Fine tuning filter stage for lower THD or noise probably does not yield significant or audible improvements as both SNR and distortion (THD & IMD) figures are already excellent. Much better than I anticipated and in fact at least on par with what I measure with ES9038Q2M. Chapeau, Marcel!
Some of the remaining SQ issues don't sound like they will show up very well on FFTs. Doesn't sound like HD, and doesn't sound like noise floor. Sounds a lot like some line level cables, and or like sometimes what is heard with DA. Of course, DA produces a linear distortion according the RC ladder network model, but one that arguably could time-smear music transients.
Regarding choice of op amps for best measurements, that may not be the same as the best choice for SQ.
A couple of interesting forum posts quotes:
"the OPA627 metal had a wire to the can too, so it was really shielded!
Also at this old days, a BB engineer told me, that the chip was build on a glass substrate, otherwise the better figures could not be fulfilled as on plastic."
Don't have a link for that one, only saved the text in a file. Attributed to @HpW
Second one is this:
"As the person who actually developed many of those audio parts listed in a previous post, I thought I'd chime in. When I was defining TI's audio products I released:
OPA1688, OPA1622, INA1650/1, INA1620, OPA1678/9, OPA1692, and had started the OPA1671, OPA1656 and OPA1637 before moving to lead our high speed amplifiers team. Another engineer released or is releasing those last 3.
The following products were developed 100% specifically for audio: OPA1622, INA1620 INA1650/1651, OPA1678/9, OPA1692, OPA1656. The only device with a similar variant sold for other applications is OPA1656 (OPA2156). But the input stage architecture of OPA2156 is actually different, and it is laser trimmed. The two devices do not share a mask set.
The products are not on an "audio specific" process flow. They use our high voltage SiGe process fab'd in Germany, which is also used by a number of other HV bipolar amplifiers.
I can honestly say we never revised a product due to the results of a listening test. We did however make a revision on a product when I found behavior on the Audio Precision I did not like.
OPA1688 and OPA1671 do have non-audio variants.
OPA1637 is an audio version of a very high precision product called THP210 which is about to be previewed on ti.com. While high precision applications drove the specs of that product, I will say that I kept audio applications in mind while defining the AC performance.
Also, despite what I've read here, all of these developments made financial sense for TI and we're quite happy with the investment
And no, the cost of an 8 inch mask set is not prohibitive. I would argue that transistor feature size is a much larger factor in mask cost, and we're not building op amps on <45nm CMOS here...
--
Without giving away details, 1611 has slightly better distortion but slightly less capacitive load drive vs OPA211. It was a tradeoff.
----------------------------
National did develop parts specifically for audio: the LME series. I spent time with the product manage for the line and heard the details of the development. However the hard reality is that they have mostly been killed off which supports somewhat the issue around business potential.
Today's audio specific parts with $$$ potential are amps with on board DSP's and Bluetooth SOC's. Even those are becoming commoditized killing the profit potential.
AKM does have a golden ear on staff (I have met him) but that is an outlier as well."
https://www.diyaudio.com/forums/the...lowtorch-preamplifier-iv-129.html#post6068063
Also seems to me from memory that one of the TI engineers noted that plastic op amp packages are different from metal cans in that the epoxy comes into contact with silicon and can mechanically stress it. IIRC the other thing someone pointed out is that the air gap in metal cans is better than contact with plastic.
Anyway, once built and active filter using OPA1612 that sounded good. There was no RFI/EMI coming into it and its just a good sounding op amp for audio. Some other op amps for whatever reasons are not recommended for audio by IC manufacturers. They don't really tell us everything they know about why that is.
Regarding choice of op amps for best measurements, that may not be the same as the best choice for SQ.
A couple of interesting forum posts quotes:
"the OPA627 metal had a wire to the can too, so it was really shielded!
Also at this old days, a BB engineer told me, that the chip was build on a glass substrate, otherwise the better figures could not be fulfilled as on plastic."
Don't have a link for that one, only saved the text in a file. Attributed to @HpW
Second one is this:
"As the person who actually developed many of those audio parts listed in a previous post, I thought I'd chime in. When I was defining TI's audio products I released:
OPA1688, OPA1622, INA1650/1, INA1620, OPA1678/9, OPA1692, and had started the OPA1671, OPA1656 and OPA1637 before moving to lead our high speed amplifiers team. Another engineer released or is releasing those last 3.
The following products were developed 100% specifically for audio: OPA1622, INA1620 INA1650/1651, OPA1678/9, OPA1692, OPA1656. The only device with a similar variant sold for other applications is OPA1656 (OPA2156). But the input stage architecture of OPA2156 is actually different, and it is laser trimmed. The two devices do not share a mask set.
The products are not on an "audio specific" process flow. They use our high voltage SiGe process fab'd in Germany, which is also used by a number of other HV bipolar amplifiers.
I can honestly say we never revised a product due to the results of a listening test. We did however make a revision on a product when I found behavior on the Audio Precision I did not like.
OPA1688 and OPA1671 do have non-audio variants.
OPA1637 is an audio version of a very high precision product called THP210 which is about to be previewed on ti.com. While high precision applications drove the specs of that product, I will say that I kept audio applications in mind while defining the AC performance.
Also, despite what I've read here, all of these developments made financial sense for TI and we're quite happy with the investment
And no, the cost of an 8 inch mask set is not prohibitive. I would argue that transistor feature size is a much larger factor in mask cost, and we're not building op amps on <45nm CMOS here...--
Without giving away details, 1611 has slightly better distortion but slightly less capacitive load drive vs OPA211. It was a tradeoff.
----------------------------
National did develop parts specifically for audio: the LME series. I spent time with the product manage for the line and heard the details of the development. However the hard reality is that they have mostly been killed off which supports somewhat the issue around business potential.
Today's audio specific parts with $$$ potential are amps with on board DSP's and Bluetooth SOC's. Even those are becoming commoditized killing the profit potential.
AKM does have a golden ear on staff (I have met him) but that is an outlier as well."
https://www.diyaudio.com/forums/the...lowtorch-preamplifier-iv-129.html#post6068063
Also seems to me from memory that one of the TI engineers noted that plastic op amp packages are different from metal cans in that the epoxy comes into contact with silicon and can mechanically stress it. IIRC the other thing someone pointed out is that the air gap in metal cans is better than contact with plastic.
Anyway, once built and active filter using OPA1612 that sounded good. There was no RFI/EMI coming into it and its just a good sounding op amp for audio. Some other op amps for whatever reasons are not recommended for audio by IC manufacturers. They don't really tell us everything they know about why that is.
Last edited:
I have used OPA1612 in many devices. E.g. my ES9038Q2M uses that for I/V and all my ADC have OPA1612s as input buffers. However I don't agree that it is audibly superior to e.g. OPA2210. Such sweeping statements are baseless as it depends on the circuit and implementation.
It is not possible to tell what shows or doesn't show in FFTs without seeing the FFT. So no need to repeat that ad nauseam.
It is not possible to tell what shows or doesn't show in FFTs without seeing the FFT. So no need to repeat that ad nauseam.
Merely pointing out that removing one op amp improved the sound. Widened the sound stage, among other things. Question is how/why?
Also, there is still a certain SQ difference between Andrea dac and Marcel dac. Question is why?
IMHO most likely the answers will turn out to be some collection of whatever isn't being measured very well.
Also, there is still a certain SQ difference between Andrea dac and Marcel dac. Question is why?
IMHO most likely the answers will turn out to be some collection of whatever isn't being measured very well.
You are entitled to your opinion but repeating it does not make it "the truth". And without measuring you cannot know if it can be measured.
Everything physical can in principle be measured. And everything is physical. So the question might arise as to what could be we be measuring that we aren't? How could we do some meaningful non-PSS measurements? Should be possible. To some extent it has already been discussed before. How about leveraging the Hilbert Transform for envelope detection of a pulsed or stepped amplitude sine wave? It can be done, I'm just not crazy about doing all the work myself. This is a hobby, not paid work.
Attachments
Hi Marcel,
When testing both your Dac’s, simulated distortion figures were quite comparable to the measurements, so in this case the rudimentary models are obviously not that bad to give a reasonable indication.
Hans
By the way, regarding noise simulations, Burkhard Vogel recently wrote a book about DIY op-amp noise macromodelling, Slopes and Levels, Spice Models to Simulate Vintage Op-Amp Noise. He started making his own noise models because he found that some of the models of the manufacturers were way off. I wrote a book review you can find here: https://audioxpress.com/article/boo...spice-models-to-simulate-vintage-op-amp-noise
Sometimes, I get this fuzzy feeling about these diy measuring instruments cobled from freebies/cheap stuff. At some stage after all these listening tests we may need to package everything and send it to ASR
Mark, please allow me to explain: the discovery was made after the original board was made and already ordered to be sent to me. I could have modified it before sending it to you but held back a bit just in case the mods on a known good board created other problems. So now, the results coming from you is an important validation of listening tests vs theoretically analysis/measurements
Get yourself a Cosmos E1DA (grade A) ADC and some appropriate software. The notch filter will also help. This will come pretty close to AP.
Thanks, very interesting but why bother when ASR can do it for free with AP. That’s what I was thinking when all improvements and listening tests to confirm completed, this could be the next stage. So, in this way we have two references, listening and measurement to make informed decisions.
Hi Terry,
Thx for the tip, very interesting piece of hardware, although what Bohrok made and the Picoscope + LNA that I used aren’t exactly inferior either.
@acko,
The problem here is that the RTZ Dac performance is dependent on the hardware splitting the original .dsf file in Left, Right and Bitclock.
It is so to say only half of a Dac, who’s potential is completely dependent on the splitter.
See the big differences in the various measurements so far, not because of the test equipment, but because of the different ways the incoming signal was made.
And looking at the figures, the combo with the Amanero seems to my opinion not to be the best choice.
Sending this Dac without .dsf splitter to ASR would be like sending a car without engine to a car test magazine.
Hans
Yes, agree we need to complete the system build first (USB/Spliter, clocks etc included)and test everything before sending out anywhere else. It just caught my attention when ESS sent their latest 9039PRO chips for evaluation and tests done on their EVB were using an AP machine. Why didn’t they use their own 9822 (grade A)ADC and some software?
Attachments
Last edited:
I'm all for sending the dac to ASR for testing and review. Maybe you could send your RTZ dac and Markw4 could send his Andrea Mori's dac.
Cosmos E1DA (as well as any ES9822PRO based ADC) suffers from "ESS IMD hump" and high order artefacts. This makes e.g. IMD sweeps of well measuring dacs useless. Here is some discussion about the issue:
https://www.diyaudio.com/community/threads/return-of-the-ess-hump.391267/
https://www.diyaudio.com/community/threads/ess-hump-strikes-back.394086/
Last edited: