As I don't have a profound knowledge about phono preamp circuitry (yet, may or may not change), that compact design is indeed appealing.
Would love to see performance characteristics and comparisons to the OPA1656 design you're working on !
Hi 1093i3511
I'm also very curious to compare these two versions.
At the moment I'm waiting to receive the prototypes of the MarcelvdG version. That will be the first one I can try.
I still have some difficulty with the availability of two smd components for the version with OPA2192, but I should solve it soon.
The last component I'm missing is the transformer.
And unlike what I said previously, I think I'll will buy the Talema 2x15V 10VA version.
Powering only the Preamp stage, I suppose less than 10% of the 333 mA it can deliver will be needed.
With the advantage of a lower cost and smaller size, compared to a 15-25VA.
I know, I did it all by myself...🙂
I'm also very curious to compare these two versions.
At the moment I'm waiting to receive the prototypes of the MarcelvdG version. That will be the first one I can try.
I still have some difficulty with the availability of two smd components for the version with OPA2192, but I should solve it soon.
The last component I'm missing is the transformer.
And unlike what I said previously, I think I'll will buy the Talema 2x15V 10VA version.
Powering only the Preamp stage, I suppose less than 10% of the 333 mA it can deliver will be needed.
With the advantage of a lower cost and smaller size, compared to a 15-25VA.
I know, I did it all by myself...🙂
Was also thinking about using an opa2192 in a preamp, any special quirks for these new tangled CMOS chips?
Hi Tashi1,
Sorry, but I don't understand what you mean with:"any special quirks for these new tangled CMOS chips".
About the RIAA Pre with OPA2192, it is a design that was suggested by Nick Sukhov and it is an op-amp that I have never tested.
I should start printing this PCB in the next few days, so for now I can't make any comments.
About the RIAA Pre designed by MarcelvdG with High Pass Filter I have already received the pcbs, but I am waiting for the transformers for the power stage. Within the next week I should be able to put everything together and perform the first "raw", tests.
I will update the post as soon as I see that it works.
(my PCB, not the MarcelvdG Design...)
Sorry, but I don't understand what you mean with:"any special quirks for these new tangled CMOS chips".
About the RIAA Pre with OPA2192, it is a design that was suggested by Nick Sukhov and it is an op-amp that I have never tested.
I should start printing this PCB in the next few days, so for now I can't make any comments.
About the RIAA Pre designed by MarcelvdG with High Pass Filter I have already received the pcbs, but I am waiting for the transformers for the power stage. Within the next week I should be able to put everything together and perform the first "raw", tests.
I will update the post as soon as I see that it works.
(my PCB, not the MarcelvdG Design...)
OPAx192 is nice but not really a universal audio opamp: https://www.diyaudio.com/community/...really-not-a-good-choice.428152/#post-8023192
Gotcha, be interested to hear how it goes.
I was just referring to the fact that using CMOS opamps for audio seems to be a fairly recent thing (I could be wrong), and the OPAx192 is chopper stabilised which means it uses some strange tricks to achieve its remarkable offset at such a cheap price.
Low offset isn't a requirement for audio (which is AC coupled), and chopper-stabilization creates spectral artifacts which might be audible. If you were monitoring a thermocouple however it would be great choice.
CMOS opamps have only come down in 1/f noise in the last few decades, and still lag behind JFET and bipolar in this respect. They usually have good bandwidth and output drive capability.
CMOS opamps have only come down in 1/f noise in the last few decades, and still lag behind JFET and bipolar in this respect. They usually have good bandwidth and output drive capability.
No, it is factory-trimmed for low offset, similar to the expensive laser resistor trimming just very much cheaper. The process is called e-Trim: https://www.ti.com/lit/ab/sbot037d/sbot037d.pdf
The non- e-Trimmed version of the OPA192 (still with low offset) is the OPA197 at basically half the price.
Chopper-stabilized introduces no 1/f-noise which is the other main asset besides low offset.
While the OPA192/197 is exceptional at "true DC" it has tons of 1/f noise and thus is very drifty near DC. The JFET-based opamps like 1642 and 827/828 are much better, and the latter have low offset, too and all have good modern specs overall (what you mentioned already).
You’re absolutely right, apologies for the confusion, I was thinking of the OPA189
Last edited:
OPA192/197 are not a chopper and have no more 1/f noise than typical JFET-input OPA627\1655
Yes, see my post above.
I was referring to 828 or 1642, not CMOS 1655 or grandfather's 627.
OPA828: 60nV/rtHz @ 0.1Hz
OPA192: 300nV/rtHz @ 0.1Hz
OPA828: 0.34uVpp
OPA192: 1.3uVpp
OPA1642 is even better with 50nV/rtHz @ 0.1Hz and ~0.3uVpp:
Nick, can you explain what is a "transadmittance cascade at the output" and show some examples with schematics please,
Last edited:
Regarding 1/f voltage noise: if it is the RIAA- and A-weighted total noise you are interested in, and the voltage noise is partly white and partly 1/f, what matters is the noise density at about 1162 Hz. For a given noise voltage density at 1162 Hz, you end up with the same total weighted noise no matter what part is white and what part 1/f. For RIAA- and ITU-R 468 weighting, it's 2897 Hz.
The noise weighting curves drop faster with decreasing frequency than the RIAA playback curve increases.
The noise weighting curves drop faster with decreasing frequency than the RIAA playback curve increases.
See Marcel`s post #136 (plus keep in mind a 7950 us tau4 https://www.patreon.com/posts/vazhno-po-povodu-97802586 ). And also 1655 is FET-input, not CMOS
Last edited:
Yes, see here [ https://imrad-com-ua.translate.goog..._sl=ru&_x_tr_tl=en&_x_tr_hl=ru&_x_tr_pto=wapp ] and here Transadmittance (or transfer admittance) is the AC equivalent of transconductance [ https://en.wikipedia.org/wiki/Transconductance ] and here [ https://www.patreon.com/posts/otsenite-tekst-90783025 ]
The OPA1655 is the single version of the OPA1656, which is a CMOS op-amp according to https://www.diyaudio.com/community/...ormance-cmos-audio-op-amp.335416/post-5732359 The datasheet only says "FET", but then again, a MOSFET is also a FET.
I don't want to interrupt this analysis of the specifications related to the context (which is interesting), but I would like to share one of the next steps.
We are at the point where we have 10 PCBs of Marcel's Pre RIAA circuit, for which we purchased the necessary 40 Wima MKS2 6.8 uF capacitors.
The SMD components for both Pre RIAA circuits (Marcel's and Nick's) were chosen with the tightest possible tolerances, 0.1% for the resistors (all Thin Film) and I think 2% for almost all the capacitors (always all NP0).
So, in order to maintain the best possible precision on ALL the available boards, I started by measuring all the capacitors.
You can see them in the photo below, already divided by value, from the highest to the lowest.
The measurement was made with a LCR DER EE DE-5000, which showed up to the third decimal place, but with an unstable value, if repeated the measurement two or three times.
So all the values reported are approximated to the (stable) second closest decimal place.
The green signs represent pairs of values as close as possible, for the right and left channel. As you can see, the most unbalanced pair (L/R) is 23-24, because they are extremes of the central value (the nominal 6.8 uF).
However 6.94uF is about 2% more than the nominal value, while 6.73uF is about -1%. The difference between the two is therefore 3%.
It is probably still a better match, than choosing randomly two capacitors from this lot.
Should something else be considered?
Indeed, yes.
I remembered a short comment by Marcel, in post #73.
"C6 and C16, in your schematic, are more critical than C7 and C17, so you could use the more accurate ones for C6 and C16."
At that time Marcel was referring to the schematic version 0.1, published in post #65, then replaced by the schematic 0.1b, published in post #84, where the capacitors became respectively, C17-C18 (more critical), C19-C20 (less critical).
In order to get a good average precision on all the pcbs, my neuron suggests to choose a precise pair for C17-C18 (for example 21-22, or 25-26, and a pair on the opposite end (or adjacent?) for C19-C20, for example 1-2, or 39-40.
I would continue like this for the rest of the pairings, always trying to use the most precise pairs for C17-C18.
I haven't tried all the divisions yet, but I hope we can keep the unlucky pair 23-24 for the C19-C20 position.
Does this seem like a correct reasoning to you?
We are at the point where we have 10 PCBs of Marcel's Pre RIAA circuit, for which we purchased the necessary 40 Wima MKS2 6.8 uF capacitors.
The SMD components for both Pre RIAA circuits (Marcel's and Nick's) were chosen with the tightest possible tolerances, 0.1% for the resistors (all Thin Film) and I think 2% for almost all the capacitors (always all NP0).
So, in order to maintain the best possible precision on ALL the available boards, I started by measuring all the capacitors.
You can see them in the photo below, already divided by value, from the highest to the lowest.
The measurement was made with a LCR DER EE DE-5000, which showed up to the third decimal place, but with an unstable value, if repeated the measurement two or three times.
So all the values reported are approximated to the (stable) second closest decimal place.
The green signs represent pairs of values as close as possible, for the right and left channel. As you can see, the most unbalanced pair (L/R) is 23-24, because they are extremes of the central value (the nominal 6.8 uF).
However 6.94uF is about 2% more than the nominal value, while 6.73uF is about -1%. The difference between the two is therefore 3%.
It is probably still a better match, than choosing randomly two capacitors from this lot.
Should something else be considered?
Indeed, yes.
I remembered a short comment by Marcel, in post #73.
"C6 and C16, in your schematic, are more critical than C7 and C17, so you could use the more accurate ones for C6 and C16."
At that time Marcel was referring to the schematic version 0.1, published in post #65, then replaced by the schematic 0.1b, published in post #84, where the capacitors became respectively, C17-C18 (more critical), C19-C20 (less critical).
In order to get a good average precision on all the pcbs, my neuron suggests to choose a precise pair for C17-C18 (for example 21-22, or 25-26, and a pair on the opposite end (or adjacent?) for C19-C20, for example 1-2, or 39-40.
I would continue like this for the rest of the pairings, always trying to use the most precise pairs for C17-C18.
I haven't tried all the divisions yet, but I hope we can keep the unlucky pair 23-24 for the C19-C20 position.
Does this seem like a correct reasoning to you?