Val_r
Well-known member
Hi,
Here's another hybrid mic preamp.
Please post your comments and suggestions.
Thank you,
Val.
:idea:
Here's another hybrid mic preamp.
Please post your comments and suggestions.
Thank you,
Val.
:idea:
Another hybrid mic preamp
- Thread starter Val_r
- Start date
Help Support GroupDIY Audio Forum:
[quote author="Val_r"]Hi,
Here's another hybrid mic preamp.
Please post your comments and suggestions.
Thank you,
Val.
:idea:
[/quote]
Looks pretty familiar. We used a version very similar to that in the AMR recording consoles (late '80s), Peavey Mark VIII, and sundry other Peavey models. Major difference (from memory) we used single 2SD786 low noise input devices per input (now obsolete), TL07x opamps, Rs for upper current sources, and polarity switch when provided was at mic input not pre output.
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF. I have never seen any capacitance (like the 470 pF) between the emitters. Perhaps it is used to roll off feedback factor for stability but it looks like it will add closed loop gain at very HF which could be an issue if input becomes unbalanced in presences of RF. It is also pretty common to use a shunt C and some Cs to ground at mic input. C7 across the second inverting stage will have some impact on stability and too much roll off there will lead to asymmetrical differential output at HF.
While less important, the 100uF input caps may be problematic for fault currents if inadvertently shorted (as discussed in recent threads). Also the 1k input termination is a little low for my tastes. I typically target 2k nominal (10x 150-200 ohms). With a minimum gain of 26dB+ it might also benefit from a switchable front end pad. I also routinely use anti-avalanche clamp diodes across input b-e junctions, precision Rs for phantom, etc.
Based on my experience with a similar version this topology is a solid performer and somewhat more cost effective than Transamp/Cohen topology as it can generate SE output with one less opamp (useful in large consoles and cost sensitive smaller designs). The 470P makes me a little nervous and I would instead suggest a compensation RC at opamp 1 input.
JR
PS: Kit- If by CCS you mean "cap coupled" the reason for using a coupling cap in gain leg at emitters is mainly to prevent scratchy pot noise (or switch clicks) due to Vb-e differences. With larger offsets and max gain, the input current density of + and - inputs could also be impacted but this is usually secondary to the perceptual (noise) issue. For commercial use the scratchy pot is problematic, for personal use I have no problem with DC coupled at that point.
Here's another hybrid mic preamp.
Please post your comments and suggestions.
Thank you,
Val.
:idea:
[/quote]
Looks pretty familiar. We used a version very similar to that in the AMR recording consoles (late '80s), Peavey Mark VIII, and sundry other Peavey models. Major difference (from memory) we used single 2SD786 low noise input devices per input (now obsolete), TL07x opamps, Rs for upper current sources, and polarity switch when provided was at mic input not pre output.
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF. I have never seen any capacitance (like the 470 pF) between the emitters. Perhaps it is used to roll off feedback factor for stability but it looks like it will add closed loop gain at very HF which could be an issue if input becomes unbalanced in presences of RF. It is also pretty common to use a shunt C and some Cs to ground at mic input. C7 across the second inverting stage will have some impact on stability and too much roll off there will lead to asymmetrical differential output at HF.
While less important, the 100uF input caps may be problematic for fault currents if inadvertently shorted (as discussed in recent threads). Also the 1k input termination is a little low for my tastes. I typically target 2k nominal (10x 150-200 ohms). With a minimum gain of 26dB+ it might also benefit from a switchable front end pad. I also routinely use anti-avalanche clamp diodes across input b-e junctions, precision Rs for phantom, etc.
Based on my experience with a similar version this topology is a solid performer and somewhat more cost effective than Transamp/Cohen topology as it can generate SE output with one less opamp (useful in large consoles and cost sensitive smaller designs). The 470P makes me a little nervous and I would instead suggest a compensation RC at opamp 1 input.
JR
PS: Kit- If by CCS you mean "cap coupled" the reason for using a coupling cap in gain leg at emitters is mainly to prevent scratchy pot noise (or switch clicks) due to Vb-e differences. With larger offsets and max gain, the input current density of + and - inputs could also be impacted but this is usually secondary to the perceptual (noise) issue. For commercial use the scratchy pot is problematic, for personal use I have no problem with DC coupled at that point.
bcarso
Well-known member
CCS = constant current sources I think John.
I agree that given the small voltage swings there is little gained (and noise increased) using CCSs. Power supply rejection may be enhanced a bit, but just use good quiet power supplies.
EDIT part number corrected: Recommend 2SA1316 for the input devices.
I agree that given the small voltage swings there is little gained (and noise increased) using CCSs. Power supply rejection may be enhanced a bit, but just use good quiet power supplies.
EDIT part number corrected: Recommend 2SA1316 for the input devices.
Samuel Groner
Well-known member
Havin current sources instead of simple resistors will have some positive effect on CMRR, though the effect will likely be small for this design. I'd think about using resistors from the +48 V supply here as these will provide lower noise than the current sources yet better CMRR than the resistors from the +15 V supply.
In addition to this, providing the bias current for a NE5532 through a 1 M resistors looks unhappy--I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further. R44 seems not to be needed?
Samuel
In addition to this, providing the bias current for a NE5532 through a 1 M resistors looks unhappy--I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further. R44 seems not to be needed?
Samuel
[quote author="bcarso"]CCS = constant current sources I think John.
I agree that given the small voltage swings there is little gained (and noise increased) using CCSs. Power supply rejection may be enhanced a bit, but just use good quiet power supplies.
Recommend 2SA1306 for the input devices.[/quote]
To quote Homer Simpson... Doh! Thanks, that actually fits the letters too.
Agreed, as stated I used resistors in my version. Main benefit from current sources would only be found at much lower gain than this version is capable of. Poor matching of those Rs or output Cs of two current sources may cause HF CMR term. A single current source could also feed both transistors through a pair of matched resistors which could make the current source noise common mode to both, but for this case current sources add complexity with no real benefit.
A subtle difference between Rs and current source(s) to + supply in this topology vs. Transamp/Cohen is that while - supply noise cancels out CM at first opamp, + supply noise would show up in differential outputs CM and rely upon quality of later differential to remove. This is even more of an issue for low cost application using just one opamp for SE feed. Of course pretty simple to passively filter out +PS noise at those two resistors.
JR
PS: The low noise device I mentioned, in addition to being obsolete is NPN so they wouldn't drop in even if you could get them.
I agree that given the small voltage swings there is little gained (and noise increased) using CCSs. Power supply rejection may be enhanced a bit, but just use good quiet power supplies.
Recommend 2SA1306 for the input devices.[/quote]
To quote Homer Simpson... Doh! Thanks, that actually fits the letters too.
Agreed, as stated I used resistors in my version. Main benefit from current sources would only be found at much lower gain than this version is capable of. Poor matching of those Rs or output Cs of two current sources may cause HF CMR term. A single current source could also feed both transistors through a pair of matched resistors which could make the current source noise common mode to both, but for this case current sources add complexity with no real benefit.
A subtle difference between Rs and current source(s) to + supply in this topology vs. Transamp/Cohen is that while - supply noise cancels out CM at first opamp, + supply noise would show up in differential outputs CM and rely upon quality of later differential to remove. This is even more of an issue for low cost application using just one opamp for SE feed. Of course pretty simple to passively filter out +PS noise at those two resistors.
JR
PS: The low noise device I mentioned, in addition to being obsolete is NPN so they wouldn't drop in even if you could get them.
[quote author="Samuel Groner"]Havin current sources instead of simple resistors will have some positive effect on CMRR, though the effect will likely be small for this design. I'd think about using resistors from the +48 V supply here as these will provide lower noise than the current sources yet better CMRR than the resistors from the +15 V supply.
In addition to this, providing the bias current for a NE5532 through a 1 M resistors looks unhappy--I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further. R44 seems not to be needed?
Samuel[/quote]
I agree with your speculation that his design might be happier with a bifet opamp in first position, maybe both.
Regarding DC offset in 5532, using the same value Rs at + and - inputs should result in a first order cancellation of input bias current generated voltages. Changing just one resistor may make matters worse, not to mention reducing negative power supply rejection ratio.
To that same end, the second 5532 needs a 5K in series with it's + input to compensate for it's input bias current and deliver a nominally 0V DC output, but at that lower impedance the magnitude of the error will be smaller so perhaps OK to ignore.
JR
In addition to this, providing the bias current for a NE5532 through a 1 M resistors looks unhappy--I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further. R44 seems not to be needed?
Samuel[/quote]
I agree with your speculation that his design might be happier with a bifet opamp in first position, maybe both.
Regarding DC offset in 5532, using the same value Rs at + and - inputs should result in a first order cancellation of input bias current generated voltages. Changing just one resistor may make matters worse, not to mention reducing negative power supply rejection ratio.
To that same end, the second 5532 needs a 5K in series with it's + input to compensate for it's input bias current and deliver a nominally 0V DC output, but at that lower impedance the magnitude of the error will be smaller so perhaps OK to ignore.
JR
G
Guest
Guest
Continuing inventing wheels, you may look at Yamaha MX-400 mic pre. It was made pretty clever, as the whole console... Very wide dynamic range, minimum active elements in signal path, optimal controls.
Val_r
Well-known member
[quote author="JohnRoberts"]Looks pretty familiar.
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF.[/quote]
Thanks for the replies!
Could you pls clarify the RC network? Do you have an example?
[quote author="bcarso"]Recommend 2SA1306 for the input devices[/quote]
I used 1085 for RS has them regularly in stock.
Using CCS's:
distortion should benefit.
[quote author="Samuel Groner"]I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further.[/quote]
Do you have another opamp to suggest?
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF.[/quote]
Thanks for the replies!
Could you pls clarify the RC network? Do you have an example?
[quote author="bcarso"]Recommend 2SA1306 for the input devices[/quote]
I used 1085 for RS has them regularly in stock.
Using CCS's:
distortion should benefit.
[quote author="Samuel Groner"]I guess there was a FET input opamp there once? Having an AC-bypassed 5k resistor from noninverting input to ground for the second opamp should improve output offset further.[/quote]
Do you have another opamp to suggest?
[quote author="Val_r"][quote author="JohnRoberts"]Looks pretty familiar.
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF.[/quote]
Thanks for the replies!
Could you pls clarify the RC network? Do you have an example?[/quote]
Schematics for MarkVIII and AMR consoles are published in the owner's manuals. Similar preamps were used in many Peavey models across the line.
Exact value of RC will depend upon specifics of opamp used. R shunt between + and - opamp inputs reduces available open loop gain, C in series with R selected so it only reduces gain high above audio band frequencies. This is one of several common stabilization techniques used.
[quote author="Val_r"]Using CCS's:
distortion should benefit.[/quote]
Any mismatch in these resistors can cause a CM voltage gain and a lower level CM distortion term due to different changing Vb-e due to operating current differences. Well matched Rs will mainly interact with closed loop and open loop gain.
Using two uncorrelated current sources means their noise will combine in the mic preamp outputs. One current source evenly split between the two will be coherent and so it's noise would cancel out. There would be some improvement from shorting the two current source emitters together as that would force the noise voltage to be the same but mismatched Vb-e between the two would cause unequal outputs. If you prefer to use a current source I would use one and force sharing with precision resistors at it's output as I earlier suggested.
[quote author="Val_r"]Do you have another opamp to suggest?[/quote]
TL07x series used in most designs I refer to. 2K drive capability may be inadequate for low impedance output. There are many newer bifets on the market that would drop in there limited by how much you feel like spending.
JR
A few comments on stability... Since the second inverting opamp is effectively inside the overall feedback loop for the first, it’s delay needs to be compensated for. The approach that I am familiar with is a RC network between + and - inputs of the first opamp to reduce open loop gain at very HF.[/quote]
Thanks for the replies!
Could you pls clarify the RC network? Do you have an example?[/quote]
Schematics for MarkVIII and AMR consoles are published in the owner's manuals. Similar preamps were used in many Peavey models across the line.
Exact value of RC will depend upon specifics of opamp used. R shunt between + and - opamp inputs reduces available open loop gain, C in series with R selected so it only reduces gain high above audio band frequencies. This is one of several common stabilization techniques used.
[quote author="Val_r"]Using CCS's:
distortion should benefit.[/quote]
Any mismatch in these resistors can cause a CM voltage gain and a lower level CM distortion term due to different changing Vb-e due to operating current differences. Well matched Rs will mainly interact with closed loop and open loop gain.
Using two uncorrelated current sources means their noise will combine in the mic preamp outputs. One current source evenly split between the two will be coherent and so it's noise would cancel out. There would be some improvement from shorting the two current source emitters together as that would force the noise voltage to be the same but mismatched Vb-e between the two would cause unequal outputs. If you prefer to use a current source I would use one and force sharing with precision resistors at it's output as I earlier suggested.
[quote author="Val_r"]Do you have another opamp to suggest?[/quote]
TL07x series used in most designs I refer to. 2K drive capability may be inadequate for low impedance output. There are many newer bifets on the market that would drop in there limited by how much you feel like spending.
JR
Val_r
Well-known member
[quote author="JohnRoberts"]
TL07x series used in most designs I refer to. 2K drive capability may be inadequate for low impedance output. There are many newer bifets on the market that would drop in there limited by how much you feel like spending.
JR[/quote]
I was thinking of the AD712, which claims a lower impedance driving capability ...
TL07x series used in most designs I refer to. 2K drive capability may be inadequate for low impedance output. There are many newer bifets on the market that would drop in there limited by how much you feel like spending.
JR[/quote]
I was thinking of the AD712, which claims a lower impedance driving capability ...
Samuel Groner
Well-known member
OPA2134 looks like a nice fit.
Samuel
Samuel
Samuel Groner
Well-known member
Only at very low frequencies. Within the audio range, the 1 M resistors are shunt by the collector impedance, R10 and R11.
Samuel
[quote author="Samuel Groner"]
Samuel[/quote]
To expand upon your answer, the 0.68uF cap in series with the 1K collector loads will diminish their ability to shunt the 1M and degrade noise performance starting above 2 kHz. It will still only rise to a few K ohms dropping though the sensitive mid band which isn't horrible for a 5532 but there is no reason I can think of to not use a larger value electrolytic (I routinely did). While I also routinely used a BiFet opamp there.
Then again I never took a final output right from that circuit point, so the drive capability of those particular parts was not significant for my use. I avoided simple differential outputs. Typically for balanced outputs I used the popular cross connected feedback approach that gain corrects and tolerates (one) shorted output seamlessly. I didn't have the luxury of knowing what my customers would plug into and it was always my fault if something didn't work, no matter what they did wrong.
JR
Only at very low frequencies. Within the audio range, the 1 M resistors are shunt by the collector impedance, R10 and R11.
Samuel[/quote]
To expand upon your answer, the 0.68uF cap in series with the 1K collector loads will diminish their ability to shunt the 1M and degrade noise performance starting above 2 kHz. It will still only rise to a few K ohms dropping though the sensitive mid band which isn't horrible for a 5532 but there is no reason I can think of to not use a larger value electrolytic (I routinely did). While I also routinely used a BiFet opamp there.
Then again I never took a final output right from that circuit point, so the drive capability of those particular parts was not significant for my use. I avoided simple differential outputs. Typically for balanced outputs I used the popular cross connected feedback approach that gain corrects and tolerates (one) shorted output seamlessly. I didn't have the luxury of knowing what my customers would plug into and it was always my fault if something didn't work, no matter what they did wrong.
JR
bcarso
Well-known member
I get 234 Hz as the -3dB point with 1k. And the problem is noise increasing at lower frequencies, not higher.
BTW the shot and excess noise in 5532 bias current is considerably dominant over the noise in the 1M's. There is little data on it, but the minimum it can be for full shot noise is 310fA per sq rt Hz if the magnitude is 300nA. And typically there is a significant rise at lower frequencies in the noise for bipolars (see Motchenbacher and Fitchen's curves for example).
BTW the shot and excess noise in 5532 bias current is considerably dominant over the noise in the 1M's. There is little data on it, but the minimum it can be for full shot noise is 310fA per sq rt Hz if the magnitude is 300nA. And typically there is a significant rise at lower frequencies in the noise for bipolars (see Motchenbacher and Fitchen's curves for example).
[quote author="mediatechnology"]
I can think of one John and that's tolerance with the 0.68 uF most likely being available in 5 and 10% at reasonable cost. So maybe a little better (but not much) LF CMR vs. a larger electrolytic with smaller Rs?
Still no real justification for the 5532 though if 1Ms are used.[/quote]
I agree and don't like 5532 in that location.
WRT CMR I base my premise on an ASSumption that typical external CM noise sources are mains related and typically 50-60Hz and higher. While my calculations today appear to be a little suspect (yes Brad I fat fingered that 2.3K pole is indeed @ 230Hz). A quick look at CMRR @ 50 Hz for 10% .68 uF mismatch vs. 6.8 uF with 80% mismatch calculates out to 63dB for the former and 73 dB for the latter (You may want to check my math :?: ). Since I don't expect two capacitors within the same production run to actually vary 80% I'd anticipate much better results than that from the larger caps in practice. Of course this hinges on my premise that significant CM noise sources are 50 Hz or higher.
If there is significant 1/F noise in -15V rail and/or positive current sources were changed to feed inputs CM there may be lower frequency CM signals to worry about that could change the calculus, but in those cases there is merit to filter those noises at their source.
JR
I can think of one John and that's tolerance with the 0.68 uF most likely being available in 5 and 10% at reasonable cost. So maybe a little better (but not much) LF CMR vs. a larger electrolytic with smaller Rs?
Still no real justification for the 5532 though if 1Ms are used.[/quote]
I agree and don't like 5532 in that location.
WRT CMR I base my premise on an ASSumption that typical external CM noise sources are mains related and typically 50-60Hz and higher. While my calculations today appear to be a little suspect (yes Brad I fat fingered that 2.3K pole is indeed @ 230Hz). A quick look at CMRR @ 50 Hz for 10% .68 uF mismatch vs. 6.8 uF with 80% mismatch calculates out to 63dB for the former and 73 dB for the latter (You may want to check my math :?: ). Since I don't expect two capacitors within the same production run to actually vary 80% I'd anticipate much better results than that from the larger caps in practice. Of course this hinges on my premise that significant CM noise sources are 50 Hz or higher.
If there is significant 1/F noise in -15V rail and/or positive current sources were changed to feed inputs CM there may be lower frequency CM signals to worry about that could change the calculus, but in those cases there is merit to filter those noises at their source.
JR
[quote author="bcarso"]I get 234 Hz as the -3dB point with 1k. And the problem is noise increasing at lower frequencies, not higher.
BTW the shot and excess noise in 5532 bias current is considerably dominant over the noise in the 1M's. There is little data on it, but the minimum it can be for full shot noise is 310fA per sq rt Hz if the magnitude is 300nA. And typically there is a significant rise at lower frequencies in the noise for bipolars (see Motchenbacher and Fitchen's curves for example).[/quote]
Guilty as charged... 234 Hz it is...
I agree that my choice of words was not much better than my math. I specifically added the "dropping" verbiage to clarify that noise got worse as freq dropped but I wasn't very clear about what was dropping (freq not impedance).
Perhaps I shouldn't post on less than 2 cups of coffee. mea culpa.
JR
BTW the shot and excess noise in 5532 bias current is considerably dominant over the noise in the 1M's. There is little data on it, but the minimum it can be for full shot noise is 310fA per sq rt Hz if the magnitude is 300nA. And typically there is a significant rise at lower frequencies in the noise for bipolars (see Motchenbacher and Fitchen's curves for example).[/quote]
Guilty as charged... 234 Hz it is...
I agree that my choice of words was not much better than my math. I specifically added the "dropping" verbiage to clarify that noise got worse as freq dropped but I wasn't very clear about what was dropping (freq not impedance).
Perhaps I shouldn't post on less than 2 cups of coffee. mea culpa.
JR
[quote author="mediatechnology"]Hey John your CMR calcs are with the assumption that the 1M reistors stay the same value correct?
My thinking was that with larger Cs the R values would typically be lowered from Bifet-friendly 1M into typical 5532 10K teritory. Say maybe from 1M to 10K as we change C from 0.68 to 47uF. My point was the 1M/0.68 (5%) combo would probably match better CMR-wise than 10K/47uF (+50% -20%) combo.
But I haven't calculated it.[/quote]
I'm going to pass on any more calculations today than I absolutely need to make , but I would also offer that dropping those 1M resistors down to 10k would hurt the ability of the inner loop to keep the input devices running constant current for lowest distortion, etc.
The 5532 IMO is the wrong part for that slot.
JR
My thinking was that with larger Cs the R values would typically be lowered from Bifet-friendly 1M into typical 5532 10K teritory. Say maybe from 1M to 10K as we change C from 0.68 to 47uF. My point was the 1M/0.68 (5%) combo would probably match better CMR-wise than 10K/47uF (+50% -20%) combo.
But I haven't calculated it.[/quote]
I'm going to pass on any more calculations today than I absolutely need to make
, but I would also offer that dropping those 1M resistors down to 10k would hurt the ability of the inner loop to keep the input devices running constant current for lowest distortion, etc. The 5532 IMO is the wrong part for that slot.
JR
mikep
Well-known member
[quote author="bcarso"]
Recommend 2SA1306 for the input devices.[/quote]
really? are those high power, in a TO220 package?
Recommend 2SA1306 for the input devices.[/quote]
really? are those high power, in a TO220 package?
bcarso
Well-known member
[quote author="mikep"][quote author="bcarso"]
Recommend 2SA1306 for the input devices.[/quote]
really? are those high power, in a TO220 package?[/quote]
Yep typo. 2SA1316.
Recommend 2SA1306 for the input devices.[/quote]
really? are those high power, in a TO220 package?[/quote]
Yep typo. 2SA1316.
Similar threads
Latest posts
-
-
-
-
-
-
-
[Build] CAPI BT50, 500 Series EQ (a la 550), Official Support Thread
- Latest: audiophil_studio
-
