This is a thread with everything discussing PCM58, and excellent chip which deserves it's own forum.
I want to resume some of my findings right now after months of juggling with many iterations of this chip in single ended dacs from China.
I will post some %%% in front of information which I could verify almost 100% by trial and error, oscilloscope and signal analysis.
%%% the signal must be 44, 48, 88 khz 16 or preferably 24 bits to the AK4118 receiver or it generates digital noise in the audio band. I don't believe any degradation from using higher resolution recording to downsample as the jitter/noise is the major quality factor here.
%%% the dac can be adjusted with trimpots, MSB, to 4th bit. It is a cumbersome process which doesn't necessarily translate to better audio quality, it is IMO necessary for the MSB and second bit. But further is no need, as the music is so complex it drowns any improvements and other factors are more important.
%%% Like the TDA1541, I find that 4X OS is the best by far, and 18bits only no less no more.
%%% In such a DAC the fidelity of the rendering is very subjective and imo not the reason why one would use such a DAC. I found futile to compare recordings/source (in DeltaWave) to assess the quality of the rendition. I did some visual checks of the waveforms screenshots vs originals and there is no visual differences. Mostly, one should focus on what details can be heard, and how much secondary voices (and complexity) are standing on their own or drowned, and how well the audible speech is rendered (are 'Rrrrs' realistic, Fss), and finally how deep and accurate is the soundstage.
more to come. And possible comparisons with PCM1701.
I want to resume some of my findings right now after months of juggling with many iterations of this chip in single ended dacs from China.
I will post some %%% in front of information which I could verify almost 100% by trial and error, oscilloscope and signal analysis.
%%% the signal must be 44, 48, 88 khz 16 or preferably 24 bits to the AK4118 receiver or it generates digital noise in the audio band. I don't believe any degradation from using higher resolution recording to downsample as the jitter/noise is the major quality factor here.
%%% the dac can be adjusted with trimpots, MSB, to 4th bit. It is a cumbersome process which doesn't necessarily translate to better audio quality, it is IMO necessary for the MSB and second bit. But further is no need, as the music is so complex it drowns any improvements and other factors are more important.
%%% Like the TDA1541, I find that 4X OS is the best by far, and 18bits only no less no more.
%%% In such a DAC the fidelity of the rendering is very subjective and imo not the reason why one would use such a DAC. I found futile to compare recordings/source (in DeltaWave) to assess the quality of the rendition. I did some visual checks of the waveforms screenshots vs originals and there is no visual differences. Mostly, one should focus on what details can be heard, and how much secondary voices (and complexity) are standing on their own or drowned, and how well the audible speech is rendered (are 'Rrrrs' realistic, Fss), and finally how deep and accurate is the soundstage.
more to come. And possible comparisons with PCM1701.
Last edited:
%%% Adjust the first bit (MSB) at -80dbv for lowest 2nd harmonics, then second bit to lower 3rd harmonics. It will give you the most dynamic sound and fidelity. The frequency can be 100hz but use a high resolution THD meter, or raise the frequency.
%%% The miracle of this chip is not in the single figures or Thd at one signal at -60db, realistically when you hit between -45/-55 db and there is no higher order it is ready to sing. If properly installed expect an harmonic at -110db and noise floor -130db when trying square waves.
%%% The most important is the digital filter
There are many options of digital filter for HIFI, DP1706, DP1704, SM5847. The SM5847 is by far the one to use. The DP1704 will adjust maybe to lower THD and give incredible dynamics. But to my ear the SM5847 was leagues better. It excels at square-waves forms dynamics at low frequency. The DP1704 has 70db between noise floor and peaks if you start the square at say 30hz. When you use the SM5847 the noise floor is 90db below (20db improvement), and wins at the sound stage and everything, I am lacking the words. (note: that could be attributed to the improved 5V regulator as well but I doubt it, if someone can obtain the same dynamics with another digital filter please let me know).
This is just personal notes/ personal preferences:
IV, the schematic recommends 3k feedback resistor for opamps such as NE5532, if you use modern ones your results may vary. You can lower the resistor to 1.5K and use a 2.2nf filter cap instead of 1.1nf. I find that OPA604 was the best sounding I/V*.
buffer filter stage: I liked the musicality of the AD8597**.
,*opamps tested thoroughly in all positions and combinations (4x of each) : ad8597, op134, op1611, NE5532 modern and vintage, note that the ad8610, op610, 627, 735, 797(which is as good, just sold it), were tried before choosing the ad8597 with many different dacs, (Tda1541, CS4397, AK4358).
%%% The miracle of this chip is not in the single figures or Thd at one signal at -60db, realistically when you hit between -45/-55 db and there is no higher order it is ready to sing. If properly installed expect an harmonic at -110db and noise floor -130db when trying square waves.
%%% The most important is the digital filter
There are many options of digital filter for HIFI, DP1706, DP1704, SM5847. The SM5847 is by far the one to use. The DP1704 will adjust maybe to lower THD and give incredible dynamics. But to my ear the SM5847 was leagues better. It excels at square-waves forms dynamics at low frequency. The DP1704 has 70db between noise floor and peaks if you start the square at say 30hz. When you use the SM5847 the noise floor is 90db below (20db improvement), and wins at the sound stage and everything, I am lacking the words. (note: that could be attributed to the improved 5V regulator as well but I doubt it, if someone can obtain the same dynamics with another digital filter please let me know).
This is just personal notes/ personal preferences:
IV, the schematic recommends 3k feedback resistor for opamps such as NE5532, if you use modern ones your results may vary. You can lower the resistor to 1.5K and use a 2.2nf filter cap instead of 1.1nf. I find that OPA604 was the best sounding I/V*.
buffer filter stage: I liked the musicality of the AD8597**.
,*opamps tested thoroughly in all positions and combinations (4x of each) : ad8597, op134, op1611, NE5532 modern and vintage, note that the ad8610, op610, 627, 735, 797(which is as good, just sold it), were tried before choosing the ad8597 with many different dacs, (Tda1541, CS4397, AK4358).
Last edited:
Hi,
I like the AD8597/99 in DEDICATED buffer or sligthy gain position as well (after a dedicated I/V). However it is too slow as I/V (bandwidth, slew rate- even with 16 bits only) and current noisy on its input for that task.
IME the digital filter is not the most important and NOS can give very good results (I prefer it to any bad filters). Oversampling can be made by some source, after x2 I can't hear the difference while I still can hear any caps or resistor swaps. passive filter is your best bet. Layout is important I don't loose too much time pn those chineese pcbs but to get that chip.
YMMV.
I like the AD8597/99 in DEDICATED buffer or sligthy gain position as well (after a dedicated I/V). However it is too slow as I/V (bandwidth, slew rate- even with 16 bits only) and current noisy on its input for that task.
IME the digital filter is not the most important and NOS can give very good results (I prefer it to any bad filters). Oversampling can be made by some source, after x2 I can't hear the difference while I still can hear any caps or resistor swaps. passive filter is your best bet. Layout is important I don't loose too much time pn those chineese pcbs but to get that chip.
YMMV.
Thanks Diyiggy!
I am responding to what you said.
In my schematic the buffer opamp is between the filter and the output, it is a classic filter. It is unity gain. I am not sure why you put in capital 'dedicated' ? Correct observation it is not the best for i/v.
I know that you can send a signal which is already in high bit rate , like 88khz? is this what you mention by x2?
but is this signal 18 bits? I noticed a lot of big anomalies if feeding the dac with more than 18 bits. The datasheet mention that bits simply get discarded but it is not true in practice unless I am missing out something. I was feeding the Filter with 24 bit then sent it as 24, 22, 20, and finally 18 bits. It played music no problem but on test signals I had extreme artifacts showing up. Only 18bits were suitable.
The digital filter and OS is up to your preference. I am simply mentioning that for me it was the most important factor, more than the chip, regulators, opamps, all together combined. And I could measure the effect on approximation of square waves starting at lower Fq.
The filter is passive, but for the impedance and isolating it from the output the 'buffer' opamp is placed there it also does a job with a feedback capacitor.
Maybe it is the regulator which is affecting the dynamic contrast at lower frequencies and the digital filter is for nothing.
Do you agree this would need more research, I don't want to remove the digital filters because these are already shaky because I redid the receiving module and broke some legs then solder new ones. I corrected a link in the PCB too with the input of the digital filter. So I am not sure I want to break more legs of a 100$ chip and spend and hour fixing it!
What I have in mind instead is to purchase a third pcm58 board then try it with the SM5847 filter and stock regulator. Then insert 1701 dac chip to see if there is any improvement. It is a direct substitute.
I am responding to what you said.
In my schematic the buffer opamp is between the filter and the output, it is a classic filter. It is unity gain. I am not sure why you put in capital 'dedicated' ? Correct observation it is not the best for i/v.
I know that you can send a signal which is already in high bit rate , like 88khz? is this what you mention by x2?
but is this signal 18 bits? I noticed a lot of big anomalies if feeding the dac with more than 18 bits. The datasheet mention that bits simply get discarded but it is not true in practice unless I am missing out something. I was feeding the Filter with 24 bit then sent it as 24, 22, 20, and finally 18 bits. It played music no problem but on test signals I had extreme artifacts showing up. Only 18bits were suitable.
The digital filter and OS is up to your preference. I am simply mentioning that for me it was the most important factor, more than the chip, regulators, opamps, all together combined. And I could measure the effect on approximation of square waves starting at lower Fq.
The filter is passive, but for the impedance and isolating it from the output the 'buffer' opamp is placed there it also does a job with a feedback capacitor.
Maybe it is the regulator which is affecting the dynamic contrast at lower frequencies and the digital filter is for nothing.
Do you agree this would need more research, I don't want to remove the digital filters because these are already shaky because I redid the receiving module and broke some legs then solder new ones. I corrected a link in the PCB too with the input of the digital filter. So I am not sure I want to break more legs of a 100$ chip and spend and hour fixing it!
What I have in mind instead is to purchase a third pcm58 board then try it with the SM5847 filter and stock regulator. Then insert 1701 dac chip to see if there is any improvement. It is a direct substitute.
yup x2, x4, etc = sampling rate.
So what is the op amp used for the I/V ?
Would be interresting to know why they putted a cap in the feedback of the AD8597 used as buffer ? to color more and add a -6 dB. Or it is simply a bad design stock 😉
Nah the pcm58 is not that expensive ??? Or take two AD1862.... There are people whom sold them in USA on Ebay iirc.
So what is the op amp used for the I/V ?
Would be interresting to know why they putted a cap in the feedback of the AD8597 used as buffer ? to color more and add a -6 dB. Or it is simply a bad design stock 😉
Nah the pcm58 is not that expensive ??? Or take two AD1862.... There are people whom sold them in USA on Ebay iirc.
The cap is standard for filters, I can post schematic if needed. The -3db is around 34500 hz (measured with scope and computer browser signal generator) . I have a 700khz amp right after so I cant take any chances with the filter, and it should sound better as it reconstruct the wave form , they look perfect both at the end and at the input of the i/v opamp, checked with 70mhz scope. I let the system run 24/7. I have Zero noise even at 100% volume, everything way under 100db unweighted.
I/V : I find that Op604 was just perfect. I am happy using vintage ne5532 as well. They didn't made much difference on performance, just the dac is a little happier with higher slew rates but works with ne5532 as well, maybe you lose 0.5 bit of resolution, nothing to raise alarm. The very fast and revealing opams didn't necessarily sounded better which was a big surprise for me, I expected to rank them 1611, 604, 134, 8597 and family , ne55, my ears decided otherwise after many swaps.
Not the pcm, the digital filter is costly, I would not worry for the quality of the pcm chip as it can be adjusted.
I/V : I find that Op604 was just perfect. I am happy using vintage ne5532 as well. They didn't made much difference on performance, just the dac is a little happier with higher slew rates but works with ne5532 as well, maybe you lose 0.5 bit of resolution, nothing to raise alarm. The very fast and revealing opams didn't necessarily sounded better which was a big surprise for me, I expected to rank them 1611, 604, 134, 8597 and family , ne55, my ears decided otherwise after many swaps.
Not the pcm, the digital filter is costly, I would not worry for the quality of the pcm chip as it can be adjusted.
Last edited:
I try to readjust with the 4 trimpots at different volumes, the problem is to get it right at all volume levels.
I have around -52 db 4th, and 1-3 harmonics are -56.... AT max volume this gives around -105db , which is exceptional and can only be attained with the trimpots.
I have around -52 db 4th, and 1-3 harmonics are -56.... AT max volume this gives around -105db , which is exceptional and can only be attained with the trimpots.
To make the fast op amps sounding good, you must work on the HF attenuation at the output when needed, but you do want higher bandwidth enough for the dac chip. It's also about the power supplly and the pcb layout/passive parts. Yppu see too much fast op amps on so-so pcb layout. That's a chineese cheap made fast pcb, that's why you experienced better sound with the slowish op amps I surmise. You do some attenuation at the output and the input with the filter but with those pcbs and poor inductance of the tht decoupling for HF, it is maybe not enough for the fast op amps.
I realised that and now do my pcbs to push the thing higher and very happy with fast op amps for I/V and if needed add "slowish" buffer to help the I/V stage. It is not always needed according the I/v op amp, while the buffer stage can voice the sound too... trade off trade off !
https://www.diyaudio.com/community/attachments/20250413_180117-jpg.1449188/
I realised that and now do my pcbs to push the thing higher and very happy with fast op amps for I/V and if needed add "slowish" buffer to help the I/V stage. It is not always needed according the I/v op amp, while the buffer stage can voice the sound too... trade off trade off !
https://www.diyaudio.com/community/attachments/20250413_180117-jpg.1449188/
I will take this into consideration...
On the good sounding board I removed the 3.3R resistors in the power lines which feeds the dac pins before the decoupling capacitor of the DAC. I reduced the values of capacitors and installed a lot of 0.47uf G-NX HiQ, (for the opamp as well).
I read the building notes of the DAC and at deployment time no Opamp could get the 17 bit of data right, 16 at most.
I calculated the dac needed at minimum 8Mhz Opamp to get the resolution right. And that is the input filter they recommend just at the DAC I out leg in // with feedback capacitor. The 200ns settling time is what the designer talk about and said no opamp back then can match this speed, its 5Mhz, I don't know the exact math. They recommend 3.3k feedback resistor. I used 1.5k, maybe I will remove it all to see.
I saw they sell opamp adaptors with integrated bypass capacitors...
On the good sounding board I removed the 3.3R resistors in the power lines which feeds the dac pins before the decoupling capacitor of the DAC. I reduced the values of capacitors and installed a lot of 0.47uf G-NX HiQ, (for the opamp as well).
I read the building notes of the DAC and at deployment time no Opamp could get the 17 bit of data right, 16 at most.
I calculated the dac needed at minimum 8Mhz Opamp to get the resolution right. And that is the input filter they recommend just at the DAC I out leg in // with feedback capacitor. The 200ns settling time is what the designer talk about and said no opamp back then can match this speed, its 5Mhz, I don't know the exact math. They recommend 3.3k feedback resistor. I used 1.5k, maybe I will remove it all to see.
I saw they sell opamp adaptors with integrated bypass capacitors...
It doesn't sound bad like in the other dac with 1611. In this one it sounds different but I can recognize a little the qualities of the opamp. However it didn't wow me. It was good/ok. I placed back the opa604 and wow again.
Maybe the op604 is simply better sounding for PCM58 i/v ?
I could place again the op604 in the other dac to see if I hear the same quality.
Maybe the op604 is simply better sounding for PCM58 i/v ?
I could place again the op604 in the other dac to see if I hear the same quality.
Last edited:
I placed the same opamps in the the other dac. Just good sound, then played the other one, big spacious sound, deep soundstage, details that goes on for ever, detailed bass, no shrill voices, just details, textures.