I would like to conclude this topic on the impedance of an average MM generator as follows:
The validity range of this regression analysis from six to five support points extends over two decades. These approximations are based on my old AT420E/OCC. I have taken the values directly from the table of values for an equivalent voltage source:
0,1kHz
397,7 Ohm + j 327 Ohm
10kHz
7k824 Ohm + j 21k564 Ohm
This is the source impedance that our input of the MM-EQ sees - in terms of magnitude |0k515 Ohm| to |22k94 Ohm| ! For Marcel, however, we would now have to use the power source model - the current.
Disclaimer
All errors are due to my manual process and my attention span.
Regards,
HBt.
- Rp(f) = 1,0589*10^-2 [sec/kOhm] * f^0,94012 [1/sec] & Lp(f) = 3,3862 [^sec/H] * f^-0,2359 [1/sec]
- Rs(f) = 8,4975*10^-4 [sec/kOhm] * f [1/sec] + 0,1264 [kOhm] & Ls(f) = 0,4994 [sec/H] * e^(-2,5993*10^-5 * f) [1/sec]
The validity range of this regression analysis from six to five support points extends over two decades. These approximations are based on my old AT420E/OCC. I have taken the values directly from the table of values for an equivalent voltage source:
0,1kHz
397,7 Ohm + j 327 Ohm
10kHz
7k824 Ohm + j 21k564 Ohm
This is the source impedance that our input of the MM-EQ sees - in terms of magnitude |0k515 Ohm| to |22k94 Ohm| ! For Marcel, however, we would now have to use the power source model - the current.
Disclaimer
All errors are due to my manual process and my attention span.
Regards,
HBt.
Last edited:
Caution
The MM generator is not quite as easy to approximate as Marcel conveniently does. Mechanically, dynamically and electrically the tactile transducer is a very interesting system, in combination with the tone arm and the accelerations, it becomes really exciting.
If we want to be precise, without counting peas, there needs to be a bit more input from the experts and authors.
We are far too superficial when it comes to the mechanism of the noise phenomenon alone. But you have to leave the church in the village - and you can make simplifications and assumptions and shift sources and summaries with impunity. Incidentally, this is standard practice here. And that's the end of this excursion and maneuver for me personally for the time being.
Who dares to try the overkill variant? Only 19 transistors are to be processed per channel - BC550C and BC560C, preferably the real ones, old ones from a pool. Selected for max hfe and pairwise similarity with minimal deviation.
The subsequent measurement and auditory check will show us whether noise of all types and colors was a reject criterion.
Questioning and checking is the correct procedure.
Regards,
HBt.
PS
We can discuss the physical phenomenon of "Schrotrauschen" (shot noise) further elsewhere - the correlations and causes.
The MM generator is not quite as easy to approximate as Marcel conveniently does. Mechanically, dynamically and electrically the tactile transducer is a very interesting system, in combination with the tone arm and the accelerations, it becomes really exciting.
If we want to be precise, without counting peas, there needs to be a bit more input from the experts and authors.
We are far too superficial when it comes to the mechanism of the noise phenomenon alone. But you have to leave the church in the village - and you can make simplifications and assumptions and shift sources and summaries with impunity. Incidentally, this is standard practice here. And that's the end of this excursion and maneuver for me personally for the time being.
Who dares to try the overkill variant? Only 19 transistors are to be processed per channel - BC550C and BC560C, preferably the real ones, old ones from a pool. Selected for max hfe and pairwise similarity with minimal deviation.
The subsequent measurement and auditory check will show us whether noise of all types and colors was a reject criterion.
Questioning and checking is the correct procedure.
Regards,
HBt.
PS
We can discuss the physical phenomenon of "Schrotrauschen" (shot noise) further elsewhere - the correlations and causes.
If anyone would like to have the DATS V3 raw data, I would be very happy to send it to them - I haven't looked at it in detail myself, but I'll do that in a minute.
Just a thought, it is an electromagnetic transducer, shouldn't a loudspeaker expert be able to model a good and valid equivalent circuit diagram of the transducer, from 30Hz to 15kHz ?
Dear Chestnutspread,
how are you doing today - are you able to sort out your thread and the project and categorize it for yourself?
I hope we can meet in real life one day and have a comprehensive exchange.
Best wishes and get well soon,
HBt.
PS
I use to overview the AT420E because you told us you're a DATS user. That is the only reason - quick and dirty.
Just a thought, it is an electromagnetic transducer, shouldn't a loudspeaker expert be able to model a good and valid equivalent circuit diagram of the transducer, from 30Hz to 15kHz ?
Dear Chestnutspread,
how are you doing today - are you able to sort out your thread and the project and categorize it for yourself?
I hope we can meet in real life one day and have a comprehensive exchange.
Best wishes and get well soon,
HBt.
PS
I use to overview the AT420E because you told us you're a DATS user. That is the only reason - quick and dirty.
The self-noise of the generator (tactile transducer) is, with good will, expressed as SNR and fully weighted, a maximum of |81dB|. This means for our preamplifier, if we were to regard it as a measuring amplifier and our ears were to analyze the sampled record - that its own SNR must be at least 3 to 10 times better, i.e. -91dB to -101dB.
And here, unfortunately, the simplest equivalent network is still the basis. So, the usual one.
The situation can now be improved slightly with an active electronic load situation.
But what is this "electrically cooled 47k load" really all about? And can we use it in the OREAD-EQ?
Let's come back to the hypothetical SNR of the original SUPRA, better than |86dB|. Since I own an original setup, I could check this claim once - but not this year. Unfortunately, I know from experience with this EQ that it is practically noiseless. With the current OREAD it is not known, and it also follows a slightly different concept than the SUPRA - similarities and prejudices just because IC in the input stage was chosen really large (from 2*500µA) and these were spontaneously divided into 2 times 5 times 100µA are not allowed.
Measurements must be made here, regardless of whether they confirm or refute the theoretical considerations. The resulting sources simply cannot correlate, so it is not possible to form a simple sum as a large node - comparable with just one transistor.
There's no getting around a metrological check, no matter how many specialist attachments are pulled out of the briefcases.
If they confirm one item or another, so much the better. But you can't let them spoil the attempt.
Best wishes,
HBt.
And here, unfortunately, the simplest equivalent network is still the basis. So, the usual one.
The situation can now be improved slightly with an active electronic load situation.
But what is this "electrically cooled 47k load" really all about? And can we use it in the OREAD-EQ?
Let's come back to the hypothetical SNR of the original SUPRA, better than |86dB|. Since I own an original setup, I could check this claim once - but not this year. Unfortunately, I know from experience with this EQ that it is practically noiseless. With the current OREAD it is not known, and it also follows a slightly different concept than the SUPRA - similarities and prejudices just because IC in the input stage was chosen really large (from 2*500µA) and these were spontaneously divided into 2 times 5 times 100µA are not allowed.
Measurements must be made here, regardless of whether they confirm or refute the theoretical considerations. The resulting sources simply cannot correlate, so it is not possible to form a simple sum as a large node - comparable with just one transistor.
There's no getting around a metrological check, no matter how many specialist attachments are pulled out of the briefcases.
If they confirm one item or another, so much the better. But you can't let them spoil the attempt.
Best wishes,
HBt.
I'm alive and well doing. Just in progress of backward engineering what happened in this thread. I'm missing a good week and it kind of exploded. I see here some valid points marked out especially regarding the topic about design philosophies for lowest noise. That's really interesting, because there is a lot of stuff I simply wasn't aware of and need to be considered for a proper result.
Feeling thrown back into my end-apprenticeship days with getting into the details of pnp and npn simple from the datasheets and making the design choices by the hard readable diagrams!
Regarding the source, I own a MC AT-F7 and the MM Ortofon OMB10. For this phono pre my only data is coming from the Ortofon which is doing a very good job especially for not so well LP pressings.
Nothing fancy:
Output voltage: 4mV/1kHz
DC Resistance :1kOhm
Inductance:580mH
Recommended Amp-Resistance:47kOhm
Recommended Amp-Capacity:200-500pF
I'm not sure if going this route, the maximum matched pairs are really the main criteria. Surely there is an "offset" between both channels but I think we do have some tolerances that are acceptable.
I just checked with a German Onlineshop that the BC560 costs 5ct so a bulk of 100 ain't the issue.
On the other hand - is there a better option?
I have to admit that I never selected the transistors on the three SUPRAs I've built - out of the bag, directly into the PCB. There were never any problems in these three cases. All static measurements, the potentials, were fine. Due to today's impracticalities, I would at least determine hfe and then do some sorting.
But about your MC, that's a real pleasure, you have an absolute insider tip with this AT
. I love the OC9-Series and the workinghorse Denons DL-103.
With the EQ, it's exactly the suggested topology that excites me, nothing more. The chosen components and operating points are (within limits!) completely free, and this thread is actually intended as a collection point for exactly that, isn't it?
Other suggestions regarding component selection and dimensioning are very welcome and desired, from every corner of our planet.
Just mentioning the jFet at the beginning is not enough collaboration on the project for me personally. It's not art and it's not constructive if we have already decided on a complete BJT story. We all know that it also works with a jFet pair at the input - whether the topology with five active components is then also suitable for AllInOne equalization is another matter for the time being.
The exciting topic is precisely the concrete exchange, the suggestions of other designers - in this case, the topology.
And simply being brave enough to try out the overkill should also have a certain appeal. Maybe it works well, maybe not ---> the learning curve!
But about your MC, that's a real pleasure, you have an absolute insider tip with this AT
. I love the OC9-Series and the workinghorse Denons DL-103.With the EQ, it's exactly the suggested topology that excites me, nothing more. The chosen components and operating points are (within limits!) completely free, and this thread is actually intended as a collection point for exactly that, isn't it?
Other suggestions regarding component selection and dimensioning are very welcome and desired, from every corner of our planet.
Just mentioning the jFet at the beginning is not enough collaboration on the project for me personally. It's not art and it's not constructive if we have already decided on a complete BJT story. We all know that it also works with a jFet pair at the input - whether the topology with five active components is then also suitable for AllInOne equalization is another matter for the time being.
The exciting topic is precisely the concrete exchange, the suggestions of other designers - in this case, the topology.
And simply being brave enough to try out the overkill should also have a certain appeal. Maybe it works well, maybe not ---> the learning curve!
It depands -
this is the trill
The topology is fixed at the moment, or in Question?!
Is your MM a good system? It kind of reminds me of disco days gone by.
Why don't you clamp on DATS V3, set the sweep level to -6dBu and write an impedance curve, please send me the data set - I find this system very strange.
Sorry, especially because you also have a Mercedes in the garage in comparison, the AT-MC.
But it's fun- not an HighEnd-Project ...
Why don't you clamp on DATS V3, set the sweep level to -6dBu and write an impedance curve, please send me the data set - I find this system very strange.
Sorry, especially because you also have a Mercedes in the garage in comparison, the AT-MC.
But it's fun- not an HighEnd-Project ...
If we bring the project to a successful conclusion, I will lay a "test record", which we will then use to get to the bottom of the overall system, starting from the source.
I'm already looking forward to the final measurements - tangible evidence of the reality at hand.
Greetings,
HBt.
Psst
Have I already told about my late father-in-law, a real "Deutsche Grammophon" (in Hanover's Podbielski Street) man? No, ok, that's a clear case for our buddy Haudegen ..!
I'm already looking forward to the final measurements - tangible evidence of the reality at hand.
Greetings,
HBt.
Psst
Have I already told about my late father-in-law, a real "Deutsche Grammophon" (in Hanover's Podbielski Street) man? No, ok, that's a clear case for our buddy Haudegen ..!
Last edited:
BC560C vs. 2SCxxxx
In his famous article, Marcel suggests the 2SC2545,46,47 series as a low-noise transistor. By the way, his test circuit for the article is really cleverly designed.
The question now is, would this Hitachi BJT be our transistor of choice in the first stage of the OREAD? Our open loop gain of the two-stage must achieve at least 80dB. Nick Sukov's EQ is a three-stage one, is that correct?
In his famous article, Marcel suggests the 2SC2545,46,47 series as a low-noise transistor. By the way, his test circuit for the article is really cleverly designed.
The question now is, would this Hitachi BJT be our transistor of choice in the first stage of the OREAD? Our open loop gain of the two-stage must achieve at least 80dB. Nick Sukov's EQ is a three-stage one, is that correct?
Last edited:
I wouldn't call it any good - it's working is the better term. It remained in a box beside the turntable for the possibility the AT-F7 fails. For MM I would go further in the Audiotechnica world and would try the famous Shibata AT-VM95SH.
For the moment my analogue enjoyment is amplified by the Ifi Zen Phono on battery supply (thankfully it's running on simple 5V/>500mA!)
We are talking about OVA2637 OPs with theoretically distortion at 0.0001%.
The THD is about -110db according to the datasheet.
Never opened the box to check on the SMDs but I guess it's a comparable view as on all the small form factor phono pre you can find out there.
I do have a test record from the German mag "Stereophile"/Stockfish Records with a couple of nice test-tracks:
Simply the best tool I got on hand for a quick test.
Regarding tests with DATS, well I think that needs to wait till the holiday season is over and the table is allowed for technical usage again.
Just need to check that everything is well wired and the test should be easily done - can be a bit flimsy with the crocodile clamps on the test-probes
- testtone 1kHz 0db
- pink noise -6db mono
- pink noise -6db Out of phase
- pink noise -6db left/right
- 3150Hz -10db
- (enjoy the) silence
Simply the best tool I got on hand for a quick test.
Regarding tests with DATS, well I think that needs to wait till the holiday season is over and the table is allowed for technical usage again.
Just need to check that everything is well wired and the test should be easily done - can be a bit flimsy with the crocodile clamps on the test-probes
We do have three modules inputstage - RIAA - outputstage.
Surely this thread is for seeing how far this can be pushed but on an kind of economic level.
The main goal is to not only have a theoretical ideal phono pre but also be able to build this one and enjoy some hours with a whiskey tumbler
We don't want to replace an op or a tube schematic with a discrete only solution, just checking what can be used as state of the art and fit the BOM.
Regarding the mentioned jfet for the input - maybe I didn't read it yet but is there a valid point to not go this route?
The good ol' BCs are well suited for the outputstage and can of course be used on the input but is it the best choice we have on hand?
On paper the JFETs are the best choice for now but somehow I do have the feeling I'm missing something!
In the meantime, I looked at the raw data of the DATS record and determined Ls at exactly 90°: 479mH and 473mH, Rdc was equal to 385Ohm in both cases, a full decade later an angle of 90° appears again ... with 47kOhm in parallel 220pF (total capacitance) we will therefore only achieve an upper cut-off frequency of just over 15kHz. However, if L20k=297mH plus R20k=17k12Ohm, 17k12 +j 37k322 (taken from the approximation) = |41.06kOhm| ..!
Well, you don't have to worry about 100% correct termination --> 47pF plus the connection capacitances, and 47 real kiloohms in parallel and the scenario fits for my Audio Technica MM - I think.
Let's build the EQ-Amp.
Bye,
HBt.
Well, you don't have to worry about 100% correct termination --> 47pF plus the connection capacitances, and 47 real kiloohms in parallel and the scenario fits for my Audio Technica MM - I think.
Let's build the EQ-Amp.
Bye,
HBt.
Sidestep
Since the dynamic transducer can also be modeled as an equivalent two-pole with parallel elements (the internal resistance of the generator is quite high anyway), let's assume a current source. At 20kHz, this results in a 1/33.53kOhm -j 1/41.14kOhm complex conductance. Expressed as a noisy voltage source |26kOhm| or as a current source ---> 0.8pArms/sqrt(Hz). At 1kHz it is 1/6k1 -j 1/4k171 ---> |3k443Ohm| ---> 2.2pArms/sqrt(Hz). At this point at the latest, the confusion should be perfect, the terminating resistor of 47kOhm is already taken into account in my calculation. We have to refer to something, why not to f=1kHz and with the resulting impedance that our base sees, and then we are with an rbb of about 10Ohm as a standard value, at an optimal collector current less than or equal to 200µA. Unfortunately, the whole thing is not only dependent on hfe alone, but the whole back and forth does us no good.
Try it out.
Question,
How do we achieve an upper cut-off frequency of greater than 15kHz?
Since the dynamic transducer can also be modeled as an equivalent two-pole with parallel elements (the internal resistance of the generator is quite high anyway), let's assume a current source. At 20kHz, this results in a 1/33.53kOhm -j 1/41.14kOhm complex conductance. Expressed as a noisy voltage source |26kOhm| or as a current source ---> 0.8pArms/sqrt(Hz). At 1kHz it is 1/6k1 -j 1/4k171 ---> |3k443Ohm| ---> 2.2pArms/sqrt(Hz). At this point at the latest, the confusion should be perfect, the terminating resistor of 47kOhm is already taken into account in my calculation. We have to refer to something, why not to f=1kHz and with the resulting impedance that our base sees, and then we are with an rbb of about 10Ohm as a standard value, at an optimal collector current less than or equal to 200µA. Unfortunately, the whole thing is not only dependent on hfe alone, but the whole back and forth does us no good.
Try it out.
Question,
How do we achieve an upper cut-off frequency of greater than 15kHz?
The comparability
Since each pickup will differ from other competitors, a specific determination method must be agreed upon to ensure comparability (of the measured values).
For example, the noise bandwidth and the resolution bandwidth, but also the level and, above all, the source impedance. Should it be "reell" or "komplex", zero ohms or 4k7 ... which is it?
How did you determine noise voltage distances according to the HiFi DIN in the past? What did the pickup dummy impedance of the 5mV MM source look'ed like?
Thel (Thomas Hartwig) gives the following resulting SNR for a very good test record with approx. 50dB distance between the signal and the noise:
PreAmp-EQ
60dB -> 47.5dB; 70dB -> 49.1dB; 80dB -> 49.6dB all unweighted.
Since each pickup will differ from other competitors, a specific determination method must be agreed upon to ensure comparability (of the measured values).
For example, the noise bandwidth and the resolution bandwidth, but also the level and, above all, the source impedance. Should it be "reell" or "komplex", zero ohms or 4k7 ... which is it?
How did you determine noise voltage distances according to the HiFi DIN in the past? What did the pickup dummy impedance of the 5mV MM source look'ed like?
Thel (Thomas Hartwig) gives the following resulting SNR for a very good test record with approx. 50dB distance between the signal and the noise:
PreAmp-EQ
60dB -> 47.5dB; 70dB -> 49.1dB; 80dB -> 49.6dB all unweighted.
Last edited:
Well, that was Nick Sukhov's point about avoiding the potentially very unpleasant shot noise - and of course because his own design includes a jFet differential amplifier as the first stage. I would like to refrain from this kind of solution strategy for the time being. For various reasons
.The more we fidget around, the more I tend to try the overkill variant with nineteen BJTs - if only to see real measured values, a testboard as a layout exists. With unevenly distributed quiescent currents. Nobody else dares.
Greetings,
HBt.
"Tipp:"
https://www.tacet.de/main/seite1.php?language=de&filename=production.php&bestnr=02101
Last edited:
The manufacturer's specifications for your Ifi ZEN Phono promise absolute high-end qualities and even a subsonic filter with its own artificial intelligence.
SNR |80dB| unweighted, based on an output voltage of 2Vrms -> 5.66Vpp (balance). And all this with a Vu of 36dB +/- 1dB for MM systems. Then we can carry out a plausibility check, 63.1 * 5mVrms = 315.5mV.
20*log(1V / 0.3155V) = 10dB, unfortunately we have to subtract these 10dB from the 80dB. What remains is an unweighted SNR_mm of just under 70dB.
The built-in DC-DC converter and the microcontroller will also contribute to the music enjoyment.
That's how you praise it.
Last edited:
https://ifi-audio.com/products/zen-phono/
Does anyone have a circuit diagram of the ifi ZEN at hand?
Sorry,
but the advertising is so cool - I'm laughing under the table, advertising is the "A" and "O".
I would like to test the equalizer (myself).
At first I didn't understand why you wanted to build an EQ yourself and develop it yourself - now I do.
Does anyone have a circuit diagram of the ifi ZEN at hand?
Sorry,
but the advertising is so cool - I'm laughing under the table, advertising is the "A" and "O".
I would like to test the equalizer (myself).
At first I didn't understand why you wanted to build an EQ yourself and develop it yourself - now I do.