Konnichiwa,
The Data is not mine, I just came across it and it squared with own experience, which where a lot less rigerous....
😉
As for running ECC88's hotter, I have done so, in this case in MC input stages, but I still found the noise, even with specially selected exemplars past what I was really happy with.
In practice, in MM sensitive stages I have usually found an ECC83 on the input quieter. Another item to note is that Valves are noisy enough that their noise dominates over that of fairly high value resistors.
Sayonara
Brian Beck said:
The Data is not mine, I just came across it and it squared with own experience, which where a lot less rigerous....
😉
As for running ECC88's hotter, I have done so, in this case in MC input stages, but I still found the noise, even with specially selected exemplars past what I was really happy with.
In practice, in MM sensitive stages I have usually found an ECC83 on the input quieter. Another item to note is that Valves are noisy enough that their noise dominates over that of fairly high value resistors.
Sayonara
Konnichiwa,
Hmmm, I think with a number of published Phonostage Designs, all in the public domain, all commented to the rationales behind them I am quite constructive.
I have build more phonostages than I care to remember.
I find the following to hold true:
Simplicity invariably sounds better than complexity
The best part is no part
THD is overrated as quality measure
Never ever use the same valve twice
If you can reduce the number of stages, do so
My little "Valve El Cheapo" is the result of consequently applying these principles. I could have made it more complex, with better measurements, but only at the cost of a worse sound quality.
I am happy to answer any questions to specific items I have not explained elsewhere. If you like, why not analyse my design and try to understand why it is the way it is?
Sayonara
Giaime said:
Hmmm, I think with a number of published Phonostage Designs, all in the public domain, all commented to the rationales behind them I am quite constructive.
I have build more phonostages than I care to remember.
I find the following to hold true:
Simplicity invariably sounds better than complexity
The best part is no part
THD is overrated as quality measure
Never ever use the same valve twice
If you can reduce the number of stages, do so
My little "Valve El Cheapo" is the result of consequently applying these principles. I could have made it more complex, with better measurements, but only at the cost of a worse sound quality.
I am happy to answer any questions to specific items I have not explained elsewhere. If you like, why not analyse my design and try to understand why it is the way it is?
Sayonara
Konnichiwa,
Sorry about this. It was not as bad a few years back when I put the stuff there. I never had the time to move and update the stuff, I have better things to do.
Sayonara
anatech said:
Sorry about this. It was not as bad a few years back when I put the stuff there. I never had the time to move and update the stuff, I have better things to do.
Sayonara
Kuei Yang Wang said:
I thought you were originally german Thorsten. You should now this:
"Warum soll man es einfach machen wenn mann es so wunderschön komplizieren kann?"
Just joking. 🙂
Whether you use a 12AX7 or a 6DJ8 in a phono front end, you have to be careful with associated resistors and gain versus RIAA EQ allocation to achieve the best noise potential for each tube. At the highest audio frequencies where shot noise dominates, the noise generated by a 6DJ8 (biased to 15mA) will approach the noise generated by a 300 ohm resistor (called Req) if said resistor were placed at the tube’s input. Input means grid or cathode. Tubes such as the 6C45 and 437A can approach 100 ohms Req. The 12AX7’s Req will be in the 2000 to 3000 ohm range. Grid stoppers and un-bypassed cathode resistors will add to the tube’s Req noise. In many RIAA stages, I’ve seen a large grid stopper, probably in an attempt to filter RF with the Miller capacitance, and/or an un-bypassed cathode bias resistor. The noise contributed by these resistors will add (by root-sum-square of voltages) to the tube’s noise. If these resistors are larger than 300 ohms (in the case of a 6DJ8), then you’ve just wasted the noise capability of this tube. As an example, if you use just a 300 ohm grid stopper with a 6DJ8, you’ve doubled the noise power at the higher frequencies (+3dB). For lowest noise, you must use the lowest value of grid stopper that is practical (or a ferrite bead - boo, hiss…) and either bypass the cathode resistor or use some other means of bias (batteries, grid bias, LEDs, etc.) so that the cathode is grounded at AC, if not at DC.
When using passive equalization, the filter stage between the first and second tubes will typically have a large series resistor in an RC low pass. This noise source, even coming after the gain of the first stage, can swamp attempts at low noise. The EQ resistors need to be as small as possible, balanced against the ability to drive these lower impedances by the preceding stage. In the first stage, higher gain drives the signal (with first stage noise already added) up higher, so that next stage noise sources have less effect. Here the 12AX7’s mu of 100 is an advantage (if headroom is duly considered), although it is possible to mitigate this advantage using 6DJ8s or similar. Because of the required losses at high frequencies from the RIAA EQ filter sections, the noise contributed by the subsequent tubes matter in overall noise performance, too.
In a typical NFB RIAA stage, the feedback is applied to the un-bypassed cathode resistor of the first stage. Here again, these resistors add to the noise, and can undo high gm selections.
So it’s not so straightforward to say that tube X is always quieter in an RIAA stage than tube Y. It depends greatly on implementation. I will say that if an all-out noise assault is planned, it can be achieved best, with all due design considerations, with very high gm, moderately high mu tubes. Hence the rarified status of tubes such as the 417A/5842, 437A, 6C45, EC8010, EC8020, 3A/167M, triode connected 7788, D3A, E55L, etc. The ubiquitous 6DJ8/ECC88 is a reasonable and cheap compromise, offering fairly high gm, and moderate mu.
The design of an RIAA stage is one of the most difficult audio challenges. It’s a battle among conflicting variables: noise, gain, overload headroom and equalization. I once spent weeks working on unplanned issues in an all out design with a friend, battling unforeseen things like larger-than-expected grid current into MC cartridges. RIAA stages have been known to drive men to drink, or in the more desperate cases, to the use of input transformers.
When using passive equalization, the filter stage between the first and second tubes will typically have a large series resistor in an RC low pass. This noise source, even coming after the gain of the first stage, can swamp attempts at low noise. The EQ resistors need to be as small as possible, balanced against the ability to drive these lower impedances by the preceding stage. In the first stage, higher gain drives the signal (with first stage noise already added) up higher, so that next stage noise sources have less effect. Here the 12AX7’s mu of 100 is an advantage (if headroom is duly considered), although it is possible to mitigate this advantage using 6DJ8s or similar. Because of the required losses at high frequencies from the RIAA EQ filter sections, the noise contributed by the subsequent tubes matter in overall noise performance, too.
In a typical NFB RIAA stage, the feedback is applied to the un-bypassed cathode resistor of the first stage. Here again, these resistors add to the noise, and can undo high gm selections.
So it’s not so straightforward to say that tube X is always quieter in an RIAA stage than tube Y. It depends greatly on implementation. I will say that if an all-out noise assault is planned, it can be achieved best, with all due design considerations, with very high gm, moderately high mu tubes. Hence the rarified status of tubes such as the 417A/5842, 437A, 6C45, EC8010, EC8020, 3A/167M, triode connected 7788, D3A, E55L, etc. The ubiquitous 6DJ8/ECC88 is a reasonable and cheap compromise, offering fairly high gm, and moderate mu.
The design of an RIAA stage is one of the most difficult audio challenges. It’s a battle among conflicting variables: noise, gain, overload headroom and equalization. I once spent weeks working on unplanned issues in an all out design with a friend, battling unforeseen things like larger-than-expected grid current into MC cartridges. RIAA stages have been known to drive men to drink, or in the more desperate cases, to the use of input transformers.
Brian,
Score 1 for the 'X7. No grid stopper needed.
I know NADA about RIAA network design. I gravitate towards circuits with "proven" EQ networks, especially the "ancient and honorable" RCA design (attached). Replacing the bias network of the 1st gain block with a HLMP6000 LED, ala Bottlehead, seems reasonable. Grid leak bias in the 2nd gain block, as has been discussed previously, is the way to go.
Score 1 for the 'X7. No grid stopper needed.
I know NADA about RIAA network design. I gravitate towards circuits with "proven" EQ networks, especially the "ancient and honorable" RCA design (attached). Replacing the bias network of the 1st gain block with a HLMP6000 LED, ala Bottlehead, seems reasonable. Grid leak bias in the 2nd gain block, as has been discussed previously, is the way to go.
Attachments
Konnichiwa,
And one must remember that at MM Level inputs the impedance/resistance of the cartridge is one of the largest noise sources. For a MM Cartridge the around 0.5 to 1H inductance allows the noise of the 47K input resistor to assert itself to a degree progressively with rising frequency, if using a MC Stepup transformer the stepped up impedance of the cratridge becomes quite large and again the 47K input resistors own noise if shunted to only a modest degree (example - 10R cartridge, 1:20 voltage stepup, effective impedance for noise purposes, 4KOhm).
If active stages are used for MC stepup Valves are more or less out, they are just so hard work and still hum and hiss like a pit of snakes and bears, for the rest, most other active MC stages also have several KOhm output impedance, which again becomes the equivalence impedance for noise purposes.
So, usually loosing sleep over RIAA resistor values and cathode resistors is on the wrong track.
Sayonara
Brian Beck said:
And one must remember that at MM Level inputs the impedance/resistance of the cartridge is one of the largest noise sources. For a MM Cartridge the around 0.5 to 1H inductance allows the noise of the 47K input resistor to assert itself to a degree progressively with rising frequency, if using a MC Stepup transformer the stepped up impedance of the cratridge becomes quite large and again the 47K input resistors own noise if shunted to only a modest degree (example - 10R cartridge, 1:20 voltage stepup, effective impedance for noise purposes, 4KOhm).
If active stages are used for MC stepup Valves are more or less out, they are just so hard work and still hum and hiss like a pit of snakes and bears, for the rest, most other active MC stages also have several KOhm output impedance, which again becomes the equivalence impedance for noise purposes.
So, usually loosing sleep over RIAA resistor values and cathode resistors is on the wrong track.
Sayonara
Konnichiwa,
Actually, the RCA RIAA is proven to be pretty inaccurate. It is probably one of the last networks/schematics that could be recommended.
At 20Hz using "zero tolerance" values the LF is attenuated nearly 1db, at 20KHz your 1.5db down, at 150Hz you are halve a db up, going flat again at 1KHz. All of that with the "official" load of 220K/180pF.
Those are quite audible differences. If your load is worse the deviations are worse too.
Plus, the RCA Network has almost 25db insertion loss at 1KHz, meaning you need to find an extra 6db gain over the standard RIAA curve which invariably pushed up distortion and noise, especially thedistortion woy
Compare the "Valve El Cheapo" network, it stays, using generically obtainable values within +/-0.1db 20Hz - 20KHz into a 50K/220pF load, which is much more in line what you find in common valve line stages. Even allowing for the maximum toterance deviation for all components (monte carlo analysis, worst case scenario) the added deviations are only +/- 0.15db.
Insertion loss is 19db, so 6db less gain are needed for the overall identical gain.
This in turn allows the use of circuitry that is more linear, giving around -66db 2nd HD @ 0.5V out (varies with valves) while the RCA design has 10db more 2nd HD. Both stages arguably have a simple and mainly 2nd HD harmonic structure, the 10db advantage of the "Valve El Cheapo" remains at +14db overload, the maximum expected under the IEC standards for LP's. The levels of HD and IMD in the RCA stage push into regions where they will be able to audibly alter the harmonic pattern of the recording (which has a fairly HD to start with) audibly and as we deal with multiplication effects only to the worse.
Linearising the input stage by using a unbypassed cathode resistor contributes most to the improvement and is allowed by using a superior design of RIAA Network. Note that this network came to my notion in Phono Preamplifiers designed by Hiroyaso Kondo, so it is not my invention, it is just a much superior way, IME of doing things.
Also, I really dislike, on subjective ground, LED biasing in low current stages (e.g. ECC83), this problem diminishes as currents go up, as the dynamic impedance of the LED becomes less nonlinear, but still, I end up ripping LED's out fairly wherever I put them into cathode circuits. Next to LED's I find the cheapest chinese electrolytic capacitor much preferable, not that one is forced to use electrolytics at all in my own designs.
It also pays to note that all stages are selected to give the noise levels apropriate. In the second stage the higher (low frequency) noise of the ECC88/6DJ8 is less relevant, as it receives already a signal significantly amplified over the cartridge signal (12db @ 1KHz and 24db @ 100Hz after the RIAA attenuation is accounted for). So the ECC83 in the input domiantes the noise.
While the RIAA Impedance is fairly high, it's noise contribution is not that great. At 100Hz the "look back" impedance of the ECC88 Grid is around 270KOhm, at 1KHz it is 74KOhm. This translates to a johnson noise of around 3uV at 100Hz and 1.5uV at 1KHz (as noisebandwidth is less).
Still it's contribution is actually more than that of the ECC88 itself, however if we take the noise contribution of the Input valve it works out to significantly less.
Starting with the noise, presuming that the around 6db local degeneration reduce LF noise by equal amounts, while increasing midband noise by around 3db I'll use an average noise of 20nV|/Hz @ 100Hz and 10nV|/Hz @ 1KHz, which will be close enough. This gives 0.28uV RMS @ 100Hz and 0.45uV at 1KHz. For reference, Johnson noise of a 4K resistor comes to 0.12uV at 100Hz and 0.36uV at 1KHz, for 47K in parallel with 0.7H it is 0.037uV at 100Hz and 0.38uV.
The input noise (which is a virtuel measure anyway) is amplified by factor of around 37, so at 100Hz we have around 10uV at the input to the RIAA Network, which attenuates this noise by around 6db, so at the input to the ECC88 stage we have 5uV noise from the input stage, while at 1KHz we have 17uV on the input of the RIAA, which after the RIAA is attenuated by 19db to around 2uV on the input of the ECC88 stage.
This shows our (fairly quiet) ECC83 as the noise limiting factor up to 1KHz, but the noisefactor (the degree by which the ECC83 increases the overall remains in the region of 3...5db or so, which means the noise generated by the phonostage overall is such that the noise of the cartridge (not the LP) is increased by around 3...5db, which means we are way below any surface noise.
The above may help to illustrate why my cheap phonostage looks like it does and indeed qualified as very close to "no compromise". My 600 Ohm T-Network equalised Pentode phonostage is of course a lot less compromised, but even the Valve El Cheapo has only minimal levels of compromise, as it is.
Sayonara
Eli Duttman said:
Actually, the RCA RIAA is proven to be pretty inaccurate. It is probably one of the last networks/schematics that could be recommended.
At 20Hz using "zero tolerance" values the LF is attenuated nearly 1db, at 20KHz your 1.5db down, at 150Hz you are halve a db up, going flat again at 1KHz. All of that with the "official" load of 220K/180pF.
Those are quite audible differences. If your load is worse the deviations are worse too.
Plus, the RCA Network has almost 25db insertion loss at 1KHz, meaning you need to find an extra 6db gain over the standard RIAA curve which invariably pushed up distortion and noise, especially thedistortion woy
Compare the "Valve El Cheapo" network, it stays, using generically obtainable values within +/-0.1db 20Hz - 20KHz into a 50K/220pF load, which is much more in line what you find in common valve line stages. Even allowing for the maximum toterance deviation for all components (monte carlo analysis, worst case scenario) the added deviations are only +/- 0.15db.
Insertion loss is 19db, so 6db less gain are needed for the overall identical gain.
This in turn allows the use of circuitry that is more linear, giving around -66db 2nd HD @ 0.5V out (varies with valves) while the RCA design has 10db more 2nd HD. Both stages arguably have a simple and mainly 2nd HD harmonic structure, the 10db advantage of the "Valve El Cheapo" remains at +14db overload, the maximum expected under the IEC standards for LP's. The levels of HD and IMD in the RCA stage push into regions where they will be able to audibly alter the harmonic pattern of the recording (which has a fairly HD to start with) audibly and as we deal with multiplication effects only to the worse.
Linearising the input stage by using a unbypassed cathode resistor contributes most to the improvement and is allowed by using a superior design of RIAA Network. Note that this network came to my notion in Phono Preamplifiers designed by Hiroyaso Kondo, so it is not my invention, it is just a much superior way, IME of doing things.
Also, I really dislike, on subjective ground, LED biasing in low current stages (e.g. ECC83), this problem diminishes as currents go up, as the dynamic impedance of the LED becomes less nonlinear, but still, I end up ripping LED's out fairly wherever I put them into cathode circuits. Next to LED's I find the cheapest chinese electrolytic capacitor much preferable, not that one is forced to use electrolytics at all in my own designs.
It also pays to note that all stages are selected to give the noise levels apropriate. In the second stage the higher (low frequency) noise of the ECC88/6DJ8 is less relevant, as it receives already a signal significantly amplified over the cartridge signal (12db @ 1KHz and 24db @ 100Hz after the RIAA attenuation is accounted for). So the ECC83 in the input domiantes the noise.
While the RIAA Impedance is fairly high, it's noise contribution is not that great. At 100Hz the "look back" impedance of the ECC88 Grid is around 270KOhm, at 1KHz it is 74KOhm. This translates to a johnson noise of around 3uV at 100Hz and 1.5uV at 1KHz (as noisebandwidth is less).
Still it's contribution is actually more than that of the ECC88 itself, however if we take the noise contribution of the Input valve it works out to significantly less.
Starting with the noise, presuming that the around 6db local degeneration reduce LF noise by equal amounts, while increasing midband noise by around 3db I'll use an average noise of 20nV|/Hz @ 100Hz and 10nV|/Hz @ 1KHz, which will be close enough. This gives 0.28uV RMS @ 100Hz and 0.45uV at 1KHz. For reference, Johnson noise of a 4K resistor comes to 0.12uV at 100Hz and 0.36uV at 1KHz, for 47K in parallel with 0.7H it is 0.037uV at 100Hz and 0.38uV.
The input noise (which is a virtuel measure anyway) is amplified by factor of around 37, so at 100Hz we have around 10uV at the input to the RIAA Network, which attenuates this noise by around 6db, so at the input to the ECC88 stage we have 5uV noise from the input stage, while at 1KHz we have 17uV on the input of the RIAA, which after the RIAA is attenuated by 19db to around 2uV on the input of the ECC88 stage.
This shows our (fairly quiet) ECC83 as the noise limiting factor up to 1KHz, but the noisefactor (the degree by which the ECC83 increases the overall remains in the region of 3...5db or so, which means the noise generated by the phonostage overall is such that the noise of the cartridge (not the LP) is increased by around 3...5db, which means we are way below any surface noise.
The above may help to illustrate why my cheap phonostage looks like it does and indeed qualified as very close to "no compromise". My 600 Ohm T-Network equalised Pentode phonostage is of course a lot less compromised, but even the Valve El Cheapo has only minimal levels of compromise, as it is.
Sayonara
Please Thoersten,
link or attach the schematic you're talking about. If not, that's like talking of castles in air.
link or attach the schematic you're talking about. If not, that's like talking of castles in air.
I am somewhat confused here... It would appear to me that your phono stage is more complicated than it needs to be. Isn't simple better in terms of audio? Wouldn't the somewhat antique schematic posted on page 3 deliver more than adequate performance?
If you looking for gain and low noise wouldn't the 6AU6/5687 Aikido with an RIAA EQ circuit work? Aren't we looking for frequency compensation at 50hz,500 hz, and 2122 hz? I'm by no means an expert so this is just an idea. I am thinking that less may be better?
Ducking for cover now....
If you looking for gain and low noise wouldn't the 6AU6/5687 Aikido with an RIAA EQ circuit work? Aren't we looking for frequency compensation at 50hz,500 hz, and 2122 hz? I'm by no means an expert so this is just an idea. I am thinking that less may be better?
Ducking for cover now....
burnedfingers said:
This is a bit OT but...
The biggest audio lie 😉
Simple is better, because you can sell for 1000$ an amp that has 50$ in parts instead of 150$... 😉
If simple is better why Tektronix scopes are so complicated? Maybe because electronic IS science, and science has nothing to do with esotherical guru's claims?
Sorry for the OT
those are the things that get me extremely angry.(ps the antique schematic in page 3 is for table top radios and grammphones, not for hi-fi phono preamps, the technical analysis that Thoersten did some posts ago about that circuit is very correct about it)
It would appear that every designer/engineer has a different take on which RIAA circuit is correct and how their design is going to correct the problem. Personally I still think the simple is better.
When the parts count is too high it just doesn't seem to sound as good. Just my opinion.....
Yes, I agree with you, but RIAA curve is very precise, and you can have only poor sound quality if you don't follow it very close.
HOW to follow it is the problem... not if it has to be followed or not 😉
HOW to follow it is the problem... not if it has to be followed or not 😉
Doesn't that bring up another problem? If your speaker response isn't correct it doesn't make a hill of beans difference how good your RIAA response is.
Konnichiwa,
I should think you are quite familiar with the design, having discussed it in length in other threads....
Sayonara
Giaime said:
I should think you are quite familiar with the design, having discussed it in length in other threads....
An externally hosted image should be here but it was not working when we last tested it.
Sayonara
I remember it very well, but even other people are reading this thread and they might not know it. 😉
Konnichiwa,
That is unfortunatly not the case. Simplicity has been proven better each and every time, ALL ELSE BEING EQUAL. I have been there and done that. Using two stages, with additional negative feedback to get the same gain invariably sounds less pure than only one of apropriate gain.
I follow Albert Einstein on this: "All things should be as simple as possible, but no simpler." Complexity needs be used if it is essential, but not for complexity's sake.
Actually, the actual (ac) gain circutry in these 'scopes is pretty simple, the complexity is added because the same gain conditions must be present all the way down to DC with minimal drift, something which is not an issue in audio, for example.
Once you condense the circuitry down to that which would be required for equivalent AC performance the impressive and brutal simplicity will become obvious.
Sayonara
Giaime said:
That is unfortunatly not the case. Simplicity has been proven better each and every time, ALL ELSE BEING EQUAL. I have been there and done that. Using two stages, with additional negative feedback to get the same gain invariably sounds less pure than only one of apropriate gain.
I follow Albert Einstein on this: "All things should be as simple as possible, but no simpler." Complexity needs be used if it is essential, but not for complexity's sake.
Giaime said:
Actually, the actual (ac) gain circutry in these 'scopes is pretty simple, the complexity is added because the same gain conditions must be present all the way down to DC with minimal drift, something which is not an issue in audio, for example.
Once you condense the circuitry down to that which would be required for equivalent AC performance the impressive and brutal simplicity will become obvious.
Sayonara
Thoersten,
we have two strongly different points of view for what is "semplicity".
For semplicity I meant the lack of knowledge of the many "self thaught" audio guru out there. The lack of interest in deeply understand how things works, and thinking that in audio the best things have been already invented and done.
Those people like to put togheter parts and sell amps for $$$$.
Probably you for "simplicity" mean to get the same results with less parts, and this implies knowledge, engineering practice, not "hi-end proclaims" or "esotherical myth" or "no nfb at all".
I'm very pleased you have such a good idea of simplicity, but sorry you're one of 3 / 4 people in the world..
Now I have to go, Italy - France is starting now 😀
we have two strongly different points of view for what is "semplicity".
For semplicity I meant the lack of knowledge of the many "self thaught" audio guru out there. The lack of interest in deeply understand how things works, and thinking that in audio the best things have been already invented and done.
Those people like to put togheter parts and sell amps for $$$$.
Probably you for "simplicity" mean to get the same results with less parts, and this implies knowledge, engineering practice, not "hi-end proclaims" or "esotherical myth" or "no nfb at all".
I'm very pleased you have such a good idea of simplicity, but sorry you're one of 3 / 4 people in the world..
Now I have to go, Italy - France is starting now 😀
- Status
- Not open for further replies.