MAX5318 vs DAC9881 vs AD5780 or AD5781...

[whippoorwill11]

Is anyone familiar with Maxim\'s MAX5318 18-bit DAC please? I hardly ever see it used but it\'s significantly cheaper than competing chips from Texas Instruments or Analog Devices.

Judging from its datasheet alone it\'s not quite as good; but, it seems to me, not by enough to account for the price difference. I\'m wondering if there\'s some other issue with the MAX5318 which I should be wary of. I\'d especially like to hear from anyone who\'s applied its digital gain and offset: did they cause any linearity problems?

(Please save the \"Never buy Maxim\" meme for some other thread. Here I\'m interested only in technical details. Thanks.)

 

[server]

On Sun, 1 Nov 2020 13:26:33 -0800 (PST), whippoorwill11
<whippoorwill11@gmail.com> wrote:

>Is anyone familiar with Maxim\'s MAX5318 18-bit DAC please? I hardly ever see it used but it\'s significantly cheaper than competing chips from Texas Instruments or Analog Devices.

If you don\'t need speed, take a look at DAC1220. 20 bits monotonic, 15
PPM linear.

Judging from its datasheet alone it\'s not quite as good; but, it seems to me, not by enough to account for the price difference. I\'m wondering if there\'s some other issue with the MAX5318 which I should be wary of. I\'d especially like to hear from anyone who\'s applied its digital gain and offset: did they cause any linearity problems?

(Please save the \"Never buy Maxim\" meme for some other thread. Here I\'m interested only in technical details. Thanks.)

Well, proceed at your own risk.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[whippoorwill11]

If you don\'t need speed, take a look at DAC1220. 20 bits monotonic, 15
PPM linear.

Thanks John, and for the quick response. I\'ll look into it.

 

[Flyguy]

On Sunday, November 1, 2020 at 1:26:39 PM UTC-8, whippoorwill11 wrote:

Is anyone familiar with Maxim\'s MAX5318 18-bit DAC please? I hardly ever see it used but it\'s significantly cheaper than competing chips from Texas Instruments or Analog Devices.

Judging from its datasheet alone it\'s not quite as good; but, it seems to me, not by enough to account for the price difference. I\'m wondering if there\'s some other issue with the MAX5318 which I should be wary of. I\'d especially like to hear from anyone who\'s applied its digital gain and offset: did they cause any linearity problems?

(Please save the \"Never buy Maxim\" meme for some other thread. Here I\'m interested only in technical details. Thanks.)

The AD5780 has significantly better noise spectral density (8 vs 24 nv/root hz) and differential nonlinearity error (1 vs 2 lsb) than the MAX5318 but is similar in other respects.

 

[whippoorwill11]

> The AD5780 has significantly better noise spectral density (8 vs 24 nv/root hz) and differential nonlinearity error (1 vs 2 lsb) than the MAX5318 but is similar in other respects.

Thanks, yes the AD5780\'s a nice chip all right. I\'ve looked hard at it but this application\'s a bit too cost-sensitive. I have another in mind for the future which will need that extra performance - perhaps then.

 

[server]

On Tue, 3 Nov 2020 00:24:22 -0800 (PST), whippoorwill11
<whippoorwill11@gmail.com> wrote:

The AD5780 has significantly better noise spectral density (8 vs 24 nv/root hz) and differential nonlinearity error (1 vs 2 lsb) than the MAX5318 but is similar in other respects.

Thanks, yes the AD5780\'s a nice chip all right. I\'ve looked hard at it but this application\'s a bit too cost-sensitive. I have another in mind for the future which will need that extra performance - perhaps then.

Can you tell us more about the requirements? There are lots of ways
one could make a cheap 20-bit DAC.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[whippoorwill11]

Can you tell us more about the requirements? There are lots of ways
one could make a cheap 20-bit DAC.

Not really relevant to this thread TBH John but thanks for your interest. Eighteen bits is a sweet spot: the controlled element can\'t resolve to 20 bits so the extra resolution would be wasted anyway. It\'s for precise control of a DC signal, with a very long interval between updates. Main considerations are linearity, stability, temperature drift and, unfortunately, cost. The MAX5318 appears to be the best all-around candidate; I just didn\'t want to choose it then find that the reason it\'s cheaper yet not widely used is some awful technical shortcoming known to everyone but me :-D

 

[John Larkin]

On Tue, 3 Nov 2020 15:14:55 -0800 (PST), whippoorwill11
<whippoorwill11@gmail.com> wrote:

Can you tell us more about the requirements? There are lots of ways
one could make a cheap 20-bit DAC.

Not really relevant to this thread TBH John but thanks for your interest. Eighteen bits is a sweet spot: the controlled element can\'t resolve to 20 bits so the extra resolution would be wasted anyway. It\'s for precise control of a DC signal, with a very long interval between updates. Main considerations are linearity, stability, temperature drift and, unfortunately, cost. The MAX5318 appears to be the best all-around candidate; I just didn\'t want to choose it then find that the reason it\'s cheaper yet not widely used is some awful technical shortcoming known to everyone but me :-D

I was thinking that a bad DAC, or even an attempt at a bad DAC, could
be disciplined by a delta-sigma ADC. D-S ADCs are phenomenal and
cheap.

We pay $4.33 for an ADS1246, a pretty good 24-bit ADC with an internal
voltage reference.

 

[whippoorwill11]

On Wednesday, 4 November 2020 at 13:13:17 UTC+13, John Larkin wrote:

I was thinking that a bad DAC, or even an attempt at a bad DAC, could
be disciplined by a delta-sigma ADC. D-S ADCs are phenomenal and
cheap.

Spooky - we may be on parallel tracks I\'d left space on the board for a higher-resolution ADC in case the results justify reading the DAC, to discipline it as you suggest. I\'m confident though that an R-2R on its own, like the ones discussed so far, is the right starting point. After all, for precision work at DC the likes of TI and AD suggest only those, strings or MDACs. I\'d use an MDAC here, just as I\'d use a better-quality R-2R - the linearity of an 18-bit string DAC isn\'t good enough - but cost is the issue.

 

[Flyguy]

On Tuesday, November 3, 2020 at 5:18:14 PM UTC-8, whippoorwill11 wrote:

On Wednesday, 4 November 2020 at 13:13:17 UTC+13, John Larkin wrote:
I was thinking that a bad DAC, or even an attempt at a bad DAC, could
be disciplined by a delta-sigma ADC. D-S ADCs are phenomenal and
cheap.

Spooky - we may be on parallel tracks I\'d left space on the board for a higher-resolution ADC in case the results justify reading the DAC, to discipline it as you suggest. I\'m confident though that an R-2R on its own, like the ones discussed so far, is the right starting point. After all, for precision work at DC the likes of TI and AD suggest only those, strings or MDACs. I\'d use an MDAC here, just as I\'d use a better-quality R-2R - the linearity of an 18-bit string DAC isn\'t good enough - but cost is the issue.

It seems like you are paying for a fast settling time (usec) when you don\'t need it. As John suggests, you could servo control a cheap DAC to get the resolution and linearity you need.

 

[server]

On Tue, 3 Nov 2020 17:18:06 -0800 (PST), whippoorwill11
<whippoorwill11@gmail.com> wrote:

On Wednesday, 4 November 2020 at 13:13:17 UTC+13, John Larkin wrote:
I was thinking that a bad DAC, or even an attempt at a bad DAC, could
be disciplined by a delta-sigma ADC. D-S ADCs are phenomenal and
cheap.

Spooky - we may be on parallel tracks I\'d left space on the board for a higher-resolution ADC in case the results justify reading the DAC, to discipline it as you suggest. I\'m confident though that an R-2R on its own, like the ones discussed so far, is the right starting point. After all, for precision work at DC the likes of TI and AD suggest only those, strings or MDACs. I\'d use an MDAC here, just as I\'d use a better-quality R-2R - the linearity of an 18-bit string DAC isn\'t good enough - but cost is the issue.

24 bit audio DACs are dirt cheap. Close a loop around one of them with
a delta-sigma ADC. Total about $6, which includes a fairly good
reference.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[whippoorwill11]

On Wednesday, 4 November 2020 at 19:51:46 UTC+13, Flyguy wrote:
> It seems like you are paying for a fast settling time (usec) when you don\'t need it. As John suggests, you could servo control a cheap DAC to get the resolution and linearity you need.

In true newsgroup fashion, this thread hasn\'t seen an answer to its topic and is instead being dragged further and further away, by posts based on uneducated guesses about the OP\'s needs and goals. So this is it from me.

Sure I\'m getting a faster settling time than I need - so what? Settling time\'s your consideration not mine; it\'s nothing I\'ve decided to pay for. I\'m also getting a better noise figure than I need - and I\'m no more interested in that than in settling time. They come with the package, that\'s all, like an unnecessary paint job on a decent car.

Good luck with servoing a cheap - or, for that matter, expensive - DAC to get better resolution, no matter how tight the resolution of the ADC you monitor it with. If your hi-res ADC tells you your 18-bit DAC needs adjusting, you get to adjust it by one LSB out of 18 bits and no better. Anything else is a Cloud Cuckoo Land attempt to create DAC bits out of thin air. What you gain from servoing is the ability to measure what the DAC\'s really doing and correct it. You need a better ADC so your measurements aren\'t meaningless, not because it can magically impart its higher resolution to the DAC.

What that costs you is expense and complexity (both of which may be worth paying) and, in this application, one or two other performance factors which are more important and which I\'ve said before I\'m not going into. If making a poor, cheap DAC adequately mimic an expensive precision one were as simple in every application as adding a cheap ADC, the combination would long ago have killed the market for standalone precision DACs. For some reason AD, TI, Maxim and others still think they have a point and sell millions of them to suckers like me. But hey, warn them that you know better; maybe they\'ll pay you a fortune not to blow the lid on their nice little earner.

 

[server]

On Wed, 4 Nov 2020 12:27:07 -0800 (PST), whippoorwill11
<whippoorwill11@gmail.com> wrote:

On Wednesday, 4 November 2020 at 19:51:46 UTC+13, Flyguy wrote:
It seems like you are paying for a fast settling time (usec) when you don\'t need it. As John suggests, you could servo control a cheap DAC to get the resolution and linearity you need.

In true newsgroup fashion, this thread hasn\'t seen an answer to its topic and is instead being dragged further and further away, by posts based on uneducated guesses about the OP\'s needs and goals. So this is it from me.

Aren\'t you the OP?

Sure I\'m getting a faster settling time than I need - so what? Settling time\'s your consideration not mine; it\'s nothing I\'ve decided to pay for. I\'m also getting a better noise figure than I need - and I\'m no more interested in that than in settling time. They come with the package, that\'s all, like an unnecessary paint job on a decent car.

Good luck with servoing a cheap - or, for that matter, expensive - DAC to get better resolution, no matter how tight the resolution of the ADC you monitor it with. If your hi-res ADC tells you your 18-bit DAC needs adjusting, you get to adjust it by one LSB out of 18 bits and no better. Anything else is a Cloud Cuckoo Land attempt to create DAC bits out of thin air. What you gain from servoing is the ability to measure what the DAC\'s really doing and correct it. You need a better ADC so your measurements aren\'t meaningless, not because it can magically impart its higher resolution to the DAC.

Dacs are commonly dithered to get sub-LSB resolution, and 24 bit audio
DACs are cheap anyhow.

You\'re sure in a bad mood. I suppose you could consider paying for
engineering.

What that costs you is expense and complexity (both of which may be worth paying) and, in this application, one or two other performance factors which are more important and which I\'ve said before I\'m not going into. If making a poor, cheap DAC adequately mimic an expensive precision one were as simple in every application as adding a cheap ADC, the combination would long ago have killed the market for standalone precision DACs. For some reason AD, TI, Maxim and others still think they have a point and sell millions of them to suckers like me. But hey, warn them that you know better; maybe they\'ll pay you a fortune not to blow the lid on their nice little earner.

Bad attitude. Bye.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Lasse Langwadt Christensen]

onsdag den 4. november 2020 kl. 21.36.06 UTC+1 skrev jla...@highlandsniptechnology.com:

On Wed, 4 Nov 2020 12:27:07 -0800 (PST), whippoorwill11
whippoorwill11@gmail.com> wrote:

On Wednesday, 4 November 2020 at 19:51:46 UTC+13, Flyguy wrote:
It seems like you are paying for a fast settling time (usec) when you don\'t need it. As John suggests, you could servo control a cheap DAC to get the resolution and linearity you need.

In true newsgroup fashion, this thread hasn\'t seen an answer to its topic and is instead being dragged further and further away, by posts based on uneducated guesses about the OP\'s needs and goals. So this is it from me.

Aren\'t you the OP?


Sure I\'m getting a faster settling time than I need - so what? Settling time\'s your consideration not mine; it\'s nothing I\'ve decided to pay for. I\'m also getting a better noise figure than I need - and I\'m no more interested in that than in settling time. They come with the package, that\'s all, like an unnecessary paint job on a decent car.

Good luck with servoing a cheap - or, for that matter, expensive - DAC to get better resolution, no matter how tight the resolution of the ADC you monitor it with. If your hi-res ADC tells you your 18-bit DAC needs adjusting, you get to adjust it by one LSB out of 18 bits and no better. Anything else is a Cloud Cuckoo Land attempt to create DAC bits out of thin air. What you gain from servoing is the ability to measure what the DAC\'s really doing and correct it. You need a better ADC so your measurements aren\'t meaningless, not because it can magically impart its higher resolution to the DAC.

Dacs are commonly dithered to get sub-LSB resolution, and 24 bit audio
DACs are cheap anyhow.

and they usually work by delta-sigma modulating only a few discrete levels

 

[Flyguy]

On Wednesday, November 4, 2020 at 12:27:15 PM UTC-8, whippoorwill11 wrote:

On Wednesday, 4 November 2020 at 19:51:46 UTC+13, Flyguy wrote:
It seems like you are paying for a fast settling time (usec) when you don\'t need it. As John suggests, you could servo control a cheap DAC to get the resolution and linearity you need.

In true newsgroup fashion, this thread hasn\'t seen an answer to its topic and is instead being dragged further and further away, by posts based on uneducated guesses about the OP\'s needs and goals. So this is it from me.

Sure I\'m getting a faster settling time than I need - so what? Settling time\'s your consideration not mine; it\'s nothing I\'ve decided to pay for. I\'m also getting a better noise figure than I need - and I\'m no more interested in that than in settling time. They come with the package, that\'s all, like an unnecessary paint job on a decent car.

Good luck with servoing a cheap - or, for that matter, expensive - DAC to get better resolution, no matter how tight the resolution of the ADC you monitor it with. If your hi-res ADC tells you your 18-bit DAC needs adjusting, you get to adjust it by one LSB out of 18 bits and no better. Anything else is a Cloud Cuckoo Land attempt to create DAC bits out of thin air. What you gain from servoing is the ability to measure what the DAC\'s really doing and correct it. You need a better ADC so your measurements aren\'t meaningless, not because it can magically impart its higher resolution to the DAC..

What that costs you is expense and complexity (both of which may be worth paying) and, in this application, one or two other performance factors which are more important and which I\'ve said before I\'m not going into. If making a poor, cheap DAC adequately mimic an expensive precision one were as simple in every application as adding a cheap ADC, the combination would long ago have killed the market for standalone precision DACs. For some reason AD, TI, Maxim and others still think they have a point and sell millions of them to suckers like me. But hey, warn them that you know better; maybe they\'ll pay you a fortune not to blow the lid on their nice little earner.

I am sorry for having wasted my time answering a question for such an ungrateful asshole such as yourself - it WON\'T happen again!