I/V stage for ES9028Q2M

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,
I am trying to design an I/V stage for ES9028Q2M DAC. I was first considering the Zen style, using a common gate amplifier to present low impedance to the DAC. But soon I have encountered problems:
40C40u9kELgV84epUoePi0mIPhOEW5oObHcGVb8I5dzibZ0wL06r9Qg3GnhrG5Y77unCsrzyL88Ff9EV_sTlMIzXD1D0bfACPzxITfMSh8i7mgxn9rhurHZR4hIzn8O6GQsoKMGpmWm8KkDPd3KTfdClM5Gq-Z945MYp5XbysklMCGYaPKLruEcvRkAk4OIcci69d-mVGOwOcr1AQy4sqP80yveeeWFYzGAOyjUbilCtiNIP4WIYKcX9Jjqwpeq8oJ9mCoE6GdMBgGExluH0yvnTWi8qKMZb6T0wjV_JKlAwdmgToO1xsEKPV2HUZ4QDNgSDOlwoCz9bNZaMW2d2GmfT43WgBvdAxiZlscbMBVSxpjEu6CiJLMi-95fysJ4vB9TEwe7E23b3hdNFRJPQDoyGozJllN0hwo8CJIXIZTFzZXEqseB17TxS1DQy6gcepxfdwfIfdKajC4dp181RjAf9vj-ypcfRkts1YCp1_hrylP0sXHENLm8ea12oyy4d3zXwf9lH-B-qPdEkCU5DgyXv6dt5mCXNsZ_DQcsXOc_Utr1bOUQsglZ8YIXX93h1WaJDFIRiXRwurLm-h70fNHY0eH1ncGyEKb7DH6sJRWoSHir5Va7vH6QeyVfanxHre8XLew8cHIa16WUKhtgBmtCotdH5YVTG=w1640-h866-no


First, in order to really get low input impedance, the transistor needs to be biased to saturation region, this requires a certain amount of Vds (> Vgs-Vth), assume 4V. In addition, the bias current should be much larger than the DAC output current to make the DAC output a 'small signal' (keep the bias point stable). The ES9028Q2M can put 7mA out, the bias current of the common gate should be around 70mA. With 150R resistor, the voltage drop is 10.5V. The third one is the DAC output is NOT 0V, it is AVCC/2, 1.65V in my case. Thus the input common mode of the common gate should match this voltage. A resistor is not possible since the resistance will be limited and this will draw some of the DAC output currents to ground. An LDO may be the only solution. This further push the supply of the amplifier to as high as 16V. I would rather NOT to have such a high voltage rail around a mainly 3.3V circuit, let alone the required 2X LDOs for each channel.

Here is my question, why don't people use low noise high bandwidth fully differential amplifier from ADI or TI to implement the I/V stage? Like following:

K0vrfK7L-EI1DPVdPgvtwN2CQWV0Jzh5PYeQjJ6g4MTb6mV4bljVP4TXuanwOEf5Tn7OTMG9ijWfPWHcwZiGoHm8EAMENUaMOlDLfJNZKIpsLCj9JtcCwETjAwBhGYG2xiR_W4prNTY_byia6rwXqVSbX6u6zRFksBk2MzWThee_Qy-PvTYrr-YorDDViR_2aacDXhFh4XykoRLyzo7JS_FXAtX_q9sSTtP6u6EjPdKpY-Xm-0ijdvIl-XxdRwJd51OTsUODk5h_kVWMWnn7tNdhyVuVekdHcE0HafxzYiNMw2GywKiDvikECPsYz9xiilxSplJogAGfMDntQ9-yrOlipkUSfMeO7OY7yZ_6H6AOTxsoJzIIBAMdtClIxge4MTttswkJ49LtLAZK5u5Ov2DyfD5X3iMB8-qHg-naYQ8MTkRplbBEBH_5QSZ6CGvahzQxLiY6_yr6pX2U4GUc-TTvFgT9KNwBLZ18UYDa2cdolo1jw8URcIxEqfLYOOmOO7NGgmlq3x6AcszaOA5-_JJqhkUERTruCMGZoXEsifqvibNrguIXa2cEfRJIGAf50ODw9k2lgeAosjWBT6EZSXgOevBdH8BwnDyX8ALb_Z8D4dfBlSJFJNi4oP9baxAZPhML1_klhGj7nQFkHuqQAPBKyM7Qq7f5=w1215-h1100-no

For example, AD8138 from ADI is a nice amplifier. It has a 300MHz+ bandwidth, 1000+ V/us slew rate, very low noise, very low distortion, works with 3.3V single rail and adjustable input/output common mode. What is the drawback to using such an amplifier to implement the I/V stage?
 
From a quick look at the datasheet, the 'very low noise' description only applies above the audio band. There does look to be rather a high noise corner frequency. In comparison to a discrete I/V stage, a disadvantage for an IC would be its classAB output stage, leading to noisy power supplies. Incidentally, to avoid the need for higher voltage rails you can adopt current source biassing for the common-gate transistor. This would allow only the signal current to pass through the I/V resistor rather than the whole of your 70mA.
 
From a quick look at the datasheet, the 'very low noise' description only applies above the audio band. There does look to be rather a high noise corner frequency. In comparison to a discrete I/V stage, a disadvantage for an IC would be its classAB output stage, leading to noisy power supplies. Incidentally, to avoid the need for higher voltage rails you can adopt current source biassing for the common-gate transistor. This would allow only the signal current to pass through the I/V resistor rather than the whole of your 70mA.

I still need an opamp as I want to keep the supply to 3.3V across the board. The entire supply would now enough to provide enough Vds for most MOSFET. After reviewing more opamps, I would like to use LT1819 instead the ADI one. It has 400MHz bandwidth, 2500V/us slew rate. Noise wise, 6nV/sqrt(Hz) after corner frequency and ~40nV/sqrt(Hz) at 100Hz, should be able to claim low noise this time.

Tentative schematic:
1EMOthJUwmkGaKRNkcYMIU1SaP8CZKKEnWD-CV8QW4Iis_pEr3Pn7mQxVGsjOk4hf7b9A2lP-TwJjSua2gqXqcvWONiCCYtaJW5JX_h0dCxWV6PbZE1TQ1KQbyYscPtK4ha1mndczK98zCBJOWDsxttJrkxGpiutdEIwfC6PA6WLkIlA-G-Lkj2nmlrFlWibOO-KwrzuQjSTktmdtDcX1xDrbPwCCzVyMd8_mWkf2a7E6wmSzuV5LbMWmiRpIuXp0CFPWJKLTVoi3P1ZbLYCUHczWuFCpkjFxDqzc61MfUf-hyCTyDMHjHOcxuWbS1nDZ02wB8qGwzMb_sltAQEm3z9SSW66SyZ-pf95rmwwZMBxty2C6Gl5Zl3t1pxKe_Id-dZIUUNN6nhUwlKVowBBHDnmdDANUCk-6-GUckaY9CKhbBpTJMcVWGeWfLTF4CdR4KPXoqG2tRh0Sm9S8_zthXTjqDzh7zAIFnpRM6HsVVsXQREqMtYmnhBbKpQ-G0YFz39ZGFkPH_tUPb75chPmm3HkJqpM02Cvklx7Q72xszvnMIoUrEO1kjqLHUdQV60G_xxlSZTMjPeDEMlbgZ0Gj8QXvfNjS08BV2sBjy0=w921-h853-no


Transient simulation with 7mAp-p current input @1KHz:
vSDaLim_mn2ElIVG2IVRpnI6TZZ7x-evmhidulh15ZUDb9qsEPUnMv0csccQ_Ahs6tQJQmpQVhCzoNTgJ-3yDoSlXzaNk9qffRrLt4QA46kbPrJxP0tzZH2eg-EGhdUcfpoq5DUSnAvmB0txB7RiF2CqRpYAcDWfrcbGfWi83WA01W_Gh4Ejwt5d_tgpAIXdJrgy5X-5O8EdfsEZFRj3no_DKnEdjXFtuSZWs0K47xZTUF2g7hZAY3Ve-YhTrVMNNzc8aNTHAVoTYzXfB1eWkmxDikFs-Ym0E4VVVU8LqCv8Wg4gbN1LBqpZMF1UL35nviFmz4SAx3jJTNHb6JFk2J5XWWVxi9FWPH9E5rKr8yZfA1rdZIZWBIJs9EyjoGuJjSgy29eT7MpXe3v1tQFdnqiLGbix49UetkxDWt601n059zKIOFsemNkR8eJ5OVy6ZDBskKZb6q2p44s0XDaY4_CfW5aYRf2l5KnmG2N196Q2oRCuk_LL7Hhb30d5hSAmFxGQOyP98FvcZl2oUA8juOrMrlJ5FMLDiVB0oYiVU4_fdHnT2piX10bqmc1b4jGGq2BVeGPSBuCp_QLusPvkoxC5aoTgv2D8jxmq19o=w673-h412-no


AC simulation (Unity gain stable!!):
l2HYNQ52SHgdeVpPLanm7eeC7o_xJdHA0IQ6OWcQYhBdmprMFypZeWb_txKgC4lM3XsrltO6IfSWUWtTDFv_yKKB2iONtOFDqdFNOGb-GWEn0J4XMiruoGlxEX7zes6z20Zo9BEV3Uta_IShOT50yo-A6RODwSMsbJHyMnq1TEMoclozIBbjo5cqeeCuLM6QE4YsL0Fr5DXCHkrp0gBAoJRw36--J7jK0BOoVCmzA_4wAiuQqYqxH4rLTOqT0DrwyM-lLqSd1KKW31Iersu8LQoZKpucTkETzpwG5wH3CyJOgmqmSXIukVdMTrHr7qKNhF7zoQoXNaub1ieEolWaAW5isHxRFFW78-XTFC-L9XjHgGDzibfrqsOpMwVWYE6ZFwct208jH31qvNsof_6T35eRJuf0j-i5tuwK7C-3pbriWtQCmoEc-8FM7R58Wn15BLTHsEa0jg1OHSi32yOIlbCp3510YcfCxArXueE2u2OL1SU1uGjovUyhO7hRuYNy9-Voo2CjL3EACQ5cy4e1HItDmpyFgG0BcHRYJidsGOs8LyiM_BPgbbeTihCCkRwS78uRq8w1OPVt3_6-t7LOMUIe2LhQpIr_91VQS9w=w847-h474-no
 
Gandolf, go look for something like the 9018 or 9016 Datasheet online, the spec design is essentially a LME49720 configured as I/V's, pretty much classic design.

Also, the parts you're suggesting aren't low noise at audio bandwidth, maybe something like the OPA1632, but feel free to simply use a dual opamp instead.

It's also not difficult to make a center-tap reference from a precision resistor divider and a capacitor (for stability). If you look for JensH's posts in one of the 9038q2m threads, he's got a circuit. Scott Wurcer threw in a nice general form (and some things to think about) for an I/V in the AD797 http://www.analog.com/media/en/technical-documentation/data-sheets/AD797.pdf pg 17.

There's a bevy of ways to do this. A quick/but not ridiculously quick low-noise general purpose opamp would be excellent. If you want to go nuts, look at an OPA1612 or better ADA4898. Mind your layout though. :)
 
Gandolf, go look for something like the 9018 or 9016 Datasheet online, the spec design is essentially a LME49720 configured as I/V's, pretty much classic design.

Also, the parts you're suggesting aren't low noise at audio bandwidth, maybe something like the OPA1632, but feel free to simply use a dual opamp instead.

It's also not difficult to make a center-tap reference from a precision resistor divider and a capacitor (for stability). If you look for JensH's posts in one of the 9038q2m threads, he's got a circuit. Scott Wurcer threw in a nice general form (and some things to think about) for an I/V in the AD797 http://www.analog.com/media/en/technical-documentation/data-sheets/AD797.pdf pg 17.

There's a bevy of ways to do this. A quick/but not ridiculously quick low-noise general purpose opamp would be excellent. If you want to go nuts, look at an OPA1612 or better ADA4898. Mind your layout though. :)

Thanks for the information. I do understand AD8138 is kind of noisy, but even LT1819 is not quiet enough? OPA1612's spec is indeed very impressive, but the high supply voltage requirement and the cost are holding me back. OPA1632 does not have the unity gain stable feature that I am also looking for. What about OPA1662? Its noise performance is kind of in between, I am not sure if its 17V/us slew rate is too slow.

By the way, I was trying to search for the posts by JensH but couldn't find it. Would you mind post it here? Thank you very much.
 
Chinese ES9018K2M I2S DAC

And the posts nearby will give you something to chew on.

LT1819 comes in a funky package--make sure you can solder it. Cannot see what it brings to the table that couldn't better be solved by slower precision opamp (cough, LME49720?). There's absolutely no need for 400 MHz bandwidth, and you're sacrificing elsewhere to get it. What sort of output are you looking for? You can readily run these guys down towards their minimum voltage, although some of their performance may be (marginally) compromised. Also, I'd think even the more conservative members would consider a 200 kHz power bandwidth (back out your slew rate) a sufficient overprovisioning.
 
Thanks for the information. I do understand AD8138 is kind of noisy, but even LT1819 is not quiet enough? OPA1612's spec is indeed very impressive, but the high supply voltage requirement and the cost are holding me back. OPA1632 does not have the unity gain stable feature that I am also looking for. What about OPA1662? Its noise performance is kind of in between, I am not sure if its 17V/us slew rate is too slow.

By the way, I was trying to search for the posts by JensH but couldn't find it. Would you mind post it here? Thank you very much.

Check out ADA4896/ADA4897 if you are still hung up on bandwidth. It can't handle more than 10V on the supply but since you want to run 3.3 it should be fine.

LT1819 is still noisier than the best choices. It's even noisier than good JFET input amps like OPA1642 and OPA827.

Also, I didn't look too closely at those pics you posted, but make sure you are using resistor values that the op-amp can actually drive.
 
Chinese ES9018K2M I2S DAC

And the posts nearby will give you something to chew on.

LT1819 comes in a funky package--make sure you can solder it. Cannot see what it brings to the table that couldn't better be solved by slower precision opamp (cough, LME49720?). There's absolutely no need for 400 MHz bandwidth, and you're sacrificing elsewhere to get it. What sort of output are you looking for? You can readily run these guys down towards their minimum voltage, although some of their performance may be (marginally) compromised. Also, I'd think even the more conservative members would consider a 200 kHz power bandwidth (back out your slew rate) a sufficient overprovisioning.

Ok, I see your point. Soldering wise there is no problem... I have a small Chinese made reflow oven, successfully soldered QFN before. The reason why I am in favor of LT1819 is that it is very stable, 90-degree phase margin all the way. LME49720 and OPA1612 requires > 5V supply so they are out of question. OPA1662 works with 3.3V, but its phase margin is merely 58 degree, makes me nervous. chris719 suggests ADA4896, it seems like a good candidate. Man, they are EXPENSIVE.
 
Check out ADA4896/ADA4897 if you are still hung up on bandwidth. It can't handle more than 10V on the supply but since you want to run 3.3 it should be fine.

LT1819 is still noisier than the best choices. It's even noisier than good JFET input amps like OPA1642 and OPA827.

Also, I didn't look too closely at those pics you posted, but make sure you are using resistor values that the op-amp can actually drive.

Yeap ADA4896 looks very promising.

Simulation setup:
MCAylTWyZDmtwcN70BdVi9zRlarVJEKNJhGyQDRAOC4trLdT0kVnnDGW6QBKJQgGtjo4pPufZ0s8HhzEsVJ-EU7nxiVLfjXnmIF60lp-hx579MgXabjl0TAHWvkkVhEjyPf6JbWnIgeqvN5srmmTa_dSV2gJiGWQg6FNM_UZR4vwPOQw3VtYfVnLNbvjsOjqdf3rsnqm3O3JGBUHEJZVVSFIlks2SMxHc3HfmhHutUDdfx7h1m-xWMyWUmExy0RZia-aB3a0riJ7mqcpl_CQh62e5YREdR_H_x9XHwLi6332JqJiqUxOab1q4ztdmpkePRipLNs1VmeE5N-KU8S5Twh2y5wZXXu6kqada6INvfgec2oWIZ371TMxCz9eqyYdcB4q-O6i8bv1KijfiZM7i2VI50himktO92k5LhdxTnfheVVaeA8AStYqAx9NTACtw8FP7hE5kVUrKLd2A4qnIZhQWNFfQbe3A0vtpYsukkctpxwgbWezpk10roc3G24ArscbkwhR1Banqk5hQW_8MYHis84JbAOqW2DqvyJG-c4Wak7hj5TxGh8L1jCqibJc1QYGKXM3XY-wrem-_JXwXAx7xP-6_5CD0GIsf04=w659-h357-no


Transfer function:
Xo4Hal0F_HZPyDwB4NBpQIOnFDi8nN2aDsNJiYexE5nEdvoT-euER59sUNIbzMqEICgCFHM36TBaelNMNuZG4Mm1gmbOCyrfSRuUPbLHkXl8w3_wq85Ljpar05jDq1oqf4BYiFUnLn1M0nTiwXOoL7hdYTTI6s1NCp9BtLSA5AwTkK7xzF-ReG0z7JLiOPmDuWrMRZ-xKT1OC7Jr7j8YhrO_mkXNJw4ITGj75mzHS0ld3JaTL_IWDQbzRfifHlQshGv4L22SX8gclHoORs-pXJ1sKshVlHcu9SoLUz6bfpqfWcbHJdZyuCU7_iXev_lgBzd2NDd5u1j9TzeMhv64nYtOWVFyBbvnsLTpXQRVX0K9ak_aJ_AANVwdNUU0XSRiekqr7f4gEY67A4NUm-2a6fHRw_LRULnNPgCOdVA0LxPq1BtS-9k6OQ6mR8jaDg3kafXjXA_ZB9yUU1mBtwNQFgR8KfyVfsTgdWropNKFRsSv-SGEjBM7DlM2GTQwBkqzsyg1puW6UdL_wgeGbrfy3riwYbLVXnaRebOX8ccqVmKCtYTiukD_Q__TFCoPXgRNd_IiW3tz6D7Phm1Z6QAtd8us6BJwMjI68ulxrPw=w979-h593-no


Transient with 7mAp-p @10KHz:
pdBAbAIrGWyfqLWVleM9-AkB4T_axLSti5NXjub2KTejnmsYUtw52iDd9OY6QOLpE_azcxC6NfeRCQ3CwjoJ1Mp9JIGqgeXuwfe8jZ9270LX-GEPWHSOMiG6sWCWjNSVpm4NqVHLvsB8dbrSR3s7ZNmlOu0C1dNkscEX58bE7IAN0QWUnj7kvPcAf-0Ag4cvJoJ8uHAmM2ZIau3Fp8Zsg9G9lneMd34BFVQk17OM68eEQEXygbKpPpaQxZs1-LHu0efJszsxu-ddNJY2bxXUlOCCSV1G4_OLD6ll6BZ_Pg6gHpugGRIDBs82QVfSKhLHn0S8xXUoAO3xLUlOeGtUJxBSW4b7omY1FTCIbNKWpF8PNo5r9V-8pIGApexNirvheQg1Mui0gYrZoICbCXqYDh6gyatA_JMjf6M28kdomB4OFHGlYNF86Z1bVvlxAJYO0npRBT95GFF0WUxLw2bzUAvNKoGV4SyvD_jfGbZ1Es3SJHF5ArOQCv7n86166vWSxSaU-O5yoLpwTpGl51cl9uzLuJDyfanPhpyooSMCwdlKZcUAZ9c9cdnQG2v0AOtfom7uorzYxBnbd8f4AFG8P9dhDmBG0aaBQZvZbUo=w1072-h593-no
 
Try to remove R2 and set C1 to some 100 pF.
Can you give the output impedance of the DAC?

No, it does not work, phase margin becomes negative. R2 is needed to limit the feed-forward current from input to output at high frequency. Otherwise, the RHP zero will severely degrade the phase margin.

I am sorry, the details of the DAC is protected by the NDA.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.