ZIP: 223
Title: Reduce Block Target Spacing from 25 to 20 Seconds
Owners: Dev Ojha <[email protected]>
Evan Forbes <[email protected]>
Status: Draft
Category: Consensus
Created: 2026-11-04
License: MIT
Discussions-To: <https://forum.zcashcommunity.com/t/proposal-lower-zcash-block-target-spacing-from-25s-to-20s/59035>
The key words “MUST”, “MUST NOT”, and “SHOULD” in this document are to be interpreted as described in BCP 14 1 when, and only when, they appear in all capitals.
The terms “block chain”, “consensus branch”, and “network upgrade” follow ZIP 200. 2 The character ยง refers to sections of the Zcash Protocol Specification. 3
“Block target spacing” means the interval between blocks targeted by the difficulty adjustment algorithm. Actual block intervals vary.
The “25-second baseline” means the rules after NU7, including ZIP 218’s 25-second target spacing and ZIP 237’s subsidy definitions. 4 5
An “orphan block” is a valid mined block that does not remain on the selected chain. The “orphan rate” is the fraction of all mined blocks that become orphan blocks.
“Reach 90%” means the mean time from a block’s solution until 90% of the nodes have verified it.
MB means 1,000,000 bytes.
This proposal reduces block target spacing from 25 to 20 seconds in a network upgrade after NU7. It builds on ZIP 218 and lowers its per-block shielded action limits and global shielded budget by 20%, from 330 to 264 Orchard-protocol actions and from 300 to 240 Sapling inputs and outputs. 4 6 The block size limit is unchanged.
The change reduces the expected wait for a first confirmation by 20%. Shielded throughput, worst-case light wallet bandwidth, and trial decryptions per day are unchanged, apart from 25% more block headers. It adjusts the difficulty averaging window, issuance parameters, and duration-based node defaults to the new spacing. Shorter inclusion intervals can expose finer transaction timing information.
A 20-second target reduces the expected wait for a first confirmation by five seconds. ZIP 218 chose its shielded action limits to bound worst-case light wallet sync bandwidth and trial decryptions per day. 4 Scaling those limits with the target spacing keeps these bounds, so mobile and other light wallets pay only for the additional block headers. This proposal evaluates that incremental trade-off from ZIP 218’s baseline.
ZIP 218 identifies timing and anchor-depth privacy effects. 4 Shorter blocks give observers finer timing information without changing what individual shielded transactions reveal.
A three-block anchor depth spans an expected 60 seconds instead of 75. A rollback past the anchor invalidates the transaction. A replacement spending the same notes exposes the same nullifiers, linking the two transactions. Ordinary orphan-block measurements do not bound rollback depth.
The smaller per-block action limits do not change the maximum shielded throughput per unit time. Under sustained demand for a pool, ZIP 218’s observation that the limits can delay transactions and make their timing more distinguishable continues to apply. 4
This section applies to the post-NU7 rules of ZIPs 218 and 237. 4 5 Implementations MUST retain all pre-activation rules at heights below the new activation height.
Define the following constants:
\[\mathsf{TwentySecondActivationHeight} : \mathbb{N}\] \[\mathsf{TwentySecondPoWTargetSpacing} : \mathbb{N} := 20 \text{ seconds}\] \[\mathsf{TwentySecondSpacingRatio} : \mathbb{Q}_{>0} := 25/20 = 5/4\]
A separate deployment ZIP will assign \(\mathsf{TwentySecondActivationHeight}\) for each network. It MUST be strictly greater than that network’s \(\mathsf{NU7ActivationHeight}\kern-0.05em\textsf{.}\) No NU7 activation constant changes.
For a height \(h \in \mathbb{N}\) at or above this activation height, define:
\[\mathsf{PoWTargetSpacing}(h) := \mathsf{TwentySecondPoWTargetSpacing}.\]
For \(h < \mathsf{TwentySecondActivationHeight}\kern-0.05em\textsf{,}\) retain ZIP 218’s definition of \(\mathsf{PoWTargetSpacing}\kern-0.05em\textsf{.}\) In particular, \(\mathsf{PostNU7PoWTargetSpacing}\) remains 25 seconds.
The difficulty adjustment algorithm (DAA) averages over 128 blocks after activation, replacing ZIP 218’s 102-block window. The larger block count preserves approximately the same expected averaging duration at the shorter target spacing.
Define:
\[\mathsf{TwentySecondPoWAveragingWindow} : \mathbb{N} := 128.\]
For \(h \geq \mathsf{TwentySecondActivationHeight}\kern-0.05em\textsf{,}\) define:
\[\mathsf{PoWAveragingWindow}(h) := \mathsf{TwentySecondPoWAveragingWindow}.\]
For \(h < \mathsf{TwentySecondActivationHeight}\kern-0.05em\textsf{,}\) retain ZIP 218’s averaging window, including 102 blocks after NU7 activation. All uses of this function in the difficulty algorithm use the height of the block under consideration.
The averaging window includes pre-activation blocks when its lookback crosses the activation height. Implementations MUST NOT reset it at activation. ZIP 218’s difficulty formulae therefore use:
The median-time span, damping factor, adjustment bounds, and proof-of-work limit remain unchanged. Implementations MUST apply the existing difficulty algorithm with these height-dependent inputs from the activation block; there is no separate difficulty reset.
Define:
\[\mathsf{TwentySecondHalvingInterval} : \mathbb{N} := \mathsf{PostNU7HalvingInterval}\cdot 5/4 = 6{,}300{,}000.\]
For the following formula, let \(A_{20} = \mathsf{TwentySecondActivationHeight}\kern-0.05em\textsf{,}\) \(A_7 = \mathsf{NU7ActivationHeight}\kern-0.05em\textsf{,}\) and \(A_B = \mathsf{BlossomActivationHeight}\kern-0.05em\textsf{.}\) These are nonnegative integer heights. For \(h \geq A_{20}\kern-0.05em\textsf{,}\) define:
\[\mathsf{Halving}(h) := \left\lfloor \frac{A_B-\mathsf{SlowStartShift}}{840{,}000} +\frac{A_7-A_B}{1{,}680{,}000} +\frac{A_{20}-A_7}{5{,}040{,}000} +\frac{h-A_{20}}{6{,}300{,}000} \right\rfloor.\]
The denominators are the pre-Blossom, post-Blossom, post-NU7, and 20-second halving intervals, respectively. For \(h < A_{20}\kern-0.05em\textsf{,}\) retain ZIP 218’s halving function, including its slow-start cases. Implementations MUST preserve the fractional progress through the halving interval at activation. They MUST NOT restart an interval at \(A_{20}\kern-0.05em\textsf{.}\)
For \(h \geq A_{20}\kern-0.05em\textsf{,}\) replace ZIP 237’s scheduled subsidy with:
\[\mathsf{ScheduledBlockSubsidy}(h) := \left\lfloor \frac{\mathsf{MaxBlockSubsidy}}{2\cdot3\cdot(5/4)\cdot 2^{\mathsf{Halving}(h)}} \right\rfloor.\]
The factors 2, 3, and \(5/4\) are the Blossom, NU7, and 20-second spacing ratios, respectively. The result is a nonnegative integer number of zatoshi. Implementations MUST use the exact rational ratio \(5/4\) and apply the floor only to the complete expression. Multiplying an already-rounded 25-second subsidy by \(4/5\) can produce a different answer.
For example, at halving index 3 the scheduled subsidy changes from 26,041,666 zatoshi to 20,833,333 zatoshi. Flooring \(26{,}041{,}666\times4/5\) would incorrectly produce 20,833,332 zatoshi.
For \(h \geq A_{20}\kern-0.05em\textsf{,}\) redefine ZIP 237’s \(\mathsf{HalvingInterval}(h)\) to return \(\mathsf{TwentySecondHalvingInterval}\kern-0.05em\textsf{.}\) Retain its earlier cases for \(h<A_{20}\kern-0.05em\textsf{.}\) Its existing fraction formula then gives:
\[\mathsf{BLOCK\_SUBSIDY\_FRACTION}(h) = \frac{\lfloor 6{,}931{,}680{,}000 / 6{,}300{,}000\rfloor}{10^{10}} = \frac{1100}{10^{10}}.\]
Retain ZIP 237’s definition:
\[\mathsf{BlockSubsidy}(h) := \mathsf{ScheduledBlockSubsidy}(h) + \mathsf{AdditionalBlockSubsidy}(h).\]
Implementations MUST carry the existing NSM Value Balance through activation. They MUST NOT reseed it or recompute historical balances under the new spacing. ZIP 237’s additional-subsidy ceiling and balance-update rules continue to apply.
Open specification item: define the NSM reissuance trigger for an activation before reissuance starts. ZIP 259’s fixed trigger heights use the 25-second issuance curve. The amended trigger must use the 20-second curve and reissuance fraction. Its reference reserve must use cumulative scheduled issuance through \(h-1\kern-0.05em\textsf{,}\) retaining each historical era’s integer rounding. If reissuance starts before this upgrade, retain that activated state.
Define the following constants:
\[\mathsf{TwentySecondGlobalShieldedBudget} : \mathbb{N} := 264\] \[\mathsf{TwentySecondOrchardProtocolBlockActionLimit} : \mathbb{N} := 264\] \[\mathsf{TwentySecondSaplingBlockIOLimit} : \mathbb{N} := 240\]
Each value is the corresponding ZIP 218 constant multiplied by \(4/5\kern-0.05em\textsf{,}\) which is exact for all three. 4
For each block at height \(h \geq A_{20}\kern-0.05em\textsf{,}\) ZIP 218’s per-pool limits and global shielded budget MUST be enforced with these constants in place of \(\mathsf{GlobalShieldedBudget}\kern-0.05em\textsf{,}\) \(\mathsf{OrchardProtocolBlockActionLimit}\kern-0.05em\textsf{,}\) and \(\mathsf{SaplingBlockIOLimit}\kern-0.05em\textsf{,}\) respectively. That is:
The definition of shielded cost, the per-pool structure of the limits, and the 2 MB block size limit are unchanged. For blocks at heights \(h < A_{20}\kern-0.05em\textsf{,}\) ZIP 218’s limits of 330, 330, and 300 continue to apply.
A transaction that by itself would violate any of these limits cannot be included in any block at a height \(h \geq A_{20}\kern-0.05em\textsf{.}\) This includes transactions that satisfy ZIP 218’s limits, such as one with 300 Orchard actions.
Nodes SHOULD NOT admit such a transaction to their mempool, or relay it, if the candidate next block’s height is at or above \(A_{20}\kern-0.05em\textsf{.}\) Nodes SHOULD evict such transactions from their mempool when the candidate next block’s height reaches \(A_{20}\kern-0.05em\textsf{.}\) Wallets SHOULD NOT create such a transaction for a target height at or above \(A_{20}\kern-0.05em\textsf{.}\)
Retain ZIP 218’s threshold of \(18 \cdot \mathsf{PoWTargetSpacing}(h)\) seconds,
evaluated with the new spacing. At and after \(A_{20}\kern-0.05em\textsf{,}\) the threshold is
therefore \(18 \times 20 = 360\) seconds, rather than ZIP 218’s 450 seconds.
A block qualifies only if its timestamp is strictly more than this threshold
after the preceding block’s timestamp. Thus a gap of 360 seconds does not
qualify, while a gap of 361 seconds does. In that case the block’s nBits
MUST equal \(\mathsf{ToCompact}(\mathsf{PoWLimit})\kern-0.05em\textsf{,}\) following ZIP 218.
This rule does not affect Mainnet.
Existing timestamp-validity bounds and local-clock tolerances remain unchanged. The Testnet minimum-difficulty exception remains subject to those rules.
When converting a height range to expected elapsed time, implementations SHOULD use the target spacing applicable to each block. The expected time from an accepted tip at height \(a\) to height \(b>a\) is \(\sum_{h=a+1}^{b} \mathsf{PoWTargetSpacing}(h)\kern-0.05em\textsf{,}\) including the new spacing for the activation block. Policies defined as elapsed durations SHOULD retain those durations across activation.
Wallets SHOULD increase the default expiry delta from 120 to 150 blocks. Choose the default using the transaction’s target height, which is the next block height after the chain tip. Use 150 when that height is at or above \(A_{20}\kern-0.05em\textsf{;}\) otherwise retain the prior default. Compute expiry relative to the target height as specified by ZIP 218. Wallets SHOULD preserve an explicit user-configured delta across activation.
Implementations SHOULD use the following post-activation defaults where they expose the corresponding settings. The table scales duration-based values by \(5/4\) relative to ZIP 218. Larger rollback windows conform.
| Constant | 25–second baseline | 20–second recommendation |
|---|---|---|
MAX_REORG_LENGTH |
At least 600 | At least 750 |
MAX_BLOCKS_IN_TRANSIT_PER_PEER |
48 | 60 |
BLOCK_DOWNLOAD_WINDOW |
3072 | 3840 |
MIN_BLOCKS_TO_KEEP |
864 | 1080 |
The consensus coinbase maturity remains 100 blocks. This proposal leaves
TX_EXPIRING_SOON_THRESHOLD at 3 and
NETWORK_UPGRADE_PEER_PREFERENCE_BLOCK_PERIOD at 1728, following ZIP 218.
Wallets SHOULD retain ZIP 315’s recommended anchor depth of 3 blocks. 7 Its expected duration changes from 75 to 60 seconds.
ZIP 218 chose its shielded action limits so that worst-case light wallet sync bandwidth does not increase relative to the pre-NU7 protocol. 4 Retaining those per-block limits at 20 seconds would raise the maximum number of shielded actions per day, and with it worst-case compact sync bandwidth and trial decryptions, by 25%. This proposal instead scales the limits by \(4/5\kern-0.05em\textsf{,}\) so the daily maxima are unchanged. All three ZIP 218 values are multiples of 5, so no rounding is needed.
Using ZIP 218’s compact-field model, which excludes serialization and transport overhead, with \(86{,}400/20 = 4{,}320\) blocks per day:
| Metric | 25–second baseline | 20–second proposal |
|---|---|---|
| Blocks per day | 3,456 | 4,320 |
| Max Orchard + Ironwood actions/block | 330 | 264 |
| Max Sapling IOs/block | 300 | 240 |
| Max Orchard + Ironwood actions/day | 1,140,480 | 1,140,480 |
| Orchard or Ironwood compact BW/day | 168.79 MB | 168.79 MB |
| Sapling compact BW/day | 120.27 MB | 120.27 MB |
| Compact block headers/day | 0.31 MB | 0.39 MB |
| Worst-case total BW/day | 169.10 MB | 169.18 MB |
| Worst-case trial decrypts/day | ~2.3M | ~2.3M |
The binding constraint remains a block filled with Orchard or Ironwood actions: \(264 \times 148 \times 4{,}320 + 90 \times 4{,}320 = 169.18\textsf{ MB/day}\kern-0.05em\textsf{.}\) The only increase is the 25% more compact block headers, approximately 78 kB per day. The trial decryption count, counting both internal and external incoming viewing keys as in ZIP 218, is \(2 \times 264 \times 4{,}320 = 2{,}280{,}960\) per account per day, identical to ZIP 218.
These figures describe worst-case compact sync bandwidth when blocks fill the shielded action budget. They do not establish a daily download requirement for every wallet.
Maximum shielded throughput is also unchanged. Two-action Orchard or Ironwood transactions remain limited to \(\lfloor 264/2 \rfloor / 20 = 6.6\) transactions per second, and four-IO Sapling transactions to \(\lfloor 240/4 \rfloor / 20 = 3.0\) transactions per second, as in ZIP 218. The benefit of this proposal is lower latency, not more shielded capacity.
The worst-case shielded verification work per block falls by 20%, while the worst-case shielded verification work per unit time is approximately unchanged. Batch verification amortizes some cost across a block, so slightly smaller blocks may verify somewhat less efficiently per action. Fixed per-block costs, such as header validation, chain state updates, and propagation round trips, increase by 25% per unit time. Syncing blocks produced during a given time offline requires processing 25% more headers.
The 2 MB block size limit is unchanged, so the maximum transparent data per unit time increases by 25%. This affects full node bandwidth and storage, but not light wallets that download only compact blocks.
ZIP 218 measured orphan rates on a distributed test network with full 2 MB blocks. 4 8 We repeat that methodology with 80 miners of equal hash rate in 12 regions: New York, San Francisco, Toronto, London, Amsterdam, Frankfurt, Bangalore, Singapore, and Sydney. Blocks hold two-action Ironwood transactions up to ZIP 218’s action limit, then transparent transactions up to 2 MB.
With every node running Zakura, full blocks had these orphan rates:
| Spacing | Orphan / mined blocks | Orphan rate (95% interval) | Reach 90% |
|---|---|---|---|
| 25 s | 7 / 356 | 1.97% (0.96โ4.00%) | 0.81 s |
| 20 s | 11 / 389 | 2.83% (1.59โ4.99%) | 0.69 s |
ZIP 218’s devnet measured 4.86% at 25 seconds. The 20-second blocks averaged 245 Ironwood actions, against the proposed limit of 264. Zakura announced each block to 32 peers immediately and to the rest after 500 ms.
To check the test network against Mainnet, we also ran a mixed network at 25 seconds: 56 Zakura nodes with Zakura 1.8.5’s block announcement settings and 24 Zebra 7.0.0 nodes. We chose the current release of each implementation and a Zebra share above Mainnet’s, where Zebra nodes mined about 17% of recent blocks. ZIP 218 models the orphan rate as \(0.55 \times\) reach 90% \(/ T\kern-0.05em\textsf{.}\) With small blocks, the test network, the model, and Mainnet agree:
| Network | Mean block size | Orphan rate | ZIP 218 model |
|---|---|---|---|
| Mixed test network | 66 kB | 0.66% (2 / 303) | 0.87% |
| Mainnet, since the NU7 activation day | 18 kB | 0.60โ0.86% | |
| Mixed test network | 2 MB | 3.78% (13 / 344) | 3.7% |
The Mainnet rates come from two observers and are lower bounds. 9
With full blocks, the mixed network had nearly twice the orphan rate of the Zakura-only network at 25 seconds. Part of that difference comes from Zakura 1.8.5 announcing each block to every peer at once, which Zakura is changing. Our runs with Zebra Andale found two other causes. Zebra relays a verified block to a third of its peers. Zebra takes 19 ms to deserialize a 2 MB block, and Zakura takes 2.5 ms. Before full blocks are common, other node implementations, notably Zebra Andale, need these relay optimizations.
ZIP 218’s 102-block window at 25 seconds spans an expected 2,550 seconds. The equivalent count at 20 seconds is 127.5 blocks. This proposal rounds up to 128, giving 2,560 seconds, about 0.39% longer.
This proposal preserves the averaging duration because mining hash rate can vary. Reducing that duration would change the DAA’s response to hash rate changes as well as changing block target spacing.
The interval change preserves its expected duration:
\[6{,}300{,}000\times20=5{,}040{,}000\times25\text{ seconds}.\]
The halving formula retains progress through the 25-second era. Scaling the unrounded scheduled subsidy by \(4/5\) preserves the expected issuance rate before rounding. Integer rounding changes the realized schedule.
The NSM fraction changes from \(1375/10^{10}\) to \(1100/10^{10}\kern-0.05em\textsf{,}\) also a factor of \(4/5\kern-0.05em\textsf{.}\) ZIP 237’s ceiling and evolving balance mean this does not establish exact equality of reissuance per unit of wall-clock time.
Open specification item: amend the funding-stream end heights and address period calculations in ZIPs 207 and 214 for a second spacing transition. 10 11 Preserve fractional progress through the active address period and define integer-boundary rounding. Keep ended streams ended and ensure recipient lists cover every extended period. The deployment height and schedule amendments must be settled together; changing NU7 constants retroactively is invalid.
Deploy this proposal in a separate network upgrade after NU7. ZIP 259 defines NU7’s deployment; this proposal does not replace that deployment. 6 A separate deployment ZIP will assign this upgrade’s activation heights and consensus branch IDs.
Before this ZIP advances to Proposed status, resolve the funding-stream and NSM reissuance-height specification items. Complete the activation and rounding analysis against the selected deployment parameters.