have you also tried the AD8620 ? It has something for it (buffer, filter)... but for I/V I liked the opa1612 a lot for an I/V task ...
looking forward your review about this new one over the 1656...a very good one for ears indeed.
Last edited:
Some years have passed since this post and now the OPA1633 which seems to have in general outstanding performance, but would it be suitable for I/V input front end for MC cart at 5uA nominal input bias current?
OPA211 is an stunning op-amp for current application, with it's low input bias high BW and low noise coupled with very low distortion. Unfortunately, not fully balanced.
OPA1637 seems a great choice but the GBW at 40dB is low. As a front end at 40dB, the opamp would see less than 1Vpp so the supply range for a suitable amp wouldn't need to be 36V indeed.
Are there more choices in 2024?
Jan, I also don't understand this, the THP210 is identical to OPA1637 same BW and same slew rate. Were you referring to a different part? The post above shows your measurements on the eval board for the 1637.
Coincidentally, I recently built a test board with SMD components to evaluate the OPA1632 and new OPA1633.
Together with a test board for high GB product OP-AMP, which is difficult to operate stably with DIP conversion boards and TH components.
The evaluation has not yet been completed at present, but the input conversion noise in the audio band has been measured with the following circuit,
OPA1632 1.76nV/√Hz
OPA1633 1.20nV/√Hz
The measurement method was to measure the noise voltage at the output through a 20 kHz steep LPF, divide by the noise gain (=201) and bandwidth √20k, then subtract the contribution of the 10 Ω thermal noise (0.4 nV√Hz) due to R1.
The old 1632 value is slightly higher than would be expected from the datasheet graph, but the new 1633 is a 3 dB improvement over the 1632. Currently, the highest specification for ordinary OP-AMPs is 0.9 nV/√Hz (AD797, LT1028 and others), except for special ones like the LMH6629 . Considering that two of them in a balanced configuration will give around 1.3 nV/√Hz, the OPA1633 seems to be one of the best choices, provided that no discrete components are used. SSM2019 and others are also possibilities, but I don't know as I don't own one.
In addition, in the above circuit, the voltage at both ends of R1 (10 Ω) was measured to be
OPA1632 70 µV
OPA1633 83µV
This is the data of one sample each, so it is only for reference, but I think it is the DC voltage applied to the cart in the case of direct connection.
Last edited:
Thank you for this contribution. OPA1633 has really impressive specs considering this is a balance OPAMP and having 2 of differential low noise OPAMP would bring to a total higher total noise than 1633.
My application is MC front end, I am concerned about the rather high input bias current of this opamp.
If it needs 5uA to properly bias, what happens when the cart is playing a low passage and delivering maybe 3-4uA!? Shouldn't we employee an opamp with much lower input bias for this application?
Question: why do you have OPA1611 cascading? Is it there to increase the gain for better measurements? Or do you have another preamp in front to help you capture the low noise reading?
Lastly, by DC voltage applied to the cartridge in case of direct connection, do you mean if the cart is connected between one of the diff inputs and ground there is a small DC offset?
My application is MC front end, I am concerned about the rather high input bias current of this opamp.
If it needs 5uA to properly bias, what happens when the cart is playing a low passage and delivering maybe 3-4uA!? Shouldn't we employee an opamp with much lower input bias for this application?
Question: why do you have OPA1611 cascading? Is it there to increase the gain for better measurements? Or do you have another preamp in front to help you capture the low noise reading?
Lastly, by DC voltage applied to the cartridge in case of direct connection, do you mean if the cart is connected between one of the diff inputs and ground there is a small DC offset?
Most of the input bias current is supplied from the output via the feedback resistors (R2, R3 in my schematic). The only current flowing into the cart should be the input offset voltage mentioned earlier (voltage applied to R1)/(cart winding resistance).
Also, if the SSM2019 etc. is used in a transimpedance circuit, i.e. I/V, it will be used with an emitter input (base common circuit), in which case a relatively large current may flow during transient conditions when the power supply is switched on and off, but in the case of a fully differential amplifier it is the base input. However, this is not a problem with fully differential amplifiers, as they use the base input.
My voltmeter does not support balanced inputs. Therefore, a 1x BAL to UNBAL conversion circuit is required. Because the front-end gain is set to a large 201x. which overwhelms the noise in the BAL-UNBAL conversion circuit.
Even if one side is grounded or completely floating, the slight current mentioned above will still flow.
thank you again. I am not sure about most of the input current will be supplied by the feedback resistor. This is I/V current is supplied to and from the feedback path to generate proper output voltage. if you do kirkoff at the node you will have the input current is injected and will split into input biasing and output path. This can constitute a problem as I mentioned earlier. You got me thinking with the feedback path I will think over it longer.
Another thing that got me thinking when reflecting on cartridges, is that a generator current has near infinite output impedance, in this case the cartridge has, let's say, 10ohm like in the case of your R1 the transimpedance amplifier is really configured as non-inverting.
Regarding input offset voltage/current is unavoidable with any circuitry OP/AMP or discrete, but op amp with uV input offset and low input current, cannot possibly consitute a problem for any cartridge. Granted we have to minimize that by choosing a suitable opamp, this is why I am struggling in this case, which op amp is mostly suited for MC application.
I have ordered 1637 and 1633 along with 211 to connect two of them, I want to attempt making measurements on different configurations.
Another thing that got me thinking when reflecting on cartridges, is that a generator current has near infinite output impedance, in this case the cartridge has, let's say, 10ohm like in the case of your R1 the transimpedance amplifier is really configured as non-inverting.
Regarding input offset voltage/current is unavoidable with any circuitry OP/AMP or discrete, but op amp with uV input offset and low input current, cannot possibly consitute a problem for any cartridge. Granted we have to minimize that by choosing a suitable opamp, this is why I am struggling in this case, which op amp is mostly suited for MC application.
I have ordered 1637 and 1633 along with 211 to connect two of them, I want to attempt making measurements on different configurations.
Curiosity, what did you use to measure the output noise? I am not sure even AP can clearly measure down to 1nV/sqrtHz without using a low noise preamp one front.
I have tried to clarify the measurement circuit and method so that such questions do not arise.
It would be impossible even for the AP to directly measure noise around 1 nV/√Hz.
Again, the noise gain at the front end is multiplied by 201, so if the input refer noise density is 1nV/, the noise density at the output is measured as 201nV/√Hz.
The noise on my noise voltmeter (which I made myself) is 0.749uV measured through a 20kHz LPF. The noise density is therefore 5.3 nV/√Hz, which is sufficient to measure 201 nV/√Hz.
Incidentally, the measured noise voltage at the output of the measurement circuit (with 20 kHz LPF) is
OPA1632 51.3 uV
OPA1633 35.9 µV
*When measuring with the noise gain of the OP-AMP increased in this way, it should be noted that
If an OP-AMP with a small GB product is set to a large noise gain, the frequency response may already be attenuated at the upper end of the measurement band.
In this case, the measurement will be made with a value less than the actual noise.
For example, if an OP-AMP with a GB product of 3 MHz (e.g. TL071) is set to a gain of 100 times, a -3 dB attenuation at 30 kHz is expected and there is already non-negligible attenuation up to 20 kHz, resulting in a measurement with a value less than the actual noise.
Thanks for clarifying. I have just received OPA1633 and built an extremely small protoboard to test the performance of the OPA as I/V.
I have one question: if you short the input R1=0ohm, shat happens at the output of the OPA if you put a scope on it?
In my case if I have 10ohm at the input I have +2mV at OUT+ and -2mV at OUT-, but if I short the input, I have a very large offset at the output.
I have one question: if you short the input R1=0ohm, shat happens at the output of the OPA if you put a scope on it?
In my case if I have 10ohm at the input I have +2mV at OUT+ and -2mV at OUT-, but if I short the input, I have a very large offset at the output.
Last edited:
R1, R2 and R3 set the gain, Setting R1 to 0 sets infinite gain, so the output saturates because the input offset is not zero. With 1k/10R/1k the gain is 201, so an input offset of 20uV will give an output offset of 4mV - this agrees well with the input offset spread histogram in the datasheet - 20uV would be pretty normal for one of these.
Thank you Mark. You make total sense with your observation and it was my same exact thought minus considering the actual 20u offset that would explain the 4mV output.
With this in mind, if I wanted to use this as I/V for the cartridge, let’s take as example Lyra Atlas Lambda SL which has a whopping 1.5ohm impedance we get an output offset if about 3.5V!! How can any MC cartridge be even considered as a generator current and be coupled with a transimpedance stage without incurring into this problem that has been daunting me for a while now trying to solve?
How can I null the input offset given 1633 doesn’t have a NULL pin?
With this in mind, if I wanted to use this as I/V for the cartridge, let’s take as example Lyra Atlas Lambda SL which has a whopping 1.5ohm impedance we get an output offset if about 3.5V!! How can any MC cartridge be even considered as a generator current and be coupled with a transimpedance stage without incurring into this problem that has been daunting me for a while now trying to solve?
How can I null the input offset given 1633 doesn’t have a NULL pin?
Shorting R1 results in no feedback, so it is natural that this should happen. Mine will also do so.
Even then, the internal common mode feedback (feedback to make the two outputs symmetrical around Vcom) will still work, so one will swing heavily + and the other -. In the case of MC input, the gain is so large that it will be a blast.
Take care.
Ah, Mark already answered that.
Even then, the internal common mode feedback (feedback to make the two outputs symmetrical around Vcom) will still work, so one will swing heavily + and the other -. In the case of MC input, the gain is so large that it will be a blast.
Take care.
Ah, Mark already answered that.
Yes Mason, the reason why there is differential voltage at the output is very clear.
My observation is how can an MC device be considered a current source and therefore operate purely on an I/V stage!? Short answer is no it is not. It will convert the transimpedance gain stage to a non-inverting confifuration instead with its low resistance.
Operating the non inverting configuration with an MC cartridge is really wild given the large variation in phono cartridges resistance, that can vary by 10-50X, thus the total gain.
Also the risk of shorting the inputs and having a large DC offset at your speakers that will destroy them is very dangerous.
On the contrary I do see the benefits in terms of noise because of the very low input impedance of the amplifier.
I wonder what manufacturers of current phonos do consider a current stage and how they deal with the aforementioned issues.
My observation is how can an MC device be considered a current source and therefore operate purely on an I/V stage!? Short answer is no it is not. It will convert the transimpedance gain stage to a non-inverting confifuration instead with its low resistance.
Operating the non inverting configuration with an MC cartridge is really wild given the large variation in phono cartridges resistance, that can vary by 10-50X, thus the total gain.
Also the risk of shorting the inputs and having a large DC offset at your speakers that will destroy them is very dangerous.
On the contrary I do see the benefits in terms of noise because of the very low input impedance of the amplifier.
I wonder what manufacturers of current phonos do consider a current stage and how they deal with the aforementioned issues.
Yes, one of the advantages of the transimpedance system is that the output voltage tends to be almost identical even if the cart changes.
(To be honest, I am sceptical about transimpedance phono inputs and balanced input systems. I am not evaluating the OPA1632/1633 with a phono front end in mind).
It seems to me that phono using RCA, XLR or other connectors has a much higher chance of opening up compared to the chance of shorting out. (Unless the input is a FET, if the input is directly connected, whether it is a transimpedance amplifier or not, it will still produce significant noise when open.)
Yes and no, not because the input impedance is lower, but because the transimpedance method is advantageous in that the cart coil doubles as the Rg of the feedback resistor, so there is no thermal noise for the Rg.
I don't think many manufacturers use the transimpedance method, but recently some have adopted it. They are prohibitively expensive.
I do not know what method those manufacturers are dealing with.
However, with regard to the input offset, from the following,
In my previous post I mentioned that the voltage applied to R1 is 70µV with the OPA1632. This is precisely +70µV with reference to -IN. The output offset of the OPA1632 was then approximately -14mV with reference to -OUT.
In other words, an offset of -gain times the input offset appears at the output. This was also the case for the OPA1633.
By adding a DC servo circuit that suppresses the output offset to almost 0 V, as shown in the circuit diagram below, it would automatically reduce the input offset to negligibly close to 0 as well. (Thought only, no experimentation)
Any further discussion may not be appropriate for this thread. Throwing out opinions and questions on Analogue Sources might catch the attention of many members and get the opinion of the wise. On the other hand, you might get confused by posts that don't fit the context.
*The polarity of the DC servo was reversed in the schematic of the first submission. The schematic has been corrected.
Last edited:
Thanks for your response. Maybe I will open a different thread time permitted on the location you suggested.
I will have to think about your answer and forumalte a new thread, in the meantime thanks.
Below some measurements on OPA1633. Output distortion (quite high), 40dB gain (diff input diff output) Rs=40ohm Rf=2K. CF=220pF.
And noise profile diff gain of the circuit 200 to SE output, provides 1.2nV/sqrtHz
I have it on a protoboard, I am sure there is external noise pick up and I am also picking up 60Hz and Multiples.
Does the distortion profile look correct? I would have expected a lower distortion at 40dB.
I will have to think about your answer and forumalte a new thread, in the meantime thanks.
Below some measurements on OPA1633. Output distortion (quite high), 40dB gain (diff input diff output) Rs=40ohm Rf=2K. CF=220pF.
And noise profile diff gain of the circuit 200 to SE output, provides 1.2nV/sqrtHz
I have it on a protoboard, I am sure there is external noise pick up and I am also picking up 60Hz and Multiples.
Does the distortion profile look correct? I would have expected a lower distortion at 40dB.
The same circuit was set up and measured. However, my measuring instrument does not support balanced, so it is a single-ended input. Because of the 100x gain, the in-phase component is negligible and should have almost no effect.
The balanced to unbalanced conversion circuit has a gain of 0.5x. This is because single-ended outputs above 10Vrms are clipped. Because of the high gain of the preceding stage, distortion and noise in this circuit are negligible.
10Vrms is measured on the differential output of the OPA1633. In practice, the 5Vrms output of the OPA1611 is measured.
At 1kHz output 10Vrms, THD+N = 0.00089% (BW = 100kHz), almost only 3rdHD.
Through a 20 kHz LPF, the noise was reduced to THD+N = 0.000773%.
The noise is calculated to be 138nV/√Hz, from the data sheet 1.1nV/√Hz @10kHz and thermal noise in 40Ω, a noise gain of 101x (signal gain is 100x but noise gain is 101x). At least at 10 kHz, this would be around 150 nV/√Hz.
Your noise data is a bit questionable, but I was a bit concerned when I looked at the datasheet, but again the 1/f knee is high and there seems to be a lot of low-frequency noise, and given that the RIAA raises the low frequencies, it may not be suitable for MC phono frontend.
The balanced to unbalanced conversion circuit has a gain of 0.5x. This is because single-ended outputs above 10Vrms are clipped. Because of the high gain of the preceding stage, distortion and noise in this circuit are negligible.
10Vrms is measured on the differential output of the OPA1633. In practice, the 5Vrms output of the OPA1611 is measured.
At 1kHz output 10Vrms, THD+N = 0.00089% (BW = 100kHz), almost only 3rdHD.
Through a 20 kHz LPF, the noise was reduced to THD+N = 0.000773%.
The noise is calculated to be 138nV/√Hz, from the data sheet 1.1nV/√Hz @10kHz and thermal noise in 40Ω, a noise gain of 101x (signal gain is 100x but noise gain is 101x). At least at 10 kHz, this would be around 150 nV/√Hz.
Your noise data is a bit questionable, but I was a bit concerned when I looked at the datasheet, but again the 1/f knee is high and there seems to be a lot of low-frequency noise, and given that the RIAA raises the low frequencies, it may not be suitable for MC phono frontend.
I want to first off all thank you so much for all the measurements you have performed for me. I deeply have appreciated it. Thank you!
I am going to analyze your response in great detail and respond better later.
Your observation about the 1/f noise is extremely on point for MC application.
You make me think the AD797 might be more suited for this application, I can use two of them for a fully differential operation.
Is the 1.1nV/sqrtHz noise figure in the datasheet for the 1633 SE or BAL? I don’t see it being specified.
Lastly, as MC fist 40dB stage, the output will work in the 50-200mV at 1KHz so the noise figure is what is going to determine the TDH+N so probably on my proto I am picking up excessive noise this why I have that large distortion at lower levels.
I am going to design a prototype with few options to test and measure so once I receive it I will be able to post results.
I am going to analyze your response in great detail and respond better later.
Your observation about the 1/f noise is extremely on point for MC application.
You make me think the AD797 might be more suited for this application, I can use two of them for a fully differential operation.
Is the 1.1nV/sqrtHz noise figure in the datasheet for the 1633 SE or BAL? I don’t see it being specified.
Lastly, as MC fist 40dB stage, the output will work in the 50-200mV at 1KHz so the noise figure is what is going to determine the TDH+N so probably on my proto I am picking up excessive noise this why I have that large distortion at lower levels.
I am going to design a prototype with few options to test and measure so once I receive it I will be able to post results.
Surely its a differential amp however its used, so the differential input pair determine the input-referred noise the same either way?
You can find it listed here.
It's not listed separately in SE or differential, but as far as I've measured the noise in #24's circuit, there is no change when R1 10Ω is split into 5Ω + 5Ω and the midpoint is connected to GND.
If you already have one, the AD797 is a good.
Here is the noise data of an OP-AMPs with a spec of 0.9 nV/√Hz that I measured previously.
They are all much the same, except for already discontinued LME49990.