP to program. It's a big step using the more capable devices like the 145x/6x series. That's my experience.
@ernperkins Can you clarify where you got these 0.5V max numbers from? My unit most definitely performs about as the datasheet specifies, which is 0.9V RMS at full scale. That's a normal, above the -10dBV standard output voltage for a line level device like a DAC.
I added a balanced headphone amp to a 1701 output to create a balanced output drive for the main power amp. The headphone amp has its own negative rail supply converter, to get the headroom needed.
You are correct, the max DAC voltage is 0.9 V. I shouldn't trust my faulty memory. I would still contend that is low compared to many DACs whose outputs are 2 or 4 V. In the end it just means you need select an amplifier with the correct gain to match the 0.9 V max output.
I was also wrong when I stated the DAC's THD + Noise as -83 dB. that's for the ADCs. The DAC's THD + Noise is -90 dB. So in theory you could gain up to 7 dB in THD + Noise by bypassing the 1701 ADCs.
Yes, 0.9 Vrms is the max output. Here's a clarification:
You are correct, the max RMS DAC voltage is 0.9 V. I shouldn't trust my faulty memory. I would still contend that is low compared to many DACs whose outputs are 2 or 4 V RMS max. In the end it just means you need select an amplifier with the correct gain to match the 0.9 V RMS max output.
You are correct, the max RMS DAC voltage is 0.9 V. I shouldn't trust my faulty memory. I would still contend that is low compared to many DACs whose outputs are 2 or 4 V RMS max. In the end it just means you need select an amplifier with the correct gain to match the 0.9 V RMS max output.